|Home | About | Journals | Submit | Contact Us | Français|
Growth in a surface-attached bacterial community, or biofilm, confers a number of advantages. However, as a biofilm matures, high-density growth imposes stresses on individual cells, and it can become less advantageous for progeny to remain in the community. Thus, bacteria employ a variety of mechanisms to control attachment to and dispersal from surfaces in response to the state of the environment. The freshwater oligotroph Caulobacter crescentus can elaborate a polysaccharide-rich polar organelle, known as the holdfast, which enables permanent surface attachment. Holdfast development is strongly inhibited by the small protein HfiA; mechanisms that control HfiA levels in the cell are not well understood. We have discovered a connection between the essential general protein chaperone, DnaK, and control of C. crescentus holdfast development. C. crescentus mutants partially or completely lacking the C-terminal substrate binding “lid” domain of DnaK exhibit enhanced bulk surface attachment. Partial or complete truncation of the DnaK lid domain increases the probability that any single cell will develop a holdfast by 3- to 10-fold. These results are consistent with the observation that steady-state levels of an HfiA fusion protein are significantly diminished in strains that lack the entire lid domain of DnaK. While dispensable for growth, the lid domain of C. crescentus DnaK is required for proper chaperone function, as evidenced by observed dysregulation of HfiA and holdfast development in strains expressing lidless DnaK mutants. We conclude that DnaK is an important molecular determinant of HfiA stability and surface adhesion control.
IMPORTANCE Regulatory control of cell adhesion ensures that bacterial cells can transition between free-living and surface-attached states. We define a role for the essential protein chaperone, DnaK, in the control of Caulobacter crescentus cell adhesion. C. crescentus surface adhesion is mediated by an envelope-attached organelle known as the holdfast. Holdfast development is tightly controlled by HfiA, a small protein inhibitor that directly interacts with a WecG/TagA-family glycosyltransferase required for holdfast biosynthesis. We demonstrate that the C-terminal lid domain of DnaK is not essential for growth but is necessary for proper control of HfiA levels in the cell and for control of holdfast adhesin development.
Communities of surface-attached bacteria, called biofilms, account for the majority of cells on the surface of our planet. Biofilms allow bacteria to engage in metabolic cooperation, share genetic information, and afford protection from chemical and physical stresses in the environment. Moreover, organic biopolymers and ions accumulate at surfaces through a process known as conditioning (1,–4); thus, the ability of a bacterial cell to adhere can confer a nutritional advantage, particularly in oligotrophic environments where nutrients are extremely scarce (5). Clearly, the capacity to transition from a free-living, unicellular lifestyle into a surface-attached multicellular community is an important adaptive feature of bacterial physiology that increases the breadth of niches a species can inhabit.
The Gram-negative bacterium Caulobacter crescentus thrives in dilute, freshwater ecosystems and can permanently adhere to a number of chemically diverse surfaces via a polar adhesin called the holdfast (6,–10). The holdfast organelle is a polysaccharide-rich matrix which is synthesized via a Wzx/Wzy-dependent secretion system and attached to the cell via localized outer membrane proteins (11,–16). Given that holdfast-mediated surface attachment is permanent, it is not surprising that the decision to elaborate a holdfast is a highly regulated process that is controlled by cell cycle, environmental, and physical signals. Newborn swarmer cells are motile and lack holdfasts, consistent with a role in dispersal. During development, or when a swarmer cell encounters a physical surface, holdfasts are elaborated as C. crescentus trades motility for replication and differentiates into a stalked cell (17). A key regulator of holdfast development in Caulobacter is a small, hydrophobic protein, HfiA, which responds to both cell cycle and environmental cues and functions as a potent inhibitor of holdfast synthesis (18).
To understand how HfiA functions to inhibit holdfast development, we previously performed a genetic selection for mutants that were insensitive to HfiA, i.e., mutants in which surface adhesion was unaffected by hfiA overexpression (Fig. 1B). This strategy led to the identification of HfsJ, a putative WecG/TagA family glycosyltransferase that is necessary for holdfast synthesis. Genetic and biochemical data support a model in which HfsJ is directly inhibited by HfiA (18) (Fig. 1A). From this same genetic selection we also mapped two independent mutant alleles of dnaK, a highly conserved gene encoding an ATP-dependent general protein chaperone that is essential for C. crescentus viability (19, 20). To date, these adhesive dnaK mutant strains have remained uncharacterized.
