We report the identification of an ATP-binding cassette (ABC) transporter and an associated large cell-surface protein that are required for biofilm formation by Pseudomonas fluorescens WCS365. The genes coding for these proteins are designated lap for large adhesion protein. The LapA protein, with a predicted molecular weight of ∼900 kDa, is found to be loosely associated with the cell surface and present in the culture supernatant. The LapB, LapC and LapE proteins are predicted to be the cytoplasmic membrane-localized ATPase, membrane fusion protein and outer membrane protein component, respectively, of an ABC transporter. Consistent with this prediction, LapE, like other members of this family, is localized to the outer membrane. We propose that the lapEBC-encoded ABC transporter participates in the secretion of LapA, as strains with mutations in the lapEBC genes do not have detectable LapA associated with the cell surface or in the supernatant. The lap genes are conserved among environmental pseudomonads such as P. putida KT2440, P. fluorescens PfO1 and P. fluorescens WCS365, but are absent from pathogenic pseudomonads such as P. aeruginosa and P. syringae. The wild-type strain of P. fluorescens WCS365 and its lap mutant derivatives were assessed for their biofilm forming ability in static and flow systems. The lap mutant strains are impaired in an early step in biofilm formation and are unable to develop the mature biofilm structure seen for the wild-type bacterium. Time-lapse microscopy studies determined that the lap mutants are unable to progress from reversible (or transient) attachment to the irreversible attachment stage of biofilm development. The lap mutants were also found to be defective in attachment to quartz sand, an abiotic surface these organisms likely encounter in the environment.
Pseudomonas fluorescens WCS365, a natural soil isolate that is employed as a biological control agent against plant pathogenic fungi (Geels and Schippers, 1983; Simons et al., 1996), has been used to study the molecular genetic basis of biofilm formation. Previous work has shown that a site-specific recombinase, a two component regulatory system, the synthesis of certain amino acids, the O-antigen of lipopolysaccharide, and type IV pili are important for P. fluorescens WCS365 to colonize tomato roots (Simons et al., 1997; Dekkers et al., 1998a,b,c; Camacho-Carbajal, 2001). However, in their natural environment, bacteria are also likely to adhere to abiotic surfaces such as soil particles.
Transposon generated mutations that render P. fluorescens WCS365 defective for attachment to a variety of abiotic surfaces, both hydrophobic (plastic) and hydrophilic (glass) were identified previously (O’Toole and Kolter, 1998b). Here we report the characterization of one class of these biofilm-defective mutants. We have identified an ATP-binding cassette transporter and a large, cell-surface associated protein that are required for P. fluorescens biofilm formation on abiotic surfaces in both static and flow cell systems. These genes are also required for robust biofilm formation on quartz sand, which serves as a model for surfaces typically encountered by soil pseudomonads. Our analyses suggest that the genes encoding this ABC transporter are conserved among sequenced soil pseudomonads, but are absent from pathogenic Pseudomonas strains. We discuss possible roles for this ABC transporter and the cell-surface associated protein in biofilm development.
Initial molecular characterization of mutants defective in biofilm formation
Previous experiments had identified a set of transposon mutations in P. fluorescens that render these strains unable to form a biofilm (O’Toole and Kolter, 1998b). The biofilm-defective mutant strains fell into two broad classes based on their ability to be rescued by changing growth conditions. Class I mutants could be rescued by growth in medium supplemented with certain amino acids, organic acids, and/or exogenous iron. Class II mutants were unable to make a biofilm under any growth condition tested (O’Toole and Kolter, 1998b). The studies here focus on this second class of mutants.
We identified the genes disrupted by the transposon insertion in Class II mutant strains by determining the DNA sequence flanking the transposon, either through arbitrary-primed PCR or by sequencing the region adjacent to a cloned transposon fragment (O’Toole et al., 1999). These DNA sequences were then compared with sequence from the P. putida KT2440 and P. fluorescens PfO1 genome projects. Because the P. putida KT2440 genome is annotated (Nelson et al., 2002) it was used to predict open reading frames as well as to assign putative functions to the genes disrupted in the P. fluorescens WCS365 mutants.
Eight transposon mutants were analysed, and based on extensive sequencing of the DNA flanking the transposon insertions (Fig. 1A and data not shown), all transposons were found to map to genes located in close proximity to each other on the chromosome. Four independent transposon insertions mapped to an open reading frame designated lapA. This open reading frame had been initially identified in strain mus-24, a transposon mutant of P. putida KT2440 defective in adhesion to corn seeds (Espinosa-Urgel et al., 2000). The four transposon insertions mapping to lapA in P. fluorescens WCS365 (lapA18, 51, 53 and 62) are located close to the 5′ end of the gene (corresponding to Domain 2 of the protein, Fig. 1B), whereas the insertion in P. putida mutant mus-24 is close to the 3′ end of the gene (corresponding to Domain 4 of the protein, Fig. 1B). The remaining four mutations analysed mapped to an adjacent gene cluster we have designated lapEBC. The lapB gene is defined by one transposon insertion just 5′ of the start codon (lapB84) and a second transposon located in the middle of the gene (lapB52). The lapC gene (lapC87) and the lapE gene (lapE83) are each defined by one transposon insertion, the lapC insertion is just 5′ of the start codon of lapC, while the lapE insertion is in the middle of the gene. The gene order as shown in Fig. 1A was confirmed in P. fluorescens WCS365 by either sequencing across the junctions of genes or using PCR with primers whose design was based on the P. fluorescens WCS365 DNA sequence (data not shown).
