ABSTRACT
In Escherichia coli O157:H7 strain ATCC 43895, a guanine-to-thymine transversion in the csgD promoter created strain 43895OR. Strain 43895OR produces an abundant extracellular matrix rich in curli fibers, forms biofilms on solid surfaces, invades cultured epithelial cells, and is more virulent in mice than strain 43895. In this study we compared the formic acid-soluble proteins expressed by strains 43895OR and 43895 using one-dimensional sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis and identified two differentially expressed proteins. A 17-kDa protein unique to strain 43895OR was identified from matrix-assisted laser desorption ionization-time of flight analysis combined with mass spectrometry (MS) and tandem MS (MS/MS) as the curli subunit encoded by csgA. A <10-kDa protein, more highly expressed in strain 43895, was identified as the Lpp lipoprotein. Mutants of strain 43895OR with disruption of lpp, csgA, or both lpp and csgA were created and tested for changes in phenotype and function. The results of this study show that both Lpp and CsgA contribute to the observed colony morphology, Congo red binding, motility, and biofilm formation. We also show that both CsgA and Lpp are required by strain 43895OR for the invasion of cultured HEp-2 cells. These studies suggest that in strain 43895OR, the murein lipoprotein Lpp indirectly regulates CsgA expression through the CpxAR system by a posttranscriptional mechanism.
Escherichia coli serotype O157:H7 is estimated to cause more than 73,000 illnesses per year in the United States (26). From 1982 to 2002, beef products (e.g., ground beef) were the most common vehicle for food-borne outbreaks associated with serotype O157:H7 and were more than twice as common as produce-associated outbreaks during that same period (26). Despite concentrated efforts to improve sanitation and dressing procedures within slaughter and processing plants, contamination with and persistence of E. coli O157:H7 within plants remains a paramount problem. In addition to instances of illness, during 2007 alone, E. coli O157:H7 contamination led to the recall of more than 27 million lb. of beef (1).
Biofilm formation can enhance the persistence of food pathogens on plant and processing surfaces and serve as a source of product contamination (21). Within the gut, biofilms may develop as coexisting commensal and pathogenic strains of E. coli (2, 4). The production of curli fimbriae and the exopolysaccharide cellulose are arguably the most common contributors to biofilm formation by Escherichia coli and Salmonella spp. In human commensal E. coli isolates, 46 to 79% of the isolates produced curli, cellulose, or both components depending on temperature (2). Those strains expressing either one or both of these components had medium to high biofilm-forming capabilities.
In E. coli and Salmonella spp., the coexpression of curli and cellulose leads to an aggregative colony phenotype (rdar; red, dry, and rough) when grown on medium containing Congo red dye (32, 40). Inactivation of the cellulose operon in these strains results in a brown (bdar) colony phenotype, inactivation of the curli operon generates a pink (pdar) colony phenotype, and inactivation of both the curli and cellulose genes renders colonies smooth and white (saw). Curli fimbriae and cellulose production in E. coli are complex and involve many regulators (13, 24). CsgD is a key transcriptional regulator which controls the transition from the planktonic to the multicellular state through regulation of both curli and cellulose production (29). CsgD expression from a promoter located between the divergent csgDEFG and csgBA operons is controlled by several regulatory proteins, including positive regulators (i.e., OmpR and MlrA) and negative regulators (i.e., CpxR) (24). The two-component Cpx system, composed of CpxA (sensor) and CpxR (regulator), controls stress responses and cell adaptation to the environment, including biofilm formation (9).
In a previous study, we identified two phase-variable, Congo red-binding strains (43895OR and 43894OR) derived from E. coli O157:H7 strains ATCC 43895 and 43894, respectively (36). Both Congo red-binding variants produced an rdar phenotype which could phase switch back to a saw phenotype in the parent variants. Expression of the rdar phenotype in each strain was accompanied by a single base switch in the −10 region of the csgD promoter that resulted in a sequence identity that more closely resembled the consensus σ70-dependent promoter sequence. The Congo red-binding variants invaded cultured epithelial cells as efficiently as Salmonella enterica serovar Typhimurium and enteroinvasive E. coli, while the saw variants were no more invasive than E. coli strain K-12 (37). In addition, the rdar variant of one strain, 43895, was more virulent in a mouse model, produced larger amounts of biofilm on various surfaces, and was highly resistant to certain antimicrobials (30, 35, 37). In a similar study, expression of a cloned curli operon in E. coli K-12 increased adhesion to and invasion of eukaryotic cells (14). Later, it was shown that O157:H7 strains expressing curli also adhere in greater numbers to HEp-2 cells (20).
The objective of this study was to gain a better understanding of the components and regulatory mechanisms utilized in the transition between strains 43895 and the invasive, more-virulent, biofilm-forming rdar variant, 43895OR. Ultimately, these findings may help determine if a transient existence within the rdar state is a mechanism used by strain 43895 to enhance virulence and persistence in the host or the environment. We used a comparison of formic acid-soluble proteins extracted from 43895 and 43895OR to identify two differentially regulated proteins. In addition to confirming earlier observations that 43895OR, but not 43895, expresses CsgA (35), we also determined that Lpp is differentially expressed in these strains. E. coli Lpp is a major outer membrane lipoprotein which exists in two forms: one-third of Lpp is linked to peptidoglycan and two-thirds is unbound in the cell envelope (18). Lpp maintains cell surface integrity by connecting the outer membrane to murein (34). Studies in S. Typhimurium showed that Lpp deletion reduces the invasion of HEp-2 cells, motility in soft agar, and virulence in mice (11, 31), suggesting that Lpp expression may promote those phenotypes. Mutant strains of 43895OR with inactivation of lpp, csgA, or both csgA and lpp were constructed to study their effects on colony morphology, cell invasion, motility, and biofilm formation.