DnaK interacts with a large number of protein substrates (21) to facilitate de novo protein folding, membrane protein targeting, and refolding of denatured proteins (22,–25). Substrate recognition is largely based on physiochemical properties. DnaK recognizes short exposed hydrophobic regions, consistent with a role in refolding denatured proteins (26,–28). DnaK has an N-terminal ATPase domain followed by a C-terminal substrate-binding domain (Fig. 1C and andD).D). The substrate-binding domain is further divided into a β-sheet domain, which forms the substrate binding pocket, and an alpha-helical lid domain at the extreme C terminus (27). The conformation of the substrate binding domain and its affinity for substrate depend on the state of the ATPase domain. In the ATP-bound state, the lid is open and affinity for substrate is low. Upon ATP hydrolysis to ADP the lid closes on the substrate-binding domain as shown in Fig. 1D, and the affinity for substrate is dramatically enhanced (29,–31). ATP-dependent regulation of substrate affinity is independent of the C-terminal helical lid, although this domain enhances the residence time of substrates on DnaK, which reduces protein aggregation and facilitates substrate folding (32, 33).
Both dnaK mutations identified in our genetic selection resulted in truncation of the alpha-helical lid at the extreme C terminus. Given that HfiA is a largely hydrophobic protein, this result suggested that DnaK facilitates stabilization of HfiA and, thus, its ability to function as an inhibitor of holdfast synthesis. Indeed, we demonstrate that truncation of the DnaK lid domain results in increased surface attachment and holdfast development in both hfiA overexpression and wild-type backgrounds. Further, we show that these enhanced adhesion phenotypes are accentuated by deletion of the entire helical lid domain. Consistent with the elevated adhesion phenotypes, the steady-state level of an HfiA fusion protein is reduced in these dnaK mutant strains. Our results support a model in which DnaK, as a general molecular chaperone, plays an important role in stabilizing HfiA in the cell and thereby affects the capacity of C. crescentus to adhere to surfaces in a regulated manner. Our results provide evidence that the nonessential lid domain of DnaK plays a crucial role in ensuring HfiA function as an environmental and cell cycle-controlled regulator of holdfast development and surface attachment.
C. crescentus cells were cultured in either PYE (0.2% peptone, 0.1% yeast extract, 1 mM MgSO4, 0.5 mM CaCl2) or M2 salts (49), supplemented with 0.15% xylose as the carbon source, at 30°C. To induce expression from the Pxyl promoter in cells grown in PYE, the culture was supplemented with 0.1% xylose. Antibiotics were used at the following concentrations in liquid and solid medium, respectively: kanamycin, 5 and 25 μg/ml; chloramphenicol, 1 and 1.5 μg/ml; nalidixic acid, 20 μg/ml. Escherichia coli cells were cultured in LB at 37°C. The following antibiotics were used in both liquid and solid medium: kanamycin, 50 μg/ml; chloramphenicol, 20 μg/ml.
C. crescentus DNA was amplified with KOD Xtreme hot-start polymerase (EMD Millipore) and supplementing reactions with 5% dimethyl sulfoxide (DMSO). Restriction sites for cloning were added to the ends of the primers. Amplified products were digested with appropriate restriction enzymes (New England BioLabs) and ligated into similarly digested, phosphatase-treated, and gel-purified plasmids using T4 DNA ligase (New England BioLabs). Plasmid ligations were transformed into E. coli Top10 (Life Technologies, Invitrogen). All cloned products were sequence confirmed.
Plasmids used and generated in this work are listed in Table 1. Primer sequences used to amplify cloned sequences are also listed in Table 1. Point mutant and null alleles were generated using an overlap extension PCR strategy (50) and unique restriction enzyme cut sites in the outermost primer sequences. Overlapping primers are indicated with a plus at the end of the name. Amplified alleles were ligated into the corresponding restriction sites in pNPTS138 or pNPTS138-CAT. Overlap extension PCR was used similarly to generate fusions in pAF508 and pAF509. First, HfiA starting at genome position 938,855 (the original predicted start) or position 938,825 (a later start codon) was amplified and cloned into the SacI and NheI sites of pXCHYN-6 to generate xylose-inducible mCherry-HfiA fusions. Sequences corresponding to the hfiA promoter and to the mcherry-hfiA fusion were amplified from chromosomal templates or these plasmid templates, respectively, and then joined by overlap extension PCR and cloned into the NdeI and NheI sites of pXGFPC-6. This excised the green fluorescent protein (GFP) gene from the plasmid and inserted a PhfiA-mcherry-hfiA fusion. The 5′ end, the hfiA gene in pAF508, is 24 bp longer than that in pAF509, which generates a short linker in the amino acid sequence of the fusion.
