Oral Microbiota Composition and Antimicrobial Antibody Response in Patients with Recurrent Aphthous Stomatitis

Recurrent aphthous stomatitis (RAS) is the most common disease of the oral mucosa, and it has been recently associated with bacterial and fungal dysbiosis. To study this link further, we investigated microbial shifts during RAS manifestation at an ulcer site, in its surroundings, and at an unaffected site, compared with healed mucosa in RAS patients and healthy controls. We sampled microbes from five distinct sites in the oral cavity. The one site with the most pronounced differences in microbial alpha and beta diversity between RAS patients and healthy controls was the lower labial mucosa. Detailed analysis of this particular oral site revealed strict association of the genus Selenomonas with healed mucosa of RAS patients, whereas the class Clostridia and genera Lachnoanaerobaculum, Cardiobacterium, Leptotrichia, and Fusobacterium were associated with the presence of an active ulcer. Furthermore, active ulcers were dominated by Malassezia, which were negatively correlated with Streptococcus and Haemophilus and positively correlated with Porphyromonas species. In addition, RAS patients showed increased serum levels of IgG against Mogibacterium timidum compared with healthy controls. Our study demonstrates that the composition of bacteria and fungi colonizing healthy oral mucosa is changed in active RAS ulcers, and that this alteration persists to some extent even after the ulcer is healed.


Humoral Response to Bacterial and Fungal Antigens
Pure culture of Candida albicans was isolated from stool samples and determined by paired-end Sanger sequencing of ITS1 (100% identical in 551 bp with GenBank KP131671.1). C. albicans was cultured for 4 days at 37 • C under aerobic conditions on Malt extract agar made with 4% malt extract (Sladovna Bruntal, Bruntal, Czech Republic) and 1.5% agar (Dr. Kulich Pharma, Hradec Kralove, Czech Republic). Candida tropicalis was obtained from The National Institute of Public Health (NIPH, Prague, Czech Republic); it was isolated from bronchoalveolar lavage and determined by auxanogram [15].
After cultivation, bacterial and fungal cultures were centrifuged (13,500 rmp, 20 min, 4 • C), pellets were collected and washed three times with distilled water. Washed bacterial cultures were kept on ice, and afterwards subjected to pressure of 1500 psig using a French Press machine (SLM Instruments FA-078 French Pressure Cell Press, American Laboratory Trading, East Lyme, CT, USA). The lysing procedure for bacteria was repeated three times as previously described [16]. Washed fungal cultures were lysed by bead-beating four times in Lysing Matrix 210 Y tubes using FastPrep homogenizer for 60 s at 6.5 m/s (both MP Biomedicals, Irvine, CA, USA). All lysates were freeze-dried in a freeze-drier (LYOVAC GT 2, Leybold Heraeus, Cologne, Germany) and stored in aliquots at −28 • C.

ELISA
For ELISA, plates (96F NuncTM MaxiSorpTM, Thermo Fisher Scientific, Waltham, MA, USA) were coated overnight with lysates at concentrations of 1 g/L for bacteria and 1 mg/L for fungi. Plates were thoroughly washed with a wash buffer (1×PBS with 0.05% Tween 20) after every incubation step. Subsequently, each well was blocked with 1% BSA in PBS for 1 h. After washing, patients' sera were applied for 2 h (dilution 1:200 for bacteria; and 1:800 for antifungal and 1:3200 for antifungal IgG. Subsequently, anti-IgG and anti-IgA secondary antibodies in a dilution of 1:2000 and 1:3000, respectively, were applied for 2 h (for bacteria, anti-Hu IgG-HRP antibody γ chain and anti-Hu IgA α chain (AFF) PEROX; both from The Binding Site Ltd., Birmingham, UK; and for fungi, Peroxidase AffiniPure F(ab') Fragment Goat Anti-Human IgG and Peroxidase AffiniPure F(ab') Fragment Goat Anti-Human Serum IgA, α chain specific; both from Jackson ImmunoResearch Europe Ltd., Cambridge House, St. Thomas Place, UK). Finally, a substrate solution was added (3, 3 , 5, 5 -tetramethylbenzidine in citrate buffer with H 2 O 2 ; Merck, Darmstadt, Germany) and the plates were incubated for 5 min in the dark. Optical density was measured at 450 nm and 650 nm on a spectrophotometer (Multiskan Ascent Plate Reader, Lab Systems, Labsystems Diagnostics Oy, Vantaa, Finland). To compare the levels, one reference serum was applied as a reference on each ELISA plate and its mean value of OD (450-650 nm) was used as an arbitrary unit (1AU).

