In this study we have combined high-resolution atomic force microscopy with light microscopy to study the roles of curli, cellulose and BapA for the morphology of bacteria grown as colonies on agar plates and within a biofilm on a solid surface in liquid. Under both growth conditions bacteria were visible on the AFM images as tightly associated cellular units, resembling a tissue-like structure. This characteristic appearance was to a lesser extent visible on scanning electron micrographs and images of ultra-thin sections  as well as on confocal scanning laser microscopy  previously taken of plate grown cells, most probably due to differences in the way of sample preparation and imaging technology.
We noticed that the bacteria grown at the air-interface were rounder, smaller in size and more tightly attached to each other, whereas the biofilm-associated bacteria grown in liquid medium retained a more elongated structure. We were able to detect extracellular matrix, predominantly consisting of curli fibres, as a layer on and between the cells grown on agar. In the liquid biofilm assay curli fibers were visualised as thin fimbrial structures at the edges of the biofilms and on and between the bacteria by the AFM, resembling largely the electron micrographs of curli published by Chapman et al. .
Noticeably, no striking differences in morphology were observed between the curli mutant MAE14 and the CsgD mutant under either condition. Thus, in contrast to curli, cellulose expression could not be directly visualised by AFM. One explanation for this finding could be that the sensitivity of cellulose to the washing and drying procedure makes its visualisation difficult [31,32]. Another possibility is that cellulose is expressed at relatively low levels and that it is not secreted into the environment. Cellulose might be embedded in the LPS layer and thereby alter the ability to adhere to the surface and to other cells. Our light microscopy data from the biofilm assay show that both curli and cellulose are necessary for the formation of a biofilm within 24 h, wherein curli seem to be more important for the formation of the early cell aggregates than cellulose.
The results from the time course of biofilm growth show that the wild type began to form aggregates and to bind firmly to the surface first after curli production had started between 8 and 16 h. Thus, curli seem to be indispensable for the formation of cell aggregates and for making strong attachment to a surface.
BapA has recently been discovered as a large secreted protein in Salmonella Enteritidis, which is loosely associated with the cell surface and required for pellicle formation at the air-liquid interface . Noticeably, the results from our biofilm assays show that neither the biofilm and colony morphology nor the ability to form a biofilm within 24 h on a submerged surface was altered in a bapA mutant in S. Typhimurium. Since BapA is a large surface protein one would expect that the disruption of BapA would cause changes in the cell surface appearance. Similar to the wild type the surface of this mutant was characterized by an orange-peel appearance that already has been discussed in earlier studies [26,33].
An interesting finding was that in some mutants, but not in the wild type, pili-like structures occurred at the early time points of biofilm formation. These pili-fimbriae were most abundantly found in the curli mutant MAE14 and to a lesser extent in the bapA and the cellulose mutants. This finding leads us to suggest that expression of pili, curli and other surface structures might be coregulated.
Several immobilisation strategies have been described in the literature for studying bacteria with AFM, including chemical fixation, air-drying, several kinds of coating or the use of porous membranes [29,34-37]. However, most of them interfere with the natural ability of bacteria to attach to a surface and are therefore not suitable for the study of biofilms. In our study the AFM operations were performed on plate grown colonies and on biofilms. Since colonies are well attached to the agar plate and grow at the air-interface, the colonies could be directly imaged with the AFM, ensuring a simple and non-destructive way of sample preparation, which to our knowledge has not been reported elsewhere. For imaging of the biofilms, grown in a liquid, we air-dried the samples prior to AFM analysis. AFM imaging of dried bacterial samples is well established [26,38-40], leading to the strong attachment necessary for the AFM operation. Moreover, AFM on dried bacteria has been demonstrated to allow for high-resolution imaging of flexible and moving extracellular structures such as flagella, fimbriae and biofilm matrix components , which would be difficult to detect in liquid. According to previous studies drying does not seem to distort or destroy the bacterial morphology . Being aware that the drying process could induce changes in the overall biofilm structure, we analysed the biofilms before and after drying with the light microscope. We noticed that drying caused the formation of small colonies on the surface, when bacteria were loosely attached to the surface, e.g. in some mutants or at the early time points (Fig. 4 and Fig. 6). This should however not affect the parameters we have studied, such as curli, flagella and pili expression as well as cell size and shape.
Our work demonstrates the power of AFM as a useful tool to monitor the expression of genes involved in bacterial morphology during biofilm formation. The power of the methodology would become even more significant, once a way has been found to image biofilms in their hydrated state under water with the AFM. However, there are several challenges that have to be overcome, such as strong interactions between the cells and the scanning tip as well as difficulties in the immobilisation of the hydrated biofilms .