Morphotypes on agar plates at high-resolution
Curli and cellulose are the predominant matrix-compounds in Salmonella enterica serovar Typhimurium (S. Typhimurium) biofilms  (Fig. 1). The disruption of both or either of these components leads to distinct changes in colony morphology on Congo Red agar plates  (Fig. 2A). To analyse these changes at high resolution with AFM, colonies of S. Typhimurium UMR1 and its mutated derivatives were carefully transferred onto a cover slip and subsequently analysed by AFM in contact mode. Our data show that the colonies consisted of small tightly associated cells with a roundish cell shape (approx. 1.2 μm in diameter) (Fig. 2B and Fig. 3). Figure 2B shows that the surface of the wild type and in particular of the mutant MAE52, a strain overproducing curli and cellulose due to a point mutation in the promoter region of the csgD gene, was covered by a layer of extracellular material. This material was not visible on a mutant lacking the global regulator CsgD (MAE51) and a CsgBA mutant (MAE14), lacking expression of the major curli subunit CsgA and the surface-exposed CsgB nucleator (Fig. 2B). Subtle differences were observed between the wild type and a bcsA mutant (MAE222), disrupted in the gene encoding the bacterial cellulose synthase. Recently, Latasa et al. have demonstrated that besides curli and cellulose, CsgD also coordinates the production of BapA, a surface protein required for biofilm formation in Salmonella enterica serovar Enteritidis. We therefore tested the effect of a bapA mutation on colony and cell morphology of Salmonella Typhimurium (Fig. 2A and 2B). Our data show that the wild type and the mutant MAE619 were indistinguishable regarding colony morphology both on a Congo red agar plate as well as on our high-resolution AFM images. This result agrees with the earlier observation that a BapA deficient strain produced similar levels of cellulose and curli as the wild type .
Imaging of Salmonella biofilms
We wanted to analyse the different morphotypes also on biofilms grown on an abiotic surface in a liquid. We allowed the bacteria to form biofilms on the mineral surface mica, which was submerged in a rich growth medium. For immobilisation the samples were air-dried at room temperature prior to AFM and light microscopy analysis. To follow possible changes induced by the dewetting and drying processes, we also analysed the biofilms before drying in their hydrated state with the light microscope. Figure 4 shows that after 24 h the mica surface was entirely covered with biofilm by the wild type strain UMR1. The AFM data show that the biofilm consisted of tightly associated bacteria, similar to the colonies on the agar plate. We noticed however that the cells were longer (approx. 1.7 μm in length) and not as shrunken and roundish as the cells grown at the air-interface on agar (Fig. 2B and Fig. 3). At the edges of the biofilm flagella (approx. 20 nm in height) and some other thinner fimbrial structures (approx. 5 nm in height) could be detected. At some locations extracellular matrix was also seen on and between the cells in the biofilm, but to lower extents than in the colonies grown on agar. The topography data indicate that the biofilm consisted of multiple layers and that the thickness varied between different areas (Fig. 5).
Impact of curli, cellulose and BapA on 24 h biofilms
With the aim to better understand how curli, cellulose and BapA contribute to biofilm formation on submerged slides, we compared the wild type biofilms with those formed by the curli, cellulose and BapA mutants. Figures 4 and 5 show that the curli and cellulose overproducing strain MAE52 formed a thicker and denser biofilm compared to the wild type. The AFM images showed more fimbrial structures at the edges of the cells than in the wild type (Fig. 4). Like with the bacteria grown on the agar plate, large amounts of extracellular material were detectable on and between the biofilm-associated cells. In contrast to MAE52, the csgD mutant MAE51 was deficient to form a biofilm (Fig. 4 and Fig. 5). Only individual cells and insular very small cell aggregates could be observed after 24 hours. The AFM images showed flagella, but no fimbriae or other apparent biofilm components. Also the csgBA mutant failed to form an area-wide biofilm within 24 h, instead, a layer of loosely attached individual bacteria was seen in the hydrated sample (Fig. 4). During the dewetting process these loosely attached cells were arranged into branched colonies on the surface. Similar, but more irregular colonies were observed on the dried biofilm samples of the cellulose mutant MAE222 (Fig. 4). However, light microscope images of the hydrated biofilm show that, in contrast to MAE14, this mutant partly retained the ability to form three dimensional loosely attached cell aggregates. This led us to suggest that curli are more important for the formation of the initial cell aggregates than cellulose. Like in the csgD mutant MAE51, we could not detect any of the thin fimbrial structures in the curli mutant MAE14, demonstrating these structures were made up of curli. On the other hand, wild type levels (or even slightly more) of the extracellular material could be observed in the cellulose mutant MAE222. Probably, this material mainly consists of curli. However, we cannot rule out that another so far uncharacterized polysaccharide, which has previously been suggested to exist , might be one of the compounds of the extracellular material.
Similar to the agar plates we were not able to detect a significant difference between a bapA mutant (MAE619) and the wild type in our liquid biofilm assay (Fig. 4 and Fig. 5). In contrast to the curli and cellulose mutant, the bapA deficient strain formed a dense biofilm within 24 hours and curli expression appeared to be the same as in the wild type.
Monitoring biofilm expression over time
The fact that we were able to monitor the expression of curli and flagella in Salmonella biofilms, prompted us to follow the expression of the extracellular structures during the growth of the biofilm. Biofilms of strain UMR1 were grown for 4, 8, 16 and 24 h and were analysed by microscopy as described above. After 4 h individual cells and only a few sporadically dispersed cell aggregates were observed on the mica surfaces covered with water (Fig. 6). The AFM images revealed that the bacteria were distinctly elongated (up to 4 μm in length) at this stage (Fig. 3 and Fig. 6). Many of them were in the process of dividing as the appearance of division septa indicated. Flagella were detectable in moderate numbers, but no other extracellular structures could be observed. After 8 h a layer of individual cells, which however did not form any aggregates, covered the surface. We assume that the cells were loosely attached to the surface because the dewetting process arranged the cells into a pattern of small periodically dispersed colonies. The AFM data show that the cell length of bacteria grown for 8 h was significantly decreased (approx. 1.7 μm) and no division septa were visible, indicating that the growth rate had decreased. Compared to the 4 hour time point flagellar expression was clearly increased. After 16 h the formation of three dimensional cell aggregates, firmly attached to the surface, had started. At this time point we could also begin to see large amounts of extracellular material that we earlier concluded had curli as the major constituent. After 24 h a confluent three dimensional biofilm was formed on the surface as described above.
Biofilm formation in mutated strains over time
To investigate the effect of overexpression of curli and cellulose on the time course of biofilm formation, we compared biofilms of MAE52 grown for 4, 8, 16 and 24 h to the wild type (Fig. 7). The light microscope images show that the cells started to form aggregates already within 4 h, apparently tightly bound to the surface as they were hardly influenced by the drying procedure. Growth of the aggregates rapidly gave rise to an area-wide biofilm on the surface. AFM analysis revealed that flagella and large amounts of biofilm matrix were produced at all time points (Fig. 7). Noticeably, MAE52 exhibited a more irregular surface structure with indentations than the wild type. We also analysed biofilm formation at the earlier time points for mutants MAE51, MAE14, MAE222 and MAE619 (summarized in Table 2). Interestingly, we were able to detect some other pili-like fimbriae in the curli deficient mutant MAE14 after 4 and 8 h (Fig. 8). Though less abundant these pili were also seen in the bapA mutant MAE619 and the cellulose mutant MAE222, but at no time point in the wild type. The height (6 nm) and the length (over 1 μm) of these features are in accordance with the properties of Type 1 pili, previously characterized by Korhonen et al. .