MATERIALS AND METHODS
Bacterial strains, media, phase-switching assays, and motility assays. E. coli strain 43895OR is a phase-variable, curli-producing strain isolated from a freezer stock of the noncurliated strain ATCC 43895 (CDC EDL933) obtained from the American Type Culture Collection (ATCC; Rockville, MD) (36). The promoter of strain 43895OR, compared to strain 43895, contains a guanine-to-thymine transversion at base −9 from the csgD transcriptional start site, which results in increased csgD transcription (36). Bacterial strains were maintained on Congo red indicator (CRI) plates or YESCA plates (16). A single colony of strain 43895OR was passed in brain heart infusion (BHI) broth (Becton Dickinson, Sparks, MD) for 5 days at 37°C and red (43895OR) and white (43895) variants were isolated on CRI plates. Stocks of the variants were frozen in BHI broth containing 15% glycerol and used to initiate the experiments in this study. The propagation of strains on agar plates and during experiments requiring incubations greater than 24 h were performed at ≤30°C to minimize the phase switching of strain 43895OR to the saw variant. TransforMax EC100 (Epicentre Biotechnologies, Madison, WI) was used as the host for all cloning steps unless otherwise specified and was propagated in BHI or Luria-Bertani (LB) broth (Becton Dickinson). To test the time required to induce phase switching, Congo red-binding strains were passed (1:100) daily in BHI broth and maintained at 37°C with shaking at 160 rpm. Strains were plated daily on CRI agar to identify the passage day on which saw colonies first appeared. The mean minimal day required for phase switching of three independent cultures of each tested strain was analyzed by analysis of variance and Bonferroni least significant difference (LSD) analysis. The starting cultures of strains tested in swimming and surface spreading assays were prepared in LB broth made without salt (LB-NS) and incubated for 48 h at 25°C without shaking. Swimming assays were performed as described by Burkart et al. (3). Plates containing 1.0% tryptone (Becton Dickinson), 0.5% NaCl, and 0.3% agar were hardened for 2 hours at 25°C and inoculated by stabbing the center of each plate with the tapered end of a toothpick wetted in the appropriate 48-h culture. The diameters of colonies grown at 25°C were plotted at 16, 19, and 24 h. The starting cultures for complementation studies contained 10 μg/ml tetracycline and incubations were shortened to 24 h to ensure maintenance of plasmids pLPP322 and pBR322. Colony diameters were compared at 18 h. The surface spreading of strains was tested on 1.0% tryptone, 0.5% NaCl, and 0.6% agar prepared with or without 0.5% glucose and dried overnight at room temperature. Starting cultures were spotted (2 μl) on the surface and colony diameters were recorded after incubation for 48 h at 25°C. For both swimming and surface spreading assays, the means of the zones of migration in millimeters were calculated from the results of a single measurement of five independent samples from each strain at each time point. The data were analyzed by analysis of variance, and the means were separated using the Bonferroni LSD mean separation technique.
Construction of mutant and reporter strains and β-glucuronidase assays.The csgA gene was PCR amplified from strain 43895 using primers 5′-AAGCTTGATAACAGCGTATTTACGTGGG and 5′-GGATCCCAACTTCGTCAAAGCAATGGG, cloned into plasmid pCR2.1-TOPO (Invitrogen Corporation, Carlsbad, CA), and moved to plasmid pBluescript II KS(+) (Stratagene, La Jolla, CA) at the BamHI/HindIII sites. The kanamycin resistance gene from plasmid pUC4K was transferred to the unique EcoRI site in plasmid pBluescript II KS(+)::csgA, and the interrupted csgA with flanking sequences was reamplified by PCR using Platinum Taq DNA polymerase high fidelity (Invitrogen). The blunt PCR products were cloned into the SmaI/SacI sites of plasmid pCVD442 (8). The interrupted csgA was integrated into strain 43895OR by homologous recombination as described by Donnenberg et al. (8) to create strain 43895OR-CsgA. The complete promoter and lpp gene from strain 43895 were amplified using primers Lppfor (5′-AGAGAATTCACCGCATCTGTTCACGTCCTG) and Lpprev (5′-AGAAAGCTTGCGCCATTTTTCACTTCACAGG) and cloned into the EcoRI/HindIII sites of plasmid pBR322, producing plasmid pLPP322. The spectinomycin resistance cassette from plasmid pIC156 (33) was cloned into the BamHI/XbaI sites of plasmid pBCSK+, excised at the XbaI/ClaI sites, and transferred to corresponding sites in pLPP322. The resulting plasmid, containing a 96-bp deletion in the lpp promoter and open reading frame, was used as template with the Lppfor/Lpprev primers to generate a PCR product that was digested with BglI and used as a linear template for chromosomal recombination using the Quick and Easy gene deletion kit (Gene Bridges GmbH, Heidelberg, Germany). Deletion of lpp in strains 43895OR and 43895OR-CsgA produced strains 43895OR-Lpp and 43895OR-C/L, respectively. The bcsA gene in strain 43895OR was deleted according to