Charles W. Kaspar, Department of Bacteriology, 1550 Linden Drive, University of Wisconsin-Madison, Madison, WI 53706, USA. E-mail: firstname.lastname@example.org.
This study analysed the growth and survival of 18 strains of the six serotypes of non-O157 Shiga toxin-producing Escherichia coli (STEC) (O26, O45, O103, O111, O121 and O145) most frequently implicated in human illness and compared them with Escherichia coli O157:H7 strain ATCC43895.
Methods and Results
The data from growth in Luria–Bertani broth (LB)-HCl (pH 4·0, 4·5, 4·8), LB-lactate (pH 4·5 and 4·8) and LB-NaCl (5%, 7%) were fitted to modified Gompertz equations to enable quantitative comparisons across strains and media conditions. Serogroup O45 strains had growth rates that were equal to or significantly greater than the O157:H7 control strain in all growth conditions tested. The growth rate was independent from the maximum growth achieved, but three strains (103A, 121A and 45B) had significantly faster growth and greater maximum cell densities in LB-NaCl 5% (strain 103A), LB-HCl pH 4·0 (strain 121A) and LB-NaCl 7% (strain 45B). Survival in LB-HCl pH 3·0 of four strains (103C, 111B, 26B and 26C) was significantly greater and five strains (26A, 45A, 111A, 121A and 145A) were significantly reduced in comparison with the O157:H7 control strain. None of the STEC strains had greater survival in LB-NaCl 12% than the O157:H7 control strain. A significant association was found between the exponential phase, but not stationary phase, RpoS level and survival of STEC.
Some STEC strains had growth or survival properties that exceeded those of the O157:H7 control strain, but none of the non-O157 STEC had both significantly greater growth and survival properties. STEC survival was associated with the exponential-phase RpoS level.
Significance and Impact of the Study
Results from this study define the variability in growth and survival of STEC strains that will be useful defining food product formulations and process interventions to control STEC. The presence of exponential phase σs expands the significance of this alternative sigma factor.
Some strains of Shiga toxin-producing Escherichia coli (STEC) are human pathogens that can cause haemorrhagic colitis and in some cases may progress to haemolytic–uraemic syndrome (Besser et al. 1999; Gyles 2007). In the United States, serotype O157:H7 strains are most frequently associated with illness, causing approximately 63 000 cases per year (Scallan et al. 2011), while non-O157 STEC are responsible for approximately 113 000 infections annually (Scallan et al. 2011). More than two hundred STEC serogroups have been identified, and over 100 of these have caused sporadic cases or outbreaks (Brooks et al. 2005; Johnson et al. 2006). Six non-O157 STEC serogroups (O26, O111, O103, O121, O45 and O145) are most commonly associated with human disease, accounting for 71% of isolates (Brooks et al. 2005). Not all non-O157 STEC cause illness in humans, indicating that virulence factors in addition to Shiga toxin production are necessary for infection (Coombes et al. 2008). Other genes involved in pathogenesis include those for type III secretion, intimin (eae), translocated intimin receptor (tir) and enterohemolysin (E-hly) (Gyles 2007; Coombes et al. 2008; Vanaja et al. 2010). In general, non-O157 STEC are less virulent than E. coli O157:H7, but a recent European outbreak of E. coli O104:H4 centred in Germany demonstrated the capacity of non-O157 STEC to cause severe disease (Bielaszewska et al. 2011; Rubino et al. 2011).
Outbreaks and cases of STEC have been linked to the consumption of contaminated food such as ground beef, raw milk, produce, water and by contact with an infected animal or person (Brooks et al. 2005; Rangel et al. 2005; Gyles 2007). Acidic foods, like apple cider and fermented meats, have also been implicated in outbreaks (Besser et al. 1993; Tilden et al. 1996; Ethelberg et al. 2009), which led to studies that found some STEC O157:H7 strains are acid tolerant (Miller and Kaspar 1994; Conner and Kotrola 1995; Cheville et al. 1996) and able to survive the fermentation and drying steps in the production of fermented meats (Clavero and Beuchat 1996; Hinkens et al. 1996; McQuestin et al. 2009). Central to the dissemination and survival of STEC in unfavourable conditions is the general stress response system that is regulated by the alternative sigma factor RpoS (σs), encoded by rpoS (Cheville et al. 1996; Boor 2006; Klauck et al. 2007). RpoS plays an important role as cells switch from growth to survival and cell protection (Loewen et al. 1998; Hengge-Aronis 2000; Klauck et al. 2007) and is a master regulator of hundreds of genes that are part of the RpoS regulon or subnetworks (Klauck et al. 2007). The steady-state level of RpoS is tightly controlled at transcription, translation and post-translationally (Lange and Hengge-Aronis 1994; Loewen et al. 1998; Klauck et al. 2007). Genes within this regulon impart protection from a variety of stresses including pH, oxidative and osmotic stresses (Cheville et al. 1996; Loewen et al. 1998; Hengge-Aronis 2000; Klauck et al. 2007) and conditions encountered during animal passage (Price et al. 2000); therefore, the steady-state level of RpoS may be a useful predictor of the survival properties of an organism.