DNA Extraction
Extraction of total DNA was done using the DNeasy PowerBiofilm Kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. Thawed swab samples were thoroughly vortexed, nipped off swabs were aseptically removed, and samples were centrifuged for 10 min at 13,000 rpm. The precipitate was used for DNA isolation, including homogenization of the sample with beads (6.5 m/s, 1 min, Fast Prep, MP Biomedicals, Irvine, CA, USA) and omitting the incubation of the bead tubes at 65 • C for 5 min.

Sequencing Data Analysis
The amplicon sequencing data were processed with SEED v2.1 [17]. Paired-end reads were joined using fastq-join [18]. In fungal data, the ITS region was extracted using ITSx [19]. Chimeric sequences were detected using the UCHIME algorithm, deleted and clustered using UPARSE at a 97% similarity level in USEARCH 8.1.1861 [20]. The most abundant sequences were chosen as one representative strain per cluster. The 8,451,285 sequences of bacterial 16S rRNA were clustered into 2,488 different operational taxonomic units (OTUs), determined at a 97% similarity threshold after excluding singletons. Taxonomic classification was based on the bacterial 16S rRNA database eHOMD, version 15.1 (doi: 10.1128/mSystems.00187-18). Fungal read clustering was based on 1,648,130 sequences and resulted in 1780 OTUs at a 97% similarity threshold after excluding singletons. Fungal taxonomy was assigned using the UNITE Community (2017) [21]. Raw demultiplexed sequencing data, with sample annotations, were deposited in the NCBI database under the access code PRJNA521448. Singletons were excluded from all analyses. Alpha diversity within samples was determined using richness (number of OTUs in the sample), Chao1 and Shannon indices, and evenness after subsampling to the same sequence depth in each sample. Beta diversity, or the differences between microbial communities, were ordinated and plotted using principal coordinate analysis (PCoA). Statistical significance was assessed using Permutational Multivariate Analysis of Variance (PERMANOVA) and Analysis of Similarities (ANOSIM). To determine the discriminative features for bacterial and fungal taxonomic profiles of communities, the LEfSe analysis tool was employed (unique identifier OMICS_07818) [22].

Bacteria-Fungi Correlation
We explored co-occurrence of bacterial and fungal OTUs using the Markov Chain Monte Carlo (MCMC) simulation model implemented in the R package BaNOCC (Schwager et al. 2017). The analyses were conducted separately for the Active, Passive, and Control cohorts. Our initial simulation for all OTUs exhibited frequent spurious correlations. This was caused particularly by rare OTUs. Consequently, only OTUs represented by at least 1% of reads in at least three samples were included in the final simulations to resolve this issue. The final simulations were run using eight independent MCMC chains, where the number of iterations was set at 500, with burn-in period and thinning interval set at 100 and 10 steps, respectively. Model convergence was checked using the Rhat statistic. Correlation coefficients deviating from zero (p < 0.05) were visualized using heatmaps.