The lap genes are required for biofilm formation
To further define the role of the lap genes in biofilm formation, and to assess the defects in the various lap mutant alleles, we carried out the detailed analysis of biofilm formation.
To characterize the kinetics of biofilm formation we performed a time-course study. The extent of biofilm formation was determined by measuring crystal violet (CV)-stained biomass accumulating on the walls of the microtitre dish over 24 h (O’Toole et al., 1999). The wild-type strain reaches maximum biofilm formation in this assay by 10 h after which the extent of biofilm formed decreases, then remains steady, until the end of the assay period at 24 h (Fig. 2A). The lapB52 and lapE83 mutants were deficient in attachment over the entire 24 h period (Fig. 3A); similar biofilm results were seen for the lapA51 and lapC87 mutants (data not shown). These data demonstrated that the lap mutant strains were not simply delayed in the initiation of biofilm formation. The planktonic growth of all these lap mutants was identical to the wild type (not shown). These results are consistent with those for the P. putida lapA (mus-24) mutant that is also unable to form a biofilm on plastic or glass, and has no planktonic growth defect (Espinosa-Urgel et al., 2000).
Several approaches were utilized to demonstrate that the mutations identified above caused the defects in biofilm formation. First, independent mutations in the lap genes conferred identical phenotypes in several assays, providing strong genetic evidence that these mutations were responsible for the observed biofilm defects. Second, generalized transduction was utilized to mobilize one or more alleles in each lap gene into a wild-type genetic background. Of the over 40 transductants assayed, all conferred the documented antibiotic resistance and biofilm phenotypes of the parental strains, demonstrating 100% linkage between the transposon insertions and the observed phenotypes. Finally, numerous attempts were made to clone the lapEBC region into various vectors (pGEM, pUCP18, pSMC32 and pME6000), however, we were unable to successfully clone this locus. These data suggest that providing the lapEBC genes in multiple copies may be toxic to the cell. Despite the inability to perform complementation assays, the genetic data presented here demonstrate that the lap genes are required for biofilm formation by P. fluorescens WCS365.
Monitoring early attachment in a static system with phase-contrast microscopy
Microscopic analysis of plastic tabs confirmed the results of the CV assay presented above. To visualize the attachment during the early stages of biofilm formation, the wild-type and mutant strains lapA51 and lapB84 were allowed to attach to plastic tabs for up to 5 h (described in the Experimental procedures) then examined by phase-contrast microscopy. At early time points (less than 1 h) there was no difference in attachment between the wild and the lap mutants. For example, in a representative experiment assessing attachment at 40 min, an average of 134 (±40) wild-type cells were attached per field viewed by phase-contrast microscopy compared to 135 (±34) lapC mutant bacteria per field (eight fields were analysed for each strain). However, by 5 h postinoculation, there was a clear difference in biofilm formation between the wild-type and lap mutants. The wild type attached to the surface and formed organized microcolonies comprised of hundreds of cells by 5 h (Fig. 2B). In contrast, the lapA51 and lapB84 strains formed only small clusters of bacteria (5–20 cells), but did not establish the larger microcolonies typical of the wild type. Similar results were obtained for the lapE83 and lapC87 mutants (data not shown).
To determine if lap mutants that attached to the tabs at later time-points were due to a secondary mutation that rescued the lap mutant defect, these cells were removed from the tab by sonication and tested for biofilm formation. Upon retesting, all the lap mutant cells removed from the tabs still had a biofilm formation deficiency identical to that of the original strains tested (data not shown). Therefore, the residual adherence displayed by the lap mutants was unlikely to be caused by the accumulation of a second, compensatory mutation.
Analysis of biofilm formation in a flow cell
To analyse development of a mature biofilm, we grew the wild type and the mutants in a flow cell system. Flow cells provide a constant influx of fresh nutrients, thus sustaining the continued development of the biofilm over many days. Biofilms grown in the flow cell form the characteristic architecture comprised of large macrocolonies surrounded by fluid-filled channels after 1–2 days of incubation.
We examined biofilm formation at 1, 2 and 3 days. As shown in Fig. 3A, the wild type bacteria established a dense monolayer of cells by day 1, and then began developing microcolonies and eventually macrocolonies by days 2 and 3. The macrocolonies formed by the wild-type strain were visible by the naked eye throughout the flow chamber by day 3 (Fig. 3C). In contrast, the lapB52, lapA51, and lapE83 mutants showed a severe defect in attachment to the surface on day 1 (Fig. 3A). The surface of the flow cell was only sparsely covered at this time point. By day 2, the mutant bacteria had formed some small microcolonies, along the edge of the flow cell, which is generally subjected to slower medium flow. The few microcolonies formed on day 2 continue to develop and by day 3 formed some small macrocolonies. Similar results to those observed for the lap mutants shown in Fig. 3 were also observed for lapC87 (data not shown). The lapA (mus-24) mutant of P. putida KT2440 is also defective in biofilm formation in a flow cell system with an architecture similar to the P. fluorescens WCS365 lap mutants (data not shown).
