|Home | About | Journals | Submit | Contact Us | Français|
Adhesion to both abiotic and biotic surfaces by the gram-negative prothescate bacterium Caulobacter crescentus is mediated by a polar organelle called the “holdfast,” which enables the bacterium to form stable monolayer biofilms. The holdfast, a complex polysaccharide composed in part of N-acetylglucosamine, localizes to the tip of the stalk (a thin cylindrical extension of the cell wall and membranes). We report here the isolation of adhesion mutants with transposon insertions in an uncharacterized gene cluster involved in holdfast biogenesis (hfs) as well as in previously identified polar development genes (podJ and pleC), and the holdfast attachment genes (hfa). Clean deletions of three of the four genes in the hfs gene cluster (hfsDAB) resulted in a severe holdfast biogenesis phenotype. These mutants do not bind to surfaces or to a fluorescently labeled lectin, specific for N-acetylglucosamine. Transmission electron microscopy indicated that the hfsDAB mutants fail to synthesize a holdfast at the stalk tip. The predicted hfs gene products have significant sequence similarity to proteins necessary for exopolysaccharide export in gram-negative bacteria. HfsA has sequence similarity to GumC from Xanthomonas campestris, which is involved in exopolysaccharide export in the periplasm. HfsD has sequence similarity to Wza from Escherichia coli, an outer membrane protein involved in secretion of polysaccharide through the outer membrane. HfsB is a novel protein involved in holdfast biogenesis. These data suggest that the hfs genes play an important role in holdfast export.
The dimorphic bacterium Caulobacter crescentus undergoes programmed developmental changes during its cell division cycle, differentiating from a nonreplicative, motile stage (swarmer cell) to a sessile, replicative stage (stalked cell) (3). After an obligatory gap period, which is approximately one-third the length of one cell cycle, the swarmer cell sheds its flagellum and forms a stalk at the pole that once held the flagellum. The stalked cell elongates and divides, giving rise to the two different cell types. Primarily inhabiting freshwater and marine environments dilute in nutrients, Caulobacter forms biofilm monolayers on abiotic and biotic surfaces (29). Stable attachment to surfaces requires an adhesive holdfast. Attached to the tip of the stalk, the holdfast can be used to maintain associations with various surfaces, but never with the cell surface of another Caulobacter. Holdfasts can often be seen at the center of a group of cells (called a “rosette”) coordinated by their stalks (20).
The holdfast appears at the incipient stalked pole early during the differentiation of swarmer cells into stalked cells, resulting in its positioning at the tip of the stalk (9). Competition experiments with fluorescently labeled lectins, proteins with a specific binding affinity for a polysaccharide moiety, demonstrated that the holdfast is composed in part of oligomers of the sugar residue, N-acetylglucosamine (GlcNac) (13). The fluorescently labeled lectin fluorescein isothiocyanate-wheat germ agglutinin (FITC-WGA) was shown to bind specifically to the holdfast (13). The holdfast is resistant to both glycolytic and proteolytic enzymes, except for lysozyme and chitinase, which are specific for oligomers of GlcNac (13).
A previous screen undertaken to identify adhesion-deficient mutants of C. crescentus CB2 identified four regions in the genome potentially responsible for holdfast biogenesis (14). Mutants with insertions mapping to one of these regions, the holdfast attachment (hfa) operon (10), failed to bind to surfaces but also did not attach their holdfasts to the stalk efficiently, a defect known as “holdfast shedding.” Studies of hfa expression suggest that the holdfast attachment proteins are preloaded in swarmer cells. The expression of the hfaA promoter occurs maximally in the swarmer compartment of the predivisional cell, despite the fact that swarmer cells do not have exposed holdfasts (9).
Genes involved in polar development are also thought to affect holdfast biogenesis. Mutations in pleC and in podJ affect the ordered development of C. crescentus, leading to the absence or inactivation of the polar structures. Mutants with mutations in pleC and in podJ do not exhibit rosette formation, which can be an indicator of the presence of the holdfast (33). Nonetheless, the direct observation of the absence of a holdfast in these mutants has not been made.