Strains used in this study are described in Table 2. Plasmids were transformed into C. crescentus strains by electroporation or triparental conjugation as described in reference 18 and selected on solid media containing appropriate antibiotics. The presence of the plasmid was confirmed by PCR. A two-step double recombination strategy based on sucrose counterselection with sacB was used to generate allele replacement strains. For a detailed description, see reference 18. Briefly, the first step of recombination entails selection for integration of the pNPTS138-derived plasmid on medium containing kanamycin or chloramphenicol as appropriate. After a short period of nonselective growth (4 to 20 h), cells in which a second recombination event resulted in plasmid excision were selected by growth on medium containing 3% sucrose. Alleles were confirmed by sequencing gene-specific PCR products. We used this strategy both to replace wild-type alleles with mutant alleles and to replace mutant alleles with the wild type, thereby restoring a native locus.
The replicating plasmid, pMT805-hfiA, was cured from the dnaK(I587::Tn5) strain by continuous passaging in the absence of chloramphenicol for 2 to 3 days. Periodically cells were serially diluted and plated on nonselective medium to identify well-isolated colonies. The resulting colonies were patched on medium both with and without chloramphenicol to identify clones which had become sensitive to chloramphenicol.
The selection strategy to identify mutants that are insensitive to hfiA overexpression was described in reference 18. Briefly, cells overexpressing the holdfast inhibitor, hfiA, from a high-copy-number plasmid were grown in polystyrene dishes. Cells that were sensitive to HifA are nonadherent cells and were washed away with sterile water. The rare cells that elaborate holdfasts and adhere to the surface served as the inoculum and populate the culture when fresh medium is added. Each selection was enriched with siblings of a small number of winning mutants; thus, this selection strategy was conducted several times to identify a collection of independent mutants. In a pilot selection, we mutagenized FC1935 (CB15/pMT805-hfiA) with the EZ-TN5 R6Kgammaori/KAN-2 transposome kit (catalog no. TSM08KR; Epicentre). The resulting mutagenized library was evaluated in our selection screen. The insertion sites in selected clones were mapped using rescue cloning of the EZ-TN5 transposon by following the manufacturer's protocol. Subsequent selections relied on spontaneous genetic lesions, which were mapped with whole-genome sequencing as described in reference 18.
Overnight starter cultures were diluted 1:10, outgrown for 4 to 5 h, and then diluted to an optical density at 660 nm (OD660) of 0.001 (for PYE) or 0.01 (for M2X) in 1 ml of fresh medium in a 24-well culture dish. The lid of the dish was sealed with strips of AeraSeal (Excel Scientific) and grown at 30°C with shaking at 150 rpm overnight to saturation (16 to 18 h for PYE, 20 to 22 h for M2X). Unbound cells were vigorously washed away under tap water. The attached cells in each well were stained with 1.25 ml of 0.01% (wt/vol) crystal violet dissolved in water, with shaking for 5 to 15 min. Unbound stain was removed by vigorous washing with water. Bound stain was extracted with 1.5 ml 100% ethanol, with shaking for 10 min. One hundred microliters of the ethanol extract was diluted in 500 μl of ethanol, and the absorbance at 575 nm was measured with a Genesys20 spectrophotometer (ThermoFisher Scientific).
Holdfasts were detected as described in reference 18. Briefly, starter cultures were diluted to a calculated OD660 of 0.0005 to 0.001 so that after ~14 h of growth the culture density would reach ~0.05 to 0.1 OD660. Growing cells to low density minimizes cell-cell adhesion and ensures that all cells are born into nearly similar, nutritionally replete conditions. Five hundred microliters of cells was incubated at room temperature with 5 μg/ml wheat germ agglutinin (WGA)-Alexa Fluor 594 conjugate (Life Technologies, Molecular Probes) for 5 to 10 min. Cells were then diluted and washed with 1 ml water or medium, collected by centrifugation for 3 min (14,000 × g), and resuspended in 20 to 30 μl of remaining liquid in the tube. Cells were imaged as described below. For each sample, ~500 cells were manually scored for the presence or absence of polar fluorescence.