Quantitative analysis of biofilm structure
The images in Fig. 3A show a striking difference in the architecture of the biofilms formed by the wild-type strain and the lap mutants. To quantify the biofilm formed by each strain we utilized the comstat program (Heydorn et al., 2000a). comstat converts the digital information of images (i.e. shown in Fig. 3) into quantitative parameters representing various aspects of biofilm architecture.
The information required for calculating the quantitative parameters of the biofilm is acquired by obtaining optical sections in the z-plane of the biofilms, followed by deconvolution of these images with the OpenLab software package. Biofilms grown for two days in the flow cell were chosen in order to capture the initial stages of biofilm maturation. Twelve image series in the z-plane (a z-series) for each strain effectively captured the heterogeneous structure of each biofilm. Each z-series involved capturing an image every 0.5 µm starting at the attachment substratum and moving up to a final height of 50 µm above the surface. The results of the comstat analysis are displayed in Table 1. The average thickness of the biofilm formed by the wild type (10.8 µm) was ∼ 30-fold greater than the biofilm formed by the mutant (0.37 µm). The bio-volume analysis also demonstrated a similar trend with the average volume of wild type at 9.93 µm3µ−1 m2 and the mutant registering 0.33 µm3µ−1 m2. Furthermore, the wild type covered approximately 34% of the substratum while the mutant occupied only 5%. The surface to volume ratio was calculated as a measure of the amount of biomass exposed to media. The wild type averaged 0.17 µm2µ−1m3 and the mutant 0.64 µm2µ−1m3. Large structured biofilms like those formed by the wild type tend to produce a lower surface to volume ratio than do individual cells. These quantitative data are consistent with the biofilm formation phenotypes revealed by the microscopic images of initial attachment and later biofilm development for wild-type and mutant strains.
Table 1. .COMSTAT: Quantitative analysis of biofilm structure.
The standard deviation for each value is shown in parentheses.
Average thickness (µm)
Bio-volume (µm3 µ−1m2)
Substratum coverage (%)
Surface to volume ratio (µm2 µ−1m3)
The lap mutants are defective for the transition from reversible to irreversible attachment during biofilm development
Microscopy observations from the static and flow cell experiments suggested that the lap mutants were defective in an early stage of biofilm development. To further characterize the biofilm formation defect of the lap mutants we compared the bacterial attachment of the lapC mutant to the wild type over the first 8 h of biofilm development in a flow cell using time-lapse, phase-contrast microscopy. Images were acquired every 3 s over a period of 5 min. A series of representative images illustrating attachment of the wild type and the lapC mutant are shown in Fig. 3B, and the time-lapse movies have been posted on the web (http:www.dartmouth.edugotoolehinsamovieshinsamovies.html).
Both the lapC mutant and the wild type are able to anchor one pole of the cell to the surface (polar attachment). Bacteria attached in this fashion are known as ‘reversibly attached’ because they can readily detach from the surface. The wild-type bacteria eventually become firmly anchored to the surface along the long axis of the cell in a process referred to as ‘irreversible attachment’ (Zobell, 1943; Marshall et al., 1971; van Loosdrecht et al., 1990; Jensen et al., 1992; Fletcher, 1996). The bottom panel of Fig. 3B shows a field of approximately 10 wild-type bacteria which have adhered and remain unmoving over the course of the 12 s in which these images were captured. Time-lapse microscopy over a period of 5 min shows that the wild type is capable of prolonged stable interactions with the surface (see web site above). These data suggest that the wild-type strain is able to make stable interactions with the surface even under conditions of flow. Furthermore, a majority of the wild-type bacteria are attached to the surface across the long axis of their cell body (which serves here as the functional definition of ‘irreversible attachment’).
In contrast to the development observed for the wild-type strain, the majority of lapC mutant bacteria appear unable to progress to the irreversible attachment phase of biofilm formation. As shown in Fig. 3B and the web supplement, many of the lapC mutant bacteria are still anchored by their pole at this time-point and can still be observed moving, spinning rapidly, and/or frequently detaching from the cell surface. After extended incubation in the flow cell, the lapC mutant is able to form small microcolonies, possibly at sites where a cell was occasionally able to tightly attach during initial colonization. Similar results were observed for the other lap mutants (data not shown).
We also quantified the extent of reversible vs. irreversible attachment for the wild type and the lapC mutant. By 8 h postinoculation 94% (±4.5%) of the wild-type bacteria were irreversibly attached to the surface. That is, these bacteria were attached by the long axis of the cell body and did not move over the course of the 5 min time period in which the images for the time-lapse movies were acquired. Only 6% (±4.5%) of wild-type cells were attached by one pole and continued to move during this period. In contrast, for the lapC mutant, only 12% (±4%) of were irreversibly attached, whereas 88% (±4%) were attached by one pole and continued to move for at least some period during this 5 min interval.