In order to identify genes required for holdfast biogenesis, we screened a C. crescentus CB15 transposon mutant library for the absence of attachment to cellulose acetate. Twenty mutant strains with severe adhesion deficiencies were further phenotypically characterized by cell morphology, lectin binding (FITC-WGA), bacteriophage sensitivity, and motility. Based on the results of this analysis, the adhesion-deficient mutants were divided into three classes: developmental mutants (podJ and pleC), holdfast attachment mutants (hfa), and a novel class of holdfast biogenesis mutants (hfs, for holdfast synthesis). The hfs genes are the first nonregulatory genes shown to be required for holdfast synthesis or export. The proteins predicted to be encoded by these genes are an inner membrane-associated periplasmic protein (HfsA), an outer membrane lipid-modified secretin (HfsD), and a novel protein (HfsB). These proteins bear significant sequence similarity to polysaccharide export proteins of gram-negative bacteria, suggesting that they are required for the export of the holdfast polysaccharide.
The bacterial strains and plasmids used during the course of this study are listed in Table Table1.1. All Caulobacter strains were cultured with peptone-yeast extract (PYE) medium (20) or Hutner base, imidazole, buffered glucose-glutamate (HIGG) minimal medium (21) at 30°C. The antibiotics kanamycin and nalidixic acid (20 μg/ml each) were used to supplement the Caulobacter media as necessary. Escherichia coli strains were cultured with Luria-Bertani (LB) medium at 37°C. LB medium was supplemented with kanamycin at 50 μg/ml when necessary. In order to create a library of adhesion mutants, a mini-Tn5lacZ2 transposon mutagenesis of C. crescentus CB15 and CB2A was performed as previously described (8). Aliquots of mutagenized cells were stored at −80°C until plated. Transduction of kanamycin resistance markers from the transposon mutants into C. crescentus CB15 was performed as previously described (5).
Genomic DNA isolated from the adhesion mutants was used as a template for arbitrarily primed PCR (AP-PCR) to determine the location of each mini-Tn5lacZ2 insertion (8, 18). DNA sequencing was performed on an ABI 3700 automated DNA sequencer at the Institute for Molecular and Cellular Biology, Indiana University, Bloomington. C. crescentus genome sequence data (16) were acquired from The Institute for Genomic Research (TIGR). Sequence analysis was performed with the Genetics Computer Group GCG Wisconsin package, v.10 (Accelrys), and BLAST program (National Center for Biotechnology Information). Signal sequence and localization predictions were made with PSORT (15). Transmembrane prediction was performed with TMHMM (30a).
Southern blot analysis (2) was used to confirm the locations of the mini-Tn5lacZ2 insertions. Southern blot probes were prepared with the Roche DIG High Prime digoxigenin-UTP labeling kit. A 2.1-kb hfa probe, extending from 230 bp upstream of the hfaA start codon to 612 bp downstream of the hfaD start codon, was PCR amplified from the pRJ41 template with the oligonucleotides hfaA215Pst and hfaC2263HindIII. A 2.0-kb hfs probe, extending from 509 bp upstream of the hfsD stop codon to 661 bp downstream of the hfsA start codon, was PCR amplified from CB15 genomic DNA with the oligonucleotides 5gumBC1367 and 3gumBC1568. A 3.0-kb pleC probe, extending from 188 bp upstream of the pleC start codon to 281 bp downstream of the stop codon, was PCR amplified from CB15 genomic DNA with the oligonucleotides pleCfor and pleCrev. A 2.0-kb podJ probe, extending from 1,839 bp upstream to 175 bp downstream of the start codon, was PCR amplified from CB15 genomic DNA with the oligonucleotides 5podJ2508 and 3podJ4522. The oligonucleotide sequences are available from the authors upon request. For Southern hybridization, genomic DNA prepared from the insertion mutants was digested with appropriate enzymes as follows. A 6-kb fragment containing the hfa operon was released from the restriction enzyme digestion of genomic DNA with PstI. A 4.93-kb fragment containing the hfs operon was released following the digestion of genomic DNA with HpaI and ClaI. A 3.13-kb fragment containing pleC was released after digestion of genomic DNA with SalI. A 5.85-kb fragment containing podJ was released after the digestion of genomic DNA with SfiI. Transposon insertions mapping to these fragments were identified by Southern hybridization as shifts in their molecular weight compared to that of the wild type by QuickHyb (Stratagene) and anti-DIG-UTP alkaline phosphatase Fab fragment (Roche) as described by the manufacturers.