Cells were spotted between a glass slide and no. 1 coverslips and imaged with a Leica DM5000 upright microscope in phase contrast and fluorescence modes. We used an HCX PL APO 63×/1.4-numeric-aperture Ph3 objective and an EL6000 external mercury halide lamp (Leica) as a fluorescence excitation source. Standard filter sets were used to detect WGA-Alexa 594 (Chroma set 41043). Images were captured with an Orca-ER digital camera (Hamamatsu) using LAS-X (Leica).
Overnight starter cultures were diluted to a calculated OD660 of 0.0003 to 0.0005 so that after 16 to 20 h of growth the cells would reach a density of an OD660 of 0.10 to 0.16. For any individual experiment, variation in cell density was limited to ±0.02 absorbance unit. Cells were harvested by centrifugation (1 min at ~21,000 relative centrifugal force). The volume of culture was normalized so that the number of cells harvested was equivalent to collecting 4 ml of cells at an OD660 of 0.1 (i.e., we harvested cells from 2.5 to 4 ml of the culture depending on the actual OD at the time of collection). Cells were resuspended in 50 μl of Triton-based detergent buffer (50 mM Tris-HCl, pH 7.4, 1 mM MgCl2, 150 mM NaCl, 0.1% Triton X-100) supplemented with cOmplete mini protease inhibitor cocktail tablets (Roche) and 0.5 mg/ml DNase I. To each sample, 100 μl of loading dye (150 mM Tris, pH 6.8, 6% SDS, 30% glycerol, ~0.001% bromphenol blue, 0.75% β-mercaptoethanol) was added before heating to 95°C for 5 min. After samples cooled to room temperature, 15 μl of each sample was loaded onto a Mini-Protean TGX 4 to 20% gradient gel (Millipore). Proteins were separated at 200 V for ~60 min and transferred to an Immobilon-P polyvinylidene difluoride (PVDF) membrane (Bio-Rad) in 25 mM Tris base, 190 mM glycine, 20% methanol at 100 V for 75 min at 4°C. Membranes were blocked with 5% milk in TTBS (10 mM Tris, pH 7.5, 150 mM NaCl, 0.05% Tween 20). Membranes were incubated overnight with either a 1:2,000 dilution of polyclonal rabbit anti-mCherry antisera (provided by Patrick Viollier) or a 1:5,000 dilution of polyclonal rabbit anti-DnaK antisera (provided by Christina Jonas) at room temperature, washed 3 times with TTBS, incubated with a 1:5,000 dilution of monoclonal goat anti-rabbit antibodies conjugated to horseradish peroxidase (HRP) (Thermo Scientific) for 1 h, and washed 5 times. The secondary antibody was detected with SuperSignal West Femto maximum sensitivity substrate (Pierce) imaged using a ChemiDoc MP system (Bio-Rad). Bands were detected and quantified using Image Lab software (Bio-Rad).
We previously described a suppressor selection strategy to identify the downstream targets of the holdfast inhibitor protein, HfiA (18). Briefly, we selected for mutants that were insensitive to hfiA overexpression, as evidenced by their ability to elaborate holdfasts and attach to a polystyrene surface. In addition to identifying HfiA-suppressing mutations in hfsJ (18), this selection yielded two independent suppressing mutations at the 3′ end of dnaK. First, in a pilot screen using a Tn5 mutagenized pool, we identified a strain carrying a Tn5 transposon insertion at nucleotide 9924 of the C. crescentus chromosome (NCBI reference sequence NC_011916.1), corresponding to codon I587 of dnaK. This insertion results in the truncation of the final 44 residues of the C-terminal lid (Fig. 1C and andD)D) and the addition of 11 codons before a stop codon is reached. In a subsequent screen for spontaneous mutants, we isolated an additional mutant (named B5A) carrying two spontaneous mutations in dnaK. The first mutation is a nonsense mutation in codon K568, resulting in the truncation of the final 64 residues of DnaK (Fig. 1C and andD),D), and the second, which would have resulted in the nonsynonymous change I579V, is after the nonsense mutation and thus not manifested in the protein sequence. For simplicity the dnaK(K568*, I579V) allele is referred to as dnaK(K568*) here. In the presence of an intact hfiA overexpression plasmid, both of these isolated dnaK mutant strains exhibit elevated surface attachment compared to their wild-type parent as measured by crystal violet staining (Fig. 2). As B5A harbors multiple polymorphisms compared to the wild type (Table 3), we tested whether the dnaK(K568*) allele was sufficient to account for the increase in surface adhesion. We generated a strain with only the dnaK(K568*) allele and transformed this strain with the same hfiA overexpression plasmid used in our initial screen. This strain phenocopied the original B5A suppressor strain exhibiting the same increased surface adhesion compared to the wild type (Fig. 2). Furthermore, to evaluate if any other mutations in the B5A background affect surface attachment, we restored the mutant dnaK locus in the B5A background to the wild-type allele, leaving all other mutations in the background intact. For comparison, we then added the same hfiA overexpression plasmid used in the screen. This B5A dnaK+ strain phenocopied the wild type (Fig. 2). We conclude that the dnaK mutations are the primary determinant of the elevated adhesion phenotype in the B5A mutant strain.