Sampling Site and Disease Status Have an Effect on the Observed Microbiota Diversity
The oral cavity harbors many distinct microbial communities [14]. This diversity stems from the oral cavity heterogeneity and the interrelationships between its different anatomic structures [23]. Thus, we focused on the overall composition of oral microbiota isolated from five distinct sites in the oral cavity, namely, lower labial mucosa (LL), lower jaw mucosa (LM), tongue (T), upper jaw mucosa (UM), and buccal mucosa (CH). To detect possible interference of different niche competitors, we focused on both bacterial and fungal communities.
Bacterial alpha diversity at all five sites of oral cavity together was characterized by significantly more OTUs (Number of OTUs) with a higher, yet not significantly so, number of rare OTUs (Chao1) in Active and Passive RAS cohorts over healthy controls. Abundance and even distribution of species (Shannon index), as well as Evenness index alone, were similar in both cohorts ( Figure S1). When we looked at the sampling sites alone, we found that the sampling site (LL, LM, T, UM, CH) and the disease status of an individual, e.g., Active or Passive RAS, healthy control, both have a meaningful effect on the observed beta diversity of bacterial communities ( Figure S2). Fungal alpha diversity indices did not reveal any significant differences between Active RAS, Passive RAS, or healthy controls at all five sampled sites of oral cavity ( Figure S3). Fungal beta diversity was, however, strongly impacted by the sampling site, similarly to bacterial diversity ( Figure S4).

Descriptive Analysis of Bacterial Microbiota in the Oral Cavity of RAS Patients with Special Emphasis on the Lower Labial Mucosa
The most pronounced differences between healthy controls and Active and Passive patients were found in the lower labial mucosa (LL); therefore, we further focused on the LL in our detailed analysis. RAS patients were found to have an increased alpha diversity over healthy controls, not only at the ulceration site, but also in other tested areas of the oral cavity, i.e., around the ulcer and at the contralateral unaffected site ( Figure 1A). Beta diversity in both the Active and Passive RAS patients was significantly more similar to each other than to healthy controls ( Figure 1B). Taxonomic profile of the bacterial community revealed the highest abundance of Streptococcus oralis in healthy controls, further decreasing in Passive and Active RAS patients ( Figure 1C); the same finding was obtained also in other sampled areas of oral cavity (see Figure S2). LEfSe analysis focused on disease status unveiled discriminative association of the genera Lachnoanaerobaculum, Cardiobacterium, Leptotrichia, Fusobacterium, and the class Clostridia with Active RAS patients, whereas the genus Selenomonas was significantly associated with Passive RAS patients ( Figure 1D).
Microorganisms 2019, 7, 636 6 of 14 ( Figure S2). Fungal alpha diversity indices did not reveal any significant differences between Active RAS, Passive RAS, or healthy controls at all five sampled sites of oral cavity ( Figure S3). Fungal beta diversity was, however, strongly impacted by the sampling site, similarly to bacterial diversity ( Figure S4).

Descriptive Analysis of Bacterial Microbiota in the Oral Cavity of RAS Patients with Special Emphasis on the Lower Labial Mucosa
The most pronounced differences between healthy controls and Active and Passive patients were found in the lower labial mucosa (LL); therefore, we further focused on the LL in our detailed analysis. RAS patients were found to have an increased alpha diversity over healthy controls, not only at the ulceration site, but also in other tested areas of the oral cavity, i.e., around the ulcer and at the contralateral unaffected site ( Figure 1A). Beta diversity in both the Active and Passive RAS patients was significantly more similar to each other than to healthy controls ( Figure 1B). Taxonomic profile of the bacterial community revealed the highest abundance of Streptococcus oralis in healthy controls, further decreasing in Passive and Active RAS patients ( Figure 1C); the same finding was obtained also in other sampled areas of oral cavity (see Figure S2). LEfSe analysis focused on disease status unveiled discriminative association of the genera Lachnoanaerobaculum, Cardiobacterium, Leptotrichia, Fusobacterium, and the class Clostridia with Active RAS patients, whereas the genus Selenomonas was significantly associated with Passive RAS patients ( Figure 1D). Similarly to the analysis of bacterial composition, we evaluated the distribution of fungal communities in samples from the LL. Fungal taxonomic profile at the species level revealed that Cladosporium sp. dominated in healthy controls and in the Passive cohort, where it was accompanied by Itersonilia sp., whereas Malassezia sp. and Candida albicans were highly enriched in the Active RAS cohort (Figure 2A). LEfSe analysis additionally found Malassezia spp. to be significantly associated with the Active cohort. Furthermore, the species Itersonilia and Candida tropicalis were found to be discriminative features for Passive RAS patients; however, they were not observed in the whole cohort, but rather in a few individuals ( Figure 2B,C).

