Necrotrophic growth of periodontopathogens is a novel virulence factor in oral biofilms

The oral use of antimicrobial agents embedded in toothpastes and mouth rinses results in an oral microbial massacre with high amounts of dead bacteria in close proximity to few surviving bacteria. It was hypothesized that this provides the surviving pathogenic bacteria a large amount of dead microbial biomass as a nutritional source for growth (necrotrophy). This study demonstrated the necrotrophic growth of periodontal pathogens in the presence of different dead oral species. In addition, the presence of dead bacteria resulted in an outgrowth of several periodontal pathogens in complex multi-species biofilms. Additionally, upon contact with dead oral bacteria, virulence genes of P. intermedia and P. gingivalis were up-regulated (necrovirulence). This resulted in a more pronounced epithelial cytotoxicity (necrotoxicity). These findings indicate that presence of dead bacteria induce necrotrophy, necrovirulence and necrotoxicity in several oral pathogens.

many surviving bacteria. Although these survivors can recover quickly from an antiseptic assault 9 , the fate of the bacterial casualties and their impact on survivors is currently unknown and has never been explored.
Recently it was shown that certain bacterial species can use dead microbial cells as a nutritional source for their own survival. Legionella pneumophila, for example, can use the remaining dead bacteria after heat treatment as a nutritional source to persist in water piping 10,11 . This feature has been termed "bacterial necrotrophy" 11 . It is also known that starving Bacillus subtilis cells can produce toxins to kill sibling bacteria for nutrition 12 . The nutrients released by this behaviour from the killed siblings are being used to feed the surviving bacteria under conditions of nutritional stress 13 . By this, B. subtilis can prolong its survival and delay or avoid going into sporulation, which is a time and energy consuming process. This behaviour has been termed "cannibalism" 14 .
The widespread and daily use of antiseptic mouth rinses and toothpastes together with the inability for most people to accomplish a complete dental plaque removal makes that bacterial necrotrophy might be relevant for the oral ecosystem. Therefore, it is crucial to clarify if and how oral microbial ecosystems respond to sudden increases in dead bacteria.
The aim of this study was to determine if periodontopathogens can use dead bacteria as a nutritional substrate and to determine the impact of dead bacteria on bacterial virulence.  Table 1. Effect of dead bacteria on bacterial growth of periodontal pathogens. Data represent the final cell concentration of living periodontal pathogens (Log 10 CFU/mL)after 24 anaerobic incubation starting from 7.7 Log 10 CFU/mL and in presence of different concentrations of dead bacteria (mean ± standard deviation, n = 3). Control contains only living bacteria without addition of dead bacteria. *Statistically significant difference of periodontal pathogens concentration in presence of dead bacteria respect to the control (p < 0.05).

Results
Effect of dead bacteria on bacterial growth. In order to evaluate the necrotrophic behaviour of P. intermedia, a culture of P. intermedia was supplemented with dead P. intermedia. As shown in Table 1, after 24 hours, the total number of P. intermedia (CFU/mL) was significantly increased in the culture supplemented with dead P. intermedia when compared to a not-supplemented culture. This increase was dose dependent and resulted in a maximum supplemental outgrowth of P. intermedia of up to 1.46 (±0.01) Log 10 CFU/mL in the presence of dead P. intermedia. The necrotrophic behaviour of P. intermedia was however depending on the dead species. Supplementing a P. intermedia culture with either dead P. gingivalis, F. nucleatum, S. oralis or S. gordonii resulted in a significant increase of P. intermedia growth (Table 1). However, dead S. gordonii only increased the growth of P. intermedia in its highest concentration whereas dead P. gingivalis and dead F. nucleatum significantly increased the growth of P. intermedia in all used concentrations with up to 0.94 (±0.26) Log 10 CFU/mL. Also for these dead species, a dose dependent effect was observed.
Necrotrophic behaviour was also observed for F. nucleatum and P. gingivalis in the presence of dead F. nucleatum and dead P. gingivalis respectively (Table 1). However, the magnitude of the growth increase induced by supplementing these species with dead bacteria was clearly lower than for P. intermedia. All concentrations (10 9 , 10 8 and 10 7 CFU/mL) of dead P. gingivalis and dead F. nucleatum could increase the growth of P. gingivalis (Table 1).