Clean in-frame deletions of hfsA, hfsB, hfsC, and hfsD were created by homologous recombination between two fragments—one upstream and one downstream of the gene—as previously described (8, 31). The upstream and downstream fragments were cloned into the suicide vector pNPTS138, which carries a sacB cassette. hfsABamup, hfsAXhoup, hfsAXhoend, and hfsAHinend were used to PCR amplify 500-bp DNA fragments directly upstream and downstream of hfsA. hfsBBamu2, hfsBXhou2, hfsBXhoe2, and hfsBHine2 were used to PCR amplify 500-bp DNA fragments directly upstream and downstream of hfsB. hfsCBamup, hfsCXhoup, hfsCXhoend, and hfsCHinend were used to PCR amplify 500-bp DNA fragments directly upstream and downstream of hfsC. hfsDBamup, hfsDXhoup, hfsDXhoend, and hfsDHinend were used to PCR amplify the 500-bp downstream and upstream fragments of hfsD. The oligonucleotide sequences are available from the authors upon request. The upstream fragments were cut with BamHI and XhoI, while the downstream fragments were cut with XhoI and HindIII. Each pair of fragments was ligated into the BamHI- and HindIII-digested vector pNPTS138. Deletions were confirmed by colony PCR with primers used to clone fragments positioned upstream and downstream of the gene. All deletions were transduced by Cr30 (5) into a wild-type CB15 background by using the closely linked kanamycin resistance marker from strain CMS27 (34). Deletions of the hfs genes were complemented by a low-copy-number plasmid carrying each gene. An 870-bp PCR fragment, extending 28 bp from the hfsD stop codon to 102 bp from the hfsD start codon, was amplified with hfsDxbadn2 and hfsDkpnup2. The PCR product was then cloned into the low-copy-number vector plac290 between the XbaI and KpnI restriction sites. A 2.3-kb PCR fragment, extending 106 bp from the start codon of hfsA to 22 bp downstream of hfsB, was amplified with hfsAbamup2 and hfsBhindend3. The resulting PCR product was also cloned into plac290 between a BamHI site and a HindIII site. These plasmid constructs were transformed into E. coli S17-1 and mated into the various mutants. Rosette formation in liquid culture was determined by phase-contrast light microscopy.
Lectin binding assays were performed as previously described (17). Two microliters of 5-mg/ml FITC-WGA (Molecular Probes) was added to 200 μl of exponential-phase culture and incubated with gentle shaking at room temperature for 15 min. The culture was diluted fivefold with double-distilled water (ddH2O), and the cells were harvested by centrifugation at 13,000 × g for 4 min (9). The cells were resuspended in 20 μl of SlowFade B (Molecular Probes) (9). Labeled cells were examined by epifluorescence on a Nikon E800 light microscope equipped with a FITC-HYQ filter cube and 100× Plan Apo oil objective. Image capture was carried out with a Princeton Instruments cooled charge-coupled device (CCD) camera and Metamorph imaging software, v.4.5.
Electron microscopy was used to determine the nature of the adhesion defect of the mini-Tn5lacZ2 insertion mutants. Cells were mounted onto 0.25% Formvar-coated copper grids (Ted Pella) for 30 min. Each grid was washed three times in a drop of water, followed by positive staining with 7.5% uranyl magnesium acetate for 5 min. The grids were then washed again three times in a drop of water. Grids were examined with a JEOL JEM-1010 transmission electron microscope (TEM) at 60 kV.
In order to screen for adhesion mutants, disks of cellulose acetate (Apollo WO100C) were applied to patch plates of mini-Tn5lacZ2 mutants and screened as previously described (17).
As a semiquantitative method of measuring adhesion defects, approximately 50 μl of exponential-phase culture concentrated to an optical density at 600 nm (OD600) of 4.8 was spotted onto a piece of cellulose acetate (or borosilicate glass). The cells were allowed to attach for 3 min, before each drop was removed with a pipette tip. The cellulose acetate was washed thoroughly with a strong jet of cold water. The remaining adherent cells were visualized by staining the disks with 0.1% Coomassie blue (R-250) in 10% acetic acid-10% isopropanol for 3 min. Residual stain was removed by washing thoroughly in cold water.