Descriptive Analysis of Fungal Microbiota in the Oral Cavity of RAS Patients with Special Emphasis on the Lower Labial Mucosa
Similarly to the analysis of bacterial composition, we evaluated the distribution of fungal communities in samples from the LL. Fungal taxonomic profile at the species level revealed that Cladosporium sp. dominated in healthy controls and in the Passive cohort, where it was accompanied by Itersonilia sp., whereas Malassezia sp. and Candida albicans were highly enriched in the Active RAS cohort (Figure 2A). LEfSe analysis additionally found Malassezia spp. to be significantly associated with the Active cohort. Furthermore, the species Itersonilia and Candida tropicalis were found to be discriminative features for Passive RAS patients; however, they were not observed in the whole cohort, but rather in a few individuals ( Figure 2B,C).

Bacterial and Fungal Correlation Networks in RAS
We investigated co-occurence/co-avoidance between bacterial and fungal OTUs in the oral cavity of RAS patients and healthy controls. Highly supported correlations were identified for OTUs of the bacteria Porphyromonas, Streptococcus, Leptotrichia, and Haemophilus, interacting with the fungi Malassezia and Rhodotorula. Using microbial and fungal databases, the OTUs were identified as:

Bacterial and Fungal Correlation Networks in RAS
We investigated co-occurence/co-avoidance between bacterial and fungal OTUs in the oral cavity of RAS patients and healthy controls. Highly supported correlations were identified for OTUs of the bacteria Porphyromonas, Streptococcus, Leptotrichia, and Haemophilus, interacting with the fungi Malassezia and Rhodotorula. Using microbial and fungal databases, the OTUs were identified as: OTU00011 Porphyromonas pasteri; OTU00001 Streptococcus oralis; OTU00005 Streptococcus vestibularis; OTU00024 Leptotrichia HMT; OTU00002 Haemophilus prainfluenzae; OTU0004 Malassezia sp.; OTU0016 Malassezia globosa; OTU0025 Rhodotorula mucilaginosa. Active ulcers were dominated by Malassezia spp., one of which was found to have a negative correlation with Haemophilus and a positive correlation with Porphyromonas OTUs. The other Malassezia had a negative correlation with Streptococcus in active ulcers, but not in healed mucosa. However, a negative correlation of this particular Malassezia OTU with a Haemophilus OTU was observed in healed mucosa. Rhodotorula had negative associations with both Strepococcus OTUs, Leptotrichia and Haemophilus in the Active cohort. None of the observed associations were found in healthy controls (Figure 3). OTU00011 Porphyromonas pasteri; OTU00001 Streptococcus oralis; OTU00005 Streptococcus vestibularis; OTU00024 Leptotrichia HMT; OTU00002 Haemophilus prainfluenzae; OTU0004 Malassezia sp.; OTU0016 Malassezia globosa; OTU0025 Rhodotorula mucilaginosa. Active ulcers were dominated by Malassezia spp., one of which was found to have a negative correlation with Haemophilus and a positive correlation with Porphyromonas OTUs. The other Malassezia had a negative correlation with Streptococcus in active ulcers, but not in healed mucosa. However, a negative correlation of this particular Malassezia OTU with a Haemophilus OTU was observed in healed mucosa. Rhodotorula had negative associations with both Strepococcus OTUs, Leptotrichia and Haemophilus in the Active cohort. None of the observed associations were found in healthy controls (Figure 3). Color intensity shows the magnitude of the association; the darker the color, the stronger the association. Correlations that did not deviate from zero (p < 0.05) are indicated by white color. Statistical significance is based on Markov Chain Monte Carlo sampling.