Since the above described experiments were performed in a nutrient rich medium, it could be hypothesized that the effect of supplementing dead bacteria could be more pronounced in nutrient limited media. This was examined by supplementing P. intermedia with dead P. gingivalis in a nutrient limited medium BHI diluted 1/10 or 1/100 in PBS. After 24 hours of incubation, the concentration of P. intermedia in 1/10 BHI was reduced to 4.65 (±0.77) Log 10 CFU/mL, indicating starvation of P. intermedia. However, if P. intermedia in 1/10 BHI was supplemented with dead P. gingivalis, the number of surviving bacteria was 6.42 (±0.13) Log 10 CFU/mL ( Table 1). The latter was significantly (p < 0.05) more than in the not supplemented condition and points towards a higher survival rate when dead P. gingivalis was present. When the culture media was further depleted from nutrients (1/100 BHI), no surviving bacteria could be detected for both conditions (data not shown).
The fate of dead bacteria. In order to further substantiate the necrotrophic behaviour, the DNA of dead bacteria was monitored over time by qPCR and correlated to the number of consumed cells.
As shown in Table 2, the qPCR data confirmed the dose dependent culturing data showing that when P. intermedia is supplemented with dead P. gingivalis or dead S. oralis, its growth over 24 hours is increased with up to  Table 3. Vitality qPCR (Log (MO/mL) of the different species (mean ± standard deviation, n = 3) in the 14 species community in presence of living and dead bacteria in BHI-2 after 24 hours incubation. BHI-2 condition refers to a 14 species community without the addition of dead bacteria. *Designates a statistically significant increase of the bacterial concentration with respect to BHI (p < 0.05). # Designates a statistically significant decrease of the bacterial concentration in respect to BHI (p < 0.05).  Necrotrophic growth in a multispecies community. Necrotrophic activity in the presence of other bacteria was tested to check for competitive effects. An increase in the concentrations of periodontal pathogens was observed inside of the 14 species planktonic community by the presence of dead P. intermedia and P. gingivalis in BHI-2, saliva or serum (Table 3, Supplementary Table 1, Supplementary Table 2 and Supplementary  Table 3). The presence of dead P. intermedia increased the growth of living P. intermedia in BHI-2, saliva and serum (p < 0.05). F. nucleatum and A. actinomycetemcomitans increased also significantly by the addition of dead P. intermedia in BHI-2. Only the presence of dead P. intermedia increased the concentrations of P. gingivalis and F. nucleatum in saliva. Moreover the presence of dead P. gingivalis increased significantly the growth of P. intermedia, P. gingivalis and F. nucleatum in BHI-2, of P. gingivalis and F. nucleatum in saliva, and of P. intermedia, A. actinomycetemcomitans and P. gingivalis in serum.  Table 4. Primers of virulence and house-keeping genes from P. gingivalis (Pg) and P. intermedia (Pi).
The effect of dead P. intermedia and dead P. gingivalis was also examined on the rest of bacterial species in the 14 species community (Table 3, Supplementary Table 1, Supplementary Table 2 and Supplementary Table 3 Virulence gene expression. Gene expression of P. intermedia and P. gingivalis was analysed after supplementing the cultures with dead P. gingivalis and dead P. intermedia respectively. After 24 hours of growth, the P. intermedia virulence genes kpsD, phg, ecf, inpA and adpC were 1.96 to 4.21 fold up-regulated when compared to the control culture (Fig. 1A). P. intermedia heat shock protein genes htpG, dnaJ, dnaK, groeL, groeS, clpB, were also up-regulated with a factor of 4.72 to 11.62 (Fig. 1A). Similarly, supplementing a P. gingivalis culture with dead P. intermedia resulted in a significant 6.02 +/− 3.17 fold increase in prtC gene expression and a 2.94 +/− 1.46 fold increase in fimA expression (Fig. 1B).
Additionally, the effect of supplementing a P. gingivalis culture with living P. intermedia was compared. Although there were no significant differences found in rgpA, rgpB, prtC and fimA gene expression between both cultures, the P. gingivalis culture supplemented with living P. intermedia showed a 7.29 +/− 4.55 fold higher serB and a 53.37 +/− 36.37 fold higher kgp expression when compared to the P. gingivalis culture supplemented with dead P. intermedia (Supplementary Table 4).
