Dysbiosis by neutralizing commensal mediated inhibition of pathobionts

Dysbiosis in the periodontal microbiota is associated with the development of periodontal diseases. Little is known about the initiation of dysbiosis. It was hypothesized that some commensal bacteria suppress the outgrowth of pathobionts by H2O2 production. However, serum and blood components released due to inflammation can neutralize this suppressive effect, leading to the initiation of dysbiosis. Agar plate, dual-species and multi-species ecology experiments showed that H2O2 production by commensal bacteria decreases pathobiont growth and colonization. Peroxidase and blood components neutralize this inhibitory effect primarily by an exogenous peroxidase activity without stimulating growth and biofilm formation of pathobionts directly. In multi-species environments, neutralization of H2O2 resulted in 2 to 3 log increases in pathobionts, a hallmark for dysbiosis. Our data show that in oral biofilms, commensal species suppress the amounts of pathobionts by H2O2 production. Inflammation can neutralize this effect and thereby initiates dysbiosis by allowing the outgrowth of pathobionts.

promoting biofilm formation and inhibiting the adhesion of certain species 16 . Additionally, serum and blood components can have peroxidase activity 17 . This is of significance since these could interact with the H 2 O 2 produced by commensal bacteria to suppress the outgrowth of pathobionts. Clinically it is known that absence of bleeding on probing is a good marker for periodontal stability 18 . Recently, it is shown that crevicular myeloperoxidase concentrations are highly correlated with periodontal disease severity 19 .
Therefore, it can be hypothesized that commensal bacteria suppress the overgrowth of pathobionts by H 2 O 2 but some serum and blood components released during inflammation can neutralize this suppressive effect, leading to the initiation of dysbiosis.
The objective of this study is to determine the neutralizing effect of serum, hemoglobin and hemin on the inhibitory effect of the commensal bacteria towards pathobionts.

Results
Decrease of the inhibitory effect of commensals on agar plates. In order to verify the inhibitory effect of some commensal species on pathobionts and the influence of hemin, serum, hemoglobin and peroxidase, a qualitative agar-plate method was used. On agar plates, hemin, serum, hemoglobin and peroxidase significantly lowered the inhibitory effect of commensal bacteria on P. intermedia, P. gingivalis and A. actinomycetemcomitans (Fig. 1). This effect was observed for all commensal species, all substrates on all pathobionts with exception for the effect of hemoglobin on the inhibition of P. intermedia and P. gingivalis by S. mitis (Table 1). Only peroxidase was able to completely neutralize the inhibitory effect of all the commensals. In most cases, hemin lowered the inhibition of the commensals more than serum and hemoglobin.
Decrease of the inhibitory effect of commensals in dual-species planktonic cultures and biofilms. Since bacteria within the oral cavity primarily live as biofilms, which can change their behavior, the decreased inhibitory effect induced by hemin, serum, hemoglobin and peroxidase in agar-plate experiments was verified in dual-species planktonic and biofilm cultures containing a commensal species and a pathobiont. In dual-species planktonic cultures and biofilms, the inhibitory effect of S. oralis on P. intermedia, P. gingivalis and A. actinomycetemcomitans was decreased (p < 0.05) by serum, hemin, hemoglobin and peroxidase ( Fig. 2A and B). The effect was not significant for serum and hemoglobin on planktonic A. actinomycetemcomitans. For most of the dual-species experiments, adding serum, hemin, hemoglobin and peroxidase completely abolished the inhibitory effect of S. oralis resulting in numbers of pathobionts similar to the negative control in which S. oralis was not present.
S. intermedius, a non-H 2 O 2 producing species, did not show an inhibitory effect on the pathobionts ( Fig. 2C and D). Consequently serum, hemin, hemoglobin and peroxidase could not decrease an inhibitory effect on the pathobionts but they also did not increase planktonic growth and biofilm formation in dual-species experiments. However, in these experiments, serum decreased the planktonic growth of P. intermedia and dual species biofilm formation of P. intermedia and P. gingivalis (p < 0.05).