Serum Levels of IgG Against Mogibacterium timidum are Elevated in Patients with RAS
To analyze humoral response to bacteria and fungi, we measured serum levels of antibodies against bacteria Acinetobacter johnsonii, Fusobacterium nucleatum, and Mogibacterium timidum and fungi Candida albicans and Candida tropicalis in patients with Active or Passive stage of RAS and in healthy controls. In the sera from patients in both the Active and Passive stage of RAS, we found significantly increased levels of IgG against M. timidum over healthy controls (p = 0.015 and 0.003, respectively) ( Figure 4A). RAS patients did not significantly differ from healthy controls in levels of IgG against A. johnsonii and F. nucleatum or against C. albicans and C. tropicalis, although abundance of C. tropicalis was increased in RAS patients ( Figure 2B). Serum levels of specific IgA against any of the tested microbes did not show any significant changes. We only observed a tendency towards an increased anti-C. albicans IgA response in the Control cohort, and the opposite trend in IgA response against all tested bacteria and C. tropicalis in the same cohort ( Figure 4B). Color intensity shows the magnitude of the association; the darker the color, the stronger the association. Correlations that did not deviate from zero (p < 0.05) are indicated by white color. Statistical significance is based on Markov Chain Monte Carlo sampling.

Serum Levels of IgG Against Mogibacterium timidum are Elevated in Patients with RAS
To analyze humoral response to bacteria and fungi, we measured serum levels of antibodies against bacteria Acinetobacter johnsonii, Fusobacterium nucleatum, and Mogibacterium timidum and fungi Candida albicans and Candida tropicalis in patients with Active or Passive stage of RAS and in healthy controls. In the sera from patients in both the Active and Passive stage of RAS, we found significantly increased levels of IgG against M. timidum over healthy controls (p = 0.015 and 0.003, respectively) ( Figure 4A). RAS patients did not significantly differ from healthy controls in levels of IgG against A. johnsonii and F. nucleatum or against C. albicans and C. tropicalis, although abundance of C. tropicalis was increased in RAS patients ( Figure 2B). Serum levels of specific IgA against any of the tested microbes did not show any significant changes. We only observed a tendency towards an increased anti-C. albicans IgA response in the Control cohort, and the opposite trend in IgA response against all tested bacteria and C. tropicalis in the same cohort ( Figure 4B).

Discussion
To date, only a few studies have examined the specifics of microbiota composition in patients with recurrent aphthous stomatitis [6,7,24]. In agreement with these previous studies, we observed that microbiota in RAS patients differs from microbiota found in healthy individuals. We sampled patients' microbiota either from minor RAS lesions or from healed mucosa at five different sites of the oral cavity. Lower labial mucosa showed the most pronounced differences between Active or Passive RAS patients and healthy controls, which is why we focused on this particular oral site in detailed analyses. Consistent with the published literature, inflammation, i.e., the presence of ulcer, was found to be the main determining factor affecting microbiota composition [6]. Moreover, we showed that not only microbiota within the ulcer itself, but also in the surrounding area differ from healthy controls.
As a pioneer colonizer and the most represented bacterial genus in the oral cavity, Streptococcus belongs to the core of healthy oral microbiome [25]. We noticed that its abundance decreased during active ulcer manifestation and increased a little after ulcer healing. Furthermore, the abundance of S. oralis was slightly higher in healthy controls (69.8%) over RAS patients (42.6% in the Active cohort and 50.9% in the Passive cohort). Streptococcus populations may fluctuate in response to a disturbed environment, which is in line with earlier observations [7,26]. A lower abundance of Streptococcus, together with a higher abundance of Acinetobacter johnsonii, has been recently linked to a higher risk of RAS [7]. In agreement with this study, we also found a higher abundance of A. johnsonii in RAS patients over healthy controls (0.031% in the Active cohort; 0.002% in the Passive cohort; and 0.001% in healthy controls), although not significantly so. An initial imbalance in bacterial microbiota might be crucial for expansion of opportunistic pathogens, consequently leading to compositional changes in fungal microbiota and further to the disruption of oral tissue through inter-kingdom interaction [27,28]. A greater amount of Haemophilus parahaemoliticus or H. parainfluenzae found in RAS patients over healthy controls further strengthens the connection of opportunistic pathogens with the risk of . Each point represents a sample from one patient or healthy control. Statistical significance was confirmed using Kruskal-Wallis test with Dunn's multiple comparison test. Significant differences are denoted by * p < 0.05 and ** p < 0.01. Reference serum was added to each ELISA plate and its mean value of OD (450-650 nm) was used as an arbitrary unit (1 AU).