The glass coverslip binding assay was adapted from an existing protocol (17). Overnight cultures were grown in PYE medium at 30°C. In the morning, cultures were diluted in PYE medium to an OD600 of 0.15. Glass coverslips were dipped into ethanol, exposed to flame, and placed into the well of a tissue culture dish (six-well, tissue culture, treated polystyrene dish, catalog no. 3506; Corning). A 1.5-ml sample of cell culture was added to each well. The tissue culture plate was placed on a shaker (90 to 100 rpm) at 30°C and incubated for 4 to 5 h. After incubation, coverslips were rinsed with a steady stream of ddH2O from a water bottle for 10 to 20 s. The coverslips were placed cell side up in a humid chamber and incubated with 50 μl of FITC-WGA at a concentration of 0.05 μg/μl for 30 min. The coverslips were rinsed with water as described above. The coverslips were placed onto a slide with 2 μl of SlowFade A (Molecular Probes) solution (glycerol-phosphate-buffered saline), and examined by phase and epifluorescence microscopy.
The ability of cells to bind to various surfaces was measured by incubating those strains in 96-well microtiter plates. Each culture was grown in a test tube at 30°C until exponential phase was reached, at an OD600 of 0.6. Two hundred microliters of each culture was pipetted into the wells of the microtiter plate, where it was incubated for 15 min at room temperature. After allowing time for the cells to attach, a strong jet of cold water was used to displace any loosely attached cells. The microtiter plate was briefly air dried before 200 μl of 0.1% Coomassie blue (R-250) in a mixture of 10% acetic acid and 10% isopropanol was pipetted into the wells. After incubation for 15 min, each well was emptied by pipetting out the stain. The microtiter plate was washed with a strong jet of cold water and air dried.
Two microliters of overnight culture in PYE medium were stabbed into 0.3% semisolid agar PYE medium. The plates were incubated at 30°C for 3 to 5 days.
Phage sensitivity assays for determining resistance to polar caulophages, Cbk and Cr30, were performed as previously described (23).
While genes required for holdfast attachment to the stalk have been identified previously, no genes required for holdfast synthesis or export have been characterized. Therefore, we performed a screen in C. crescentus CB15 to identify genes required for holdfast biosynthesis and export. The screen used binding to cellulose acetate as its basis. Approximately 9,000 mini-Tn5lacZ2 mutants were screened for absence of adhesion to cellulose acetate, resulting in the identification of 236 potential adhesion mutants. In order to decrease the number of potential false negatives, these adhesion mutants were put through a second screen that tested the adhesion of equal numbers of cells from exponential-phase cultures to cellulose acetate. Twenty mutants were identified that did not adhere to cellulose acetate. These mutants were characterized by phase-contrast microscopy for their ability to form rosettes, overall cellular morphology, and motility (Tables (Tables22 and and3).3). Each insertional mutation was transduced into the wild-type background, and the resulting transductants were examined for cellulose acetate and lectin binding. All transductants displayed the same phenotype as the insertional mutants, indicating that the observed phenotype was the result of a single mini-Tn5lacZ2 insertion. Thirteen of the nonadherent mutants were deficient only in rosette formation, suggesting specific defects in holdfast biogenesis. Six of the mutants had additional developmental defects, suggesting that the mutations were in genes involved in the regulation of development. The site of mini-Tn5lacZ2 insertion in various mutants was determined by sequencing either AP-PCR products or products of PCRs that used a transposon primer and a primer from the suspected insertion region in cases in which the insertion had been roughly mapped by Southern hybridization.
The two mutants YB2861 and YB2868 lacked discernible stalk structures and did not form rosettes (Fig. (Fig.11 and Table Table2).2). However, upon growth in HIGG media limited for inorganic phosphate (30 μM), both strains formed stalks (data not shown). These mutants were also unable to swarm in 0.3% semisolid agar (Table (Table2).2). However, both mutants were flagellated when examined by phase-contrast microscopy with a flagellar stain (data not shown). In addition, YB2861 and YB2868 were resistant to the polar caulophage CbK, but were sensitive to Cr30 (Table (Table2).2). The caulophage CbK binds to the polar pili of swarmer cells (27), suggesting that YB2861 and YB2868 lack pili.