A number of studies have defined growth and survival parameters for E. coli, E. coli O157:H7, and some non-O157 STEC (Presser et al. 1998; Duffy et al. 2000), but there is a need for additional data with a greater diversity of STEC serogroups and strains. The aim of this study was to generate growth and survival data for the six most epidemiologically significant serotypes of non-O157 STEC, estimate growth and survival parameters and evaluate their relationship to RpoS levels for use as a possible predictor of growth and survival properties. Data were fitted to growth curve models to allow for quantitative strain-to-strain comparison across various growth conditions.
Materials and methods
Bacterial strains, plasmids and primers
The bacterial strains, plasmids and primers used in this study are listed in Tables 1 and 2. Overnight cultures of E. coli were grown aerobically in Luria-Bertani (LB; BD, Sparks, MD, USA) broth at 37°C with shaking (150 RPM). Growth and survival studies were performed using three different isolates of the six epidemiologically significant non-O157 serogroups (O26, O45, O103, O111, O121 and O145). A genetic profile of each strain was determined by Gonzales et al. (2011) using the OpenArray® System (Applied Biosystems, Carlsbad, CA, USA).
Growth in low pH and low water activity (aw) was monitored using a Bioscreen C (Growth Curves USA; Piscataway, NJ, USA). Media types included LB adjusted to pH 4·8, 4·5 and 4·0 with 6·0 mol l−1 HCl or lactic acid (85% w/w; Sigma-Aldrich, St Louis, MO, USA), and LB with 5, 7 and 9% (w/v) NaCl. Respective final molarities were 12·3, 14·7 and 20·4 mmol l−1 HCl; 14·4, 18·2 and 30·1 mmol l−1 lactate; and 0·86, 1·20 and 1·54 mol l−1 NaCl. LB with 5, 7 and 9% NaCl (final concentration) had an aw of 0·98, 0·96 and 0·95, respectively. Water activity measurements were taken at room temperature using an AquaLab 4TE as described by the manufacturer (Decagon Devices, Inc, Pullman, WA, USA). Bioscreen C 100-well plates were loaded with 0·3 ml of medium inoculated with overnight cultures at a final dilution of 1 : 100 000. Optical Density at 600 nm (OD600) was measured hourly for 96 h at 30°C, and the resulting data were collected and organized in a spreadsheet. Wells not used during experiments were filled with uninoculated medium to minimize evaporation. Growth experiments were repeated three times.
Survival was assessed in low pH and low aw media. LB adjusted to pH 3·0 with 6·0 mol l−1 HCl (37·5 mmol l−1) and LB with 12% NaCl (aw of 0·92) were used. Overnight cultures (1 : 100) were used to inoculate 10 ml of the challenge medium in a 50-ml conical tube (BD Biosciences, Sparks, MD, USA) and then incubated at 30°C. Viable cell counts were determined immediately, followed by daily viable cell counts for three days in LB-HCl (pH 3·0) and weekly counts for two weeks in LB at aw 0·92. Tryptic soy agar (TSA; BD) was used for plate counts and 0·1% Bacto-peptone (BD) was used for dilutions. A minimum detection limit of 1 CFU ml−1 for plate counts was accomplished by plating 1·0 ml of inoculated medium on duplicate TSA plates (150 mm; Fisher Scientific, Waltham, MA). Survival experiments were performed three times.
The level of RpoS for each isolate was determined in exponential phase, stationary phase and the transition from exponential to stationary phase. Each growth phase was defined as 0·2–0·4 OD600 for exponential phase, 1·4–1·9 OD600 for the transition phase and 2·9–4·0 OD600 for stationary phase. Optical densities were determined using a Thermo Spectronic BioMate 3 spectrophotometer (Thermo Scientific, Rockford, IL, USA). Growth curves and whole-cell extracts were obtained using 50 ml of LB at 30°C with shaking (100 rev min−1). At each growth phase, cells were harvested by centrifugation (16 100 g, 5 min) and resuspended to an OD600 of 1·0 that yielded approximately equal protein amounts. Total protein was extracted from the cell pellet in 200 μl of 2× tricine sample buffer [200 mmol l−1 Tris pH 6·8, 4% (w/v) SDS, 20% (v/v) glycerol, 1% (w/v) bromophenol blue, 5% (v/v) 2-mercaptoethanol].
