Various biofilm models have been developed for S. pneumoniae over the last years including sorbarod filter models [18, 19] and continuous culture reactor biofilms [17, 20–22]. Simpler models rely on biofilms formed on microtiter plates, with or without exchange of culture medium [7–10, 15, 16, 23, 24, 27, 34]. Since no comparative analysis has previously been done, in this work we compare the impact of quorum sensing in three models.
We have previously described the importance of CSP addition to culture media to obtain stable biofilm after o.n. incubation using a narrow range of CSP concentrations in a model based on low multiplicity seeding of cells [8, 34]. Here we show that pneumococci attach to surfaces during late exponential phase, and that this attachment is competence independent, while the stability of the sessile cell-community is dependent on the addition of exogenous CSP and a functional competence regulatory system. These results are in accordance with previous data on attachment to plastic surfaces influenced by sialic acid  and competence dependent late biofilm [8, 34]. Attachment during late exponential phase of growth is in accordance with many models that identify the signal for formation of sessile communities in nutrient limitation or other stresses [10, 27, 35]. The fact that late addition of CSP has no effect on biofilm stability indicates that the competence-dependent events most probably occur prior, during or early after attachment.
When assaying for competence related phenotypes in the two other biofilm models, the effects of quorum sensing were different. The second microtiter biofilm model, more frequently used in pneumococcal research, relies on incubation of high numbers of stationary-phase cells . In this model, the addition of synthetic CSP was not a necessary, however strains unable to synthesize or sense CSP were found to attach to a lower extent the surface compared to the wt. By microscopic analysis we verified that this phenotype was not due to a reduction in the number of single attached cells, but it was due to a reduction in number and size of surface attached microbial aggregates. Microcolony formation, already described as an important phenotype in pneumococcal biofilm [7, 15, 24], could be restored in comC mutant strains by addition of synthetic CSP to levels similar to wt strains.
The fact that none of the well known genes directly or indirectly regulated by competence has a direct link to attachment of biofilm underlines that effects seen in planktonic exponentially growing competent cells differ from the biofilm stabilisation phenotype seen here . There are parallelisms between our findings and recent work in S. mutans where biofilm formation was also linked to the ComCDE system , although if genomic and genetic data indicate that the S. mutans ComDE is orthologous to the S. pneumoniae BlpRH system and does not directly control transformation [33, 38]. Competence quorum sensing defects in S. mutans were found to determine reduction in biofilm biomass, and addition of CSP partially restored wt biofilm architecture . In contrast to S. pneumoniae these ComCD-dependent phenotypes were correlated to the initial stages of biofilm development .
Biofilm microcolonies are examples of non-homogeneous microbial populations. In this context, our data indicate a significant effect of the competence quorum sensing system on the capacity of pneumococci to form these aggregates. Such aggregation behaviour in a non-homogeneous population is consistent with the observed clumping in a mixture of competent and non-competent cells which depends on the release of DNA into the medium [40, 41]. Correlation of competence, cell clumping and DNA release fit well with the presence of DNA in the extracellular matrix of attached pneumococci and to subsequent sensitivity of pneumococcal biofilm to DNAse [23, 24]. The release of DNA into the extracellular matrix through the endogenous CSP pathway has also been described to have a significant impact on biofilm biomass in S. mutans . We lack a precise molecular characterisation of the events and we cannot exclude that some of the effects may be indirect and determined through an unknown regulatory pathway. The intriguing correlation of the pneumococcal serine-rich repeat protein (PsrP) to adhesion, aggregation and biofilm formation seems not to appear to show any obvious overlap to the observations reported in this work, inasmuch strain D39 and its derivatives do not carry the psrP operon . The neuraminidase upregulation found in this work is also in accordance with the observed impact of sialic acid and the nanAB regulon on pneumococcal biofilm, even if again no obvious correlation can be drawn between the two putatively involved regulatory events . In both cases, conditioning experiments may provide a useful approach to correlate phenotypes as shown in the related species S. mutans and the sialidase-positive S. intermedius [43, 44].
In contrast to the two previous models, the continuous culture biofilm model gave a different result. Here the biofilm formation is not influenced by the competence system, despite gene expression analysis of the competence genes appears to be approximately the same in all models. In contrast to the microtiter models, the reactor model demonstrates a significant impact of the capsule. Decreased attachment of encapsulated strains is in agreement with data of others which carefully documented enhanced adhesion to surfaces and biofilm formation in rough strains [19, 22, 23, 25, 45].