To further characterize the role of dnaK in modulation of cell adhesion, we generated strains harboring additional mutant alleles of dnaK. Given that dnaK is essential in C. crescentus (19, 20), we pursued a strategy to generate a minimal functional allele of dnaK. Based on the high-density Tn sequencing data of Christen et al. (20), we noted that C. crescentus tolerates Tn5 insertions at the 3′ end of dnaK, after codon 496. As such, we engineered a strain with a nonsense mutation in codon 497. Notably, this is the site of two dnaK frameshift mutations reported to suppress the phenotypic effects of dnaA overexpression (34). Among the 25 dnaK mutant alleles recovered by Jonas et al. (34) that result in reduced levels of dnaA, the frameshift at Q497 (Q497fs) is the maximally truncated allele. These two independent screens suggest that this allele, which results in the loss of the entire alpha-helical lid domain, is the maximal nonlethal truncation of the dnaK in C. crescentus (Fig. 1C and andD).D). Our strain bearing a nonsense codon at position 497, dnaK(Q497*), exhibits increased surface adhesion in the presence of the hfiA overexpression plasmid (P < 0.0001), similar to the dnaK mutant strains identified in our hfiA suppressor screen (Fig. 2).
We next tested if truncation of the DnaK substrate-binding lid affected surface attachment in an otherwise wild-type genetic background, i.e., in the absence of the hfiA overexpression plasmid. Specifically, we evaluated surface adhesion and holdfast development in the dnaK::Tn5 strain cured of the hfiA overexpression plasmid and in the two engineered dnaK 3′ lid truncation strains. In complex peptone-yeast extract (PYE) medium, dnaK mutant strains are significantly enhanced in surface attachment compared to the wild type (Fig. 3A). We note a trend of increasing adhesion that correlates with increasing truncation of the substrate-binding lid domain of DnaK. Using standard two-step recombination, we restored the dnaK locus to the wild-type allele in each of our mutant strains. This restoration of dnaK fully complemented the hyperadhesive phenotype and restored attachment to wild-type levels (Fig. 3A).
To test whether elevated bulk surface attachment correlated with an increase in holdfast development, we sought to directly measure the fraction of cells that displayed a holdfast in each of our dnaK mutant strains by staining with fluorescent wheat germ agglutinin (WGA). This lectin binds to N-acetylglucosaminyl residues and has been used to specifically label holdfast (7). In logarithmic phase in PYE medium most wild-type cells develop a holdfast (18), which makes it difficult to detect an increase in the fraction of cells with a holdfast. Thus, to assess whether the dnaK mutants show increased holdfast development, we turned to a minimal defined medium (M2-xylose) in which wild-type cells elaborate holdfast at a low frequency (18). Under this condition, we observe that truncation of DnaK results in a significant increase in the fraction of cells with a holdfast (P < 0.01 to P < 0.0001) (Fig. 3B and andC).C). Greater truncations of the C terminus correlate with an increased frequency of holdfast development, consistent with our bulk surface attachment measurements. Again, restoration of the dnaK locus to the wild-type allele fully complements holdfast development to wild-type levels (Fig. 3B).