Proteins were separated on a NuPAGE 4–12% Bis-Tris protein gel with 1× MES (2-(N-morpholino) ethane sulphonic acid) SDS running buffer (Invitrogen, Carlsbad, CA, USA) at 150 V and transferred to an Immun-Blot PVDF membrane (Bio-Rad) at 65 V for 1 h. A prestained protein standard was used (Bio-Rad, Hercules, CA). Immunoblotting was performed as described by Gentry et al. (1993) with minor adjustments. A 1 : 1500 dilution of mouse IgG anti-sigma S monoclonal (clone 1RS1) antibody (NeoClone Biotechnology, Madison, WI, USA) and a 1 : 2500 dilution of mouse IgG anti-DnaK monoclonal (clone 8E2/2) antibody (Enzo Life Sciences, Farmingdale, NY, USA) were used. Goat anti-mouse IgG peroxidase (1 : 10 000; Sigma-Aldrich) was applied as a secondary antibody. Incubations were conducted at room temperature for 1 h, and membranes were washed with phosphate-buffered saline (PBS; 137 mmol l−1 NaCl, 2·7 mmol l−1 KCl, 10·1 mmol l−1 Na2HPO4, 1·8 mmol l−1 KH2PO4, pH 7·4) supplemented with 2·5% Tween-20. Membranes were prepared for detection with Pierce ECL Western Substrate (Thermo Scientific) as described by the manufacturer and exposed to Blue Ultra Autorad Film (BioExpress, Kaysville, UT, USA). Protein bands were quantified using ImageQuant TL (GE Healthcare, ver. 7·0, Uppsala, Sweden). DnaK was used to confirm equal protein loading. A normalized ratio of RpoS to DnaK was used to compare RpoS levels between samples. Ratios were normalized to the protein extract from stationary phase E. coli O157:H7 that was loaded in each gel. Two trials were performed per strain.
A mutant strain of E. coli O157:H7 (ATCC43895) lacking rpoS was constructed using homologous recombination as previously described (Choi et al. 2000). A jointed PCR technique (Davidson et al. 2002) was used to amplify and join regions up- and downstream of rpoS. Primers KP325 and KP345 (Table 2) amplified a 521 bp region upstream of rpoS, partially located within nlpD, and primers KP328 and KP346 amplified a 540 bp region downstream of rpoS, partially located within the Z4048 region of the chromosome. Overlapping sequences designed into primers KP345 and KP346 allowed for a 1040 bp nlpD::Z4048 fragment lacking rpoS to be amplified during a second round of PCR using the initial pre- and post-rpoS DNA fragments. nlpD::Z4048 was ligated into pCVD442 (Donnenberg and Kaper 1991) at the SmaI restriction site, forming pCK2782. Transformation of pCK2782 into SY327 λ pir (Miller and Mekalanos 1988) allowed for vector amplification and storage. SM10 λ pir, which contains the tra gene necessary for conjugal mobilization of vectors with a RP4 origin of transfer (Simon et al. 1983), was transformed with pCK2782 for conjugation with ATCC43895. Transformants were selected by the addition of antibiotics in the following concentrations, ampicillin at 100 μg ml−1 and kanamycin at 30 μg ml−1. Incorporation into host chromosome was confirmed by growth on sorbitol-MacConkey agar (SMAC) supplemented with ampicillin at 40 μg ml−1. Because ATCC43895 does not contain λ pir, necessary for replication of pCK2782 that contains oriR6K (Miller and Mekalanos 1988), ampicillin resistance could only be conferred by the presence of the vector in the chromosome. ATCC43895 with rpoS::pCK2782 was grown on LB agar to allow for a second crossover event to occur and then plated on SMAC with 20% sucrose. Growth in the presence of sucrose indicates the removal of sacB-containing pCK2782 from the chromosome. Deletion of rpoS was confirmed by PCR using primers KP325 and KP328.