Cell viability assay. In order to verify whether dead bacteria can reduce cell viability on their own or by increasing P. gingivalis and P. intermedia virulence or by the combination of both, a cell viability assay was performed. As shown in Fig. 2, both living and dead P. gingivalis and P. intermedia decreased the epithelial cell viability. However, when the epithelial cells were challenged with a combination of living P. intermedia and dead P. gingivalis, the cell viability further decreased to 36.30% (±2.83) (p < 0.05) ( Fig. 2A). Also when the epithelial cells were challenged with a combination of living P. gingivalis and dead P. intermedia, the percentage of cell viability further decreased significantly to 65.87% (±3.44) (Fig. 2B).

Discussion
The data of this study shows that some oral pathogens can use dead bacteria as a supporting substrate in order to increase their growth in planktonic and biofilm conditions. Such a feature has been named necrotrophy 11 . Although this behaviour is often mentioned, it has never been shown for oral pathogens. Additionally, we show for the first time that in the presence of dead bacteria, oral pathogens increase their virulence, which we name "necrovirulence". Both aspects can be of importance since the use of toothpastes and mouth rinses, containing bactericidal agents, is a common daily practice in society. It is very well known that these antimicrobials only kill the bacteria in the superficial layers of biofilms 7 . The fate of these bacterial casualties after an antiseptic attack is however not known although these layers of dead bacteria disappear within hours 15 . Therefore it can be hypothesized that these high and dense amounts of dead bacteria could provide multiple nutritional sources, such as C, N, P, Fe, for the surviving bacteria in the same ecosystem. Such behaviour has been shown for B. subtilis and L. pneumophila 11,16,17 . Starving B. subtilis can produce toxins to kill siblings for nutrition ("cannibalism") and prolong its survival 14 . L. pneumophila, on the other hand, can use heath killed dead bacteria as a nutritional source to persist in water piping 11 .
The necrotrophic behaviour of the tested oral pathogens showed a dose and species dependency. When exposed to different concentrations of dead bacteria, a clear dose dependent effect was observed on the growth of P. intermedia and P. gingivalis. Cultures of P. intermedia and P. gingivalis showed a significant increase in growth when there were at least 10 dead cells available to 1 living cell. The necrotrophic behaviour of these pathogens resulted in a supplemental growth of up to 1.46 Log 10 CFU/mL This magnitude and the dose dependency are similar to what was reported for L. pneumophila supplemented with dead Pseudomonas putida 11 . The species specificity of the necrotrophy was shown both for the living species as well as for the dead species. In homospecific necrotrophy experiments, the growth increase for P. intermedia was always stronger than for P. gingivalis. F. nucleatum only showed a weak growth increase. These data indicate that the necrotrophic activity specifically depends on the bacterial species. In heterospecific necrotrophy experiments, different dead bacterial species increased the growth of P. intermedia in different magnitudes, depending on the used dead bacterial species. These data indicate a clear dead species preference. A similar species preference has been shown for L. pneumophila. Whereas P. putida, Escherichia coli, Saccharomyces boulardii and Acanthamoeba castellanii increased the necrotrophic growth of L pneumophila, L. plantarum and B. subtilis did not 11 . Our data showed that the growth of the Gram negative P. intermedia was much less stimulated by the dead Gram positive S. oralis and S. gordonii than by more closely related Gram negative dead P. gingivalis or P. intermedia. This could be explained by the robust cell wall structure of these organisms, which prevents nutrients from becoming readily available, as suggested by Temmerman and coworkers 11 .
In order to verify that the dead bacteria were used as a nutritional source, DNA degradation of the dead bacteria was followed by qPCR as a proxy. This technique showed a decrease in dead bacteria over time concomitant to an increase in living bacteria. Such decrease in dead bacteria was not seen in control experiments without living bacteria. These data suggest that the living bacteria were actively breaking down the dead bacteria. However when the bacterial growth medium (BHI) was diluted, the addition of dead bacteria was insufficient to increase the growth of living bacteria, although their survival was improved. It has already been described that a concentration of 10 9 CFU/mL of dead bacteria is approximately 1 mg/mL of carbon biomass 18 . However, bacterial growth medium (BHI) contains about 30 mg/mL of carbon. In terms of nutritional values, the dead bacteria represent only a low proportion of 30:1 of carbon sources compared to the bacterial growth medium (BHI). Although the dead bacteria might provide other essential nutritional sources, their effect on growth cannot be explained only by nutritional aspects. It can be hypothesized that the presence of dead bacteria can also induce a change of the bacterial phenotype in terms of growth and behaviour.