the manufacturer's protocol for the Quick and Easy gene deletion kit using forward and reverse primers, designed from the EDL933 genome sequence (accession number AE005174), that provided 60 bp of upstream and downstream sequence homology and terminated with the bcsA start and stop codons, respectively. The correct integration of all interrupted genes was verified by PCR using primers (not shown) that were designed from sequence outside of the chromosomal regions that provided homology for recombination. The lpp promoter from strain 43895 was amplified using Platinum Taq DNA polymerase high fidelity (Invitrogen) with primers 5′-AGAGAGCTCGTAAATCTGCTCGCGCAGTACG and 5′-AGAGGATCCTACCCTCTAGATTGAGTTAATCTC and cloned in front of E. coli uidA in plasmid pMLK117 at the SacI/BamHI sites (19). The promoter activities in transformed strains were compared by assaying for β-glucuronidase activity as described previously (12). Promoter activities in nanomoles of p-nitrophenol released per minute per milligram of total protein were calculated from the results of duplicate assays of three individual samples from one (24 h on YESCA agar) or two (24 h in YESCA broth and 48 h on YESCA agar) independent trials. The mean enzymatic activity of each tested strain carrying plasmid pMLK117 was subtracted from the mean activity of that strain carrying pMLK117::lpp to eliminate nonspecific background. The results were analyzed by analysis of variance using a randomized complete block design and Proc Mixed of SAS to estimate the effects and interaction means. Means were separated by the Bonferroni LSD technique.
Protein separation and measurement.Strains were grown on YESCA agar for 48 h at 25°C and analyzed for total soluble proteins as previously described (28) with modifications. Polymerized curli require treatment with formic acid prior to gel separation (5). Cell pellets collected from 4 ml of cells were suspended in sterile water to an optical density at 600 nm of 1.0, resuspended in 200 μl of 98% formic acid for 10 min on ice, vacuum dried, and resuspended in 200 μl of Laemmli buffer (Bio-Rad Laboratories, Hercules, CA). All samples were separated on a 12% bis-Tris gel (Invitrogen) as described previously (22) and stained with Bio-Safe Coomassie stain (Bio-Rad). The relative amounts of CsgA protein expressed by various strains were compared using the densitometry function of an AlphaImager (Alpha Innotech, San Leandro, CA). The densitometer measured the number of pixels in each of the Bio-Safe Coomassie-stained CsgA bands separated from the total proteins extracted from equal cell quantities of each strain as determined by plate counts.
Mass spectroscopy and protein identification.Protein bands separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) were stained with Bio-Safe Coomassie, extracted using 50% acetonitrile, 5% trifluoroacetic acid (TFA) solution from the 8 to 16% polyacrylamide gels or 12% bis-Tris gels, and digested with Trypsin Gold (Promega, Madison, WI) using the manufacturer's protocol. After digestion, samples were treated with 2% TFA to stop the trypsin activity and the resulting peptides were extracted and cleaned using a precleaned C18 ZipTip, washed with water containing 0.1% TFA, reextracted with acetonitrile-water (50:50)-0.1% TFA, and mixed with a recrystallized α-cyano-4-hydroxy-cinnamic acid matrix solution (5 mg/ml; acetonitrile-water [50:50]-0.1% TFA) to a final concentration between 100 fmol and 1 pmol/μl. Approximately 0.6 to 0.7 μl of the peptide-matrix solution was spotted in the mass spectrometer target plate. Matrix-assisted laser desorption ionization-time of flight mass spectrometry with automated tandem time of flight fragmentation of selected ions (MALDI-TOF/TOF) data for trypsin-digested proteins were acquired with a 4700 Proteomics Analyzer mass spectrometer (Applied Biosystems, Framingham, MA) as previously described (15).
Electron microscopy.The morphology of bacterial strains following growth on YESCA agar for 48 h at 25°C was studied by scanning electron microscopy (SEM). Petri dishes (10 cm) bearing bacterial colonies on agar were inverted over lids containing 9-cm-diameter, number 2 Whatman filter papers saturated with 25% glutaraldehyde solution, sealed with parafilm, and floated on a 40°C water bath for 1 hour. Next, approximately 1- by 2-cm agar strips with colonies were excised and immersed in 50% ethanol. Dehydration was gradually continued in 80% ethanol, followed by several changes of absolute ethanol. Side views of the colonies were obtained by cross-fracturing the agar strips after immersion in liquid nitrogen, using a cold surgical scalpel blade, followed by thawing in absolute ethanol and critical point drying from liquid CO2. Fractured faces of agar containing colonies were mounted on specimen stubs with carbon adhesive tabs (Electron Microscopy Sciences, Hatfield, PA), sputter coated with a thin layer of gold, and viewed in the high-vacuum secondary electron imaging mode with a Quanta 200 FEG scanning electron microscope (FEI Co., Hillsboro, OR).