All statistics were performed using R version 2·12·2 (R Development Core Team 2011). Data from growth and survival studies were loge transformed. The minimum value of growth measurements was negative: loge (0·075) = −2·59. As the model is unable to use negative values, the value 2·59 was added to all loge transformed growth measurements (y = loge (OD600) − loge (0·075). Initially, the modified Gompertz equation (Zwietering et al. 1990) was used for modelling all growth curves because each coefficient of the equation can be biologically interpreted. However, this equation does not account for a death phase (decrease in OD600 after the maximum value is reached) that can occur during growth in certain media. Two additional equations were developed based on the modified Gompertz equation to improve the model fit and accommodate the death phase. To accomplish this, growth was modelled in two phases: first, from the start until maximum growth (tmax) is reached using the original modified Gompertz equation ((G1) below), and second, the decay in value beyond the point of maximum growth. The first variation of the modified Gompertz equation ((G2) below) uses linear decay, and the second variation ((G3) below) uses exponential decay to account for a decrease plateau in OD600. The equations used are as follows:
Parameter definitions are the following: μ1 is the maximum growth rate; a1 is the asymptotic maximum growth; λ is the duration of lag phase; t is time; tmax is the time at which decay starts; μ2 is the rate of decay at time tmax; a2 is the difference in loge OD600 value from a1 to the decay plateau (Fig. 1). Subsets of growth data were made based on strain, medium and trial, and equations were modelled using nonlinear regression with the nls function from the stats library that is innate to R. Subsets were first modelled with equations (G1) and (G2) and the fit of the models were compared using a log-likelihood ratio test (Pinheiro and Bates 2000) at the 0·05 significance level. If the preferable model was derived from equation (G2), equation (G3) was then modelled and the fit of models from equations (G2) and (G3) were compared using the log-likelihood ratio test at the 0·05 significance level. Parameters μ1, a1 and λ estimated by the preferable model were extracted and used for further analysis. The goodness of fit of the models was evaluated using plots of the predicted versus fitted values and by testing for the randomness of the residuals plotted over the fitted values. Linear regression analysis was used to model parameter differences (α = 0·05) among strains where E. coli O157:H7 was defined as the reference strain. To test for the significance of parameter differences, an overall two-way analysis of variance (anova) was performed on all media and all strains in a single test. This provided significance for interactive effects, and a one-way anova was then performed for each medium separately. Pairwise comparisons among strains were only made if the one-way anova gave a P-value <0·05. The parameters μ1 and a1 were graphed as the differences to the values estimated for E. coli O157:H7 (Fig. 2).
A mixed effect linear regression model was used to estimate the negative slope of the survival data. The functions used were lme from the nlme library (Pinheiro et al. 2011) and lmer from the lme4 library (Bates et al. 2011). To accommodate repeated measurements per strain, trial was used as a random effect. A separate analysis was performed for each medium, and 95% confidence limits were calculated as follows: estimate ± standard error * t-value (Fig. 3). As some strains were below detection limits prior to experiment completion, t-values were obtained from a T-distribution where degrees of freedom were adjusted for the number of available observations (Dohoo et al. 2003). Finally, μ1, a1 and λ were predicted by RpoS level in all tested media using multiple linear regression by means of the lm function in R. As multiple comparisons between RpoS levels and growth curve parameters were made, one for each of the eight tested media, the Bonferroni method was used to adjust the significance level. A coefficient to RpoS level comparison was significant only if the P-value was less than α/n, where α is the normal significance level (0·05) and n is the number of comparisons per parameter (0·05/8 = 0·00625). Parameter estimates per strain for growth in LB-NaCl 5% were plotted for RpoS level during exponential phase for μ1, a1 and λ (Fig. 4). In addition, the survival rate (loge CFU per day) in LB-NaCl 12% (aw = 0·92) was plotted over the survival rate in LB-HCl (pH 3·0) with RpoS levels depicted by point size (Figs 5 and S2). A separate linear regression with an F-test was performed to determine the significance of a relationship between RpoS levels and survival rates in each medium.
Growth curve modelling
Growth curves were modelled with the modified Gompertz equations to enable quantitative comparisons between the biologically relevant parameter values, maximum growth rate (μ1), asymptotic maximum growth (a1) and length of lag phase (λ; Fig. 1). Modified Gompertz equation (G1) was initially used for parameter estimation, and equations (G2) and (G3) were used for growth curves exhibiting a death phase. Equation (G1) tended to underestimate the parameters μ1 and a1 in data sets that decreased in value after reaching a maximum OD600. Parameters were selected from the preferred model (α = 0·05) for linear regression analysis. Modified Gompertz equations (G1), (G2) and (G3) yielded the best fit for 85, 9 and 6% of the growth curves from the 20 STEC analysed, respectively.
The growth of 18 non-O157 STEC, E. coli O157:H7 (ATCC43895) and an rpoS deletion mutant (FRIK2787; Table 1) were monitored in 10 different media comprised of LB base with the pH modified by either HCl or lactate, or the aw changed by the addition of NaCl. None of the strains grew in two media, LB with 9% NaCl and LB-lactate (pH 4·0); therefore, these data were not included in the statistical analyses. In LB with 7% NaCl and LB-lactate (pH 4·5), strain 26A did not grow, but was included in the analysis to enable comparison among all strains. Growth curves with a death phase occurred in 63 and 42% of strains in LB-NaCl 5 and 7%, respectively. In acidified and untreated LB, a death phase occurred in 2·5% of the strains.
