The ability of upper respiratory pathogens including S. pneumoniae to persist in the nasopharynx and cause chronic disease upon the appropriate conditions may be associated with the ability to form biofilms on mucosal epithelium [7–9]. The presence of structurally complex bacterial biofilms is important because biofilms have been shown to exhibit increased resistance to host immune effectors and increased tolerance to antibiotic treatment [11, 12, 15], and therefore suggest that biofilms may contribute to the persistence of pathogens.
Previous reports have shown that S. pneumoniae can form biofilms in vitro using several different models of biofilm culture [16, 20–26] including a study where CSLM data suggested that some of the strains used in the present study differed according to structural complexity . Moreover in a pair of companion studies we have demonstrated that these strains have vastly different genomic complements  and produce significantly different disease phenotypes in an animal model of infection . We therefore wished to further characterize biofilm formation of 6 of these pneumococcal strains by investigating the kinetics of biofilm formation, biofilm matrix composition, antibiotic resistance and capsule expression.
All clinical isolates developed biofilms containing viable, adherent cells over time under static conditions in vitro. However, biofilm development was highly variable among the different isolates. A multiparameter ranking, formulated to compare biofilm formation among the clinical strains using 3 standard assays commonly used to measure biofilm formation, identified two groups: those with a high biofilm forming index (BFI); BS69, BS72 and BS75, which attached quickly and produced more biofilm than the other strains, and those with a low BFI; BS68, BS71 and BS73 which produced biofilms consisting of adherent cells in small cell clusters with few towers. Our study suggests that a multi-assay approach for the quantification of biofilm formation better addresses strain to strain variation in the context of overall biofilm heterogeneity. Bacteria are frequently categorized as being positive or negative for biofilm formation typically using only a single assay such as crystal violet staining. However, our study suggests that such a binary distinction may lead to an oversimplified conclusion. Clearly, each of the six strains formed biofilms, but to different degrees. In the absence of clinical data correlating the degree of in vitro biofilm formation with infection severity and history, a biofilm etiology, even for a poor in vitro biofilm forming strain, should not necessarily be discounted. Not only was the BFI a useful construct to provide a clearer picture of the relative biofilm forming capacity of each strain, but it was also useful in testing putative biofilm characteristics, since a higher BFI correlated with greater binding with carbohydrate-specific lectins and antibiotic resistance in vitro.
Since SEM suggested that BFI was associated with biofilm structural complexity, biofilms from each pneumococcal strain were assessed for the presence of carbohydrate in the EPS matrix using lectin binding. High BFI strains (BS69, BS72 and BS75) exhibited the most lectin co-localization with the EPS matrix, while strains with a low BFI (BS71 and BS73) demonstrated reduced lectin co-localization, suggesting that polysaccharide is associated with larger multilayered aggregates of attached cells. BS68 exhibited some biofilm towers associated with nucleic acid and larger discrete patches of lectin binding (suggesting carbohydrate) than other low BFI strains. Lectin binding did not necessarily correlate with the pneumococcal capsule, since an unencapsulated strain (R6) which formed good biofilms, also bound lectin, suggesting that carbohydrate is associated with noncapsular EPS.
We further assessed another hallmark of biofilm development; antibiotic resistance. Six day pneumococcal biofilms with a low BFI exhibited a 2-fold increase in resistance to azithromycin compared to planktonic growth. However, strains with a high BFI required significantly higher concentrations to inhibit outgrowth from biofilms. In situ examination of 6 day biofilms treated with 20 μg ml-1 of azithromycin (a concentration that was bactericidal for S. pneumoniae in epithelial cell culture)  showed large clusters of viable bacteria in high ranked BFI strains. Few viable attached cells were observed after azithromycin treatment in low ranked isolates. Although our study did not reflect a standardized minimum inhibitory concentration on biofilm bacteria due to the difficulty of assessing cell density within biofilms, these results support the hypothesis that biofilm-forming pneumococcal strains can persist in spite of antibiotic treatment and may contribute to chronic infection.
The presence of extracellular DNA (eDNA) in biofilms is now well documented in several types of bacterial biofilms including non-typeable H. influenzae, another major pathogen associated with chronic OM [33–36]. We therefore investigated if eDNA was present in the pneumococcal biofilm matrix. All strains showed evidence of eDNA shown by in situ staining of the matrix with the dsDNA stain, PicoGreen, and by a dose-dependent reduction of the pneumococcal biofilm biomass by recombinant human DNase I treatment using Pulmozyme® (dornase alfa), used clinically to treat cystic fibrosis pneumonia. DNase-treated biofilms were significantly reduced in all pneumococcal strains when treated with the clinical dose of Pulmozyme® (1 mg ml-1), exceeding 90% reduction in all but one strain. These results agree with those of Moscoso et al. who showed that DNase treatment reduced biofilms of the unencapsulated S. pneumoniae strain R6, and are consistent with the ability of S. pneumoniae to autolyse and release DNA [24, 35].
Other in vitro studies with pneumococci have shown that capsule expression was associated with decreased biofilm formation [20, 23, 24, 26] and some have reported that biofilm development may select for unencapsulated phenotypic variants [20, 24, 26]. However, S. pneumoniae is also known to phenotypically vary capsule production upon adherence to epithelial cells . Therefore we hypothesized that the capsule operon might be downregulated in pneumococcal biofilms. Real time-qPCR results indicate that cpsA is downregulated in the biofilm up to 10-fold depending on the strain, compared to planktonic cultures regardless of the strain's BFI, suggesting that capsule production is variably modulated during sessile growth. In situ immunofluorescence staining with anti-capsule type-specific antibody further demonstrated that S. pneumoniae growing in biofilms was encapsulated and capsule immunostaining was brightest in biofilm towers compared to adherent cells. These results support the role of cpsA in capsule production [37, 38]. Numerous reports demonstrate capsule phenotypic variation and show that variants with greater amounts of capsule colonize epithelial cells less efficiently [39–44]. Phenotypic variation of the capsule has been demonstrated microscopically during the initial stages of infection where adherent or invasive S. pneumoniae exhibited reduced amounts of capsule compared to cells not associated with epithelial cells . Our results showing downregulation of cpsA in pneumococcal biofilms are consistent with observations that S. pneumoniae modulates capsule production upon adherence.