qRT-PCR.The transcription levels of csgA, lpp, and cpxR were compared in various strains by quantitative real-time reverse transcriptase PCR (qRT-PCR) using the primer pairs 5′-ATCAGTACGGTGGTGGTAACTC with 5′-CCAACATCTGCACCGTTACCAC, 5′-TCTACTCTGCTGGCAGGTTGCTC with 5′-ATTGCGTTCACGTCGTTGCTC, and 5′-AACGACAACGGTTCACCGACAC with 5′-ACCGGTTAACTCCAGCGTTTGC, respectively. Strains were grown on YESCA agar for 48 h at 25°C, and RNA was extracted using TRI reagent (Molecular Research Center, Inc., Cincinnati, OH) according to the manufacturer's instructions. Contaminating DNA was removed by DNase I digestion using Turbo DNA-free (Ambion Inc., Austin, TX). The cDNA was generated from 1 μg total RNA using the SuperScript III first strand synthesis system according to the manufacturer's protocol (Invitrogen). qRT-PCR was performed on a SmartCycler II (Cepheid, Sunnyvale, CA) using 25-μl reaction mixtures consisting of 50 ng cDNA (or 50 ng RNA as a negative control), 0.2 mM deoxynucleoside triphosphates, 1× PCR buffer, 2 mM MgCl2, 0.5× Evagreen dye (Biotium Inc., Hayward, CA), 0.5 μM primers (Integrated DNA Technologies, Coralville, IA), and 1.25 U AmpliTaq Gold DNA polymerase (Applied Biosystems). Amplification of the rrsB gene, using primers 5′-GAATGCCACGGTGAATACGTT and 5′-ACCCACTCCCATGGTGTGA, was used as a control to normalize the results. The reaction mixtures were subjected to 95°C for 600 s, followed by 40 cycles of 95°C for 15 s, 64°C for 10 s, and 68°C for 20 s, with the optics on. Real-time PCR data were analyzed using an approach similar to that of Cook et al. (6) but incorporating nonlinear modeling of the logarithmically transformed fluorescence data to test for nondifferent amplification efficiencies and to determine the relative differences in expression with the associated confidence intervals. Expression values represent the average of single measurements of PCR product amplified from cDNA generated from six independent cultures of each strain.
Biofilm and invasion assays.Biofilms were generated on glass coupons in LB-NS using a crystal violet assay as previously described (35). For invasion studies, HEp-2 cells were obtained from ATCC and maintained on minimal essential medium alpha (Invitrogen) supplemented with 10% fetal bovine serum (MP Biomedicals, Aurora, OH) at 37°C in a 5% CO2 atmosphere. Invasiveness of cultured HEp-2 cells was assayed in a gentamicin protection assay (37). Recovered bacteria were enumerated by spread plating on BHI agar. The log10 CFU/ml values for the recovered strains were compared by analysis of variance with separation of means using a Bonferroni LSD technique at a P level of 0.05. Each experimental trial was performed on four independent samples from each strain and the results of three experimental trials performed on different days were compared by separating the run variability (blocks) from the error term.
RESULTS
Lpp and CsgA are differentially expressed in strains 43895 and 43895OR.The SDS-PAGE separation of total formic acid-soluble protein is shown in Fig. 1. Strains 43895OR, 43895OR-Lpp, and 43895OR-Lpp carrying pLPP322 all showed bands corresponding to approximately 17 kDa that were identified by MALDI-TOF combined with MS and MS/MS spectra analysis as CsgA monomers. Strain 43895 and the two CsgA-deficient strains (43895OR-CsgA and 43895OR-C/L) showed no visible CsgA expression. SDS-PAGE band intensity of CsgA was greater in strain 43895OR than in strains 43895OR-Lpp or 43895OR-Lpp carrying plasmid pLPP322. Strains 43895, 43895OR, 43895OR-CsgA, and 43895OR-Lpp carrying plasmid pLPP322 all had bands at <10 kDa that were not present in the strains with deletion of lpp (strains 43895OR-Lpp and 43895OR-C/L). These protein bands were identified by mass spectroscopy as processed Lpp (native Lpp following cleavage of the leader peptide). Lpp SDS-PAGE band intensity for strain 43895 was greater than that of strain 43895OR and its mutant progeny.
A bis-Tris gel of formic acid extractions from strains 43895 (lane 2), 43895OR (lane 3), 43895OR-Lpp (lane 4), 43895OR-Lpp with pLPP322 (lane 5), 43895OR-CsgA (lane 6), and 43895OR-Lpp/CsgA (lane 7), shown with molecular weight marker proteins (lanes 1 and 8). The positions corresponding to CsgA and Lpp protein migrations are shown in bold outlines.
Lpp deletion reduces CsgA expression in 43895OR.Densitometry analysis was used to estimate the relative quantity of the CsgA proteins separated from standardized amounts of each strain (gel not shown). The stained CsgA band from strain 43895OR-Lpp (29,600 pixels/band) was 63% of the density of the band separated from equivalent numbers of strain 43895OR (44,770 pixels/band). The density of the CsgA band of strain 43895OR-Lpp carrying plasmid pLPP322 (37,000 pixels/band) was 83% of the density of the 43895OR CsgA band. Lpp bands from some of the quantity-standardized strains were faint, but the strain 43895OR Lpp band was less dense than that of strain 43895 (results not shown).
lpp and csgA are differentially expressed and bcsA is not expressed in strains 43895 and 43895OR.The results of the β-glucuronidase assays comparing lpp promoter expression in strains 43895 and 43895OR are shown in Fig. 2. After growth on YESCA agar for 24 h at 25°C, lpp expression in strain 43895 was more than 11-fold greater (P < 0.05) than in strain 43895OR. After 48 h of growth, lpp expression remained almost threefold higher (P < 0.05) in strain 43895 compared to strain 43895OR. The 24-h expression of lpp by strain 43895 was twofold greater (P < 0.05) than expression at 48 h. In contrast, the 24-h lpp expression of strain 43895OR was less than at 48 h (P > 0.1) Following growth in YESCA broth for 24 h, lpp expression from strain 43895 was again more than twofold higher (P < 0.05) than expression from strain 43895OR. Expression of lpp by either strain was higher (P < 0.05) in broth at 24 h than on agar at either time point.