Figure 2 shows significant differences of coefficients μ1 (growth rate) and a1 (maximum growth) as compared to the E. coli O157:H7 control strain. In LB-HCl (pH 4·0, 4·5 and 4·8) and LB-lactate (pH 4·5 and 4·8), three strains (FRIK2787, 145C and 121B) had growth rates that did not differ significantly from the O157:H7 control strain (Fig. 2a, columns 1–5). Four strains (45A, 45B, 45C and 121A) had the same or significantly greater growth rates than the O157:H7 control strain. Specifically, strains 45B, 45C and 121A grew faster in LB-HCl (pH 4·0), while 45A and 121A grew faster in LB-lactate (pH 4·5 and 4·8). As a serogroup, O45 strains (45A, 45B and 45C) had growth rates that were equal to or significantly greater than the O157:H7 control strain in all growth conditions tested. The remaining 12 strains had slower growth rates than the O157:H7 control strain in the low-pH conditions tested. As a serogroup, all three O111 strains (111A, 111B and 111C) had significantly lower growth rates in LB-HCl (pH 4·5 and 4·8) and LB-lactate (pH 4·5 and 4·8) than the O157:H7 control strain. In contrast, a majority of the non-O157 STEC grew faster than the O157:H7 control strain in LB-NaCl 5% (13/18) and LB-NaCl 7% (9/18; Fig. 2a, columns 6–7). Three strains (121A, 121B and 121C) did not differ significantly from the O157:H7 control strain, while strains 26A and 145A had significantly slower growth rates in LB with 7% NaCl. It should be noted that in unmodified LB (pH 6·8, 1% NaCl), strains 26A, 121A, 121B and 121C had significantly slower growth rates, while the remaining 15 strains did not differ from the O157:H7 control strain. Overall, strains 26A, 121B and 121C appear to be impaired in their ability to grow in the conditions tested.
Although the growth rates of the non-O157 strains in acidified LB and LB-NaCl were varied in comparison with the O157:H7 control strain, most (13/18, 72%) of the non-O157 strains obtained an equivalent or lower maximum growth than the O157:H7 control strain in all tested media (Fig. 2b). Among the remaining five strains, three strains (45B, 145A and 121A) achieved greater maximum growth in either acidified LB or LB-NaCl but not both, and the remaining two strains, 45A and 103A, reached equivalent or greater maximum growth in all test media. All strains of serogroup O103 (103A, 103B and 103C) as well as strain 45A had maximum growth equivalent to or greater than the O157:H7 control strain in all test media.
The growth rate was independent from asymptotic maximum growth. Strain 45A was the only strain to have equivalent or greater growth rates and maximum growth than the O157:H7 control strain in the media tested. Three other non-O157 strains (45B, 121A and 103A) had greater growth rates and maximum growth than the O157:H7 control strain in one of the growth media tested. The strains and media in which this was observed were strain 121A in LB-HCl pH 4·0, 103A in LB-NaCl 5% and 45B in LB-NaCl 7%. Conversely, there were non-O157 strains with a greater growth rate but a lower maximum growth, including strain 45C in LB-HCl pH 4·0 and strains 26C and 145C in LB-NaCl 5%. In LB-lactate pH 4·8, the O157:H7 control strain had a significantly greater growth rate than 67% of the non-O157 STEC but maximum growth that did not differ from the other strains with the exception of strain 26A. Despite the ability of most (72%) non-O157 STEC to grow faster than the O157:H7 control strain in LB-NaCl 5%, only one strain (103A) obtained greater maximum growth.
Strain FRIK2787, the ΔrpoS mutant of E. coli O157:H7, did not have growth rates or maximum growth that varied significantly from the parent strain (O157:H7 control strain) in acidified LB (Fig. 2). While in LB-NaCl 5% and LB-NaCl 7%, strain FRIK2787 had significantly greater growth rates than the parent strain, and in LB-NaCl 5%, it had the fastest growth rate of all strains. As with the comparison of the other strains, the greater growth rates in LB-NaCl did not translate into greater maximum growth.