Effect of serum, hemoglobin, hemin and peroxidase on single species cultures. In order to verify if the outgrowth of the pathobionts in the dual-species experiments was due to a decreased inhibitory effect of the commensal species and not due to an increased growth or biofilm formation of the pathobionts by the presence of hemin, serum, hemoglobin or peroxidase. The effect of these substrates on pathobiont growth and biofilm formation was evaluated. Serum, hemin, hemoglobin and peroxidase did not increase growth or biofilm formation of P. intermedia, P. gingivalis and A. actinomycetemcomitans (Fig. 3). Moreover, serum and hemin induced a small, but significant reduction on P. gingivalis biofilm formation (p < 0.05). Additionally, A. actinomycetemcomitans growth was decreased by serum and its biofilm formation by hemin and peroxidase (p < 0.05). The commensals were spotted 24 hours before the pathogens in the center of the pictures. The pathogens were spotted at both sides of the commensals, at the left side plus the blood compound (serum (+ Se), hemin (+ He), hemoglobin (+ Hb) and peroxidase (+ Pe)) and at the right side without any blood compound.

Reduction of the inhibitory effect of commensals in multi-species ecologies.
Bacteria within the oral cavity are part of complex microbial ecologies. Since the observed effects in dual-species experiments might be different in more complex ecologies, the effect of hemin, hemoglobin, serum and peroxidase on simplified and complex multi-species ecologies was examined. In both biofilm models, commensal biofilms containing S. oralis, S. gordonii, S. cristatus, S. parasanguinis, S. mitis and S. sanguinis were challenged with either only 3 pathobionts (simplified ecology) or with a complex 14 species ecology (complex ecology). The commensal biofilm significantly inhibited the planktonic and biofilm concentrations of P. gingivalis and P. intermedia both in simplified ( Fig. 5A and B) and complex multi-species ecologies ( Fig. 5C and D). Its effect on A. actinomycetemcomitans concentrations was limited in simplified multi-species ecologies ( Fig. 5A and B). Although it was more pronounced in complex multi-species ecologies, it did not reach statistical significance ( Fig. 5C and D). In general, serum, hemin, hemoglobin and peroxidase decreased the inhibitory effect of the commensal biofilm on planktonic and biofilm concentrations of the pathobionts. This inhibition resulted in an outgrowth of A. actinomycetemcomitans, P. gingivalis and P. intermedia respectively of up to 2.37 (± 0.98), 4.48 (± 0.62) and 2.94 (± 0.38) log 10 CFU/ ml in complex planktonic multi-species ecologies and of up to 2.21 (± 0.46), 8.17 (± 0.14) and 3.12 (± 0.18) log 10 CFU/ml in complex multi-species biofilms. The neutralizing effect of serum and hemoglobin was less than that of hemin and peroxidase. The presence of the commensal biofilm did not only result in decreased planktonic concentrations of P. gingivalis and P. intermedia (Table 2). In the planktonic ecology, also a decreased concentration of S. salivarius and increased concentrations of S. sanguinis and S. oralis were observed when the commensal biofilm was present (p < 0.05). On the other hand, the biofilm concentrations of S. mutans, S. sobrinus and S. salivarius were decreased and the biofilm concentrations of F. nucleatum and S. sanguinis were increased (p < 0.05).
Moreover, in the complex multi-species ecologies (Table 2), the presence of serum increased the concentration of planktonic P. gingivalis (p < 0.05) and the concentrations of P. intermedia, P. gingivalis,  In contrast, the planktonic concentrations of A. actinomycetemcomitans and P. intermedia as well as the biofilm concentrations of A. actinomycetemcomitans, P. gingivalis, P. intermedia, A. viscosus and S. sanguinis were increased in the presence of hemoglobin (p < 0.05).
The addition of hemin increased the planktonic concentrations of P. gingivalis and P. intermedia but decreased the concentrations of A. viscosus, S. mutans, S. gordonii and S. mitis (p < 0.05). In the biofilms, hemin increased the concentrations of P. gingivalis and P. intermedia and decreased the concentrations of A. viscosus, A. naeslundii, S. sanguinis and S. gordonii (p < 0.05).