Comparison of the β-glucuronidase activities of strains 43895 and 43895OR containing an lpp::uidA fusion in plasmid pMLK117 and cultured at 25°C in YESCA broth for 24 h or on YESCA agar for 24 h or YESCA agar for 48 h. The enzymatic activity is reported as nanomoles of p-nitrophenol released/min/mg of total protein. Bars represent the averages of duplicate assays of three individual samples from one (24 h agar) or two (24 h broth and 48 h agar) independent trials and were analyzed by analysis of variance using a randomized complete block design.
qRT-PCR analyses comparing the expression of lpp, csgA, and bcsA in strains 43895OR and 43895, showed that csgA expression in strain 43895OR was approximately 1,000-fold higher (confidence interval [CI], 1,802.0 to 613.7) than in strain 43895, where there was essentially no expression. The cellulose synthase gene, bcsA, showed little or no expression in strain 43895OR and only minimal expression in strain 43895. The expression of lpp in strain 43895 was 2.27-fold higher (CI, 1.5 to 3.4) than that of strain 43895OR.
Lpp affects CsgA posttranscriptionally through the Cpx system.The results of the comparison of csgA, lpp, and cpxR expression levels in strains 43895OR-Lpp with pBR322 and 43895OR-Lpp carrying pLPP322 to that of the parent strain 43895OR with pBR322 are shown in Table 1. In strain 43895OR-Lpp with pBR322, lpp expression dropped more than 165-fold compared to strain 43895OR with pBR322. The lpp primers did amplify some small amount of product from the reverse-transcribed RNA of the lpp-deleted strain. This slight background could represent trace contamination from the laboratory or contamination of Taq polymerase, as has been described elsewhere (17). Transcription of csgA in strain 43895OR-Lpp with pBR322 was more than twofold higher than in strain 43895OR with pBR322. Deletion of lpp in strain 43895OR with pBR322 activated the CpxAR system by increasing cpxR expression more than 11-fold.
Relative expression levels of lpp, csgA, and cxpR in E. coli strain 43895OR-Lpp compared to those in strain 43895OR
The expression of lpp was more than sixfold greater in strain 43895OR-Lpp containing plasmid pLPP322 than in strain 43895OR with pBR322, confirming lpp transcription from pLPP322. The expression of csgA remained more than twofold greater in strain 43895OR-Lpp with pLPP322 than in the wild-type strain 43895OR with pBR322. pLPP322 restored the CsgA protein reductions in 43895-Lpp but did not reduce the increases in csgA transcription. Transcription of cpxR was returned to wild-type levels in strain 43895OR-Lpp by plasmid pLPP322, indicating that the loss of lpp is responsible for activating cpxR in strain 43895OR-Lpp.
Lpp enhances but CsgA is essential for the rdar phenotype and biofilm formation.Strain 43895OR produced a dark red colony with a dry surface on CRI plates after 24 to 48 h at 25°C (results not shown). The colonies formed dry aggregates when removed from the agar surface. Strain 43895OR-Lpp produced a red to brown, smooth colony on CRI agar which did not form dry aggregates when colonies were disturbed. Strains 43895, 43895OR-CsgA, and 43895OR-C/L all produced a smooth and white phenotype, although strain 43895OR-C/L appeared slightly darker in color, suggesting that it bound small amounts of dye. When strain 43895OR with disruption of the bcsA gene was tested on CRI agar, the colony morphology was indistinguishable from that of strain 43895OR (results not shown), confirming our expression data and earlier staining studies (35) which suggested that cellulose does not contribute to the rdar phenotype. When strain 43895OR-Lpp was transformed with plasmid pLPP322, the colony color on CRI agar at 25°C after 48 h appeared more red than colonies of strain 43895OR-Lpp carrying pBR322, but not as dark red or as dry as the colonies of strain 43895OR carrying pBR322, indicating that pLPP322 could partially transcomplement the phenotypic deficiencies of the lpp mutant strain (results not shown).
SEM demonstrated only minor differences between strains 43895OR and 43895OR-Lpp (Fig. 3). Both strains appeared as tightly packed, rod-shaped bacteria embedded in a network of extracellular matrix following growth on YESCA plates for 48 h at 25°C. While 43895OR had a regular outer membrane, the surface of 43895OR-Lpp had irregularities consistent with membrane blebbing (38). Strains 43895OR-CsgA and 43895OR-C/L were indistinguishable from strain 43895 on SEM (results not shown).
Scanning electron micrographs of colony growth on YESCA agar after 48 h at 25°C for strains 43895 (A), 43895OR-Lpp (B), and 43895OR (C). The enlarged area of panel B demonstrates strain 43895OR-Lpp membrane blebs.
