Colonization of Abiotic Surfaces

CHRISTOPHE BELOIN, SANDRA DA RE, AND JEAN-MARC GHIGO*

[SECTION EDITOR: JAMES P. NATARO]

Posted August 29, 2005

Groupe de Génétique des Biofilms, Institut Pasteur, CNRS URA 2172, 25 rue du Dr. Roux, 75724 Paris Cedex 15, France

*Corresponding author. E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

Structured communities of surface-attached bacteria are commonly referred to as biofilms (152). Growth as a biofilm is a predominant lifestyle that is recognized as a basic tenet of bacterial behavior and an important aspect of bacterial colonization and impact in most natural and artificial environments. Beside naturally available mineral, vegetable, and animal surfaces, industrial or medical abiotic surfaces such as catheters and indwelling devices are particularly prone to colonization and the development of potentially pathogenic or detrimental bacterial biofilms is therefore commonly associated with health and economic problems.

Bacterial interactions with a surface promotes novel behaviors, and bacterial physiology in a biofilm is very likely to be distinct from that of free-swimming organisms (28, 42, 152). Numerous studies aiming at providing clues about the genetic requirements underlying biofilm formation have been conducted in recent years. In the first two editions of this book, in which the literature was covered up to 1996, the word "biofilm" did not even appear. Since then more than 4,000 papers have been published on the subject, many of them (10%) using Escherichia coli as a model organism.

E. coli is a predominant species among facultative anaerobic bacteria of the gastrointestinal tract, a complex symbiotic ecosystem with the structural characteristics of a multispecific biofilm (19, 115). E. coli is a highly versatile bacterium, which exists both as a harmless gut commensal or as an intra- or extraintestinal pathogen. It is also a common colonizer of medical devices and the primary cause of recurrent urogenital infections (76).

E. coli, therefore, is a relevant model organism for the study of the molecular mechanisms underlying surface colonization. This process requires two essential steps: adhesion to a surface, followed by cell-cell adhesion counteracting the shear forces of the environment, with both steps contributing to the formation of a biofilm (Fig. 1).

Fig. 1Colonization of abiotic surface: from surface adhesion to biofilm formation. Major panel: Confocal scanning laser microscopy (CSLM) of a biofilm formed by E. coli F+ strain expressing the green fluorescent protein (GFP) after 6 h of growth in minimal medium at 37°C. Front, tilted view; bottom and side, x axis profile views. (A) Insets corresponding to scanning electron microscopy (SEM) of bacteria at early colonization stages. (B) Insets corresponding to SEM of bacteria at mature biofilm stage.

Consequently, we will attempt in this chapter to provide an overview of the current knowledge of the genetic analyses aiming at identifying factors involved in both of these two highly related biological processes, with a particular emphasis on studies performed in E. coli K-12. Since many issues related to E. coli biofilms, such as adhesin presentation and virulence, will also be addressed in other sections of this edition, the reader is also referred to them, in addition to many excellent recent reviews of the biofilm mode of growth (48, 52, 73, 84, 95, 145, 152, 167).

EXPERIMENTAL APPROACHES TO BIOFILM FORMATION

Classical experimental approaches developed for the study of agitated liquid cultures, in general, are considered as inappropriate for the analysis of the dynamic and complex nature of bacterial biofilms. This has led to the development of experimental approaches combining classical microbiology, genetic, microscopy, and biofilm experimental models (Fig. 2 and 3) (32, 33, 34).

Fig. 2Microscopic observation of E. coli biofilm formation. (A) Optical microscopy (×40) of E. coli F+ grown on PVC coupon and stained with crystal violet dye. (B) CSLM (×40) of a biofilm formed by E. coli F+ strain expressing the GFP after 10 h of growth in minimal medium at 37°C. Front, tilted view. Sides, x and y axis profile views. (C) Transmission electron microscopy (×2,000) of E. coli F+ grown on microscopy plastic coupon in a continuous-flow microfermentor (see Fig. 3). Bar, 5 μm. (D) Scanning electron microscopy (×10,000) of an environmental E. coli grown on microscopy plastic coupon in a continuous-flow microfermentor (see Fig. 3). Bar, 2 μm.

Fig. 3Experimental approaches to biofilm formation. Experimental models commonly used for the study of bacterial biofilms. (A) Microtiter plate model used to screen mutant libraries for biofilm-negative mutants, starting from a biofilm-positive E. coli strain. Biofilm rings are revealed after coloration with the nonspecific crystal violet dye. (B) Air-liquid interface biofilm pellicle formed by a Salmonella sp. standing culture. (C) Continuous-flow dynamic analysis of biofilm formed by a GFP-expressing E. coli in a flow cell. The biofilm formed after 10 h was analyzed by CSLM (see Fig. 2). (D) Continuous-flow analysis of mature biofilm in a fermentor. The right part shows biofilm formed on Pyrex glass surfaces by an E. coli F+ strain after 24 h of culture.

The amenability of E. coli to genetic analyses makes it a valuable biofilm experimental model. The use of static high-throughput models has been an important breakthrough in biofilm research. Small-scale biofilm cultures grown in 96-well microtiter plates allowed the identification of genes involved in biofilm formation on abiotic surfaces (Fig. 3A) (45, 96, 110). Although this approach is now widely used to study the initial or early stages of biofilm formation, questions addressing late stages or mature biofilm issues are commonly studied using either static or dynamic culture models, such as air-liquid-interface in tubes (Fig. 3B), flow cells (Fig. 3C), or chemostats (Fig. 3C) (49, 119, 120).

The combination of continuous-flow chamber (flow cell) cultures and confocal laser scanning microscopy represents a noninvasive analytic tool for use with strains engineered to express fluorescent proteins (14, 82, 173). These qualitative and quantitative approaches (66) have been used to study E. coli biofilm architecture (119), bacterial activity at the single-cell level within a highly heterogeneous E. coli mature biofilm (138, 147), and gene transfer (85).

Both static biofilm and flowthrough systems have proved to be amenable to a wide spectrum of molecular biology approaches from classical to genomewide genetic analysis (12, 110, 120, 134, 157). These methods have been reviewed recently and, combined with advanced microscopic analysis, they represent the tools of the trade that have been instrumental in revealing the different functions involved in E. coli biofilm formation (32, 33, 34, 48, 97, 112).

THE INITIAL ADHESION STEPS

Motility and Biofilm Formation

In one of the earliest studies using E. coli as a model system, half of the insertion mutants deficient for biofilm formation identified by Pratt and Kolter were found to perturb flagellar functions (110). The authors showed that motility itself and not chemotaxis or direct surface contact by flagella was required for formation of a biofilm. They proposed that the movement generated by the flagella enable the bacteria to reach the surface to overcome repulsive forces between the bacteria and also, possibly, to spread along the surface. This analysis was confirmed in E. coli by Genevaux and coworkers (43), and in several other gram-negative bacteria.