YB2861 and YB2868 were assayed for their binding of the fluorescently labeled lectin FITC-WGA, which specifically binds oligomers of N-acetylglucosamine, a component of the holdfast. To ensure that counted cells had stalks, we scored only predivisional cells. Wild-type strain CB15 had 78% labeling, whereas the holdfast-deficient strain NA1000 had less than 1% labeling. Both YB2861 and YB2868 had less than 1% of cells labeled with FITC-WGA, indicating a severe deficiency in holdfast synthesis (Table (Table33).
Mapping of the insertions in YB2861 and YB2868 indicated that the transposons were inserted in the pleC gene at approximately 500 bp from the end of the gene and exactly 1,523 bp from the start codon, respectively (Fig. (Fig.2).2). PleC is a dynamically localized histidine kinase required for polar development (7, 30, 33, 35). Our phenotypic characterization of these mutants matches previous findings.
Four other insertional mutants also displayed a pleiotropic phenotype. YB2863, YB2864, YB2866, and YB2867 failed to form rosettes, but otherwise their cellular morphology was wild type (Fig. (Fig.11 and Table Table2).2). Phage sensitivity assays showed that these mutants were resistant to the polar caulophage CbK (Table (Table2),2), suggesting that they lack pili (8a; this work). Swarming motility in 0.3% semisolid agar was reduced relative to that of the wild type, although cultures of the insertion mutants clearly contained motile swarmer cells (Table (Table2).2). Three mutants (YB2864, YB2866, and YB2867) had no FITC-WGA labeling of predivisional cells (Table (Table3).3). The other mutant, YB2863, displayed approximately 7% labeling (Table (Table33).
The insertions in YB2863, YB2864, YB2866, and YB2867 were mapped to the polar organelle development gene, podJ (Fig. (Fig.2).2). The insertions interrupted the gene 277 (YB2863), 800 (YB2864), 787 (YB2866), and 638 (YB2867) nucleotides (nt) after the start codon (Fig. (Fig.2).2). podJ mutants have been shown to have defects in chemotaxis, rosette formation, and sensitivity to phage CbK (33), and our results agree with these findings.
Our analysis of pleC and podJ mutants extends previous results by showing that these mutants are unable to bind surfaces and to synthesize a holdfast. Since these mutants exhibit pleiotropic defects, pleC and podJ are clearly not specifically involved in holdfast synthesis.
Seven mutants (YB2779, YB2780, YB2781, YB2782, YB2783, YB2784, and YB3738) were deficient in adhesion to surfaces, were wild type in terms of cellular morphology (Table (Table2),2), and formed rosettes at a low frequency or not at all compared to wild-type cells. These mutants frequently shed their holdfasts into the medium as measured by FITC-WGA assays (Table (Table3).3). The mutants exhibited various degrees of stalk-associated FITC-WGA labeling (Table (Table3),3), indicating that the attachment of the holdfast to the tip of the stalk was weak. This is similar to the phenotype of previously identified holdfast attachment (hfa) mutants.
We used Southern hybridization to map the transposon insertions and found that all of them mapped to the previously identified (10, 11) hfaABDC gene cluster (data not shown). All the mutants were fully complemented by plasmid placHfa7, which contains the hfaABD genes, indicating that the shedding phenotype was due to mutations in the hfaABD region and that none of the mutations had dominant effects (Table (Table3).3). PCR amplification with primers in the hfaABDC region indicated that insertions 113, 115, and 123 were in hfaB (Fig. (Fig.2).2). The fact that all seven hfa adhesion mutants identified in this screen and three hfa mutants identified previously (10) map to this gene cluster strongly suggests that there are no other nonessential genes required for holdfast attachment to the stalk.
The hfa mutants were also analyzed by using a binding assay in which cultures of cells were allowed to adhere to a glass coverslip (13, 17). Wild-type strain CB15 was able to form dense cellular biofilms. These biofilms were associated with spots of fluorescence, indicating the presence of holdfasts (Fig. (Fig.3).3). Phase-contrast microscopy of coverslips exposed to the holdfast-deficient strain NA1000 indicated that no cells remained bound, and fluorescence microscopy showed that FITC-WGA was unable to bind to the coverslips (Fig. (Fig.3).3). The cells of all of the hfa mutants exhibited very poor binding to glass; however, spots of fluorescence densely stained the glass, indicating the presence of holdfasts (Fig. (Fig.3).3). These results indicate that the hfa mutants are able to synthesize holdfast material and that this holdfast material can bind to glass; however, the holdfasts do not remain attached to the stalk.