Survival of the 18 non-O157 STEC, the O157:H7 control strain, and FRIK2787 was examined in LB-HCl pH 3·0 and LB-NaCl 12% (aw = 0·92). The slope and 95% confidence limits of the linear regression for loge CFU ml−1 over time (survival rate) are shown in Fig. 3. In LB-HCl pH 3·0, the survival of four strains (103C, 111B, 26B and 26C) was significantly greater than that of the O157:H7 control strain. While in LB-NaCl 12%, none of the non-O157 STEC strains had significantly greater survival. Six strains (111A, 121A, 145A, 26A, 45A and FRIK2787) had significantly lower survival in both low aw and low-pH media than the O157:H7 control strain. The rpoS deletion mutant (FRIK2787) had the poorest survival at pH 3·0 but was not significantly different from other strains with poor survival in LB-NaCl 12%.
Total protein was harvested from strains at three points in their growth curves (exponential, late-exponential or transitional and stationary phases) and RpoS levels were determined by Western blotting. Immunoblots from a selection of strains are presented in Fig. S1. In general, the lowest quantity of RpoS was detected in exponential phase in all strains. In 50% of the strains, the maximum amount of RpoS was detected during the transition from exponential to stationary phase. RpoS was not detected in exponential or transition phases in strain 121A, and in stationary phase, RpoS was detected at 4% of the level of the O157:H7 stationary-phase control (Table 3). RpoS was not detected in FRIK2787 or strain 145A. A nonpathogenic strain of E. coli K12 was analysed for comparison and lacked detectable RpoS in exponential or transition phases, while in stationary phase, the RpoS level was 19% of the O157:H7 stationary-phase control (Fig. S1). Protein bands smaller than RpoS (38 kDa) appeared during transition and stationary phases in some strains (26B, 103A, 103B, 103C, 145B, 145C and ATCC43895). It is unknown if these bands are the result of nonspecific binding by the anti-RpoS antibody or RpoS degradation; therefore, they were not included in the quantification of protein levels.
Table 3. RpoS levels of Shiga toxin-producing Escherichia coli strains in different growth phases
Values represent a normalized ratio of RpoS to DnaK and are the average from two independent trials. Ratios were normalized to the protein extract from stationary phase Escherichia coli 0157:H7.
0·01 ± 0·01
0·36 ± 0·05
0·54 ± 0·43
0·60 ± 0·02
0·62 ± 0·43
0·67 ± 0·36
0·26 ± 0·20
0·28 ± 0·05
0·34 ± 0·10
0·45 ± 0·26
0·83 ± 0·29
1·08 ± 0·44
0·26 ± 0·10
0·64 ± 0·02
0·47 ± 0·07
0·20 ± 0·11
0·53 ± 0·13
0·33 ± 0·04
0·47 ± 0·17
0·94 ± 0·50
0·70 ± 0·40
0·51 ± 0·22
1·33 ± 0·89
1·23 ± 0·64
0·47 ± 0·05
0·96 ± 0·18
0·77 ± 0·05
0·34 ± 0·26
0·63 ± 0·21
1·02 ± 0·13
0·27 ± 0·10
0·58 ± 0·20
0·42 ± 0·10
0·37 ± 0·08
0·52 ± 0·10
0·39 ± 0·06
0·00 ± 0·00
0·00 ± 0·00
0·04 ± 0·04
0·27 ± 0·05
0·51 ± 0·10
0·53 ± 0·06
0·33 ± 0·02
0·50 ± 0·22
0·42 ± 0·06
0·00 ± 0·00
0·00 ± 0·00
0·00 ± 0·00
0·31 ± 0·15
0·86 ± 0·30
1·01 ± 0·28
0·30 ± 0·07
0·73 ± 0·05
0·54 ± 0·11
0·29 ± 0·02
0·70 ± 0·24
0·70 ± 0·42
0·00 ± 0·00
0·00 ± 0·00
0·00 ± 0·00
Comparisons were made between RpoS levels and the growth curve parameters μ1, a1 and λ in the tested growth media. Figure 4 shows coefficients μ1, a1 and λ from growth in LB with 5% NaCl and exponential-phase RpoS levels. There was not a significant association between the maximum growth rate and exponential-phase RpoS levels in LB-NaCl 5% or any of the tested media (Fig. 4a, data not shown). There was a significant association between a1 and exponential-phase RpoS levels (Fig. 4b; P < 0·01), and strains with greater RpoS levels tended to achieve a greater maximum density. Growth in other media did not yield a significant association between a1 and exponential-phase RpoS levels (data not shown). Likewise, λ was significantly associated with exponential-phase RpoS level but only in LB-NaCl 5% (Fig. 4c; P < 0·01), wherein high levels of RpoS was related to a shorter lag phase. This association was not found in the other media tested (data not shown).