Although this suggests that the requirement for force-generating cell surface organelles is a common theme in biofilm formation, it is not an absolute requirement and nonmotile bacteria can still form biofilms under certain conditions (109). In a nonmotile strain that overexpresses the surface adhesins known as curli (see below), Prigent-Combaret and coworkers showed that flagellar motility is neither required for initial adhesion nor for biofilm development (113). In this case, it is possible that the expression of strong adhesion factors may replace force-generating movements during the initial physicochemical interactions between the adhering bacteria and the surface (28, 42, 109).

Bacterial Surface Properties

Bacterial adhesion to abiotic surfaces is likely to be highly dependent on the physicochemical and electrostatic interactions between the bacterial envelope and the substrate, which is itself often conditioned by the fluids to which it is exposed. Pleiotropic changes in cell surface integrity of the lipopolysaccharide (LPS) structure consistently indirectly affect bacterial adhesion, possibly by modifying the balance of the adhesion forces at the surface (42, 44).

Surface Appendages

Besides thermodynamic aspects, the direct contribution of surface structures to the irreversible attachment of bacteria to surfaces and the subsequent three-dimensional biofilm structure formation that follows has been amply demonstrated.

Type I Fimbriae.

Type I fimbriae (or pili) are filamentous proteinaceous adhesins commonly expressed both by commensal and pathogenic E. coli isolates (130). Type 1 pili can adhere, in a mannose-dependent manner, to a variety of receptor molecules on eukaryotic cell surfaces and are well documented virulence factors in pathogenic E. coli (see EcoSal chapters in section E. coli Adhesins). Several groups have reported that these adhesins are critical for E. coli biofilm formation on abiotic surfaces (16, 55, 92, 101, 110). Mutants in both fimA, the gene encoding the major type 1 pilus subunit, and fimH, which codes for the mannose-specific adhesin located at the tip of the pilus, have been reported to reduce E. coli initial attachment to polyvinyl chloride (PVC) and other abiotic surfaces (12, 110). These findings suggest that the type 1 pili FimH adhesin, besides binding eukaryotic mannose oligosaccharides, may also have a nonspecific binding activity to abiotic surfaces (110). The expression of type 1 pili is induced by adhesion and biofilm formation at early and late stages (12, 120, 134).

While the FimH adhesin itself could be responsible for this nonspecific binding, several studies have indicated that the expression of type 1 pili itself may affect the adhesion of E. coli to abiotic surfaces by altering the composition of the outer membrane (101). Indeed, Otto and coworkers showed that type 1 pili surface contacts mediate a decrease in the abundance of several outer membrane proteins, such as BtuB, EF-Tu, OmpA, OmpX, Slp, and TolC (104). These changes in the envelope probably affect the general physicochemical characteristics of the bacterial surface and thereby influence adhesion (104). Consistently, production of type 1 pili is up-regulated in the absence of OmpX. The decrease in the OmpX level upon type 1 pili-mediated surface contacts may therefore serve as a signal leading to a physiological adaptation in surface-associated bacteria (103).

Curli.

The proteinaceous cell surface curli has been shown to be involved in adhesion phenotypes in E. coli and Salmonella spp. (8, 16, 53, 159). Genes involved in curli production are clustered in two divergently transcribed operons: the csgBA operon, encoding the structural components of curli, and the csgDEFG operon, encoding a transcriptional regulator (CsgD) and the curli export machinery (CsgE-G). In environmental and clinical isolates of E. coli, the synthesis of curli is subject to a tight and complex regulation, whereas the expression of curli is cryptic in most E. coli laboratory strains. However, Vidal and coworkers isolated, in such a domesticated strain, a gain-of-function mutant with an increased capacity for surface adhesion and cell-to-cell interactions due to the constitutive expression of the cell surface adhesin curli (159). They determined that the hyperadhesive phenotype was the result of a point mutation in the OmpR protein that constitutes with the EnvZ protein, a two-component regulatory system that senses variations in osmolarity. This ompR allele (ompR234) leads to a more efficient OmpR-dependent activation of the csgD promoter, which stimulates curli production and biofilm formation in laboratory strains (Fig. 4) (111, 113, 159).

Fig. 4A model for the regulatory network controlling curli production. Curli production at low or moderate osmolarity results essentially from transcriptional activation of the csgD promoter by OmpR. When cells encounter high osmolarity due to an increase in salt concentration, curli production will be repressed by the response regulator CpxR. In an increase of osmolarity due to sucrose, the repression of the csg operon expression is due to the histone-like protein H-NS. The different regulations are indicated by arrows for positive controls or via a line with a bar for negative controls. This schematic representation of the regulation of csg expression results from conclusions drawn in the recent article of Jubelin et al. (74).

Beside the EnvZ/OmpR two-component system, several transcriptional regulators (CpxR, RcsCB, RpoS, H-NS, IHF, Crl) responding to different environmental conditions are involved in the regulation of curli expression through a complex network of interactions between transcription factors and the csg regulatory region (13, 29, 47, 111). This regulatory network is proposed to allow a fine tuning of curli expression that may play a role in the colonization of specific niches by E. coli (111) (Fig. 4).

Antigen 43.

Antigen 43 (Ag43) is the major phase-variable outer membrane protein found in most commensal and pathogenic E. coli. Ag43 is a self-recognizing surface autotransporter protein that has been described as affecting colony morphology, mediating autoaggregation and biofilm formation by promoting cell-to-cell adhesion leading to the three-dimensional development of the biofilm architecture (Fig. 5) (20, 57, 59, 78, 106, 134). Ag43, however, does not seem to be involved in nonspecific adhesion to abiotic surfaces (79) and was even found to be dispensable for adhesion in some experimental conditions (20, 119). The structure, function, and regulation of Ag43 are detailed by Klemm and Schembri (Chapter Type 1 Fimbriae, Curli, and Antigen 43: Adhesion, Colonization, and Biofilm Formation).

Fig. 5Ag43 expression promotes bacterial aggregation and biofilm formation. Ag43 is an adhesin that promotes bacterial aggregation in culture tube (top). This aggregation is due to increased cell-to-cell interactions (middle: SEM microscopy of mature biofilms; bar, 2 μm; inset, optical microscopy of liquid cultures, ×40). The formation of cell aggregates favors three-dimensional biofilm formation (bottom: biofilm formation in a microfermentor; see Fig. 3). (A) E. coli MG1655 that expresses only low levels of Ag43. (B) E. coli MG1655oxyR, a strain that is derepressed for the expression of Ag43.

Conjugative Pili.

