This study focused on the importance of the nutritional conditions and the structure of subgingival biofilms generated on HA discs in vitro. The alteration of the growth medium by eliminating saliva and increasing the concentration of heat-inactivated human serum affected the biofilms positively as they developed to higher thickness, were more stable and enabled the extensive proliferation of T. denticola, which were observed only in small numbers using media with low or no heat-inactivated human serum. We were able to locate all the 10 organisms by multiplex FISH in combination with CLSM. The biofilms displayed a stratified structure reminiscent of in vivo subgingival biofilms . However, in contrast to the in vivo situation, F. nucleatum was predominant in the basal layer along with streptococci of the biofilms grown in mFUM4. In biofilms cultured in iHS, F. nucleatum was detected as dispersed cells in the top layer. Earlier experiments showed that F. nucleatum has a strong dependency on streptococci, and is only able to establish along with them (data not shown). This observation is in accordance with the finding of co-aggregation studies that identified the ability of streptococci to attach to components of the pellicle, while F. nucleatum was shown to bind to the streptococci and act as a “bridging organism” for other species to colonize the biofilm . The observed difference that F. nucleatum establishes in the basal layer might very well be due to the fact, that all strains were inoculated simultaneously. If no streptococci were added to the inoculum, but added to the biofilms at a later time point, F. nucleatum did not establish in the basal layer but rather after the addition of the streptococci, forming an intermediate layer. In this case, mainly A. oris was detected as an early colonizer (data not shown). Possibly, it would make sense to add the various strains sequentially, simulating the shift from health to disease.
The growth medium affected not only the biofilm composition; it had a strong influence on the rate of biofilm formation as well. The observed delay of biofilm formation in iHS medium could be explained possibly by the blocking of adhesion receptors on the saliva pellicle. It was shown for Streptococcus pneumonia and Escherichia coli that albumin inhibits biofilm formation on various surfaces [15, 16]. It is very likely that this effect also occurs in our model during colonization of the discs. However, even though the initial attachment of the bacteria is prevented to a certain degree, all ten organisms were able to persist on the discs and were not washed away during dip-washing. Independent of the used medium, the biofilms showed a phase with a pronounced increase in thickness and bacterial abundance. This phase took about 20 h regardless of the used medium, however, the medium does affect its onset. Concluding this, it seems that a certain number of bacteria attached to the disc is required to promote “exponential” biofilm formation. Our experimental setup did not allow defining the reason(s) behind this phenomenon. Possibly, it is triggered by quorum sensing, as it was shown for several oral species that AI-2 or CSP signalling is involved in biofilm formation . Alternatively, it could be that early biofilm formation under different nutritional conditions leads to different degrees of biofilm rigidity and therefore to different levels of sensitivity to shear-forces applied during biofilm dip-washing.
The iHS medium produced significantly higher cell numbers of T. denticola per biofilm compared to mFUM4 or SAL medium. However, P. gingivalis and T. forsythia were not affected by the higher serum concentration. This is surprising, since P. gingivalis was reported to profit from gingival crevicular fluid as well as from menaquinone secreted by veillonellae , and since one of the main growth factors of T. forsythia, N-acetyl-muramic acid , should be plenty available in thicker biofilms with probably increased proportions of lysing cells. On the other hand both species are known to be quite fastidious and our data indicate that it will be necessary to optimize further media components to increase their growth rates. S. anginosus, A. oris, and V. dispar showed mathematically significant reactions to the different growth media as well. However, in neither case the differences were greater than one log, which can hardly be considered as “biologically significant”.
The biofilms proliferating in iHS medium showed a consistent structure throughout the replicates and the organisms showed interactions as they could be expected according to literature. Zjinge et al. described three different layers in in vivo subgingival samples . Our model biofilms showed differences between top- and basal layers as well, however, it was not possible to clearly define an intermediate layer. It rather seems that there is a fluent transition between top- and basal layer of the biofilms. The two layers show distinct characteristics. In the basal layer, biofilms were very compact and contained mainly streptococci, some veillonellae and large amounts of F. nucleatum and P. intermedia. The presence of P. intermedia was unexpected as it is in contrast to the in vivo situation where coccoid Prevotella species preferentially colonize the top layer in form of compact microcolonies . The top layer of the model biofilms showed a rather loose structure with a lot of EPS. V. dispar and other cocci were embedded as compact microcolonies in their matrix, while A. oris appeared as loose microcolonies, with EPS surrounding each cell. In some preliminary diffusion experiments, similar to these described by Thunheer et al. for in vitro built supragingival biofilms , it seemed that these loose regions might work as diffusion channels, allowing large molecules to reach the basal layer in less than two minutes (data not shown).