The results of crystal violet biofilm assays (Fig. 4) indicated that both strains 43895OR (mean optical density [MOD], 0.581; standard deviation, [SD], 0.103) and 43895OR-Lpp (MOD, 0.417; SD, 0.063) formed significantly more biofilm than strain 43895 (MOD, 0.023; SD, 0.003; P < 0.05), which was not different than either strain 43895OR-CsgA (MOD, 0.39; SD, 0.005; P > 0.05) or 43895OR-C/L (MOD, 0.20; SD, 0.003; P > 0.05). However, the amount of biofilm produced by strain 43895OR-Lpp was significantly less than with strain 43895OR (P < 0.05), indicating that Lpp has a positive effect on CsgA-dependent biofilm formation.
Biofilm assay results of strains 43895OR, 43895OR-Lpp, 43895OR-CsgA, 43895-C/L, and 43895 incubated with glass coupons in LB-NS broth for 48 h at 25°C. Each bar represents the mean optical density (OD) of eluted crystal violet dye measured at 590 nm ± the standard deviation from a single sample taken from each of three independent trials. Letters in brackets represent the results of a Bonferroni LSD means separation, and strains having the same letter are not statistically different from each other.
CsgA and Lpp are essential for cell invasion in 43895OR.The results of the HEp-2 cell invasion assays are shown in Fig. 5. Strain 43895OR (mean log10 CFU/ml, 3.11; SD/√n, 0.12) was recovered from gentamicin-treated HEp-2 cells in 10-fold-higher numbers than strain 43895 (mean log10 CFU/ml, 2.02; SD/√n, 0.09; P < 0.05). Strains 43895OR-CsgA (mean log10 CFU/ml, 2.22; SD/√n, 0.07) and 43895OR-Lpp (mean log10 CFU/ml, 2.31; SD/√n, 0.06) were recovered in numbers that were not different from those of strain 43895. Strain 43895OR-C/L (mean log10 CFU/ml, 1.93; SD/√n, 0.1) was slightly less invasive (P < 0.05) than strains 43895OR-CsgA and 43895OR-Lpp but was not significantly different (P > 0.05) than strain 43895. When we attempted to complement the cell invasion deficiencies of strain 43895OR-Lpp with plasmid pLPP322 under the same conditions used in the initial invasion assays, there were no differences in invasion between strains 43895OR-Lpp carrying pBR322 and strain 43895OR-Lpp carrying pLPP322 (results not shown). However, growing the strains to be tested overnight in LB broth containing 10 μg/ml tetracycline to maximize lpp transcription and to discourage loss of the plasmid resulted in the recovery of strain 43895OR-Lpp carrying pLPP322 (mean log10 CFU/ml, 2.85; SD, 0.16) from cultured cells in greater numbers (P < 0.05) than strain 43895OR-Lpp carrying pBR322 (mean log10 CFU/ml, 2.33; SD, 0.22) but in lower numbers (P < 0.05) than strain 43895OR carrying pBR322 (mean log10 CFU/ml, 3.34; SD, 0.19), indicating that plasmid pLPP322 partially complemented the invasion deficiencies of strain 43895OR-Lpp.
HEp-2 cell invasion assays comparing strains 43895OR, 43895OR-Lpp, 43895OR-CsgA, 43895-C/L, and 43895. Each bar represents the mean log10 CFU/ml ± SD/√n of bacteria recovered from gentamicin-exposed HEp-2 cells. Single counts were taken on each of four independent samples from each strain per experiment. The results of three experiments performed on different days were compared by separating the run variability (blocks) from the error term. Letters in brackets represent the results of a Bonferroni LSD means separation, and strains having the same letter are not statistically different from each other.
Lpp and CsgA have variable effects on motility.The results of the swimming motility assays are shown in Fig. 6. At each time point, strains 43895OR-Lpp and 43895OR-C/L showed significantly greater (P < 0.05) motility than all other strains but they were not different from each other. Motility of strain 43895 was lower (P < 0.05) at each time point compared to all other strains. Strains 43895OR and 43895OR-CsgA had intermediate levels of motility and were slightly different (P < 0.05) from each other at 16 and 24 h but not at 19 h. In complementation studies, strain 43895OR-Lpp carrying pBR322 produced 18-h migration zones (mean zone diameter [MZD], 11.4 mm; SD, 0.65) greater (P < 0.05) than those of strain 43895OR carrying pBR322 (MZD, 8.6 mm; SD, 0.74) or strain 43895-Lpp carrying pLPP322 (MZD, 8.3 mm; SD, 0.57). Migration was not different (P > 0.05) between strains 43895OR and 43895OR-Lpp with pLPP322, indicating that pLPP322 could fully complement the motility suppression exerted by the chromosomal lpp.
Swimming motility of strains 43895OR, 43895, 43895OR-Lpp, 43895OR-CsgA, and 43895OR-C/L in 0.3% agar at 25°C. Points represent the mean diameters of the zones of migration measured at 16, 19, and 24 h during incubation at 25°C. Means were calculated for five independent biological samples of each strain at each time point, analyzed by analysis of variance, and separated using the Bonferroni LSD mean separation technique.