Seven insertion mutants (YB2862, YB2865, YB2869, YB2870, YB2877, YB2878, and YB2879) were adhesion deficient, yet wild type in other respects (Table (Table2).2). None of the mutants bound lectin to any appreciable degree or shed holdfasts (Table (Table33).
All seven insertions were located in a previously uncharacterized gene cluster (Fig. (Fig.2).2). The first two genes in the cluster, hfsD (CC2432) and hfsA (CC2431), are divergently transcribed based on their orientation. The start codon of hfsB (CC2430) lies 78 bp downstream of the preceding gene, hfsA. The gene downstream of hfsB, hfsC (CC2429), overlaps the stop codon of the preceding gene (Fig. (Fig.2).2). This suggests that hfsB and hfsC are translationally coupled and form an operon; it is not clear if hfsA is cotranscribed with them. Three of the insertions in this region were located in hfsA, interrupting the gene 614 (YB2878), 680 (YB2865), and 1,085 (YB2869) bp after the start codon (Fig. (Fig.2).2). Three insertions were located in hfsB, interrupting the gene 77 (YB2862), 115 (YB2870), and 279 (YB2877) bp from the start codon (Fig. (Fig.2).2). Another insertion was located in the N-terminal portion of hfsD (YB2879), which interrupts the gene 240 bp after the start codon (Fig. (Fig.22).
Sequence analysis indicated that hfsA, hfsC, and hfsD are homologous to genes involved in exopolysaccharide (EPS) synthesis and export. hfsA encodes a predicted protein of 501 amino acids (aa) (GCG). This protein is 21% identical and 37% (1) similar to GumC from the plant pathogen Xanthamonas campestris. hfsB is predicted to encode a protein of 233 aa (GCG). HfsB has no homologues in the sequence databases and possesses no obvious signal sequences with which to predict its localization. hfsC is predicted to encode a protein of 422 aa (GCG). HfsC is 29% identical and 39% (1) similar to ExoQ from the plant symbiont Rhizobium melliloti. hfsD is predicted to encode a protein of 246 aa (GCG). HfsD is 32% identical and 50% (1) similar to Wza from E. coli. Wza is an outer membrane lipoprotein secretin that functions to export polysaccharides (4).
In order to rule out possible polar effects of insertions on the observed phenotype of the hfs mutants, clean in-frame deletions of each gene were created. The deletions of hfsA, hfsB, and hfsD (YB2833, YB2837, and YB2845) resulted in the same cellulose acetate adhesion defect observed with the insertion mutants. In order to determine if the adhesion defect observed with the deletion mutants and the insertion mutants was global for different types of surfaces, we performed a surface binding assay with polypropylene, polyvinylchloride, polystyrene, and borosilicate glass. We found that the hfsA, hfsB, and hfsD deletion mutants were defective in attaching to all of these surfaces (data not shown). C. crescentus CB15, however, bound noticeably to each of the surfaces, whereas the negative control NA1000 did not bind. The cellular morphology of the hfs deletion mutants was indistinguishable from that of the wild type, as determined by phase-contrast microscopy (Table (Table2).2). When examining whole-cell mounts of hfsA, hfsB, and hfsD deletion mutants by TEM, however, the stalked cells all lacked dark staining material where the holdfast normally resides (Fig. (Fig.4).4). No other morphological differences between the wild type and the deletion mutants were apparent. The fact that hfsA, hfsB, and hfsD deletion mutants lack holdfasts was bolstered by the lectin binding results. The deletion mutants did not bind FITC-WGA (Fig. (Fig.5).5). The hfsC deletion mutant (YB2841) exhibited no phenotypic differences from the wild type for binding to surfaces or to FITC-WGA (Fig. (Fig.5),5), indicating that hfsC does not have a role in holdfast synthesis and in binding to the surfaces tested. The hfs deletion mutants were transduced into a clean background in order to further rule out unlinked mutations as the cause of the holdfast biogenesis phenotype. In each case, we recovered transductants that did not form rosettes at the predicted frequency, which suggests that the deletions are the sole cause of the phenotype (data not shown). Introduction of a low-copy-number plasmid carrying hfsD fully restored rosette formation in the hfsD deletion mutant, and introduction of a plasmid carrying hfsAB fully restored rosette formation to the hfsA and hfsB deletion mutants (data not shown).