Since a change in phenotype could be induced by the presence of dead bacteria, the virulence profiles of P. intermedia and P. gingivalis were investigated at the gene expression level. Although it is well known that associations of oral pathogenic bacteria can increase their virulence, for instance P. gingivalis can enhance the attachment and invasion of T. forsythia to epithelial cells 19 , the effect of dead bacteria on bacterial virulence gene expression has never been investigated. Several well-known virulence genes of P. gingivalis and P. intermedia were selected. The data showed that these genes become up-regulated in the presence of dead bacteria. This up-regulation can be explained by the presence of certain specific molecules (inducers) in the solutions of dead bacteria. The gingipain genes rgpA, rgpB, kgp of P. gingivalis were up-regulated in the presence of dead P. intermedia. They are involved in several bacterial functions which contribute to survival and virulence. The gingipain genes of P. gingivalis play an important role in the acquisition of nutrients by degrading peptides from the host or the environment 20,21 . Their increased expression can therefore be directly related to the observed necrotrophic behavior of P. gingivalis. Additionally they are also involved in the adhesion, invasion and degradation of host tissues. These genes can also modulate the host immune response by inhibiting IL-2 accumulation in T-cells, by modulating T-cell communication and proliferation 22 . The long fimbriae gene fimA was also up-regulated. It encodes for the fimbriae structure, necessary for adhesion to and invasion of host tissues 23,24 . The up-regulation of fimA can be due to the increase in rgp gingipains, which process the precursor proteins of long fimbriae 25 . An up-regulation of the collagenase prtC gene was also observed. prtC has a role in collagen and nutritional peptide degradation 26,27 . In essence, the virulence genes of P. gingivalis which were upregulated by dead P. intermedia all related to the protein metabolism and to adhesion, invasion and destruction of host tissues. On the other hand, dead P. gingivalis upregulated in P. intermedia the virulence related genes kpsD, ecf, phg, inpA and adpC. The polysialic acid transport protein gene kpsD and the RNA polymerase ECF-type sigma factor ecf gene are involved in the production of a polysaccharide structure which can increase the resistance of P. intermedia against the host defense 28 . The expression of the hemagglutinin gene phg and the cysteine protease interpainA gene (inpA) were also increased. The pgh gene is involved in the hemolytic degradation of red blood cell to obtain hemoglobin and haem as an exogenous iron source for nutrition 29 . The inpA gene is also involved in the haem acquisition by breaking down hemoglobin 30 . It can be hypothesized that these genes are up-regulated in P. intermedia to acquire the iron compounds from inside of the dead P. gingivalis cells. Similar to the upregulated invasion genes in P. gingivalis, the invasin gene adpC, involved in the cell invasion of P. intermedia 31 , was also up-regulated. A group of heat shock proteins (htpG, dnaJ, dnaK, groeL, groeS, clpB) were also up-regulated in the presence of dead P. gingivalis. These proteins quickly up-regulate their expression under stressful conditions and can protect the bacterial proteins against external aggression 32 . In essence, dead P. gingivalis activated genes in P. intermedia which related to nutrition, protection and invasion of host tissues.
In order to verify whether dead bacteria can reduce eukaryotic cell viability on their own or by increasing P. gingivalis and P. intermedia virulence or by the combination of both, a cell viability assay was performed. It is well known that living periodontal pathogens can invade and destroy host tissues 33,34 . It also described that bacterial compounds like lipopolysaccharide (LPS), bacterial DNA and certain bacterial proteins can induce an inflammatory reaction 35,36 . However, the influence of dead bacteria (necrotoxicity) alone or in combination with living bacteria on the host tissues has never been shown. Our results showed a lower cell viability for cells challenged with living pathogenic bacteria together with dead pathogenic bacteria when compared to living pathogenic bacteria or dead pathogenic bacteria alone. The increase in cell death can be explained by the increased expression of genes involved in the adhesion, invasion and destruction of host tissues like rgpA, rgpB, kgp, fimA and prtC in P. gingivalis and adpC in P. intermedia. However, since dead bacteria alone also reduced cell viability, a combined effect of increase of virulence factors together with toxic compounds of dead bacteria like LPS, bacterial proteins and free bacterial DNA should be considered.