Clearly, the capacity of hfiA to function as an adhesin inhibitor is attenuated in strains harboring truncated dnaK alleles. This phenotype could be due to DnaK-dependent effects on downstream targets of HfiA or could be a result of reduced steady-state levels of functional HfiA in the cell. Given that (i) HfiA is a small, largely hydrophobic protein and a potent inhibitor of adhesion (18), (ii) DnaK has a high affinity for exposed hydrophobic protein regions (26,–28), and (iii) truncation of the substrate-binding lid of DnaK results in increased surface adhesion, we hypothesized that DnaK plays a role in stabilizing HfiA and, thus, controlling the level of functional HfiA in the cell. Efforts to generate polyclonal antiserum to directly detect HfiA by immunoblotting were unsuccessful due to difficulties in protein expression and purification.
Alternatively, we sought to assess the level of HfiA in the cell by generating tagged HfiA fusions. An ideal tag minimally disrupts function and enables detection by immunoblotting. We evaluated fusions with the small hemagglutinin (HA) epitope tag and with fluorescent proteins, GFP and mCherry, each expressed from an hfiA promoter on a plasmid integrated at the xylose (xylX) locus. We observed that the tag, and the position of the tag, affect the ability of the fusion protein to complement the ΔhfiA hyperadhesive phenotype. This is presumably due to differing stabilities of the fusion proteins (see Fig. S1 in the supplemental material). The HA and GFP fusions fail to complement the null phenotype in complex medium. In minimal medium, the HfiA-GFP fusion complements the null phenotype but was not detected with our anti-GFP antibody. We focused our efforts on the mCherry fusions, which were robustly detected with anti-mCherry antisera (Fig. 4C). Expression of each of the three mCherry fusions attenuated adhesion in a ΔhfiA background compared to the empty vector control and thereby functionally complemented the ΔhfiA hyperadhesion phenotype (Fig. 4B; see also Fig. S1 in the supplemental material). We note that all three mCherry fusions reduced C. crescentus surface adhesion to below wild-type levels. From this, we infer that HfiA-mCherry fusions are more stable than the untagged native HfiA protein. Indeed, the fusions that resulted in the most marked reduction in adhesion are present in higher levels in the cell as assessed by immunoblotting (Fig. 4C). The adhesive properties of the ΔhfiA strain bearing the C-terminal HfiA-mCherry fusion more closely resembled the wild type than the strains bearing N-terminal fusions (Fig. 4B). Thus, we proceeded to characterize the effects of DnaK truncation on steady-state levels of this HfiA fusion.
We introduced the plasmid containing the HfiA-mCherry fusion into the dnaK mutant strains and then assessed steady-state levels of this fusion in each dnaK mutant background by immunoblotting. Strains bearing truncated dnaK alleles exhibited reduced levels of HfiA-mCherry relative to wild-type C. crescentus in both minimal and complex media (Fig. 5A and andB).B). Mutant strains with smaller truncations of the C terminus of DnaK did not show (statistically) significantly lower levels of HfiA-mCherry, although levels of fusions did trend lower; the precision of this assay is not ideal for detection of small changes in protein levels. However, deletion of the entire alpha-helical lid domain from the C terminus of DnaK resulted in ≈40% reduction of steady-state HfiA-mCherry levels in cells grown in minimal medium (P < 0.0001) and ≈25% reduction in cells grown in complex medium (P < 0.05). This defect in protein level can be complemented by restoration of the truncated dnaK locus to wild-type dnaK (Fig. 5).
The trend in HfiA-mCherry protein levels across our DnaK truncation strains is inversely correlated with the trend in holdfast development and bulk surface adhesion (Fig. 3 and and5).5). Moreover, the dnaK mutant strains bearing the hfiA-mcherry fusion show enhanced surface attachment with increasing truncation of dnaK (see Fig. S2 in the supplemental material), similar to strains without the HfiA-mCherry fusion (Fig. 3). These observations are consistent with the function of HfiA as a holdfast inhibitor and provide evidence that protein chaperone functions can control activity of HfiA at the posttranslational level by stabilizing HfiA. In this manner, DnaK function is required for proper developmental control of the holdfast surface adhesin and regulated surface attachment.
To evaluate if DnaK truncation has HfiA-independent effects on holdfast formation, we deleted hfiA in the strains bearing dnaK(K568*) and dnaK(Q497*) alleles. Deletion of hfiA results in derepression of holdfast synthesis and holdfast elaboration on nearly all cells (18). Indeed, both dnaK(K568*) ΔhfiA and dnaK(Q497*) ΔhfiA strains displayed holdfast on almost every cell, similar to dnaK+ ΔhfiA cells (Fig. 6). These epistasis results support the model that HfiA is downstream of DnaK and that the effects of DnaK truncation on holdfast formation are dependent on HfiA. However, because nearly all ΔhfiA cells elaborate a holdfast, a further increase upon truncation of dnaK would be virtually impossible to detect. Thus, we cannot rule out the possibility that dnaK truncation has other effects on adhesion.