In this paper, we report the results of a saturating mini-Tn5 screen for Caulobacter mutants with adhesion defects. These mutants can be grouped into three classes: developmental mutants (pleC and podJ), holdfast synthesis or export mutants (hfs), and mutants that disrupt the attachment of the holdfast to the tip of the stalk (hfa).
The developmental mutants displayed additional polar defects, in addition to a lack of adhesion and an absence of a holdfast. The pleC mutants were deficient in stalk synthesis and flagellar rotation and were resistant to caulophage CbK as shown previously (33). PleC is a histidine kinase that localizes to the flagellated pole of swarmer cells and of late predivisional cells and is involved in coordinating many aspects of polar morphogenesis (35). The podJ mutants had defects in chemotaxis and were resistant to CbK, as shown previously (33), in addition to their deficiency in surface adhesion and holdfast synthesis. We have recently shown that PodJ is localized to the flagellated pole of swarmer cells, disappears from that pole in stalked cells, and localizes to the opposite pole, where it remains for the rest of the cell cycle (8a). PodJ is required for PleC localization (8a), suggesting that the holdfast synthesis deficiency of podJ mutants may be due to PleC delocalization (8a).
The hfa genes were identified previously (11) and are transcribed from a promoter upstream of hfaA (9, 10; J. Cole, D. Bodenmiller, and Y. V. Brun, unpublished observations). Nonpolar mutants of hfaA, hfaB, and hfaD produce holdfasts, as shown by their binding to fluorescently labeled lectins, but do not attach them to the stalk at wild-type levels (Cole et al., unpublished). This leads to an overall defect in adhesion. HfaB shares some similarity with the curli attachment gene, csgG, from E. coli and has been experimentally proven to be a lipoprotein (Cole et al., unpublished). Curli are proteinaceous fibers that mediate attachment to surfaces. hfaA and hfaD share no significant sequence similarity with any characterized proteins. Both HfaB and HfaD have been shown to localize to the stalk (Cole et al., unpublished), suggesting that they are directly involved in attaching the holdfast to the stalk. The hfa genes appear to be the only nonessential genes required for the attachment of the holdfast to the tip of the stalk based on the fact that 10 independent insertions have been isolated at this locus (10; this work).
The hfs genes are likely to be involved in holdfast export. hfsA, hfsB, and hfsD mutants all show a dramatic defect in cellular adhesion to a variety of surfaces. Lectin binding studies also show that hfs mutants do not shed holdfasts in the medium, further separating this new class of mutants from the hfa mutants. Lectin binding experiments and electron microscopy show the absence of holdfast material at the tip of the stalk. The gene products of the hfs operon have a high degree of sequence similarity to polysaccharide export components. HfsD resembles an oligomeric secretin, Wza, from E. coli. An outer membrane lipoprotein, Wza functions as a hexameric channel for export of capsular polysaccharide to the surface (4). Wza belongs to the outer membrane auxiliary, or OMA, family of proteins (19). HfsD also carries a potential lipoprotein sorting signal sequence at its N-terminal portion (15). In addition, based upon the presence of noncharged residues at the +2 and +3 positions of the predicted mature lipoprotein, HfsD is predicted to reside in the outer leaflet of the outer membrane (15, 24).
HfsA has sequence similarity in its 400-aa periplasmic loop to the polysaccharide transport protein, GumC, from X. campestris. GumC has been shown to be involved in the export of the EPS xanthan gum from the cytoplasm to the cellular exterior (32). Export of high-molecular-weight EPS is required for the invasion of plant hosts by X. campestris. GumC is a member of membrane periplasmic auxiliary (MPA-1) family of polysaccharide export proteins (19). One of the defining characteristics of the genes encoding the MPA-1 is the nearby presence of an OMA gene, which participates in EPS export (19). Normally, members of the MPA-1 protein family share an inner membrane topology and possess a large cytoplasmic domain containing a Walker A (GXXXXGKT/S) ATP binding motif, which is responsible for providing the energy necessary for driving the export process in the form of ATP hydrolysis (19). Although HfsA possesses the required two-transmembrane-helix topology, with a large periplasmic loop, it lacks a sizeable cytoplasmic domain (30a) and the necessary Walker A motif (1). However, two other members of the MPA-1 family also lack the relatively well-conserved C-terminal cytoplasmic domain: GumC and OtnB (19). In these cases, energy may be provided by an ABC cassette transport system. Members of the MPA-2 family, such as KpsE and CtrB, work in concert with these cytoplasmic membrane transporters (19). It is unknown how the 400-aa periplasmic loop of the MPA-1 proteins functions to facilitate transport of polysaccharide residues from the cytoplasmic membrane face of the periplasm to the outer membrane face.