Several studies investigating plasmid conjugational transfer indicate that physical contact between donor and recipient cells is highly favored within monospecific or mixed E. coli biofilms, where efficient horizontal transfer of genetic material has been demonstrated (26, 58, 83, 85, 91). Although most laboratory E. coli K-12 strains are poor biofilm formers, the introduction of the conjugative plasmid F in these strains induces the formation of a thick mature biofilm. A mutational analysis of the conjugative transfer apparatus genes demonstrated that this phenotype does not depend on the ability of the plasmids to mediate DNA transfer but it does require a functional conjugative pilus (49). The F-pilus promotes both initial adhesion and biofilm formation through nonspecific attachment to abiotic surfaces and subsequent cell-to-cell contacts, which stabilize the structure of the biofilm (49, 91, 119). Reisner and coworkers also showed that the expression of the F conjugative pilus could functionally substitute for other known adhesion factors such as type 1 pili, Ag43, or curli (119). Plasmid-mediated biofilm production is not restricted to the F plasmids and most tested conjugative plasmids directly contribute, upon derepression of their conjugative function, to the bacterial host capacity to form a biofilm (49). These studies showed that both conjugative and nonconjugative plasmids are likely to carry determinants for biofilm initiation and architecture, which in return will affect the extent of the plasmid-mediated horizontal gene transfer within the biofilm (175). This general connection between conjugation and biofilm formation is consistent with early observations showing that surface contacts positively affect the dynamics of plasmid transfer (137).

Several cell adhesins encoded by genes on plasmids or mobile genetic elements have been characterized in pathogenic E. coli (61, 76), suggesting that the extrachromosomal gene pool (plasmids and other mobile genetic elements) may also constitute an important source of adhesion factors leading to biofilm formation, influencing both the probability of biofilm-related infection and of conjugational spread of plasmid-borne virulence factors (5, 49, 91).

The Adhesive Potential of E. coli.

Genetic analyses have revealed the diversity of genetic factors in E. coli that participate in colonization and biofilm formation on abiotic surfaces. However, these factors can often be replaced functionally or overridden by others, depending on the media and growth conditions. This indicates that few, if any, of these adhesion factors are absolutely required and that E. coli can use multiple adhesion pathways. A recent study demonstrated that four previously uncharacterized E. coli genes (yfaL, yeeJ, ypjA, and ycgV), sharing homologies with the autotransporter adhesin Ag43, lead to a clear adhesion and biofilm phenotype when expressed from a chromosomally introduced inducible promoter. The deletion of the genes coding for these putative adhesins do not alter the adhesion phenotypes of the wild-type MG1655 E. coli strain under laboratory conditions (126). This suggests that these genes may be cryptic under laboratory conditions.

These studies suggest that E. coli K-12 probably possesses a wide and partly unexplored arsenal of surface adhesins with different binding specificities that are expressed under specific physiological conditions, possibly in response to different environmental cues. This "adhesion potential" is likely to be even greater in some pathogenic E. coli isolates, which often have a larger genome than E. coli K-12 (27, 107, 169).

BUILDING THE MATURE BIOFILM

The Biofilm Matrix

One of the most distinctive features that distinguishes biofilms from planktonic populations is the presence of an extracellular matrix embedding the biofilm bacteria (146, 154). Because of biofilm heterogeneity and its dynamic nature, the analysis of the extracellular polymeric substance (EPS) progresses slowly, and little is yet known about the composition of the biofilm matrix (154). The biofilm matrix is essentially composed of water (97%) but also includes exopolysaccharides, proteins, nucleic acids, lipids/phospholipids, absorbed nutrients, and metabolites (48). Secreted polysaccharides have been recognized as key elements that shape and provide structural support for the biofilm (155). To date, three exopolysaccharides have been detected in the biofilm matrix of E. coli and shown to be important for biofilm formation: colanic acid, cellulose, and β-1,6-N-acetyl-D-glucosamine.

Polysaccharidic Composition of the E. coli Biofilm Matrix.

Colanic Acid. Colanic acid is a negatively charged polymer of glucose, galactose, fucose, and glucuronic acid that forms a protective capsule around the bacterial cell. Colanic acid synthesis involves 19 genes located in the same cluster named wca (148). It is induced by the three-component system RcsC/YojN/RcsB and requires an auxiliary positive transcription regulator RcsA (36, 50). Although the signal for the sensor kinase RcsC remains uncharacterized, RcsC seems to respond to complex cues such as desiccation, osmotic stress, level of periplasmic glucans, and growth on a solid surface (38, 100, 140). A recent observation also indicates that colanic acid is induced by near-lethal levels of a subset of ß-lactam antibiotics that may exacerbate the formation and the persistence of a biofilm (128). Consistently, although colanic acid has been reported to impair initial bacterial attachment, its synthesis is up-regulated within biofilms, and its production plays a role in the development of the mature biofilm architecture (21, 54, 112, 113).

Cellulose. Cellulose, a homopolysaccharide composed of UDP units linked by β-1→4 glycosidic bonds has also been found in the E. coli matrix. Outside of the plant kingdom, cellulose has primarily been thought to be only produced by a few bacterial species such as the model organism Gluconacetobacter xylinum. The ability of cellulose to bind chemical dyes such as calcofluor gave a convenient primary screen for cellulose-producing bacteria, and surprisingly, enterobacteria like Salmonella enterica serovar Typhimurium, Salmonella enteritidis, and E. coli also produce cellulose (142, 180). Genetic analysis showed that cellulose synthesis genes are organized as two divergently transcribed operons composed of the bcsABZC and bcsEFG genes, displaying homologies with genes of the bacterial cellulose operon of G. xylinum (125, 142). In the case of Salmonella, cellulose was shown to participate, with the curli fimbriae, in the so-called rough, dry, and red (rdar) morphotype, characteristic of the multicellular behavior described for Salmonella (180). This multicellular morphotype is a widespread feature in Salmonella species (125, 142). Few studies have actually investigated cellulose synthesis in E. coli, and most of what is known of the regulation of cellulose production has been learnt from Salmonella (124, 179, 180). In general, the rdar morphotype is expressed at 28°C and not at 37°C, it depends on AgfD (CsgD in E. coli) via AdrA, and cellulose production is associated with the ability to form a rigid biofilm at the air-liquid interface (142, 180). However these characteristics vary for some serovars of Salmonella (125) and the regulation of cellulose production is highly dependent on environmental conditions. Indeed in ATM (poor) medium S. enteritidis can form a biofilm at 37°C, and the EPS is mainly composed of cellulose whose production does not depend on CsgD or AdrA (41, 142). Thus the regulation of cellulose synthesis may not be as straightforward as previously thought and more studies need to be done in E. coli before drawing general conclusions about cellulose production in Salmonella and E. coli.

β-1,6-N-Acetylglucosamine. A recent report identified a new exopolysaccharide in E. coli matrix, a linear polymer of β-1,6-N-acetylglucosamine (β-1,6-GlcNAc) whose synthesis depends on the pgaABCD locus (166). Expression of pgaABCD and β-1,6-GlcNAc production affects biofilm development by promoting cell adhesion to abiotic surfaces and to other cells. The same type of polymer is known to participate in biofilm formation in Staphylococcus aureus and Staphylococcus epidermidis (86, 88). The pgaABCD operon exhibits features of a horizontally transferred locus and is present in a variety of eubacteria. Therefore, it has been proposed that β-1,6-GlcNAc serves as an adhesin that stabilizes biofilms of E. coli and other bacteria (166).