Truncation of the DnaK lid correlates with destabilization of the HfiA-mCherry fusion and increased holdfast formation and surface attachment. However, these results do not distinguish whether the lid domain is required for wild-type function of DnaK by enabling chaperone activity or by determining DnaK protein stability in the cell. Biochemical characterization of lidless E. coli DnaK indicates that the lid promotes activity by enhancing substrate binding (32, 33). To test whether removal of the C-terminal lid domain affects DnaK stability, we evaluated steady-state levels of each DnaK allele by immunoblotting with anti-DnaK antiserum (Fig. 7). In both complex and defined media, we observe that the lid domain does not significantly affect steady-state levels of DnaK in the cell.
We envision at least two mechanistic models that are consistent with the genetic and molecular connections we observe between DnaK, HfiA, and holdfast development. In the first model, HfiA is a direct client of DnaK and the stabilization of HfiA by DnaK is compromised by deletion of the lid domain. Alternatively, the effects of DnaK on HfiA stability could be indirectly mediated by the Lon protease as follows. Defects in DnaK chaperone activity result in accumulation of misfolded proteins, which in turn stimulates Lon protease (34). Thus, Lon stimulation in strains with truncated DnaK could result in enhanced proteolysis of HfiA. To discriminate between these models, we sought to evaluate whether the observed effects of DnaK truncation on HfiA proteostasis and holdfast development require lon. To this end, we deleted lon in strains bearing the dnaK(Q497*) allele. If HfiA destabilization requires Lon activation, then the dnaK(Q497*) Δlon strain should exhibit restored HfiA stability and reduced holdfast development compared to the dnaK(Q497*) strain alone. However, holdfast development is not reduced and HfiA-mCherry is not stabilized in dnaK(Q497*) Δlon cells compared to dnaK(Q497*) lon+ cells (Fig. 8). These data are inconsistent with the model that dnaK truncation indirectly modulates adhesion development via activation of Lon protease.
Similar to previous reports (34,–36), our Δlon strains exhibit profound pleiotropic developmental defects, including a high frequency of filamentous cells with multiple constrictions arising from incomplete cell divisions and notable occurrences of y-shaped cells, presumably arising from inappropriate positioning of polar development proteins (Fig. 8). Consistent with these developmental defects, cells lacking lon can be found with mislocalized holdfasts along the sides of the cell rather than at the pole. Moreover, lon deletion increases the probability of holdfast development irrespective of the dnaK allele (Fig. 8). Thus, cells lacking both the lid domain of DnaK and the Lon protease display pleiotropic developmental defects owing to the loss of Lon and increased holdfast development similar to strains bearing only single mutations (Fig. 8). The contribution of each mutation to the dysregulation of holdfast development in the double mutant strain is difficult to discern.
Molecular chaperones fulfill a range of functions related to establishment and maintenance of the proteome by facilitating de novo folding and refolding of denatured proteins. In an unbiased forward genetic screen for hyperadhesive C. crescentus mutants, we identified two independent mutations in the 3′ end of dnaK. Strains harboring these mutations had higher levels of surface attachment and increased frequency of holdfast development. We further demonstrated that systematic truncation of the 3′ end of dnaK, which encodes the C-terminal alpha-helical lid domain, resulted in decreased levels of HfiA in the cell. This result is consistent with the elevated adhesion and holdfast phenotypes we observe in these strains.
We have previously shown that small relative changes in hfiA transcription have large phenotypic consequences with respect to the frequency of holdfast development at the single-cell level and bulk adhesion at the population level (18). Here, we provide evidence that a small relative decrease in HfiA protein level corresponds with increased frequency of holdfast development and surface adhesion, and that the C-terminal lid domain of DnaK is a molecular determinant of HfiA proteostasis under standard laboratory growth conditions. These data lend further support to a model in which HfiA can modulate C. crescentus cell adhesion in an almost binary fashion across a relatively low (≈2- to 3-fold) dynamic range of expression. Certainly, HfiA concentration alone may not fully determine its phenotypic effects. HfiA, in concert with HfsJ, may require additional layers of posttranslational control that ensure robust regulation of cell adhesion.