Discussion
Dysbiosis in oral bacterial communities is characterized by a microbial shift, which is translated in an increase of pathobionts and a decrease of commensal species 8,10,20 . It has been shown that some commensal species can suppress the growth of pathobionts by H 2 O 2 under specific environmental conditions. More specifically, the sequence of colonization and the presence of oxygen are major influencing factors 11 . Additionally, dysbiotic biofilms are enriched in virulence factors that stimulate the host inflammatory response 21 . Although it can be deduced that both absence of commensal species and the presence inflammation are key factors, the factors driving dysbiosis are unclear 4 . Moreover, as far as the authors know, dysbiosis has never been induced in in vitro multi-species biofilms. In this study, it was hypothesized that commensal bacteria suppress the overgrowth of pathobionts by H 2 O 2 but some serum and blood components released during inflammation can neutralize this suppressive effect, leading dysbiosis. It was shown on agar plates and in dual-species biofilms that H 2 O 2 production by commensal bacteria decreases pathobiont growth and colonization. Although it was not directly tested in the current study, Herrero et al. showed that the amount of inhibition on pathobiont growth is determined by oxygen availability 11 . Commensal bacteria could produce H 2 O 2 and inhibit pathobiont growth under anaerobic condition. However, H 2 O 2 production and pathobiont inhibition was significantly higher under aerobic conditions. Therefore, oxygen availability must play an important role into the transition from homeostasis to dysbiosis 11 . If a non-H 2 O 2 producing species was used in the current study, no inhibition was observed. Peroxidase and blood components neutralize the inhibitory effect of H 2 O 2 primarily by a peroxidase activity since they did not stimulate the growth and biofilm formation of the pathobionts directly. In multi-species environments, neutralization of H 2 O 2 by peroxidase or blood components resulted in 2 to 3 log increases in pathobionts which can be considered as a hallmark for dysbiosis.
The agar plate experiments showed an inhibition of the pathobionts when grown in the vicinity of the commensals. This inhibition was completely neutralized in the presence of peroxidase. These results were in concordance with previous studies 11,12,22 that also identified H 2 O 2 as the main inhibitory substance. It was observed that the magnitudes of inhibition described by Van Essche et al. were markedly smaller than the ones reported in the current study although the same bacterial strains were used 23 . This was attributed to the blood agar medium which was used in the Van Essche study and pointed towards an interference of certain blood compounds on the inhibition effect of these streptococcus species.
Gingival inflammation is characterized by an increase in GCF production, which is similar to serum, and bleeding tendency. Among many other components, blood contains serum, hemin and hemoglobin 24 but the hemolytic capacity of pathobionts can also increase their concentration 25 . It is already described that hemoglobin and hemin have a peroxidase activity 17,26,27 . P. intermedia and P. gingivalis also generate layers of haems with catalytic activity to degrade the H 2 O 2 using hemin and hemoglobin 28 . Additionally, serum and GCF contains catalytic molecules such as myeloperoxidase that neutralizes the antimicrobial effect of H 2 O 2 19 . Recently, it is shown that crevicular myeloperoxidase concentrations are highly correlated with periodontal disease severity. All these studies substantiate the observation that hemin, hemoglobin and serum decrease the inhibitory effect of the commensal species on agar plates by H 2 O 2 reduction. The effects of hemin, hemoglobin, serum and the role of H 2 O 2 were further demonstrated in dual-and multi-species biofilms. Their addition resulted in increases in pathobionts of up to 8 log values both in planktonic and in biofilm cultures. Although, it has been suggested that hemin and hemoglobin can increase the growth and biofilm formation of some pathogens, this effect was not seen [29][30][31][32] (Fig. 3). This can be explained by the qPCR method used to quantify growth and biofilm formation and the growth medium used in this study which might provide sufficient iron sources to the pathobionts so that addition of hemin and hemoglobin did not affected their growth or biofilm formation anymore. The effect of serum on pathobiont growth and biofilm formation is even less reported in literature. Biyikoĝlu et al. reported no effects of serum on A. odontolyticus, V. parvula, F. nucleatum and P. gingivalis over time but a decrease in F. nucleatum biovolume when 4 species biofilms were grown in the presence of serum 16 . These data are in accordance with the present data.
Overall, the study suggests that serum and blood compounds play an important role in the initiation of dysbiosis of oral biofilms by disrupting the inhibitory defensive barrier provided by the commensal bacteria. The data provide a hypothesis for the initiation of dysbiosis in oral biofilms (Fig. 6). In oral biofilms some commensal species can suppress the amounts of pathobionts by H 2 O 2 generation. However, when these biofilms persist over longer periods of time or become more abundant or when the susceptibility of the host changes, the resulting inflammatory reaction neutralizes the effects of H 2 O 2 , thereby allowing the outgrowth of pathobionts. Additionally, changes in the oxygen availability within the biofilm might also lower the H 2 O 2 production and subsequently contribute to the outgrowth of pathobionts.