RpoS levels from exponential and stationary phases were compared with survival in LB-HCl pH 3·0 and LB-NaCl 12%. A significant association between exponential-phase RpoS levels and survival during exposure to pH 3·0 or LB-NaCl 12% NaCl was found. Strains with greater exponential-phase RpoS levels had greater survival. A comparison of exponential-phase RpoS levels to survival in LB-HCl pH 3·0 yielded a P-value of 0·00174 and in LB-NaCl 12% a P-value of 0·00612. Conversely, when the survival of strains was compared with stationary-phase RpoS levels, no significant association was found in either LB-HCl or LB-NaCl. A comparison of stationary-phase RpoS levels with survival in LB-HCl yielded a P-value of 0·066 and in LB-NaCl 12% a P-value of 0·38 (data not shown).
Figure 5 shows survival in LB-HCl pH 3·0 and LB-NaCl 12% with exponential-phase RpoS levels (depicted by point size). Twelve strains (26B, 45B, 45C, 103A, 103B, 103C, 111B, 111C, 121C, 145B, 145C and the O157:H7 control strain) form a cluster that had significant survival in both conditions. Exponential-phase RpoS levels (ratios of RpoS:DnaK) within this cluster ranged from 20 to 60% (average of 37 ± 12%) of the O157:H7 stationary-phase level. Outside of this cluster, eight strains had exponential-phase RpoS levels that ranged from 0% to 45% (average of 17 ± 18%) of the O157:H7 control. In LB-HCl, FRIK2787 (0% RpoS) had the lowest survival rate (coordinates −11·2 loge CFU ml−1 day−1, −0·75 loge CFU ml−1 day−1), but in LB-NaCl, it had greater survival than strain 45A, which had a RpoS level of 45% of the O157:H7 control.
Thirteen non-O157 STEC strains had a lower stationary-phase RpoS level (RpoS/DnaK ratio < 0·703) than the O157:H7 control, including seven strains within a group with the greatest survival (Table 3 and Fig. S2). Strain 103C was the only strain that had a significantly greater survival in LB-HCl pH 3·0 and a greater stationary-phase RpoS level than the O157:H7 control strain. Strains 26B, 26C and 111B had significantly greater survival in LB-HCl (pH 3·0; Fig. 3) but lower stationary-phase levels of RpoS than the O157:H7 control strain.
Shiga toxin-producing E. coli are a diverse group of potential human pathogens that can be disseminated by a wide variety of foods. STEC strains comprised of the six serovars with the greatest impact on human health were compared to define the limits and variability of growth and survival in low-pH and high-NaCl (osmotic and desiccation stress) conditions. Data were fitted to growth curve models to allow for quantitative strain-to-strain comparison across various growth conditions in relation to E. coli O157:H7. Adjustments to the modified Gompertz equation to account for a death phase conferred accurate model fits to growth curve data and the subsequent extraction of coefficients for lag phase duration (λ), maximum growth rate (μ1) and maximum growth density (a1). Removing the death phase data points would not have allowed direct comparison of coefficients between strains and media conditions.
The 18 non-O157 STEC strains examined in this study varied in their growth and survival properties. Even in unmodified LB, the growth rates of four strains (26A, 121A, 121B and 121C) and the maximum cell densities in five strains (45B, 45C, 26A, 145B and 121C) were significantly lower than the E. coli O157:H7 control strain. Seventeen of the 18 strains studied grew in LB-lactate pH 4·5 and LB-HCl pH 4·0 which is similar to previous findings (Glass et al. 1992; Conner and Kotrola 1995; Presser et al. 1998). Strains 45A and 121A were unique among the non-O157 STEC strains in their ability to both grow more rapidly in and achieve greater maximum cell densities than the O157 control strain in two or more of the acidified LB test conditions. In contrast to their growth properties, strains 45A and 121A had significantly reduced survival in LB-HCl pH 3·0 in comparison with the control strain.
In agreement with results with E. coli O157:H7 (Clavero and Beuchat 1996; Presser et al. 1998), none of the strains tested grew in LB-NaCl 9% (aw = 0·95). In contrast to growth in acidified LB, a majority of the non-O157 strains had significantly greater growth rates in LB-NaCl 5% (14/19, 74% of strains) and LB-NaCl 7% (10/19, 53% of strains). Strains 103A in LB-NaCl 5% and 45B in LB-NaCl 7% also obtained a greater maximum growth density than the O157:H7 control strain. The ability of some of the non-O157 STEC to grow more rapidly in LB-NaCl 5% and LB-NaCl 7% than the control strain did not translate to better survival in LB-NaCl 12%; none of the strains survived better than the O157 control strain.