In contrast to E. coli, DnaK is essential for growth in C. crescentus (19). However, removing the alpha-helical lid domain at the C terminus of DnaK does not have a significant effect on C. crescentus growth; three independent genetic studies have identified viable strains bearing transposon insertions (20), frameshift mutations (34), or nonsense mutations (this work) in the linker that connect the lid domain to the substrate binding beta sheet domain. These results are consistent with studies of E. coli demonstrating that lid-less DnaK retains functionality, albeit at a reduced level. Specifically, C-terminally truncated DnaK maintains ATP-dependent binding of peptide substrates, undergoes substrate-activated ATP-hydrolysis in vitro, and supports replication of bacteriophage λ in vivo (32, 33). While the lid domain is not necessary for substrate binding, it does serve to enhance affinity for substrate and increase the lifetime of DnaK-substrate complexes. Our data provide evidence that loss of the lid domain of C. crescentus DnaK, though not lethal, does compromise DnaK function with regard to its role in determining cell adhesiveness. Lid-less DnaK results in reduced HfiA levels in the cell, which ultimately has a large effect on holdfast development and surface adhesion.
Our data support a model in which the small hydrophobic protein, HfiA, is a direct client of DnaK. In such a model, the HfiA-DnaK interaction is compromised in strains missing the lid domain. As a consequence, HfiA is insufficiently stabilized and cleared from the cell. However, our genetic results do not exclude the possibility of indirect effects of DnaK on HfiA stability. We note that models in which DnaK directly or indirectly affects HfiA and holdfast development are not mutually exclusive. Efforts to evaluate a model in which DnaK truncation stimulates Lon-mediated proteolysis of HfiA are complicated by the pleiotropic effects of lon deletion. In addition to developmental defects, cells lacking lon exhibit increased holdfast development. At this time, we cannot discern whether this is due to deficient hfiA transcription as a consequence of stabilization of SciP and, thus, inhibition of CtrA (36), a positive regulator of hfiA transcription (18), decreased functioning of DnaK due to an overload of misfolded proteins that cannot be cleared, or other effects of a misregulated cell cycle. When lon deletion is combined with dnaK truncation, the effect of each mutation on holdfast development becomes difficult to disentangle. Future in vitro characterization of the interactions between DnaK and cellular proteases with HfiA will be necessary to more fully understand the molecular regulation of holdfast development.
Molecular chaperones, including DnaK, are highly expressed in single-species biofilms cultivated in laboratory settings (37, 38) and in natural biofilm communities in situ (39). In a clinical context, chaperones are critical for development and maintenance of biofilms in species that cause human disease, including E. coli, Staphylococcus aureus, Streptococcus mutans, and Listeria monocytogenes (40,–43). In the marine symbiont Vibrio fischeri, DnaK is required for appropriate host colonization and biofilm formation in the host squid light organ (44). This broad reliance on DnaK in the formation of biofilms across a range of bacterial taxa suggests that proteostasis is a challenge that bacteria must overcome when residing in surface-attached communities. Indeed, chaperones have been implicated as determinants of biofilm resistance to antibiotics and other antimicrobial agents (38, 43). High expression of chaperones in biofilms mitigates the destabilizing effects that high-density growth conditions have on the bacterial proteome. As such, DnaK is an attractive target for pharmacological destabilization of biofilms (40).
However, in contrast to other studied bacterial species, where impairment of DnaK inhibits biofilm formation and maintenance, impairment of C. crescentus DnaK enhances biofilm formation. The dependence of HfiA levels on DnaK suggests that this molecular chaperone provides the cell with an additional mechanism to control holdfast development and surface adherence. Proteotoxic conditions strongly induce DnaK expression in C. crescentus (34, 45,–48). In a model where HfiA is a client for DnaK, increased levels of DnaK are predicted to stabilize HfiA and reduce the number of newborn cells that develop holdfasts. This would decrease the probability of attachment in proteotoxic environments and foster dispersal of progeny from biofilms that have entered a “stressful” phase. In short, we posit that direct control of HfiA stability by DnaK is a posttranslational regulatory mechanism that tunes adhesion in response to environmental conditions.
We thank Patrick Viollier for generously providing antiserum for mCherry and Kristina Jonas for generously providing antiserum for DnaK. We also thank Jon Henry for building the pNPTS138-Δlon plasmid.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JB.00027-16.