Methods
Bacterial strains and media. All Table 2. Log (MO/ml) of the different species (mean ± standard deviation, n = 3) that form part of the 14 species community exposed to serum, hemoglobin, hemin and peroxidase. Control condition refers to a 14 species community without the addition of the 6 commensal bacteria. *Designates a statistically significant increase of the bacterial concentration in respect to BHI (p < 0.05). # Designates a statistically significant decrease of the bacterial concentration in respect to BHI (p < 0.05). Competitive inhibition experiments were performed in Brain Hearth Infusion 2 (BHI-2) broth or agar containing Brain Heart infusion (Difco, Detroit, USA) supplemented with 2.5 g/L mucin (Sigma-Aldrich, St. Louis, USA), 1.0 g/L yeast extract (Oxoid, Basingstoke, UK), 0.1 g/L cysteine (Calbiochem, San Diego, USA), 2.0 g/L sodium bicarbonate and 0.25% (v/v) glutamic acid (Sigma-Aldrich, St Louis, USA). The bacteria were cultured under aerobic (5% CO 2 ) or anaerobic (80% N 2 , 10% H 2 and 10% CO 2 ) conditions. Optical densities were measured and adjusted using spectrophotometry (OD600, GeneQuant Spectrophotometer, Buckinghamshire, UK).
Serum and blood components. Hemin, human hemoglobin and horseradish peroxidase (Sigma-Aldrich, St. Louis, USA) were dissolved in BHI-2 at concentrations of 5 mg/mL hemin, 0.44 mg/mL hemoglobin 33 and 16 μ g/mL peroxidase. Human serum was obtained by venipuncture of a single, systemically healthy, male volunteer with no oral disease and who had not taken any antibiotics for 1 year. Peripheral venous blood was immediately centrifuged at 264 × g for 30 min at room temperature. The serum was removed and frozen at − 20 °C after aliquotation.
Ethics Statement. The use of human serum was approved by the ethical committee of the KU Leuven and registered with identifier B322201628215. The procedures were executed according to the Helsinki Declaration and the regulations of the University Hospital, which are approved by the ethical committee. The adult subject provided a written and oral consent after having explained to him the purpose of the study. The subject is aware that the results will be used in a scientific study. An informed consent was obtained from all subjects.
Antagonistic experiments on agar plates. The spotting technique was used to quantify the inhibitory effect of 6 commensal species on 3 pathobionts and to identify neutralization effects by serum and blood components 11 . An overnight culture of a commensal species was adjusted to a concentration of 10 9 CFU/mL. This solution was spotted on an agar plate and incubated under aerobic conditions. After 24 hours, an overnight culture of the pathobiont (10 9 CFU/mL) was spotted next to the commensal spot. After 48 hours of anaerobic incubation, a calibrated photograph was taken from the agar plate and the magnitude of inhibition was measured from the edge of the commensal colony to the border of the inhibited pathobiont colony using ImageJ (http://rsb.info.nih. gov/ij/download.html). Dual-species planktonic and biofilm experiments. S. oralis was selected as the model commensal species with a strong inhibitory effect on pathobionts by H 2 O 2 production. S. intermedius was used as a not-inhibiting commensal species (negative control) 11 . 10 mL of an overnight culture of these commensals was centrifuged (1438 × g, 10 minutes). The supernatant was discarded and the pellet was re-suspended in 10 mL BHI-2 broth. The density was adjusted to 1 × 10 8 CFU/ml. 6 ml of this solution was transferred to 6 wells (1 ml/well) of a 24 well-plate (Greiner, Frickenhausen, Germany) and incubated under aerobic conditions. After 24 hours, 500 μ L of BHI-2 broth, serum, hemoglobin, hemin or peroxidase was added to the cultures. Additionally, an overnight culture of a pathobiont (A. actinomycetemcomitans, P. gingivalis or P. intermedia) was centrifuged (1438 × g, 10 minutes) and re-suspended in BHI-2 (1 × 10 8 CFU/ml). 1 mL of this bacterial solution was inoculated in each well with S. oralis or S. intermedius and in an additional well containing 1.5 mL BHI-2 (negative control). After 24 hours of anaerobic incubation, 1 mL was taken from each well, centrifuged (1438 × g, 10 min), re-suspended in PBS and analyzed via vitality q-PCR. Afterwards, the remaining supernatant was removed and the biofilms at the bottom of the wells were washed with phosphate buffered saline (PBS). The biofilms were detached with 500 μ L 0,05% Trypsin-EDTA (Gibco, Paisley, UK) for 15 minutes at 37 °C, transferred to Eppendorf tubes, centrifuged (6010 × g, 10 minutes) and after discarding the trypsin, the biofilm pellets were re-suspended in 1 mL of PBS and analyzed by vitality q-PCR.