Role of the Matrix.

Although a hallmark of bacterial biofilms, the role of the matrix is not fully understood. The biofilm matrix offers a constantly hydrated viscous layer protecting embedded bacteria from desiccation. The matrix may also play a significant protective role as a diffusion barrier and a sink for toxic molecules (antimicrobials, hydroxyl radicals, and superoxide anions). The biofilm matrix could also inhibit the wash-out of enzymes, nutrients or even signaling molecules that could then accumulate locally and create some more favorable microenvironments within the biofilm (118, 146, 154). It was also recently shown that S. epidermidis matrix exopolysaccharide helps protect the cells against major components of human innate host defense (161). Since S. epidermidis exopolysaccharide is related to the β-1,6-N-acetylglucosamine found in E. coli matrix, it is therefore possible that the polysaccharides found in the EPS also prevent the recognition of the biofilm bacteria by the immune system. All these aspects of the putative roles of the matrix could contribute to the development of phenotypic resistance of pathogenic E. coli biofilms and lead to persistent infections (6, 75) (see also Hultgren et al. in chapter 150).

Besides protective roles, one of the main functions of the matrix is probably a structural one. The adhesive properties of the matrix allow the bacteria to keep in proximity to the surface and to adhere to each other. Moreover, the interactions between polysaccharides and the other components of the matrix, such as those between cellulose and curli, may participate to the three-dimensional growth of the biofilm (172). Consistently, colanic acid is required for the integrity of the three-dimensional structure of the biofilm (21), and in the case of the β-1,6-GlcNAc, a treatment with metaperiodate or a β-hexosaminidase from Actinobacillus actinomycetemcomitans (DspB), which degrade the β-1,6-GlcNAc, leads to a nearly complete disruption of the biofilm (70, 166). A similar observation was made with cellulose for E. coli strains whose biofilm could be released upon treatment with cellulase (C. Beloin, S. Da Re, and J.-M. Ghigo, unpublished results) (180).

To date, several exopolysaccharides have been found in the E. coli biofilm matrix. While these components may coexist in the matrix, our current knowledge seems to indicate that they may be subject to very distinct regulatory pathways, whose coordinated expression remains to be clarified (see also below).

REGULATORY EVENTS DURING SURFACE COLONIZATION

The identification of genes whose expression phenotypically affects surface colonization has received a lot of attention. This section is devoted to the outcomes of this approach, with a particular emphasis on the regulatory genes implicated in this process (see Table 1).

Table 1Regulatory circuits and signaling molecules involved in E. coli biofilm formation

Regulatory Events during Initial Contact between Bacteria and Surfaces

The initial contact between the bacteria and the surface mainly depends on collision probability that is governed by bacterial trajectory and cell surface interactions that can counteract natural surface repulsive forces. However, past this step, it is clear that the expression of defined adhesion factors, often tightly regulated, is required to produce productive cell surface interactions that will lead to stable biofilm formation. How bacteria know that they are on a surface is still poorly understood. Among possible inducing cues are direct physical contact, perception of extracellular signals, or gradients and bacterium-to-bacterium interactions (56). E. coli has proved to be a valuable model to investigate some aspects of this question.

Sensing the Surfaces

The cpx System Senses the Surface and Neighboring Bacteria.

The two-component regulatory cpxRA system is known to respond to envelope stresses such as overproduction and misfolding of membrane proteins and elevated pH (117). However, until recently, relatively little was known about the physiological role of envelope stress responses. Early adhesion of E. coli cells to abiotic surfaces has now been shown to activate the cpx system through a process called surface sensing (105). This abiotic surface contact induction depends on CpxR, the cognate sensor of the system and also on NlpE, an outer membrane lipoprotein, which has previously been shown to induce the cpx system when overproduced (141). The exact role of the induction of the cpx pathway remains unknown; however, a mutation in the cpxR gene lead to the destabilization of cell surface interactions (105). This result suggests that the cpx system is required for an adaptive response that is necessary to stabilize the contact of the attached cells with the surface. However, this phenotype may also be mediated by a defect in the cellular binding properties required for initial attachment.

Possible roles of the cpx system are (i) the degradation (via the activation of periplasmic proteases like DegP) of subunits of surface polymers that could accumulate in the periplasm after contact between the bacteria and the surface; (ii) the modification of the cell surface composition through the regulation of cell surface protein expression (i.e., OmpC) (24); (iii) the modulation of flagellar gene expression, cpxR being a repressor of motility and chemotaxis genes (23), two pathways that must be switched off to optimize initial attachment and to avoid leaving the recently colonized surface. Consistently, flagellar encoding genes are repressed early after the bacteria reach the surface (114, 120) and the overexpression of the flagellar genes reduces biofilm formation by E. coli (157); (iv) in addition to sensing the surfaces, the cpx system may also sense neighboring bacteria. Components of the cpx system are indeed induced in mature E. coli biofilms where most of the bacteria are in contact with one another rather than with the surface (12). The induction of the cpx pathway may therefore play an important role in the maturation of the biofilm by affecting both the initial adhesion between bacteria and the surface, and the subsequent interactions between the bacteria themselves (12, 105).

The rcs System Senses Membrane Perturbation and Mediates Envelope Remodeling.

The rcs two-component regulatory pathway is composed of the membrane-associated proteins RcsC and YojN, and the cytoplasmic response regulator RcsB. Although the signal for RcsC remains uncharacterized, several studies have shown that this sensor kinase responds to complex signals, including desiccation and changes in osmolarity (100, 140). Recently, the RcsC sensor kinase has been shown to respond to growth on a solid surface by sensing membrane perturbations and therefore to be required for normal biofilm development in E. coli (38). Ferrieres and coworkers demonstrated that the surface induction of colanic acid genes (cps or wca genes) observed in E. coli (113, 114) depended of the rcs pathway, which also represses the expression of surface appendages, such as curli, fimbrial proteins, and Ag43 (38). Flagellar gene expression was also recently described as repressed in E. coli by the rcs two-component signaling pathway (39). Considering the involvement of this system in the regulation of several biofilm formation factors, the induction of the rcs system is likely to cause a remodeling of the bacterial envelope. Therefore, the rcs pathway may well be a key regulatory system controlling the transition from attached cells to mature biofilm.

Recent studies on the modifications in the composition of outer membrane proteins during fimbria-mediated adhesion also demonstrated that such cell surface interactions indeed induce a remodeling of the bacterial envelope (104). Among proteins whose quantity is reduced upon attachment to abiotic surfaces is the OmpX protein (see section on type I fimbriae) (103).

