Exopolysaccharide Production Is Required for Development of Escherichia coli K-12 Biofilm Architecture

Isolation of an E. coli strain defective in colanic acid production.

We performed mini-Tn10cam transposon mutagenesis (14; P. N. Danese, unpublished observation) on E. coli K-12 in an effort to find mutations that rendered E. coli defective in swarming along a hard agar surface (13). Because of the parallels between the movement of bacteria along surfaces during swarming and the formation of communities of cells attached to surfaces (biofilms), we were interested in examining the effects of certain swarming defect mutations upon biofilm formation.

One of the insertion mutations isolated in the surface-swarming screen (wcaF31::cam) disrupted the wcaF open reading frame via insertion into the 107th codon of the wcaF gene, which normally encodes a protein comprised of 182 residues. Based on sequence similarity and its chromosomal location within the cps (capsule) gene cluster, wcaF has been proposed to be required for production of colanic acid (Fig. 1), an EPS produced by E. coli K-12 (18).

Consistent with this hypothesis, overproduction of RcsF, a positive regulator of colanic acid synthesis (11), in strain ZK2686 [W3110 Δ(argF-lac)U169] resulted in a mucoid colony phenotype, whereas overproduction of RcsF in strain ZK2687 (ZK2686 wcaF31::cam) had no such effect. In addition, in vitro quantification of EPS production (19) demonstrated that our wcaF mutant had severely reduced EPS production (data not shown).

CV analysis of cell attachment to PVC.

Cultures of strains ZK2686 (colanic acid-positive [CA⁺]) and ZK2687 (CA⁻) were grown in Luria-Bertani broth (LB) in polyvinylchloride (PVC) wells. After various periods of growth (Fig. 1), the planktonic cells were removed by vigorous rinsing with water, and the extent of biofilm formation of both strains was analyzed macroscopically by staining with crystal violet (CV), a dye which stains attached cells but not PVC (16). As illustrated in Fig. 2, during the early time points (17 h or less) the CV staining observed for the strain defective in colanic acid production was significantly less intense than that observed for the wild-type strain. However, the CV staining increased over time, and ultimately it more closely approximated that of the wild-type parent (Fig. 2). These initial observations indicated that the production of colanic acid affects E. coli biofilm formation but its absence does not completely abolish surface attachment. Qualitatively equivalent results were obtained when the CA⁺ and CA⁻ strains were allowed to form biofilms in minimal glucose medium (data not shown).

Fluorescence microscopy of CA⁺ and CA⁻ cells.

In order to more precisely understand the effects of colanic acid production (or the lack thereof) on E. coli biofilm formation, CA⁺ (ZK2686) and CA⁻ (ZK2686 containing cpsC::Tn10 [19], cpsE::Tn10 [19], or wcaF::cam mutations) strains were transformed with a green fluorescent protein expression plasmid and analyzed by fluorescence microscopy. The cells (either CA⁺ or CA⁻) were allowed to grow in rich medium for 72 h in the presence of a glass coverslip. Planktonic cells were removed from the coverslip by rinsing with water, and the remaining attached cells were examined by fluorescence microscopy.

The three-dimensional complexity of the wild-type E. coli biofilm can be clearly observed by examining a sequential Z-series scan of all xy focal planes of the biofilm. Importantly, a similar collection of Z-series scans of a colanic acid-defective biofilm illustrates the tightly packed cells and the absence of both significant depth and complex architecture. Video clips of these Z-series scans can be viewed at http://gasp.med.harvard.edu/biofilms/ecoli/colanic.html.

A different perspective of the structures formed on an abiotic surface (by CA⁺ and CA⁻ strains) can also be obtained by viewing a sagittal section of the attached cells (Fig. 3). As illustrated in Fig. 3c, the wild-type strain forms pillars of cells that are approximately 26 μm high. In addition, it is clear that the cell bodies within microcolonies often do not physically interact but instead appear to be suspended above the surface (Fig. 3c). This complex structure is in stark contrast to that observed in the sagittal section of the CA⁻ attached cells (Fig. 3d). This CA⁻ strain forms densely packed structures with extensive cell-surface and cell-cell interactions. Importantly, there is no evidence of the extensive depth seen with the wild-type parent (Fig. 3c). Video clips showing the rotation from the overhead (xy) to the sagittal (xz) perspective (both shown in Fig. 3) of the CA⁺ and CA⁻ attached cells can be viewed at http://gasp.med.harvard.edu/biofilms/ecoli/colanic .html. These video clips provide additional perspectives of the CA⁺ and the CA⁻ attached cells and emphasize the differences between their respective structures. We also note that this collapsed phenotype can be complemented by a plasmid-carried copy of wcaF, indicating that the phenotypes observed are due solely to the disruption of wcaF (data not shown).

Concluding remarks.

A previous report based on sequence similarity led to the hypothesis that wcaF is involved in colanic acid production (18). Specifically, based on its similarity to other genes and its chromosomal position within the cps (capsule) gene cluster, it was hypothesized that the wcaF gene product might function as an acetyltransferase during the synthesis of colanic acid (18). The evidence presented here provides genetic data demonstrating that the wcaF product is indeed required for colanic acid production.

However, the most striking finding reported here is that colanic acid is required not for initial attachment to an abiotic surface but rather for establishing the complex three-dimensional structure of an E. coli biofilm. Indeed, we have examined the initial attachment of both the wild type and the wcaF mutant to PVC via phase-contrast microscopy at times of <2 h, and we observe no difference in initial cellular attachment (data not shown), consistent with the view that colanic acid is required not for surface attachment but for biofilm architecture. This result is dramatically different from the phenotypes observed with other bacterial species. Mutations that abolish EPS production in either V. cholerae (20), S. putrefaciens (Newman and Kolter, unpublished), or S. epidermidis (15) render these strains severely defective in the very initial stages of attachment to abiotic surfaces.

In E. coli K-12, mutations that prevent EPS (colanic acid) production do not block the ability of the cells to initially attach to abiotic surfaces. In fact, E. coli K-12 strains defective in colanic acid production form bacterial films that are one to two cells in depth (Fig. 2B and 3d). These results highlight the fact that the role(s) EPS plays in the formation of E. coli biofilms is, in at least one respect, different from the role (i.e., adhesion) it plays in other organisms. However, it is important to emphasize that despite the successful attachment of E. coli CA⁻ mutants, such mutants do not display the pillars of cells that are typical of biofilms even at times at which the CV staining of the wcaF mutant equals that of the wild type (100 h). We should emphasize that we routinely rinse our biofilms before microscopic examination. Thus, it is possible that colanic acid is not required specifically for establishing biofilm architecture. Rather, it may be required to maintain biofilm architecture in the face of environmental vicissitudes, such as alterations in medium bulk flow.

Regardless of this possibility, our results demonstrate that it is not simply pili, flagella, or other cellular appendages that contribute to the strength of this complex three-dimensional biofilm structure. Rather, colanic acid is an integral part of this elaborate structure.

It is worth noting that the phenotype displayed by the CA⁻ E. coli K-12 strain is analogous to that observed with P. aeruginosa strains that are unable to produce the intercellular signaling compound N-(3-oxododecanoyl)-l-homoserine lactone (8). In both instances, the three-dimensional, complex structure of the biofilm is completely absent. Instead, the cells appear collapsed and tightly packed close to the surface. The similarity of these phenotypes suggests the intriguing possibility that intercellular signaling is an integral part of the induction and/or arrangement of colanic acid within the microcolonies. We are currently investigating this possibility.