Discussion
To date, only a few studies have examined the specifics of microbiota composition in patients with recurrent aphthous stomatitis [6,7,24]. In agreement with these previous studies, we observed that microbiota in RAS patients differs from microbiota found in healthy individuals. We sampled patients' microbiota either from minor RAS lesions or from healed mucosa at five different sites of the oral cavity. Lower labial mucosa showed the most pronounced differences between Active or Passive RAS patients and healthy controls, which is why we focused on this particular oral site in detailed analyses. Consistent with the published literature, inflammation, i.e., the presence of ulcer, was found to be the main determining factor affecting microbiota composition [6]. Moreover, we showed that not only microbiota within the ulcer itself, but also in the surrounding area differ from healthy controls.
As a pioneer colonizer and the most represented bacterial genus in the oral cavity, Streptococcus belongs to the core of healthy oral microbiome [25]. We noticed that its abundance decreased during active ulcer manifestation and increased a little after ulcer healing. Furthermore, the abundance of S. oralis was slightly higher in healthy controls (69.8%) over RAS patients (42.6% in the Active cohort and 50.9% in the Passive cohort). Streptococcus populations may fluctuate in response to a disturbed environment, which is in line with earlier observations [7,26]. A lower abundance of Streptococcus, together with a higher abundance of Acinetobacter johnsonii, has been recently linked to a higher risk of RAS [7]. In agreement with this study, we also found a higher abundance of A. johnsonii in RAS patients over healthy controls (0.031% in the Active cohort; 0.002% in the Passive cohort; and 0.001% in healthy controls), although not significantly so. An initial imbalance in bacterial microbiota might be crucial for expansion of opportunistic pathogens, consequently leading to compositional changes in fungal microbiota and further to the disruption of oral tissue through inter-kingdom interaction [27,28]. A greater amount of Haemophilus parahaemoliticus or H. parainfluenzae found in RAS patients over healthy controls further strengthens the connection of opportunistic pathogens with the risk of RAS. In addition, opportunistic pathogens Fusobacterium or Leptotrichia, whose abundance we found to be elevated in the Active RAS cohort, have been described as part of the normal oral microbiota [29,30]. Fusobacterium nucleatum has been previously connected to incidence of dental caries and periodontitis [31,32]. Furthermore, F. nucleatum can spread to the intestinal tract and is tightly associated with colorectal carcinoma development [33]. The most abundant fungal species in the Active RAS cohort belong to the genus Malassezia, which has been mainly reported as a commensal and opportunistic pathogen of the skin, but nowadays is being included also as a member of the oral cavity mycobiome [34]. Using the software tool LEfSe [35], we discovered the genus Malassezia to be highly enriched in patients with an active RAS ulcer, which is consistent with the hypothesis of microbial shift in favor of opportunistic pathogens. This observation is also in line with published data connecting the presence of Malassezia with incidence of other diseases, including dandruff, seborrheic dermatitis, and psoriasis [36][37][38][39][40]. The second most abundant species in the Active RAS cohort was Candida albicans, often described either as a commensal or opportunistic pathogen in the oral cavity [34,41]. C. albicans is believed to get virulent by modulation of co-colonizing microbiota [28]. Consistently with the published literature, we detected C. albicans in low amounts as a commensal in healthy oral cavity, with the abundance rapidly increasing in active ulcers. Another two species were found only in RAS patients, namely Itersonilia sp. and Candida tropicalis. Associated mostly with healed mucosa, both species were present to some extent also at the site of active ulcers. Itersonilia is usually described as a plant commensal or pathogen; therefore, its presence in the oral cavity is probably inconsequential in relation to RAS, since it is most likely associated with food intake [42][43][44]. C. tropicalis is considered to be the second most virulent Candida species and a very strong biofilm producer [45]. In our dataset, we found C. tropicalis to be abundant rather in few individuals with healed mucosa than in the whole cohort of RAS patient.
Fascinating hints of inter-kingdom interactions led us to perform correlation analysis and, for the first time, to explore the co-occurrence network between bacteria and fungi within the oral cavity of RAS patients and healthy controls. Apart from a negative correlation of Malassezia sp. with Haemophilus parainfluenzae and its positive correlation with Porphyromonas pasteri, we unveiled a number of associations of Rhodotorula mucilaginosa with bacterial species during active RAS. Bacterial-fungal networks might be of great importance and, in the future, might potentially serve for prognostic or diagnostic purposes in RAS.
Antibody response against specific microbes is a sign of activated adaptive immunity, which could either lead to specific pathogen clearance, or result in a chronic failure to control the pathogen [31,46]. This has been documented in chronic periodontitis, where elevated levels of IgG antibodies against Aggregatibacter actinomycetemcomitans are connected to the severity of periodontal tissue destruction [12,47]. Therefore, we investigated serum IgG and IgA antibodies against selected bacterial and fungal species, recently connected to pathogenesis of RAS or other inflammatory oral diseases [12]. Increased presence of Mogibacterium timidum in the oral cavity has been previously linked to periodontitis incidence [48]. Here, we reported, for the first time, that both Active and Passive RAS patients have elevated levels of IgG antibodies against M. timidum as compared to healthy controls, while IgA response to all and IgG response to other analyzed microbes was unchanged. This may be caused by the fact that multiple dominant antigens from these microbes may be shared with tolerated commensals or by inability of these microbes to produce immunodominant virulence factors when cultivated in vitro in artificial growth media [49][50][51]. Moreover, microbes may respond to inflammatory conditions, such as in inflamed mucosa during RAS, by producing different surface antigens, thus differing even from isolates from healthy mucosa [52,53]. Nevertheless, by preparing these microbes in vitro under standardized conditions, we were able to compare all tested groups within a single assay. Therefore, while we cannot exclude importance of antibody response to components from the other microbes which are expressed only during the infection, our results suggest a potential role of M. timidum in the pathogenesis of RAS and point at an interesting biomarker.