Consequently, it appears that contact with a solid surface induces an adaptative response in E. coli cells leading to stable adhesion. This surface-sensing mechanism appears to implicate several pathways, whose complex overlap and interplay need to be elucidated.

Sensing Microenvironments at an Abiotic Surface

The EnvZ/OmpR Two-Component Pathway Senses Surface Osmolarity.

Surfaces tend to adsorb organic molecules and acquire a conditioning film, which locally increases the osmolarity compared with the surrounding medium. Bacteria entering into contact with surfaces will therefore face a favorable growth environment, especially if the surrounding medium contains low nutrient concentrations. It has been proposed that this growth advantage may be a significant selective force driving growth on abiotic surfaces (18). The EnvZ/OmpR two-component pathway is known to respond to external osmolarity by regulating the transcription of the ompF and ompC porin-encoding genes and it increases surface adhesion in response to moderate increases in osmolarity (111). This EnvZ/OmpR activity may thus favor adhesion in zones supporting high metabolic activity. This effect is mediated by the activation of the regulator CsgD by phosphorylated OmpR. Interestingly, csgD encodes a key regulator of the FixJ family that positively regulates the production of curli and cellulose in E. coli and Salmonella and these two factors are promoting biofilm formation (see "Curli" under "Surface Appendages" and "Cellulose" under "Polysaccharidic Composition of the E. coli Biofilm Matrix," above; see also references 142 and 180). Recently, the modulation of csgD expression by modification of osmolarity was shown to be essentially dependant to the interplay between the two-component systems EnvZ/OmpR and CpxA/CpxR and the histonelike protein H-NS (Fig. 4) (74).

A recent study identified the gene bolA, as a potential regulator of biofilm formation. bolA has been described as being negatively regulated by OmpR (178). It is involved in E. coli morphogenesis and causes round-cell morphology when overexpressed (4). Furthermore, the morphological effect of bolA overexpression depends on an active ftsZ gene product (4) and bolA has been described as a regulator of cell wall biosynthetic or biosynthesis enzymes (129). In minimal medium, mutation of bolA slightly reduces E. coli biofilm formation in 96-well polystyrene microtiter plates, whereas overexpression of bolA strongly induces biofilm formation (160). These results suggest that bacterial adhesion could cause a modification of the cell structure and shape, hinting at a physiological connection between cell morphology, cell division, and biofilm formation. Surface induction of EnvZ/OmpR expression has been proposed to repress bolA and should in turn repress biofilm formation according to the results of Vieira et al. (160). This is however contradictory to the observed csgD-dependent induction of biofilm formation caused by surface sensing via EnvZ/OmpR and supposes that the csgD effect is dominant over the bolA effect in vivo.

AcP and ppGpp, Two Small Molecules Regulating Biofilm Formation.

Among small molecules that coordinate gene expression in response to environmental stimuli, two molecules have recently been identified as signals linking nutrient status to biofilm formation in E. coli: acetyl phosphate (AcP) that accumulates intracellularly in the presence of an abundant carbon source and/or a low oxygen concentration in the medium (174) and ppGpp, the molecule of the stringent response that accumulates upon nutrient starvation conditions (9). A local depletion of oxygen occurring when bacteria reach a surface is hypothesized to be the signal causing intracellular AcP levels to rise. An increase in AcP levels correlates notably with an elevated level of type 1 pili and colanic acid gene expression and with a decreased level of flagellar gene expression, favoring maturation of the biofilm (174). One hypothesis is that AcP could influence biofilm formation by acting as a phospho-donor for response regulators like FimZ, OmpR, or RcsB, that are known to control biofilm-associated genes. In addition, as in the "surface-sensing" mechanisms described above, modifications of the intracellular AcP pool could also lead to bacterial surface modifications. Indeed, genes that respond negatively to high AcP levels include these encoding outer membrane porins (ompF, ompC) and other proteins associated with, or predicted to be associated with the envelope (rbsB, b1996, yqiH, glpD, rfbX, rbsD) (174).

The effect of ppGpp on biofilm formation was assessed by comparing wild-type E. coli K-12 MG1655 with an isogenic relA spoT mutant that does not produce ppGpp (9). Clearly, in nutrient-limited conditions, the absence of ppGpp production caused a decrease in biofilm cell density, therefore signifying that the stringent response is necessary for normal development of E. coli biofilm. However, in rich Luria Bertani broth (LB) medium, where the ability to synthesize ppGpp is also crucial, the reverse was observed, that is, the absence of ppGpp production caused an increase in biofilm cell density. It is likely that some of the ppGpp effects might be linked to the stationary phase sigma factor rpoS (see below) that is positively regulated by ppGpp (46).

Central Carbon Flux Also Regulates Biofilm Formation

Catabolite repression has been recognized recently as a regulatory signal controlling E. coli biofilm formation (71). The presence of 0.2% glucose in rich medium appears to decrease the biofilm biomass. Glucose repression is partially mediated by the cAMP receptor protein, CRP. Indeed, a crp mutant displays decreased biofilm formation abilities compared with the wild type. However, this effect is more pronounced when glucose is added during the initial steps of biofilm formation than in the later stages of biofilm maturation, suggesting that catabolite repression preferentially affects components required in the early stages of bacterial adhesion. Whereas the exact step affected by glucose is unknown, this effect could be mediated by modulation of flagellar gene expression by CRP (143) and glycogen biosynthesis and turnover (via glpCPA operon), two factors that are required for optimal biofilm formation (71).

Jackson and coworkers have shown that CsrA, a protein that regulates swimming motility and glycogen biosynthesis and turnover, also represses E. coli biofilm formation (72). CsrA binds to mRNA and posttranscriptionally regulates gene expression by affecting mRNA decay rates. CsrA action is antagonized by the protein CsrB that binds to CsrA, thereby modulating its activity without modifying its intracellular levels. CsrA also activates csrB transcription thus creating a negative autoregulation loop. CsrA primarily affects biofilm formation through the repression of glycogen metabolism but its regulatory effect on the swimming-motility master regulator flhDC (168) might also explain how CsrA represses biofilm formation. Expression of csrA is sharply decreased a few hours after initiation of growth on surfaces, a profile that is compatible with the decrease of flagellar gene expression upon attachment. On the other hand, csrA expression is reactivated after maturation of the biofilm (2-day-old biofilm). In Pseudomonas putida, Sauer and Camper have shown that flagellar gene expression, while repressed in early stages, is reactivated after biofilm maturation (131). An increase in csrA expression in mature E. coli biofilm might also lead to a resumption of swimming motility. Therefore, there may be a link between increased flagellar gene expression and biofilm detachment after reinitiation of swimming motility. Accordingly, Jackson and coworkers have shown that over-expression of csrA is responsible for biofilm detachment (72).