In conclusion, the presence of dead bacteria can stimulate the growth of several oral pathogens (necrotrophy), inducing a new phenotype characterized by an up-regulation of bacterial virulence genes (necrovirulence) and an increased cytotoxicity towards host tissues (necrotoxicity). Such behaviour can also be expected from other human pathogenic species present in other microbiological ecologies such as the gut, the skin and the reproductive tract. These data should warn us for the indiscriminant use of broad-spectrum antimicrobials in conditions where there is an unspecific and incomplete eradication of the whole bacterial population. Therefore, the clinical impact of these results should be investigated.  Preparation of dead bacteria. P. intermedia, P. gingivalis, F. nucleatum, S. gordonii and S. oralis were grown on blood agar for 48 hours. These agar plates were exposed to chloroform (Acros organics, Geel, Belgium) vapors for 30 minutes. After exposure, the cultures were removed from the agar plates by a sterile loop and suspended in 3 mL BHI, BHI-2, saliva or serum. This bacterial solution was transferred to an empty Petri dish and exposed to UV light (253 nm) for 30 minutes. The effectiveness of this killing procedure was verified by plating 50 µl of the killed bacterial solutions and control solutions containing living bacteria on a blood agar plate and incubating them for 10 days under aerobic and anaerobic conditions. Absence of bacterial growth confirmed the effectiveness of the killing procedure. Additionally, the killed bacterial solutions and control solutions were subjected to conventional and vitality RT-qPCR. The absence of amplification, when using vitality qPCR on the killed bacterial solutions, additionally confirmed the absence of living bacteria (data not shown). The bacterial concentration was determined by spectrophotometer (OD600, GeneQuant Spectrophotometer, Buckinghamshire, UK) and stock solutions were frozen at −80 °C upon use.
Necrotrophic growth experiments. 3 mL of an overnight culture containing 5 × 10 7 /mL P. intermedia was supplemented with 3 mL of dead bacteria at 10 9 , 10 8 , and 10 7 CFU/mL for 24 hours. A culture without dead bacteria (3 mL fresh BHI-2) was used as a control. After 24 hours of anaerobic growth, the cultures were analyzed by plate counting and qPCR.
For the plating experiments, 100 µL from the different conditions were taken and diluted from 10 −1 until 10 −6 in physiological water containing 9 g/L of sodium chloride. These dilutions were inoculated on blood agar plates and incubated for 48 hours under anaerobic conditions. Afterwards, P. intermedia colonies were counted to calculate the final number of bacteria CFU/mL in each condition. The same setup was used to test other bacterial combinations.

Quantification of bacteria in broth by qPCR.
A conventional qPCR technique was used to quantify the total number of P. intermedia and DNA of dead P. intermedia, P. gingivalis, F. nucleatum and S. oralis. A vitality RT-qPCR was also employed to quantify only the living bacterial species in the planktonic and biofilm experiments using specific primers 37,38 . Both were performed with a CFX96 Real-Time System (Bio-Rad, Temse, Belgium). The Taqman 5' nuclease assay PCR method was used for detection and quantification of bacterial DNA (already developed for the species). Quantification was based on a plasmid standard curve.
Necrotrophic growth in multispecies communities. An overnight culture of a bioreactor derived complex multi-species co-culture of 14 species was centrifuged (1438 × g, 10 minutes) and re-suspended in BHI-2, serum or saliva (1 × 10 8 CFU/mL). 1 mL of this bacterial multi-species solution was inoculated together with 500 µl of 1 × 10 9 CFU/mL dead or living bacteria in BHI-2, saliva or serum. After 24 hours of anaerobic incubation, 1 mL was taken from each well and analysed via vitality qPCR or/and conventional qPCR. 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 µL 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 analysed by conventional and vitality qPCR.