The high abundance of T. denticola along with P. gingivalis and T. forsythia in the top layer was remarkable. The location, combined with the known high pathogenic potential of these species, might indicate a high inflammatory potential of our model biofilms. Particularly striking was to find T. denticola and P. gingivalis to colonize in close proximity, indicating some sort of metabolic dependency. This observation corresponds well with several previous studies. For example, it has been shown in a murine abscess model that the pathogenicity of P. gingivalis was significantly increased in presence of T. denticola. The result was recently confirmed in a murine alveolar bone loss model, where co-inoculation showed a strong response not only for bone loss, but also for P. gingivalis specific T cell proliferation and interferon-γ production . And in yet two other studies P. gingivalis and T. denticola had shown metabolic synergies by exchanging iso-butyric- and succinic acid  and an ability to co-aggregate with the Hgp44 domains of RgpA, Kgp and HagA acting as the key adhesins . Other organisms found in this study in highest density in the top layer but without a specific focal distribution were C. rectus, F. nucleatum and T. forsythia. In the case of C. rectus, a highly motile microaerophilic organism, this meets the expectation. In biofilms grown in iHS medium, it was not possible to detect dense colonies of F. nucleatum in the basal layer by FISH, as it was the case in thin mFUM4 biofilms. There are several factors that could explain this finding. On the one hand, Sharma et al. made the same observation in two species biofilms of F. nucleatum and T. forsythia. Using a live-dead staining, they found mainly non-viable F. nucleatum attached to the substratum, while the bacteria in the upper layer of the biofilms showed a high viability . Further, they observed synergistic growth of these organisms, which could explain the occurrence of T. forsythia together with the active F. nucleatum in the top layer of our biofilms. On the other hand, the effect could be explained by known problems of FISH staining procedure: The probes or the fixative possibly were not able to diffuse through the whole biofilm, or cells had strongly reduced number of ribosomes, leading to an extremely weak, non-detectable fluorescence. It is known that the obtained fluorescence intensity, with a few exceptions, is directly correlated with the growth rate of the target bacteria. The accessibility of the targets is controlled mainly by cell wall properties, which again require to get permeabilized by either the fixative or, in case of gram positive cells, lysozyme . As P. intermedia and streptococci were readily stained at the base of the biofilms, a hindered diffusion of the probes or fixatives through the biofilms does not seem to be the problem. The accessibility of the cells can be sorted out as well, as the signal is very clear in the top layer of the biofilm. Careful examination of the images, by enhancing the contrast settings for the general DNA staining in our samples, revealed structures at the base of the biofilms that very much resembles the well-stained colonies of F. nucleatum observed in less thick biofilms. Combined with the high abundance detected by IF, it seems that F. nucleatum was in fact present in at the base of the biofilms, however, either in a non-viable- or at least non-active state. For future experiments, it might be worth investigating new methods to increase fluorescent signals, in order to obtain a bright staining throughout the whole biofilm. Catalysed reporter deposition (CARD)-FISH , the use of helper oligonucleotides , or designing probes targeting the 23S rRNA  might be solutions. Due to the large size of the horseradish peroxidase used with CARD-FISH, it seems unlikely that this method would be appropriate, and the use of helper oligonucleotides or probes targeting the 23S rRNA seem more promising to reach stronger signals.
One of the major differences to the in vivo situation is that the model biofilms grew without the presence of an epithelial cell layer. Some of the observed differences will be caused by the lack of interactions that occur in vivo. A future project will address this circumstance and aims to incorporate an epithelial cell layer into the model system. The main difficulty in maintaining such a co-culture system is that different growth conditions that are needed to cultivate either epithelial cells or biofilms. While the strict anaerobes in the consortium of the biofilms are very sensitive to oxygen, the epithelial cells do require oxygen for growth. Further, biofilms and epithelial cells do have very different nutritional requirements. In our co-culture experiments performed so far, cells and biofilms were cultured separately and incubated as co-culture after the development of both biofilms and epithelial cells . Current experiments showed, that the biofilm consortium is still able to grow on agar plates after 48 h of co-culture, however, the viability of the bacteria was greatly reduced (data not shown).