Recently a relationship between the CsrA/CsrB system, the two-component system BarA/UvrY and biofilm formation has been discovered (127, 156). The sensor kinase BarA of E. coli was first identified as a multicopy suppressor of an envZ defect in the expression of outer membrane proteins (94) and was shown to activate the transcription of rpoS (93). UvrY was subsequently recognized as the cognate regulator for BarA (108). However, UvrY can apparently also be activated by another unknown pathway (156). CsrA can activate uvrY expression by both the BarA-dependent and BarA-independent pathways. UvrY in turn activates csrB expression, thus implementing expression of the CsrA/CsrB negative regulatory loop (156). Mutation of either barA or uvrY attenuates biofilm formation suggesting that BarA and UvrY are necessary for the development of a biofilm (127, 156). Whereas UvrY activates csrB expression, the UvrY effect on biofilm formation is mainly mediated by a still unknown CsrB-independent pathway. As BarA activates UvrY by protein phosphorylation, it might be supposed that BarA activates rpoS transcription via UvrY and that the effect on biofilm formation of both barA and uvrY is mediated by RpoS. However, a recent transcriptome analysis of all two-component regulatory system mutants of E. coli K-12 revealed that mutations of barA and uvrY had contrary effects on rpoS transcription (102). While a barA mutation slightly decreased rpoS transcription, an uvrY mutation increased it. This implies that activation of biofilm formation by BarA and UvrY might in fact involve different pathways and, consequently, it is possible that BarA activates a response regulator other than UvrY (158). There are several possible explanations for how UvrY activates biofilm formation since 123 genes are up- or down-regulated in an uvrY mutant compared with wild-type E. coli (102). As flagellar motility is increased in an uvrY mutant it is possible that UvrY could favor maturation of E. coli biofilms by decreasing swimming motility. The differences in barA and uvrY mutant phenotypes also reinforce the likelihood of a BarA-independent activation of UvrY.

Other Regulators Also Influence Biofilm Formation: H-NS and RpoS

Two regulators, H-NS and RpoS, associated with responses to environmental conditions also play a role in modulating biofilm formation. H-NS is a nucleoid-associated protein that has been shown to regulate a large number of genes in E. coli (about 5% of the E. coli K-12 genome), including numerous cell envelope components, most of them being linked to environmental stimuli including pH, oxygen, temperature, or osmolarity (30, 67). H-NS, for example, is described as an activator of flagellar gene expression (143, 144), a thermoregulator of type I fimbriae expression (31, 99), a repressor of LPS biosynthesis genes (67) and as indirect repressor of colanic acid production genes via its effect on the rcsA gene transcription (139). All these regulatory activities identify H-NS as a possible major regulator of E. coli biofilm formation. However, since a mutation in the hns gene results in a reduction of the growth rate (10), the role of H-NS in the development of mature biofilm could be highly pleiotropic. Nevertheless, H-NS was recently described to be the link between anoxia and impaired adhesion capacities of E. coli (81). Inactivation of hns counteracts increased production of LPS and flagella in response to anoxia and allows E. coli to attach to sand columns even when it is grown under oxygen-limited conditions (81).

H-NS often interferes with the expression of genes that depend on the RpoS sigma factor. This interference occurs both by competing with RpoS for binding to the promoter of these genes but also by indirectly repressing rpoS translation and stimulating RpoS turnover (64, 65). For example, the expression of curli genes is rpoS dependent but can also be mediated by sigma 70 in an hns background (98). Whereas rpoS expression in E. coli seems unchanged between cells grown as a planktonic culture in a chemostat and as a biofilm (1), the role of RpoS in biofilm formation remains controversial. Depending on the experimental setup an rpoS mutation has different effects: a strong negative effect when biofilms are formed in flow cells with morpholinopropanesulfonic acid (MOPS) minimal media supplemented with 0.2% glucose (134), a moderate negative effect when biofilms are formed in a Robbins device with glucose-limited defined GDM medium (1) or in polystyrene microtiter plate with rich CFA medium (72), and a positive effect (three- to sixfold increase in amounts of biofilm) when biofilms are formed in PVC microtiter dish with rich LB medium (17). These results pinpoint the difficulty of comparing studies performed with different experimental protocols and do not consequently allow definite conclusions to be drawn concerning the role of RpoS in the formation of E. coli biofilms.

Modulation of Biofilm Formation through Cross Talk between Extracellular Factors: Type I Fimbriae and Capsules against Ag43 in E. coli

The studies presented above indicate clearly that complex regulation mechanisms are used by bacteria to modulate the expression of different factors required for biofilm formation, especially when these factors are required at different times in biofilm development. For example, type I fimbriae are important for the initial steps of bacterium-to-surface interactions, whereas Ag43 is required later to promote bacterium-to-bacterium interactions. A number of recent papers by the group of P. Klemm clearly describe how the production and activity of these two surface organelles are coordinated. Schembri and coworkers first showed that the overexpression of type 1 fimbriae (and also of P or F1C fimbriae) abrogates the exposition of Ag43 to cell surfaces (135). This effect is mediated by a regulatory effect at the Ag43-encoding gene, flu, whose transcription is enhanced 20-fold when the genes encoding type 1 fimbriae are absent (136). Ag43 is primarily regulated through a phase variation mechanism governed by the concerted action of the Dam methyltransferase (positive regulation) and the OxyR regulator (negative regulation) whose activity is modulated by the redox status of the cell (60, 63, 162, 164) (for more details, see Chapter Type 1 Fimbriae, Curli, and Antigen 43: Adhesion, Colonization, and Biofilm Formation by P. Klemm and M. Schembri). Fimbriation does not affect Ag43 production in an oxyR background where the flu promoter is in an ON situation (135). Until recently flu was thought to be repressed by the reduced form of OxyR whereas OxyR in its oxidized form was not thought to repress flu transcription (51, 62). Two results suggest that a modification of the redox status of OxyR could explain the effect of type 1 fimbriae expression on Ag43 production: first, the addition of the reducing agent DTT counteracts the effect of the deletion of type 1 fimbriae, and second, the overproduction of flagella, which do not contain any disulfide bond, has no effect on Ag43 exposition to the cell surface (135). This suggests that the expression of organelles containing disulfide bonds, such as type 1, P, or F1C fimbriae, could affect the cellular thiodisulfide status and thus modify the redox status of OxyR and the expression of Ag43. However, this hypothesis is not consistent with recent indications that phase variation of Ag43 is independent of the oxidation status of OxyR (163). However, in addition to its role in biofilm formation Ag43-mediated autoaggregation seems to protect cells from oxidizing agents (133). Moreover, independently of Ag43 expression, the presence of fimbriae on the cell surface seems to abrogate the intercellular Ag43-Ag43 interaction that is required for autoaggregation to occur (57). The reciprocal is not true and the Ag43 or OxyR status do not appear to influence fimbriae expression (57). Hence, the correlation between Ag43 expression, type 1 pili, and OxyR status still needs to be clarified.