Foodborne pathogens have varied dissemination strategies that include growth-based or survival-based distribution depending upon the pathogen and specific conditions. Growth in suboptimal conditions would yield greater numbers so that even with poor survival some cells are likely to persist. Alternatively, robust survival properties would allow survival under adverse conditions even with little or no growth in suboptimal conditions. For example, strains 45A and 121A exhibited greater growth rates in LB-lactate and achieved a greater maximum growth in LB-HCl; however, both strains had diminished survival when compared to the majority of the STEC strains in LB-HCl pH 3·0 and LB-NaCl 12%. It is possible that these strains employ growth-based (or numbers) dissemination. Of the four strains (103C, 111B, 26B and 26C) that had greater survival than the O157:H7 control strain in LB-HCl pH 3·0, none had a significantly greater growth rate or maximum growth density in acidified LB (HCl or lactate). As an example, strain 111B had a significantly lower growth rate in all five acidified LB conditions and a lower maximum growth in four of the five low-pH conditions than the O157:H7 control strain. Thus, it appears that strain 111B employs a robust survival for dissemination. The results from these non-O157 STEC indicate that both growth-based and survival-based dissemination strategies are used.
To examine possible relationships between growth, survival and the level of RpoS, the 20 STEC strains and an E. coli K12 strain were tested for RpoS by Western blot analysis of cell extracts from cultures in exponential, transition and stationary phases. Strains were grown at 30°C because this temperature results in greater RpoS levels than at 37°C and greater stress tolerance (Robbe-Saule et al. 2008). There is little RpoS in exponential-phase, nonpathogenic E. coli (Lange and Hengge-Aronis 1994; Loewen et al. 1998). Accordingly, exponential-phase RpoS levels were the lowest, but there was a difference between most STEC strains and E. coli K12. STEC exponential-phase RpoS levels, with the exception of strains 26A, 121A and 145A, were substantial and ranged from 30 to 89% of the maximum level achieved, which occurred in either transition or stationary phases (Table 3). These findings support those of Dong and Schellhorn (2009) that reported RpoS regulation of genes in exponential-phase E. coli O157:H7.
As RpoS is thought to coordinate the cellular response to environmental stress including entrance into stationary phase, an association between growth and RpoS level was not expected. This was the case when growth parameters were compared with exponential-phase and stationary-phase RpoS levels. The one exception was in LB-NaCl 5% where maximum growth density and lag phase duration were associated with exponential-phase RpoS levels (Fig. 4b,c). The lack of association between RpoS level and growth of STEC was supported by results with FRIK2787 (ΔrpoS) that had growth similar to the parent O157:H7 strain in acidified and unmodified LB but had a significantly greater growth rate in LB-NaCl.
Overall, exponential- and stationary-phase RpoS levels were not useful predictors of growth; however, there was an association between exponential-phase RpoS level and survival. The contribution of the general stress regulon to stress tolerance and survival is well established (Hengge-Aronis 2000); therefore, an association between RpoS and survival was expected. Considering that maximum levels of RpoS are found in late exponential and stationary phases and the degree of stress tolerance in stationary-phase cells (Loewen et al. 1998; Hengge-Aronis 2000), it was thought that survival would be associated with the RpoS level in stationary-phase rather than exponential-phase cells. These findings indicate that RpoS has a role in both stationary-phase and replicating cells. As expected and previously reported, the survival of FRIK2787 was significantly impaired (Cheville et al. 1996). In future experiments, the role of exponential-phase RpoS levels in the growth and survival of STEC will be examined.
Considering the diversity among the E. coli strains used in this study and findings from previous studies (Miller and Kaspar 1994; Presser et al. 1998; Duffy et al. 2000), it is not surprising that some STEC strains grew faster or survived better in the conditions tested than the O157:H7 control strain. However, none of the non-O157 STEC tested had both significantly greater growth and significantly greater survival properties than the O157:H7 control strain. This raises the question as to whether growth or survival is more important to the dissemination of STEC and foodborne pathogens. This likely depends upon the pathogen, vehicle and modes of dissemination employed by an organism. Data from this study will be useful in defining product formulations and processing interventions (i.e. acid washes and dry-ageing of carcasses) to control STEC. Future studies will further explore the significance of exponential-phase RpoS in STEC.
This research was supported by USDA National Integrated Food Safety Initiative grant #2009-5111-20143. We thank Dr Nancy Strockbine (CDC) and Tim Monson (Wisconsin State Laboratory of Hygiene) for providing strains. We also appreciate the guidance offered by Dr. Kwang-Cheol Jeong during the construction of the rpoS deletion mutant, and the assistance of Jihun Kang, Matthew Bozile, Megan Shiroda, Emily Barker and J.T. McCrone.