Bacterial biofilms are structured communities of cells enclosed in self-produced hydrated polymeric matrix adherent to an inert or living surface (1). Formation of these sessile communities and their inherent resistance to antibiotics and host immune attack are at the root of many persistent and chronic bacterial infections (1), including those caused by Pseudomonas aeruginosa, which has been intensively studied as a model for biofilm formation (2, 3). The matrix, which holds bacterial biofilms together, is a complex mixture of macromolecules including exopolysaccharides, proteins, and DNA (4). The latter has been presumed to be derived from lysed cells and has not been thought to represent an important component of biofilm structure. However, it has been known for many years that some bacteria, including P. aeruginosa, produce substantial quantities of extracellular DNA through a mechanism that is thought to be independent of cellular lysis and that appears to involve the release of small vesicles from the outer membrane (5, 6). During studies of alginate biosynthesis in P. aeruginosa, we discovered that the majority of the extracellular material that reacted in the carbazole colorimetric assay was not exopolysaccharide but DNA [as determined by its peak absorbance at 260 nm, by electrophoretic display, and by its deoxyribonuclease (DNase) but not ribonuclease sensitivity] and therefore hypothesized that this DNA may play a functional role in P. aeruginosa biofilms. Using a tube ring assay (2), we found that addition of DNase I to the culture medium strongly inhibited biofilm formation (Web fig. 1A)(7), although not bacterial growth per se. We then investigated the effect of DNase I on biofilm formation in more detail using a flow-chamber system (8). Four flow-chamber channels were inoculated with green fluorescent protein (GFP)–tagged P. aeruginosa PAO1, and two channels each were irrigated with minimal medium with or without DNase I. The presence of DNase I in the medium prevented biofilm formation. The channels irrigated with medium without DNase I were extensively colonized after 3 days, whereas the channels irrigated with DNase I–containing medium were essentially without cells, or contained few attached cells, after the same period (Web fig. 1B).

We also investigated whether DNase I could dissolve established biofilms. To this end, we inoculated five flow-chamber channels and irrigated them with minimal medium without DNase I to allow the establishment of P. aeruginosa biofilms of varying age. At various times, the medium was shifted to medium supplemented with DNase I, and the fate of the biofilms was