Sequence analysis of the lap genes
To begin to elucidate the mechanism underlying the biofilm defect of the lap mutants, we performed a detailed analysis of the predicted proteins encoded by these genes. The complete predicted lapA gene, as deduced from the genome sequence of P. putida KT2440, is 26 kb long and would correspond to one of the largest bacterial proteins (8682 amino acids) with an estimated molecular weight of ∼888 kDa and a predicted pI of 4.1. Thus, we have named this gene lapA, for large adhesion protein (‘lapa’ is also the Spanish name for limpet, a mollusk that lives on seashore rocks and sticks firmly to the rock surface when disturbed). At this stage of the sequencing project of P. fluorescens it has not been possible to assemble the complete lapA gene, however, as detailed below, the size and structure of the protein appears to be very similar to its P. putida counterpart.
Four domains can be clearly distinguished in LapA, two of them being composed of long multiple repeats, which constitute more than three-quarters of the total length of the protein (Fig. 1C). Because no significantly similar proteins of known function could be found in the databases when LapA was compared as a whole we analysed each domain separately. Domain 1, encompassing the first 277 amino acids, contains a predicted non-cleavable N-terminal signal sequence and a transmembrane region (psort program, http:psort.ims.u-tokyo.ac.jp). The function of Domain 1 is unknown, but shows some limited sequence similarity (29% identity) with the N-terminal part of the RTX toxin of the fish pathogen Aeromonas salmonicida (GenBank accession ♯AF218037). The role of the N-terminal domain of this RTX toxin has not been elucidated. Domain 2, from amino acids 278–1178, comprises nine quasi-perfect repeats of 100 aa stretches (Fig. 1C). Sequence similarity was found with the surface protein Bap, from Staphylococcus aureus (Cucarella et al., 2001), a protein that contains 13 nearly identical repeats of 86 amino acids and is involved in biofilm formation by S. aureus. Domain 2 also shows structural similarity with the outer surface protein A of Borrelia burgdorferi (predicted with 123D+, http:123d.ncifcrf.gov; (Alexandrov et al., 1995). Separated from Domain 2 by 18 aa is Domain 3, which is also a large repetitive region, spanning 6400 aa, organized in 29 imperfect repeats of 218–225 aa (Fig. 1C). Some sequence similarity was found with the surface-associated adhesin CshA of the Gram-positive oral bacterium Streptococcus gordonii (McNab et al., 1994). CshA also shows a repetitive structure (13 repeats of 101 amino acids) and is an essential element for oral cavity colonization, participating in co-aggregation of S. gordonii with another oral microorganism, Actinomyces naeslundii (McNab et al., 1994; 1999). Domain 4 (1087 aa) contains several Ca2+-binding motifs similar to those identified in haemolysins and other secreted proteins known to participate in bacterial–eukaryotic interactions (Economou et al., 1990). This sequence analysis suggests that LapA may be a cell surface protein working as a multifunctional adhesin.
Based on sequence analysis, we hypothesize that the lapEBC genes code for an ABC transporter (Fath and Kolter, 1993; Young and Holland, 1999; Dassa and Bouige, 2001). ABC transporters involved in export generally are composed of three separate components – an inner membrane anchored ATPase, a membrane fusion protein and an outer membrane protein (Dassa and Bouige, 2001). LapB is the predicted inner membrane protein of 74 kDa, with several predicted transmembrane regions and a C-terminal ATPase domain containing the canonical Walker box motifs characteristic of this family. LapC is predicted to be 50 kDa, contains a single predicted transmembrane domain, shows similarity to toxin secretion proteins of the HlyD family, and is therefore proposed to be the membrane fusion protein. LapE is a 48 kDa protein predicted to localize to the outer membrane and is similar to AggA (50% identity and 71% similarity over the 750 bp we have sequenced), a protein that was described as a factor involved in agglutination and adherence in Pseudomonas putida strain Corvallis (Buell and Anderson, 1992). Furthermore, LapE contains a domain that is predicted to function as either an outer membrane efflux protein domain or a TolC-like domain (NCBI Conserved Domain Search). This TolC domain is characteristic of the PRT-HLY family of ABC transporters that are involved in protein export in prokaryotes (Dassa and Bouige, 2001). Most proteins exported by this subfamily of ABC transporters contain a series of glycine-rich repeats that forms a calcium-binding site (Baumann et al., 1993; Young and Holland, 1999; Dassa and Bouige, 2001). This so-called ‘repeats in toxin’ or RTX calcium-binding domain has been identified in LapA. Members of the PRT-HLY family of proteins can be further distinguished based on their C-terminal sequence. LapA is likely a member of the HLY subfamily, because it lacks the signature extreme C-terminal motif DXXV (where X is a hydrophobic residue) of the PRT subfamily members (Letoffe and Wandersman, 1992; Ghigo and Wandersman, 1994; Duong et al., 1996). Thus, we predicted that LapB is the ATP-binding element, LapC is the membrane fusion component and LapE is the outer membrane component of an ABC transporter responsible for the export of LapA.
Localization of LapE and LapA
The sequence analyses presented above allowed us to develop a model in which the LapA protein serves as an adhesin to firmly anchor these bacteria to a surface (i.e. promote irreversible attachment). Furthermore, the identification of the LapE, LapB and LapC proteins as components of an ABC transporter suggested a mechanism by which LapA could be transported out of the cell.