When the same strains were compared for surface spreading on 0.6% agar, there were strain differences as well as differences attributable to the addition of glucose to the agar (Fig. 7). On LB agar without glucose, strains 43895OR and 43895OR-Lpp had greater (P < 0.05) zones of migration than all other strains but were not different from each other. Strain 43895OR-C/L had the smallest mean migration zone but was not significantly different from strain 43895OR-CsgA and was only slightly lower (P < 0.05) than strain 43895. In the presence of 0.5% glucose, strains 43895OR and 43895OR-CsgA produced significantly larger (P < 0.05) migration zones than the strains of 43895OR with deletion of lpp (43895OR-Lpp and 43895OR-C/L), suggesting that Lpp has a positive effect on surface spreading in the variant 43895OR. However, strain 43895 in the presence of glucose had the smallest zone of migration in spite of its ability to produce Lpp. There was also a positive glucose effect on surface spreading since each individual tested strain had larger migration zones when grown on medium containing glucose compared to medium without glucose.
Surface spreading motility of strains 43895OR, 43895OR-Lpp, 43895OR-CsgA, 43895OR-C/L, and 43895 on 0.6% agar with or without 0.5% glucose following incubation for 48 h at 25°C. Bars represent the mean diameters of migration zones calculated for five independent biological samples of each strain and analyzed by analysis of variance with means separation by the Bonferroni LSD mean separation technique. Strains with the same letter in brackets or parentheses were not significantly different from each other in comparisons in agar with or without added glucose, respectively.
Deletion of lpp in strain 43895OR slows initiation of phase switching.When strains 43895OR with pBR322, 43895OR-Lpp with pBR322, and 43895OR-Lpp with plasmid pLPP322 were passed daily in BHI broth to determine the mean minimum passage (MMP) required to initiate phase switching, strains 43895OR with pBR322 (MMP, 6.3; SD, 1.2) and 43895OR-Lpp with pLPP322 (MMP, 4.0; SD, 2.6) both reverted to the curli-deficient variant after fewer daily passes (P < 0.05) than strain 43895OR-Lpp with pBR322 (MMP, 12.3; SD, 2.3). However, there was no difference in the MMP required for phase switching between strains 43895OR with pBR322 and 43895OR-Lpp with pLPP322. Plasmid pLPP322 fully restored phase switching in an lpp-deleted strain.
DISCUSSION
Our finding of higher Lpp expression in strain 43895 compared to the more-virulent strain 43895OR is contrary to findings in studies in S. Typhimurium where Lpp was required for full virulence (31). However, lpp deletion in strain 43895OR reduced cell invasiveness, indicating that Lpp does enhance virulence in that strain. In addition, lpp deletion reduced, but did not abolish, CsgA expression, which we have shown is essential for invasion. The fact that strain 43895, which does not express CsgA, has high Lpp expression indicates that Lpp is not required or sufficient to initiate CsgA expression. Apparently, while some expression of Lpp is needed to maximize CsgA expression, the differences in Lpp expression between the two variants have no effect on the phenotypes in this study. However, the expression differences from lpp promoter fusions in strains 43895 and 43895OR suggest that csgD has a suppressive effect on lpp. qRT-PCR results showed that the lower lpp expression in strain 43895OR compared to strain 43895 under the growth conditions used for the invasion studies was accompanied by nearly 1,000-fold-higher csgD expression (results not shown). Additional studies will be required to fully define the regulatory role of csgD on Lpp.
In a study that profiled the global gene expression of S. Typhimurium with deletion of both copies of lpp, Fadl (10) described a nearly fivefold decrease in cpxP expression compared to the wild-type strain. The cpxP gene encodes a periplasmic repressor of the Cpx regulon (25). Although the transcription of curli was not studied in that strain, studies have shown that CpxR is a transcriptional suppressor of csgA through regulation of both the csgD and csgB promoters (23). The results of our qRT-PCR and complementation studies verified that loss of lpp in strain 43895OR increased cpxR transcription. It would be expected that increased CpxR would suppress CsgA expression. However, although CsgA protein was decreased, csgA transcripts were elevated, indicating that the Csg operon was refractory to the transcriptional suppression by CpxR and that the CsgA protein was being reduced by a posttranscriptional mechanism. The Cpx regulatory system controls protein expression by both transcriptional and posttranscriptional mechanisms, although less is known about the posttranscriptional mechanisms (9). The posttranscriptional Cpx effects are proposed to be mediated through their activation of protein folding and degrading factors (9). Activation of the Cpx system leading to degradation of curli subunits in the periplasmic space was predicted be the mechanism responsible for a dominant negative curli-suppressing phenotype associated with the expression of a CsgG recombinant protein unable to localize to the outer membrane (27). However, it was recognized that CpxR could also be suppressing transcription of the csg operons and curli formation (27). Our results indicate that a posttranscriptional influence of CpxR can suppress CsgA and curlin expression, even in the absence of transcriptional controls on csgD. We do not know why CpxR did not suppress csgA transcription in strain 43895-Lpp. The csgD promoter of strain 43895OR contains a CpxR box overlapping an OmpR box as described for E. coli K-12 (23). CpxR binds to several sequences in the csgD and csgB promoters where it may negatively affect transcription by either interfering with activator binding (OmpR) or by blocking recruitment of polymerase to either promoter (9). Curli expression in strain 43895OR likely depends on CsgD expressed from a σ70-dependent promoter and could be refractory to activator proteins that modulate rpoS-dependent promoters. It is unclear why CpxR binding at sites overlapping the start of transcription would not suppress transcription at the csgD or csgB promoters unless transcription is being initiated from alternate sites.