Saliva and serum preparation. Unstimulated saliva was obtained for 1 hour per day from a single male volunteer during several days at least 1.5 hours after eating, drinking, or tooth-cleaning. Saliva samples, collected in sterile 50-mL polypropylene tubes, were centrifuged (30 min, 4 °C, 27,000 × g), and the supernatant was pasteurized (60 °C, 30 min) and re-centrifuged in sterile Eppendorf tubes. The resulting supernatants were dispensed into sterile 1.5 mL Eppendorf tubes and stored at −20 °C. The efficacy of pasteurization was tested by plating processed saliva samples onto blood agar (Blood agar Base I, Oxoid, Basingstoke, UK) supplemented with hemin (5 mg/mL), menadione (1 mg/mL) and 5% sterile horse blood. After 72 hours at 37 °C, no CFUs were observed on either aerobically or anaerobically incubated plates. Human serum was obtained by venepuncture 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 and saliva 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.
Scientific RepoRts | 7: 1107 | DOI:10.1038/s41598-017-01239-9 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 µl/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 +/−0.1. After 24 hours, overnight cultures of S. sanguinis, S. gordonii, S. salivarius, S. mitis, S. oralis, S. mutans, S. sobrinus, A. viscosus, A. naeslundii, P. intermedia, P. gingivalis, F. nucleatum, A. actinomycetemcomitans and V. parvula were adjusted to an OD of 1.4 and added to the bioreactor. During the first 48 hours, the medium was not replaced. After that, the medium was replaced at a rate of 200 mL/24 hours. Virulence gene expression. RNA samples were extracted from the combinations living periodontal pathogen supplemented with 10 9 dead bacteria and from the periodontal pathogen without dead bacteria after 24 hours using RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA quality was analyzed by nanodrop (Thermo Fisher Scientific, Doornveld, Belgium) and the samples were adjusted to the same concentration before transforming RNA to cDNA through PrimeScript 1 st strand cDNA Synthesis Kit (Takara, Shiga, Japan). The expression of virulence genes in the periodontal pathogens was analyzed by qPCR in respect to the house-keeping genes. The primers were extracted from previous publications 31,37,39,40 or designed based on the genome of P. intermedia and P. gingivalis available in the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/) ( Table 4).
Cell culture of human oral keratinocytes. Human oral keratinocyte cell line (HOK-18A) 38 was grown in tissue culture flasks with keratinocyte growth medium (KSFM) supplemented as indicated by the manufacturer (Gibco, Life Technologies Ltd., Paisley, Scotland). Cell culture media were refreshed 3 times a week. After trypsinization, cells were maintained in culture bottles at 37 °C humidified atmosphere containing 5% CO 2 .
Cell viability assay. Cell viability was evaluated using the XTT cell proliferation assay. XTT powder (Sigma-Aldrich, Diegem, Belgium) was dissolved in RPMI 1640 medium (Life Technologies, Gent, Belgium) at a concentration of 1 mg/mL. Phenazine methyl sulphate (PMS) powder (AppliChem, Darmstadt, Germany) was also dissolved in PBS with a concentration of 0.383 mg/mL PMS. A mixture 50 µl XTT together with 1 µl PMS was added to each well. In brief, HOK-18A 41 were seeded in 96-well plates (Grenier Bio-One, Frickenhausen, Germany) at a density of 2 × 10 4 cells per well. The cells were exposed to 5 × 10 7 CFU/mL of P. intermedia, 5 × 10 7 CFU/mL P. intermedia together with 10 8 CFU/mL of dead P. gingivalis, 10 8 CFU/mL dead of P. gingivalis and 10 8 CFU/mL living P. gingivalis for 20 h at 37 °C. The same combinations were used for P. gingivalis except for the concentrations of dead and living P. intermedia which were 10 7 CFU/mL because dead P. intermedia killed the cells in higher concentrations. After being exposed, 50 µl of XTT solution was added on the cells in each well for 4 hours. Afterwards, the formazan released by HOK-18A was measured at 450 nm using a micro-plate reader (Multiskan, Thermo Electron Corporation, Vanta, Finland). Cell viabilities are expressed as the percentages in respect to the control without treatment.

Statistical analysis.
All experiments were repeated on 3 different days. A linear mixed model was fitted with run as random factor and growing condition as fixed factor. When residual analysis by means of a normal quantile plot showed that a Log-transformation was recommended, data were Log 10 -transformed. Comparisons between growing conditions and the control were corrected for simultaneous hypothesis testing according to Dunnett. Data were analysed with S-Plus 8.0 for Linux (Tibco, Palo Alto, CA, USA).