To gain insight into the role of the LapA and LapE proteins in irreversible attachment, we examined the localization of these proteins. The LapA protein was predicted to be localized to the outer membrane or cell-surface based on its sequence similarity to known proteins. Fractionation of P. fluorescens inner and outer membranes, followed by Western blot analysis, did not detect LapA in either of these fractions (Fig. 4A, lanes labelled IM and OM). We next tested if LapA was secreted from the cell and/or was loosely associated with the cell surface. To test for a loose association of the protein to the surface of the cell, bacteria were grown as described in the Experimental procedures and 20 ml of the culture was centrifuged, then resuspended in a small volume of buffer resulting in a 50-fold concentration of the bacteria. These concentrated cells were vortexed for five seconds, the suspension re-centrifuged, and an aliquot of the resulting supernatant (designated S2) was analysed by Western blot with anti-LapA antibodies. As shown in Fig. 4A, lane S2, a large molecular weight band was detected for the wild-type strain that is absent from the lapA mutant. The exact size of the band is difficult to estimate as it runs significantly larger than the largest size marker which runs at ∼190 kDa. An identically sized band was also detected in 10-fold concentrated spent supernatant from the wild type but was absent from the lapA mutant strain supernatant. Therefore, LapA appears to be found both in the cell supernatant and in a loose association with the bacterial cell surface, but not in the outer membrane.
To ensure that the centrifugation and vortexing procedures were not rupturing the cell membrane, we also probed these same fractions with antibody to LapE, a predicted outer membrane protein (Fig. 4B, top panel). LapE was not detected in either supernatant fraction, nor was it detected associated with the IM fraction (Fig. 4B). A band corresponding to the molecular weight of LapE (48 kDa) was detected in the OM fraction, but was absent from the lapE83 mutant OM fraction. A weakly cross-reacting band that runs at a molecular weight slightly greater than LapE was present in all samples.
Sequence analysis of the LapE, LapB and LapC proteins showed similarity with ABC transporter components, and therefore suggested a possible role for this putative ABC transporter system in the secretion of the LapA protein. To address this hypothesis, we tested for the presence of cell surface-associated LapA (the S2 fraction) in the wild type and the lapEBC mutant strains. As shown in Fig. 4C (top panel) LapA was detected in the preparation from the wild type, but could not be detected in the S2 supernatant fraction of any of the lapEBC mutants. Furthermore, no LapA was detected in the supernatant of the lapEBC mutants (data not shown). These observations suggest the ABC transporter may be responsible for the delivery of LapA to the exterior of the cell. As expected, no LapA was detected in the lapA51 mutant. In contrast, LapE was present in the OM fractions of all of the lap mutant strains except for the lapE83 mutant (Fig. 4C, lower panel).
. The number of amino acids (aa) known from the genome sequence of P. putida KT4220. The gene designation assigned in the P. putida genome project are given in parentheses (http:www.tigr.org).
. Per cent identity and percentage similarity was determined by comparing each complete or partial ORF to the predicted aa sequence of the putative homologue determined from the complete P. putida KT4220 genome sequence.
. Per cent identity/similarity corresponds to alignments with partial (in parentheses) or complete ORFs obtained from each respective genome project.
The organization of the lapAEBC region is very similar in P. fluorescens PfO1 and P. fluorescens WCS365 (Fig. 1A). In P. putida KT2440, lapE (aggA) is located on a different part of the chromosome from the rest of the lap genes and is associated with a different ABC transporter (not shown). The lapA and lapBC genes are organized in a similar fashion to P. fluorescens PfO1, however, they appear in an inverted orientation relative to the flanking genes. In spite of these differences in the chromosomal arrangement of the genes, LapA seems to play a similar role in P. fluorescens and P. putida. The lapA (mus-24) mutant of P. putida KT2440 is also defective in biofilm formation in static conditions (Espinosa-Urgel et al., 2000) and in a flow cell system, with a biofilm architecture similar to the P. fluorescens WCS365 lapA mutants (data not shown). LapA is not present in P. aeruginosa or P. syringae but proteins with some similarity to those encoded by the lapEBC genes are present, although at different chromosomal locations. It is worth noting that even though lapA is not present in P. aeruginosa, the gene cluster showing the highest degree of similarity lapEBC genes is also associated with a putative large outer membrane protein of ∼ 2500 amino acids (not shown).
The lap genes are required for the colonization of quartz sand
It has been previously reported that the lap mutants of P. fluorescens are deficient for attachment to polyvinylchloride, polypropylene, polystyrene and borosilicate glass (O’Toole and Kolter, 1998). However, P. fluorescens is primarily a soil microorganism, therefore its natural substrate is most likely a variety of sands and soils. Therefore, we tested the wild type and the lap mutant strains for their ability to attach to a surface this bacterium may encounter in its natural soil environment, namely, quartz sand. We implemented assays to follow sand colonization visually and quantitatively.
To visualize bacteria attached to sand, GFP-labelled bacteria were allowed to attach to the sand for 5 h and then the sand was washed and the attached bacteria visualized by epifluorescent microscopy. Figure 5A illustrates the wild type and the lapB52 mutant attached to a grain of sand. The wild-type strain efficiently colonizes the sand particle, whereas the lapB52 strain is much reduced in its ability to colonize. Similar results were seen for the lapC87 and lapA62 mutants (data not shown).