High curli expression will activate the Cpx pathway, resulting in curli downregulation after cells enter stationary growth (23). However, CsgA accumulated to high levels in strain 43895OR, while cpxR expression remained minimal. In fact, cpxR expression was only increased following Lpp deletion. Apparently, CsgA expression, regulated by CsgD driven by a σ70-dependent promoter, is refractory to feedback suppression mediated by the Cpx system. Based on our phase-switching studies, curli expression in strain 43895OR appears to be an unfavorable state that drives a phase switch back to the curli-deficient state. The inability of CpxR or other regulators to downregulate curli when expression levels become detrimental may contribute to the unique properties of strain 43895OR.
Lpp was required for complete rdar expression in 43895OR but could not induce Congo red affinity in the absence of CsgA, indicating a regulatory rather than a direct role in dye binding. Overexpression of curli alone has been shown to be sufficient for dye binding and biofilm formation in an E. coli K-12 strain with an ompR mutation that was predicted to preferentially utilize σ70 for csgD induction (23). Likewise, the constitutive curli overexpression driven by a σ70-dependent csgD promoter in strain 43895OR may be sufficient for full rdar expression and biofilm formation provided that Lpp is functional to prevent cpxR induction.
In this study, CsgA expression was essential for biofilm formation but the effects of Lpp loss were minimal in spite of causing significant reductions in CsgA. Apparently, Lpp augmentation of CsgA expression was not required to surpass the minimum CsgA threshold required for biofilm formation. In contrast, both CsgA and Lpp were essential for HEp-2 cell invasion, indicating that the invasion phenotype may require a very high minimum threshold of curli expression that can only be attained by concurrent Lpp expression, making Lpp an essential gene for that phenotype.
It is possible that Lpp has a direct role in cell invasion rather than, or in addition to, that of a CsgA regulator. In S. Typhimurium, Lpp was required for invasion but the exact mechanism was not defined and the role of curli fibers (thin aggregative fimbriae) was not examined (31). Lipoprotein released from growing bacteria has been detected in culture supernatants (39). However, we could only detect Lpp when extraction protocols included SDS, and we could not detect Lpp in culture supernatants (results not shown). Although E. coli serotype O157:H7 is considered noninvasive, our studies suggest that under conditions where curli expression is unusually high, an invasive and more virulent phenotype could emerge. In strain 43895OR, that phenotype requires expression of constitutive σ70-dependent csgD under regulatory conditions, which favor maximized CsgA expression and stability, i.e., expression of Lpp and minimal expression of cpxR.
Decreased motility in S. Typhimurium has been described due to a loss of Lpp (31). In contrast, deletion of lpp in E. coli strains 43895OR-Lpp and 43895OR-C/L had a positive effect on motility in soft agar. This was surprising, as we have shown that cpxR is increased in 43895OR when lpp is deleted. In E. coli K-12, phosphorylated CpxR inhibits motility by suppressing transcription of flagellar (motAB) and chemotaxis (cheAW) genes, which has been suggested to be an energy-saving strategy during starvation (7, 9).
Surface spreading is a characteristic of the S. Typhimurium rdar phenotype, and a similar but less pronounced response occurs in strain 43895OR (35). In this study, csgA but not lpp had a positive effect on surface spreading in all strains on media without glucose. In the presence of glucose, Lpp had a positive effect on surface spreading in the mutant and wild-type strains of 43895OR, suggesting that flagellar and chemotaxis genes are activated in the presence of certain sugars. Interestingly, the saw variant 43895, which had the highest Lpp expression, showed the least surface spreading in the presence of glucose, suggesting that the transition from the curliated (43895OR) to the noncurliated (43895) variant adds to the regulatory complexity governing motility, possibly due to csgD expression differences. The control of surface swarming is very complex and it will take detailed studies to define the contributions of curli, Lpp, CpxR, and other regulators, such as the carbon source regulator CRP, which may be acting under these conditions.
In this study we have shown that CsgA is an essential protein for the development of the rdar phenotype, biofilm formation, and cell invasion in E. coli O157:H7 strain 43895OR. Although Lpp loss also affected these phenotypes, our results indicate that the Lpp effect is realized through its positive influence on CsgA production. These results also indicate that the phenotypic differences between 43895 and 43895OR result from high expression of curli and that augmentation of curli expression beyond a certain threshold will result in a more virulent and invasive strain. Whether such a strain exists in nature remains to be determined. However, the ability of strain 43895OR to switch back to strain 43895 raises the interesting possibility that such a state could be transient and subject to phase variation.
ACKNOWLEDGMENTS
We gratefully acknowledge the assistance of John Phillips (statistical analyses), Peter Cooke (microscopic imaging), David Needleman (DNA sequencing), Gary Richards and David Kingsley (cell culture), and Alberto Nunez and Lori Fortis (protein analyses). We also thank Bryan Cottrell, Donna Rodgers, Paul Pierlott, and Jonnee Almond for their technical assistance.
Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture.
FOOTNOTES
- Received 29 July 2008.
- Returned for modification 13 September 2008.
- Accepted 17 January 2009.
- Copyright © 2009 American Society for Microbiology