The presence of type 1 pili is not the only extracellular component to interfere with Ag43 activity. A recent study demonstrated that the presence of capsules like K1 or K5 capsules could block Ag43 functionality (132). Whereas type 1 pili production interferes with both Ag43 expression and interaction of Ag43 molecules, capsules production appears solely to shield sterically Ag43-Ag43 interaction. As a consequence encapsulated cells expressing both capsule and Ag43 are impaired in biofilm formation on a polystyrene abiotic surface compared with noncapsulated cells. Schembri and coworkers also demonstrated that capsulation could interfere with the activity of another autotransporter adhesin, the AIDA-I adhesin (132). This suggests that this type of interference may well be a widespread mechanism that is used by the bacterium to modulate its adhering properties.

Extracellular Signals within the Biofilm

A Role for AI-2 in Biofilm Formation.

The high density of cells achieved in biofilms was a stimulus to investigate what role quorum sensing could play in biofilm formation. This question has been extensively studied in Pseudomonas where an N-acylhomoserine lactone (AHL) quorum-sensing system is thought to be a key component in the maturation process leading to characteristic mushroom-shaped microcolonies and channel system (for a review, see reference 80). Unlike many other members of the Enterobacteriaceae, E. coli does not appear to produce AHLs (3). However, E. coli K-12 strains do secrete the autoinducer-2 (AI-2) quorum-signaling molecule that is encoded by genes of the luxS family and that has been regarded as a universal cell-cell communication signal (176). An lsr-like transporter system has been described recently in E. coli K-12 and this could serve to internalize AI-2 molecules (177). In contrast to what has been observed in Streptococcus mutans (89, 170), disruption of the AI-2 signaling system of E. coli does not appear either to modify biofilm maturation mediated by derepressed IncF plasmids in a flow-cell system (119) or to affect the initial adhesion steps of biofilm formation in PVC microtiter plate (15). However, a furanone-based molecule that is inhibiting the E. coli AI-2 signaling system has been shown to decrease the thickness of a E. coli biofilm formed on steel coupons or on air-liquid interfaces and to increase the percentage of dead cells within the same biofilm (121, 123). Although one cannot rule out the possibility that furanone also inhibits another important molecule for E. coli biofilm development, this suggests that AI-2 signaling could play a role in E. coli biofilm maturation under the experimental conditions used. Accordingly, a mutant of the luxS gene of serovar Typhimurium is impaired for biofilm formation on human gallstones in the presence of bile (116). The role of AI-2-mediated quorum sensing in biofilm formation in Enterobacteriaceae and, in particular, in E. coli therefore remains an open question.

SdiA, a Putative AHL Receptor Activates UvrY.

Whereas E. coli is not known to synthesize AHL and has no apparent AHL synthase in its genome, it contains the sdiA gene that encodes a protein of the LuxR family. LuxR proteins possess one domain for binding N-acylated homoserine lactones and a second domain for binding DNA. An sdiA mutant has been shown to produce threefold less biofilm than a wild-type E. coli strain (156). This effect appears to be mediated by the SdiA activation of the uvrY gene that encodes a regulator necessary for a normal biofilm development (156). While the environmental signal that permits SdiA of E. coli to regulate uvrY expression remains to be determined, the study of SdiA in Salmonella enterica reveals that this protein can behave like an AHL receptor, which may detect signals emanating from other species (90). This leaves us with the possibility that the biofilm formation abilities of E. coli can potentially be modulated by quorum-sensing AHL-signaling molecules from other species interacting with E. coli in natural environments.

Other Extracellular Signals Are Regulating E. coli Biofilm Formation.

Other extracellular signaling systems are reported to modulate biofilm formation in E. coli. For example, the spent medium from exponentially growing E. coli cells shows an inhibitory effect on the formation of an E. coli biofilm, whereas an equivalent supernatant coming from an rpoS mutant enhances E. coli biofilm formation (17). This supports the idea that RpoS is repressing the production of an extracellular factor that promotes biofilm formation. Another diffusible molecule, indole, appears to be important for E. coli biofilm formation. A mutant of E. coli for the gene tnaA that encodes the tryptophanase protein is indeed unable to produce a biofilm in 96-well polystyrene microtiter plates (25). Indole has been described as a potential extracellular signal (165). Genes necessary for indole production (including tnaA) have been shown to be induced by addition of E. coli stationary-phase supernatant (122), suggesting the existence of a complex cross-talk between different extracellular signaling pathways. The exact role of indole in biofilm development remains to be elucidated. However, two targets of indole-mediated signaling are the atsD and gabT genes whose products participate in the degradation of amino acids to pyruvate or succinate (165). Considering that the tryptophanase TnaA is able to catabolize tryptophan, cysteine, and serine into pyruvate, it may well be that the induction of the indole-signaling pathway prepares the cells for a nutrient-poor environment by producing high amounts of energy. This situation may be encountered by bacteria within the deep layers of the biofilm.

Recently, a mutation in the gene coding for a serine acetyltransferase, cysE, that catalyzes the conversion of serine to O-acetyl-L-serine (OAS) was shown to enhance biofilm formation through the reduction of the amount of an extracellular signal molecule. The authors suggest that OAS or other cysteine metabolites may play a physiological role, possibly by activating genes whose expression leads to the inhibition of biofilm formation (153).

THE BIOFILM PHENOTYPE

Planktonic and surface-attached growth modes are simple to distinguish phenotypically. These two lifestyles are thought to require or involve a different gene expression setup, leading to the expression of some of the phenotypic characteristics of the biofilm phenotype.

Gene Expression within Biofilm: Cause or Consequence?

The existence of changes in gene expression within biofilm compared with a nonbiofilm mode of growth was recognized early. It is however important to distinguish between factors required for biofilm formation and factors induced by particular biofilm conditions. Indeed, although biofilm formation requires the expression of specific factors (see "The Initial Adhesion Steps" and "Building the Mature Biofilm," above), major modifications in gene expression patterns within biofilms could also be induced by the drastic environmental changes occurring during biofilm formation. Biochemical and genetic evidence support the hypothesis that bacteria face different conditions within a biofilm as compared with planktonic growth (68, 111, 114). Indeed, biofilm bacteria are likely to be subjected to progressive microaerobic conditions, increased osmotic pressure, pH variation, and decreased nutrient accessibility. These biofilm conditions often have strong similarities with conditions that prevail in stationary phase (planktonic) cultures and, consistently, when the stationary phase character of the bacterial lifestyle within biofilm has been investigated, it has been shown that a significant part of the E. coli K-12 biofilm response involves stationary-phase-induced genes (12, 134). Whether these genes encode functions required or induced by the biofilm conditions is still an open question for most of them. However, a systematic inactivation of genes found to be overexpressed within the biofilm showed that 34 of 54 had no phenotypic consequences, neither in the early- nor late-biofilm stages. This underlines the necessity to recognize that many changes observed in biofilm gene expression could be a consequence rather than a cause of biofilm formation (12).