To quantify the number of cells attached to the sand particles, ∼1 × 109 bacteria were allowed to adhere to sand as described above. The number of surface-associated bacteria was determined by vortexing and sonicating the sand to remove attached cells, and the number of bacteria attached to the sand was determined by dilution plating (see Experimental procedures for details). As shown in Fig. 5B, there is a 10-fold difference in attachment between wild type and the lapB52 mutant. Similar results were seen with the lapA51, lapC87 and lapE83 mutants (data not shown). To determine if competition for attachment or complementation would affect this outcome, we mixed equal amounts of wild-type and mutant bacteria (a total of ∼1 × 109 cells) and allowed them to attach to the sand. After rinsing and vortexing/sonication to remove the attached bacteria, we found that approximately 80% of the bacteria attached to the sand were the wild-type strain (Fig. 5C). Similar results were seen for lapA51, lapC87 and lapE83 (data not shown). These data also indicate that the wild-type strain is unable to rescue the biofilm formation defect(s) of the lap mutants.
Cell-to-surface interaction events mark the early steps in the development of a mature biofilm. It has long been proposed that early attachment events first involve reversible attachment to a surface, marked by the transient interactions of one pole of the bacterium with a substratum (Zobell, 1943; Marshall et al., 1971; van Loosdrecht et al., 1990; Jensen et al., 1992; Fletcher, 1996). Following up on these previous studies, Sauer et al. (2002) made detailed observations of early biofilm development, and also observed reversible and subsequent irreversible attachment steps by P. aeruginosa. In the studies presented here, microscopic analysis of early attachment events in P. fluorescens WCS365 demonstrated that this pseudomonad also undergoes the same two step early attachment pathway. In the wild-type strain, cells undergo transient polar attachment followed by subsequent undefined events that lead to a so-called irreversible attachment. The lap mutant strains are capable of initial attachment to a degree that is indistinguishable from the wild-type strain, thus they appear to have no defects in reversible attachment. However, the lap mutants appear to be unable to progress normally to the irreversible attachment step of biofilm development. This phenotype is most clearly observed in time-lapse movies obtained from flow cell studies of the wild type and lap mutant strains. Our studies show that LapA and the lap-encoded ABC transporter, which is required for LapA to be exported from the cell, are required for irreversible attachment. To our knowledge, this is the first report of genetic determinants that are necessary for, and define, the transition from reversible to irreversible attachment.
Despite the apparent defect in irreversible attachment, the lap mutants can eventually form small clusters of bacteria on the surface, however, these mutant bacteria are unable to develop the architecture observed for the wild-type bacteria. We demonstrated that the attachment of these few lap mutant cells to the surface was not due to a secondary mutation, therefore the lap mutants may use an undefined pathway to irreversibly attach to a surface after prolonged incubation. Microcolonies may eventually form as a consequence of cell-to-cell adherence (suggesting that the Lap gene products may not be required for these interactions) and/or cell division.
What role does the putative lapEBC-encoded ABC transporter and the associated LapA protein play in biofilm development? One possible role for LapE, a predicted outer membrane protein with sequence similarity to the AggA adhesin (Buell and Anderson, 1992), may be as an attachment factor required for early biofilm development. However, fractionation of the lap mutants and Western analysis revealed that the LapE protein is localized to the outer membrane in all strains but the lapE mutant, suggesting that proper localization of this outer membrane protein is not sufficient for biofilm formation. In contrast, any mutation within the lapEBC cluster resulted in the loss of any detectable LapA associated with the cell surface or in the cell supernatant. These data are consistent with a model in which the lapEBC-encoded ABC transporter is required for export of LapA outside of the cell. Particularly intriguing is the identification of a putative signal sequence at the N-terminus of LapA, which is not typical of proteins transported by ABC systems (Young and Holland, 1999; Dassa and Bouige, 2001). One possibility is that this secretion signal is cryptic and not typically utilized for transport of LapA. Another possibility is that LapA is transported into the periplasm by the Sec-dependent transport system, and then delivered outside of the cell by the ABC transporter. The detailed analysis of the mechanism by which LapA exits the cell awaits future studies. Taken together, these data are consistent with a hypothesis wherein LapA, alone or in association with LapE, serves to promote stable adhesion of P. fluorescens WCS365 to a surface early in biofilm development.
A role for ABC transporters in cell-to-surface and/or cell-to-cell interactions and biofilm development has been proposed in other organisms. An ABC transporter is important for attachment and virulence of Agrobacterium tumefaciens on carrot cells (Matthysse et al., 1996). In a recent report Sauer and Camper (2001) showed that in another soil pseudomonad, P. putida, the potB gene, which codes for an ABC transporter component, is upregulated in the early stages of biofilm formation on silicone. This observation suggests a possible role for the PotB ABC transporter in early biofilm development, but this has not been demonstrated experimentally. In the oral microbe Streptococcus gordonii, an ABC transporter was shown to be important for self co-aggregation (cell-to-cell interactions) in vitro (Kolenbrander et al., 1994). Taken together, these data indicate that a role for ABC transporters in biofilm development may be conserved across Gram-positive and Gram-negative organisms, perhaps for the purpose of secreting cell surface adhesins.