Effects on the growth and biofilm formation of pathobionts. An overnight culture of a pathobiont (A. actinomycetemcomitans, P. gingivalis or P. intermedia) was centrifuged (1438 × g, 10 minutes) and re-suspended in BHI-2 (1 × 10 8 CFU/mL). 1 mL of this bacterial solution was inoculated in each well plus 500 μ l of BHI-2 broth, serum, hemoglobin, hemin or peroxidase. The wells were incubated for 24 hours under anaerobic conditions where after planktonic bacteria and the biofilms analyzed as described above. Simplified multi-species planktonic and biofilm experiments. Similar to the dual-species experiments, overnight cultures of six commensal bacteria (S. oralis, S. gordonii, S. cristatus, S. parasanguinis, S. mitis and S. sanguinis), with inhibitory effects by producing H 2 O 2 , were centrifuged, re-suspended in BHI-2 broth (1 × 10 8 CFU/mL). Equal volumes of these solutions were mixed and inoculated in 6 wells of a 24 well-plate and incubated under aerobic conditions. After 24 hours, BHI-2, serum, hemoglobin, hemin and peroxidase were added to the wells as described above. Additionally, 1 mL of an overnight co-culture of 3 pathobionts (A. actinomycetemcomitans, P. gingivalis and P. intermedia) was centrifuged (1438 × g, 10 minutes), re-suspended in BHI-2 (1 × 10 8 CFU/ml) and added to the wells. The latter co-culture was obtained from overnight cultures of the pathobionts (simplified ecology) which were centrifuged (1438 × g, 10 minutes) and re-suspended in 10 ml of BHI-2. 1 mL of each pathobiont culture was added to 7 mL of BHI-2 and incubated for 24 hours under anaerobic conditions to obtain the co-culture. The wells were incubated for 24 hours under anaerobic conditions where after planktonic bacteria and the biofilms analyzed as described above.

Peroxidase activity of blood compounds and peroxidase on H
Complex multi-species planktonic and biofilm experiments. The experimental set-up was identical to the set-up used for the simplified multi-species experiments with the exception that instead of using an overnight co-culture of 3 pathobionts, a bioreactor derived complex multi-species co-culture of 14 species (complex ecology), as described below, was used. The wells were incubated for 24 hours under anaerobic conditions where after planktonic bacteria and the biofilms analyzed as described above.
Bioreactor derived multi-species community. A multi-species community was established in a BIOSTAT B TWIN (Sartorius, Germany) bioreactor. 750 mL of BHI-2 broth was added to the vessel together with 5.0 mg/mL hemin, 1.0 mg/mL menadione and 200 μ l/L Antifoam Y-30 (Sigma, St. Louis, USA). The medium was pre-reduced over 24 hours at 37 °C by bubbling 100% N 2 and 5% CO 2 in the medium under continuous stirring at 300 rpm. pH was set at 6.7 + /− 0. Vitality q-PCR. DNA extraction and vitality q-PCR using propidium monoazide was previously described 34 . Table 3 shows primer and probe sequences used in this study.

Statistical analysis.
All experiments were repeated on 3 different days. To account for the censored character of the inhibition data from agar plates experiments, differences between treatments (with serum, hemoglobin, hemin or peroxidase) and control (without addition of blood compound or peroxidase) for the inhibition data were analyzed by means of a survival regression model for gaussian data. Comparisons between treatments and control were made for each combination of substance (blood compounds or peroxidase) and bacteria and corrected for simultaneous hypothesis testing according to Sidak. For planktonic and biofilm data, a linear mixed model was fit to model the log-transformed CFU counts using substrate (blood compounds or peroxidase) and sample type (planktonic or biofilm) as fixed factors and run as random factor. Since a residual analysis showed that the model was heteroscedastic, weights, proportional to the inverse of the predicted value, were applied. Comparisons with BHI and control were made separately by calculating the appropriate contrasts and a correction for simultaneous hypothesis testing according to Dunnett was applied. Data were analyzed using S-plus 8.0 for Linux (Tibco, Palo Alto, CA, USA).