A Common Pattern of Gene Expression in E. coli Biofilm?

Gene fusion studies suggest that the expression of up to 38% of the E. coli genome is affected by biofilm formation (114). Such evidence for differential gene expression within bacterial biofilms has been provided by recent studies using DNA arrays. These studies indicate that, actually, a lower proportion of the E. coli genome (5 to 12%) is subject to differential expression in sessile versus planktonic life (12, 120, 134). While these studies (with others conducted in other microorganisms) suggest the existence of a common pattern of gene expression in E. coli biofilms and effectively identified some genes required for biofilm formation, a detailed comparison of the genes discovered only revealed a very modest overlap between the different studies. Underscoring the difficulty of comparing analyses carried out with different strains, different experimental setups (biofilm device, medium, presence or absence of flow), different timescale, that is, with different E. coli biofilms, this also raised the possibility that global analyses are not really appropriate to deal with the extreme complexity in time and space that resides within a biofilm (11).

Spatial Heterogeneity, a Hallmark of Bacterial Biofilms

Compared with planktonically growing cultures, biofilms are characterized by their spatial heterogeneity. This spatial heterogeneity is characterized by the differential activity observed in the different areas of the biofilm and constitutes one possible explanation for the recalcitrance of biofilms to biocidal treatments. Whereas, to our knowledge, no direct measurement of bacterial activity has been reported so far in E. coli biofilms, spatial patterns of metabolic or enzymatic activities and also of growth rates measured within biofilms of Pseudomonas aeruginosa or of Klebsiella pneumoniae are probably also applicable to E. coli biofilm (68, 69, 147, 171). This spatial heterogeneity in terms of bacterial activity probably strongly influences the structure and architecture of the biofilm itself. Whereas specific structures like mushroom shape, flow channels, or flat mats can be observed within biofilms depending on nutrients, experimental systems used to grow the biofilms, and flow conditions, a recent study tends to indicate that, in E. coli, microstructures appearing during both initial cell attachment and biofilm maturation seem to present a periodic pattern of cell distribution (2). This suggests that spatial heterogeneity of biofilms might not be controlled by random events and that the elucidation of the mechanisms governing this heterogeneity might be possible.

E. coli is also capable of forming biofilm-specific structures in vivo (see Chapter Uropathogenic Escherichia coli). Uropathogenic E. coli (UPEC) was indeed demonstrated to induce the development of so-called pods containing an important E. coli intracellular bacterial community (6, 75). The same type of spatial heterogeneity observed within in vitro biofilms also appears to exist in biofilms inside the pods. Indeed, for example, immunofluorescent staining of a thin bladder section for type 1 pili and the Ag43 protein shows a strong but heterogeneous labeling (6, 75). Whether these specific structures and the existence of this spatial heterogeneity are responsible for UPEC resistance to antibiotics and immune system attacks remains to be determined.

Resistance of E. coli Biofilm to Antimicrobial Agents

Bacteria growing on surfaces are more resistant to antimicrobial agents than their planktonic counterparts (22, 35, 149, 150). Among other bacteria, sessile forms of E. coli indeed exhibit increased resistance to biocides or antibiotics (7, 37, 40, 151). The basis of such an increased tolerance still remains to be elucidated. Several hypotheses have been proposed to explain such a phenomenon, such as (i) the existence of a reduced growth rate of bacteria within the depths of the biofilm. Whereas slow growth effectively eliminates the effect of antibiotics like penicillin or ampicillin that kill only growing cells, it only reduced the activity of antibiotics that kill nongrowing cells. Another possible explanation is (ii) the exclusion properties of the extracellular polymeric matrix that would trap or modify the activity of the antibacterial agent. Again this hypothesis has been contradicted by several studies, including one performed on a UPEC biofilm that demonstrated that there where no zones of the biofilms where tetracycline failed to penetrate (151). Yet another possible explanation is (iii) the emergence of bacteria expressing a specific survival mechanism rendering them tolerant to a broad range of bactericidal factors, this mechanism being induced by the antibacterial agent itself or by the stressful conditions that prevails within the biofilm. The multiple antibiotic resistance operon marRAB was not induced in E. coli biofilm, thus ruling out the possibility that such a mechanism could explain biofilm resistance (87). It remains possible that this specific survival mechanism leads to the apparition of so-called "persister" cells that would persevere in the presence of the antibiotic and will allow the resumption of growth after the interruption of the antibiotic treatment. Whereas such persisters have been readily detected in E. coli biofilms, their existence is not restricted to sessile bacteria and can be observed in stationary-phase cultures (77).

CONCLUSIONS

The study of surface colonization and biofilm formation represents a rapidly expanding field of investigation. We are now beginning to appreciate how surface contact and growth in structured environments lead to phenotypic adjustments from the bacterial planktonic lifestyle to the sessile behavior. The use of E. coli K-12 to investigate the genetic basis of bacterial interactions with surfaces has led to the identification of a large repertoire of adhesins whose expression is subject to a complex interplay between regulatory networks. These approaches, implemented through the development of appropriate experimental biofilm models, have been instrumental in the partial unraveling of how E. coli (K-12), and also other bacteria, modulate and coordinate their adhesion and biofilm formation capacities, both in space and time. As a consequence, an integrated view of the detailed molecular events taking place during abiotic surface colonization is progressively emerging.

Here again, E. coli faithfully fulfills its time-long role as a molecular biology model. However, these studies have also revealed unsuspected communication and coordinated physiological capacities that can be used by E. coli to operate within a biofilm. Hence, there may be much more about E. coli than just being a convenient laboratory workhorse, and, even for an organism as well studied as E. coli K-12, there may still be many central biological questions to be addressed.

Understanding how E. coli K-12 behaves in complex biofilm communities will certainly contribute to an understanding of how natural commensal and pathogenic E. coli isolates develop. Such an understanding may eventually lead to biofilm control strategies and provide clues as to how this process is connected to microbial infections. However, beyond their clinical relevance, fascinating and unexplored biological resources are also likely to lie hidden in the bacterial biofilm lifestyle. The years to come may well witness a revival of interest for the biology and ecology of E. coli, per se.

Acknowledgments

We thank Søren Molin and Charles J. Dorman for helpful suggestions and critical reading of the manuscript. We thank Philippe Lejeune, Corinne Dorel, and Gregory Jubelin for the authorization to use Fig. 4. Confocal scanning micrographs were prepared by Patricia Latour-Lambert in collaboration with Romain Briandet and Marie-Noëlle Bellon-Fontaine, URBHM, INRA, Massy, France. We thank Brigitte Arbeille and Claude Lebos, Université François Rabelais –Tours, France, for preparation of the scanning and transmission electron micrographs.

C.B. and J.M.G. are supported by the Institut Pasteur and by the CNRS URA 2172 and the Fondation BNP Paribas grants. S.DR. is supported by Sanofi-Pasteur.