All of the lap mutants of P. fluorescens WCS365 and the lapA mutant of P. putida were shown to be defective for attachment to a number of plastics (polyvinylchloride, polypropylene, polycarbonate and polystyrene) as well as borosilicate glass (O’Toole and Kolter, 1998b; Espinosa-Urgel et al., 2000). Here we report that the lap mutants are also defective for attaching to quartz sand, a substrate they are very likely to encounter in the environment. The P. putida mus-24 (lapA) mutant was isolated as defective for adhesion to corn seeds, and is also impaired in biofilm formation and competitive root colonization (M. Espinosa-Urgel, unpubl. obs.). These broad phenotypic effects on attachment to biotic and abiotic surfaces caused by mutations in LapA could be explained in two ways. The interactions mediated by this protein might be somewhat non-specific, therefore LapA may act as a ‘general purpose’ adhesin. Alternatively, the multiple domains may in fact be different binding domains, each promoting adherence to a set of substrates. The repeats in LapA (Domains 2 and 3) are reminiscent of those found in adhesion proteins of Gram-positive bacteria involved in biofilm formation or in cell–cell interactions, whereas Domain 4 shows similarities with calcium-binding proteins and haemolysins, that play a role in cell–host interactions during pathogenesis.
The presence of the lap genes in three different soil isolates, and their absence from two pathogenic Pseudomonas strains, suggests that these genes may be specifically necessary for biofilm development by only a subset of pseudomonads. Current studies are exploring the extent of conservation of the lap genes and their organization in a wide variety of soil pseudomonads. To date, no ABC transporter required for biofilm formation by P. aeruginosa has been identified. This finding, along with the data presented here, suggest that although both pathogenic and non-pathogenic pseudomonads make biofilms, they may utilize distinct mechanisms to transition from reversible to irreversible attachment during early biofilm development.
Bacterial strains, plasmids, and culture conditions
Pseudomonas fluorescens was grown in Luria–Bertani (LB) or in minimal media, as specified, at 30°C. The minimal salts medium used was M63 (Pardee et al., 1959) supplemented with MgSO4 (10 mM) and either glucose (0.2%) or citrate (0.4%), or AB10 media without trace minerals (Tolker-Nielsen et al., 2000). Pseudomonas putida was grown in LB and AB10 at 30°C and E. coli was grown in LB at 37°C. Antibiotics were added at the following concentrations: (i) E. coli: gentamycin (Gm), 10 µg ml−1; chloramphenicol (Cm), 30 µg ml−1; (ii) P. fluorescens: Gm, 50 µg ml−1; kanamycin (Kn), 250 µg ml−1. Generalized transductions were performed as described (Jensen et al., 1998). Plasmid pSMC21 is derived from pSMC2 (Bloemberg et al., 1997), expresses the green fluorescent protein (GFP) under a constitutive promoter, and carries both Ap and Kn resistance markers. Plasmids pSU21 (Martinez et al., 1988), pME6000 (Itoh et al., 1988), pSMC32 (O’Toole et al., 2000b), pUCP18 (Schweizer, 1991) and pGEM (Promega) were utilized for cloning experiments.
Sequence of the DNA flanking transposon insertions was determined by arbitrary primed PCR (O’Toole et al., 1999). Selected transposon insertions were cloned to determine additional DNA sequence flanking the element. Chromosomal DNA was prepared as described (Pitcher et al., 1989), digested with EcoRI and ligated into pSU21 previously digested with EcoRI. The ligations were electroporated into E. coli JM109 electrocompetent cells, plated on LB supplemented with Cm, then replica printed onto LB supplemented with Cm and Gm. The Cmr Gmr colonies were purified, plasmid DNA was prepared, and the plasmids were sequenced with the Tn5Ext primer (O’Toole et al., 1999). Polymerase chain reaction using primers to sequence derived from P. fluorescens WCS365 was performed to confirm the gene order inferred from sequence analysis (Fig. 1A).
Initial biofilm formation was measured using the microtiter dish assay system performed as described previously (O’Toole and Kolter, 1998a,b; O’Toole et al., 1999) using minimal M63 medium with glucose (0.2%) or citrate (0.4%) as the growth substrate. The visualization of bacterial cells attached to PVC was performed as previously reported (Bloemberg et al., 1997) except cultures were incubated at room temperature for up to 5 h before analysis. The once-flow through continuous culture flow cell system was assembled as described (Christensen et al., 1999) and modified AB10, as described above, was utilized as the growth medium.
To address whether secondary mutations were responsible for allowing the lap strains to form microcolonies at later time-periods, the attached lapA mutant cells were tested for the ability to make biofilms after re-culture under planktonic conditions. The lap mutant bacteria were allowed to attach to PVC tabs for 24 h, after which the tabs were rinsed and placed into an eppendorf tube containing LB. The bacteria adhered to the tab were removed by two alternating series of vortexing (10 s) and sonication (10 s, Tabletop Ultrasonic Cleaner, FS-60, Fischer Scientific) followed by a final 10 s vortex (as described Gardener and de Bruijn, 1998). The LB and any bacteria removed from the tab were used to inoculate an LB culture that was outgrown for 24 h, and then a standard biofilm assay was performed using these cultures as an inoculum.