Conclusions
To understand the role of bacterial and fungal communities in the human oral cavity in health and disease, we need to acquire thorough knowledge of their presence and interactions. We are the first to conduct a comprehensive study mapping the overall composition of bacterial and fungal communities in patients with RAS. We found increased serum levels of antibodies against Mogibacterium sp. in RAS patients, indicating its potential role in the pathogenesis of RAS. Our results further propose an intriguing hypothesis of altered inter-kingdom interactions, taking place especially during active RAS. We suggest that a pre-existing disruption of microbial communities (dysbiosis) might allow opportunistic pathogens to prevail, leading to oral tissue damage. Future studies would benefit from analyzing larger populations of patients and healthy controls to yield more predictive power. As we observed some interesting changes in microbial associations during active RAS, a deeper analysis of microbial metabolic functions could also reveal further factors important for disease development.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/7/12/636/s1. Table S1: Characteristics of patients and controls in the study. Table S2: Sampling sites of patients and controls. Table S3: Characteristics of patients suffering from RAS and healthy controls, both sampled from the lower labial mucosa (LL). Figure S1: Bacterial alpha diversity in active ulcer of RAS patients compared with the same site after ulcer healing and with healthy controls at five merged sites of sampling. Figure S2: Beta diversity of bacterial communities in active ulcer of RAS patients compared with the same site after ulcer healing and with healthy controls at different sites of sampling. Figure S3: Fungal alpha diversity in an active ulcer of RAS patients compared with the same site after ulcer healing and with healthy controls in five merged sites of oral cavity. Figure  S4: Beta diversity of fungal communities in an active ulcer of RAS patients compared with the same site after ulcer healing and with healthy controls at different sampling sites.