HfsC has sequence similarity to ExoQ from R. melliloti. HfsC adopts a similar membrane topology to ExoQ, with 11 (rather than 12) predicted transmembrane helices spanning the inner membrane (30a). The region of highest similarity between the two proteins occurs in an ~75-aa periplasmic loop (30a) in which they are located (33% identical and 46% similar). ExoQ has been shown to be directly involved in the polymerization of a high-molecular-weight EPS, succinoglycan, which is required for nodule formation (22). An hfsC deletion mutant does not have an adhesion-deficient phenotype to surfaces such as borosilicate glass, polyvinylchloride, polypropylene, and polystyrene and binds lectin at wild-type levels. The data do not rule out an effect on binding to different surfaces, but clearly hfsC is not required for holdfast synthesis.
Finally, HfsB has no significant similarity to available protein sequences. Combined with the absence of a signal sequence or hydrophobic regions (15, 30a), it is difficult to propose a role for HfsB in holdfast biogenesis.
No biosynthetic genes were isolated in the cellulose acetate screen; however, the possibility remains that hfsB functions in this capacity. One explanation for the absence of biosynthetic mutants is that the screen is not saturated. However, we have identified multiple insertions at every locus (seven at hfa, seven at hfs, two at pleC, and four at podJ). On the basis of this evidence, we believe the screen has been saturated for knockout mutations that abolish adhesion. Another possibility is that mutations in some genes required for optimal adhesion have been missed in this screen. For example, mutations in pilus and flagellar structural genes reduce, but do not completely abolish adhesion (D. Bodenmiller and Y. V. Brun, unpublished data; Cole et al., unpublished). Similarly, it is possible that mutations that eliminate some yet unknown component of the holdfast reduce but do not abolish adhesion to the cellulose acetate used in the screen. Isolation of these potential mutants would require a screen for binding to materials with different surface chemistries. Finally, since N-acetylglucosamine is a component of both the holdfast and peptidoglycan, mutations that abolish its synthesis would be lethal and would not be represented among our collection of insertional knockouts. EPS subunits are synthesized in the cytoplasm from precursors. The EPS subunits are then attached to a lipid carrier residing in the inner membrane, which is energized to flip across the membrane to the periplasm. In a poorly understood process, the EPS subunits are oligomerized into their final form, modified if necessary, and exported to the outer membrane, where they are transported to the exterior matrix via a secretin. Such a system would take advantage of the fact N-acetylglucosamine subunits are present in the periplasm for peptidoglycan biosynthesis, where the hfs gene products could utilize N-acetylglucosamine for holdfast biogenesis.
In conclusion, we have identified the first nonregulatory genes known to be required for holdfast synthesis. Based on their sequence, we hypothesize that the gene products of the hfs cluster are involved in EPS export from the periplasmic space to the cellular exterior. This particular area of the EPS biosynthesis pathway has not been fully elucidated. HfsA might provide periplasmic export functions for holdfast polysaccharide subunits with the energy provided by an ABC transport complex. It is also possible that HfsA plays a role in the processing of holdfast subunits, which could be required for export. HfsD probably terminates the export branch of holdfast biogenesis by serving as the site of secretion to the eventual site of holdfast localization, the tip of the stalk.
We thank members of our laboratory for critical reading of the manuscript and for helpful discussions.
This work was supported by a National Institutes of Health grant (GM51986) and a National Science Foundation CAREER Award (MCB-9733958) to Y.V.B., a Beckman Scholarship to A.H., an American Society for Microbiology Undergraduate Research Fellowship to D.L., and a National Institutes of Health Predoctoral Fellowship (GM07757) to C.S.