Bacteria were grown overnight in LB and subcultured into M63 plus citrate (0.4%) at a 1:5 dilution. Sand was placed in the bottom of the well in a 24-well plate and covered with the bacterial suspension. The plates were placed on a shaker at room temperature for four hours. A sample of sand was removed from each well, placed in an eppendorf tube, and washed five times with 500 µl of M63. A sand sample could be removed at this point for visualization of bacteria attached to sand. Epifluorescent microscopy indicated that wild-type bacteria formed a monolayer of cells on the sand particles at this time under the growth conditions described (see Fig. 5). Quantification of bacteria attached to sand was performed as follows: 50 µl of M63 was added to the sand containing tube, and the bacteria were removed from the sand by two alternating series of vortexing (10 s) and sonication (10 s, Tabletop Ultrasonic Cleaner, FS-60, Fischer Scientific) followed by a final 10 s vortex (as described by Gardener and de Bruijn, 1998). Ten microlitres of suspension was removed and used for dilution plating. Bacterial counts are normalized to grams of sand assayed.
To determine the efficacy of the vortex/sonication regimen, the percentage of bacteria removed from the sand was determined. One sample of sand was treated as described above, whereas a second sample did not receive the vortexing and sonication treatment. Both treated and untreated samples were washed an additional three times with 500 µl of M63. Fifty microlitres of M63 was added to the sand post treatment, the samples were incubated for 2 h and dilution plating performed to determine the number of bacteria present (e.g. those bacteria shed from the sand and growing planktonically). These control experiments demonstrated> 99.9% of the bacteria attached to the sand are removed during the vortexing and sonication steps.
Epifluorescent and phase-contrast microscopy were performed with a Model DM IRBE microscope (Leica Microsystems) equipped with an Orca Model C4742-5 CCD camera (Hamamatsu). Images were acquired and processed on a Macintosh G4 loaded with OpenLab 3.1 software (Improvision). comstat analysis was performed as described (Heydorn et al., 2000a, b).
Protein localization and Western analysis
Samples for Western analysis were prepared as follows. Twenty millilitres of minimal M63 medium supplemented with citrate and MgSO4 were grown shaking at 30°C for ∼16 h. The cultures were centrifuged for 10 min at 6000 g and 1 ml of the supernatant was removed and TCA precipitated as described (Kunitz, 1952). The precipitated protein pellet was resuspended in 100 µl of resuspension buffer (Tris-HCl, 20 mM, pH 8 plus 10 mM MgCl2) – this sample is designated ‘S1’. The bacterial pellet from the 20 ml culture was resuspended in 400 µl of resuspension buffer and vortexed for 5 s. The samples were centrifuged for 5 min at 12 200 g and the supernatant was collected (the supernatant of this sample is designated ‘S2’). The inner and outer membranes were separated as previously described (Lohia et al., 1984) with some modifications. The cells were grown as described above, then centrifuged for 10 min at 6000 g and the pellet resuspended in 1.5 ml of PBS. A Thermo Spectronic French press mini-cell was used to lyse the cells by processing twice at 20 000 p.s.i. Next, the samples were spun at 12 200 g to pellet unbroken cells. The supernatant was removed and spun at 100 000 g for 60 min at 4°C. The pellet was resuspended in 100 µl of Hepes buffer (10 mM, pH 5.7) and incubated with 100 µg ml−1 DNAse and RNAse for 20 min at 25°C. Nine hundred microlitres of urea (4 M) was added to the samples to solubilize the inner membrane. The samples were spun at 100 000 g for 60 min at 4°C. The supernatant fraction containing the inner membrane was removed and the pellet (outer membrane) was washed with cold H2O and centrifuged for another 60 min at 4°C at 100 000 g. SDS loading buffer was mixed with each sample, followed by heat denaturation at 75°C for 10 min. The samples were resolved on a gradient polyacrylamide gel (4–15%) at 20 mA. The protein was transferred in a to a nitrocellulose membrane in transblot buffer as described (Towbin et al., 1979). Western blots were developed with ECL Western detection reagents (Amersham).
We thank Christian Weinel for providing us with P. putida sequence data prior to publication. We thank referee ♯3 for suggesting the suppressor analysis experiment. This work was supported by grants from the NSF (CAREER 9984521) and The Pew Charitable Trusts to G.A.O. G.A.O. is a Pew Scholar in the Biomedical Sciences. We also acknowledge grant BMC2001-0576 from the Plan Nacional de I + D + I to M.E.U. M.E.U. is the recipient of a grant from the Ramón y Cajal Program (MCYT).
Time-lapse movies of biofilm-grown bacteria were made from images captured every 3 s over a period of 5 min. In this experiment, the biofilm was 8 h old and the flow direction was from the bottom to the top of the image. Most of the wild-type bacteria are attached to the surface along the long length of the cell and do not move during the course of the 5 min movie. In contrast, most of the lapC mutant bacteria are attached by one pole and can be seen spinning rapidly or moving with the flow of the medium through the flow cell.