Antibiotic Minimal Selective Concentrations and Fitness Costs during Biofilm and Planktonic Growth

In this study, we used the clinically relevant uropathogenic biofilm-forming E. coli strain CFT073 to assess the MSCs and fitness costs for different antibiotics and resistant mutants. The choice of antibiotics examined was based on both the high clinical relevance of the antibiotic (trimethoprim, nitrofurantoin, and fosfomycin) for the treatment of E. coli infections and the presence of previous data on MSCs for comparative purposes between studies, e.g., trimethoprim (16, 23) and streptomycin (12, 16, 24). The choice of resistance mechanisms was based mainly on clinical relevance in E. coli (trimethoprim [dfr gene] [32], nitrofurantoin [nfsAB mutations] [33], and fosfomycin [uhpT mutation] [34]) as well as other bacteria (streptomycin [two different rpsL mutations] [35] and rifampicin [rpoB mutation] [36]).

We performed competition experiments between susceptible and resistant bacteria genetically tagged with different fluorescent markers at subinhibitory antibiotic concentrations in rich growth media (see Fig. 1A to D for a schematic outline of the experiment) (12, 16). During biofilm growth (Fig. 1A), ratios were obtained after 8 generations of growth on FlexiPegs, an in-house-modified Calgary device for biofilm growth (37). The mixes of susceptible and resistant bacteria were allowed to attach to the FlexiPegs for 3 h in the absence of antibiotics, and antibiotics were then added. After an additional 9 h of incubation with several medium changes (to reduce any potential contribution of planktonic free cells to biofilm growth), the bacteria were removed from the FlexiPegs by vortexing, and the ratio of susceptible to resistant bacteria was measured by flow cytometry (see Fig. S1 in the supplemental material). Control experiments showed that the planktonic cells present in the growth medium in the microwells where the FlexiPegs were incubated had a minimal impact on biofilm growth (see Fig. S2 in the supplemental material). Thus, when the growth of the biofilm with the continuous presence of planktonic cells in the microwells was compared with that after several growth medium changes 3, 4, 5, 6, 7, 8, 9, and 10 h after inoculation (Fig. 1A), the numbers of CFU per FlexiPeg were similar at 3, 6, and 12 h (see Fig. S2 in the supplemental material), indicating that there was little to no reattachment of planktonic bacteria to the biofilm during this time period. For planktonic growth, ratios of susceptible to resistant bacteria (see Fig. S3 in the supplemental material) were measured over 30 generations of growth (Fig. 1B) by flow cytometry, as previously described (12, 16).

For both planktonic and biofilm growth, the obtained ratios of resistant to susceptible bacteria during growth in the presence of different antibiotic concentrations were then plotted over time to obtain the selection coefficients, i.e., the slope of the curves (Fig. 1C). These calculated selection coefficients were then plotted against the antibiotic concentrations to estimate the minimal selective concentration (intercept on the x axis) and the fitness cost (intercept on the y axis) (Fig. 1D) (12). All competition experiments were performed with fluorescent dye swaps between the resistant and susceptible bacteria to correct for any potential effect of the fluorescent markers on growth and competitive ability. In total, 16 and 6 independent competition experiments were performed for planktonic growth and biofilm conditions, respectively. No measurable difference in fitness cost was observed for the two different fluorescent markers SYFP2 and dTomato during biofilm or planktonic growth (see Fig. S4 in the supplemental material).

The fitness cost determinations are based on measuring the bacterial growth reduction displayed by the resistant mutants in the absence of antibiotics. As described above, the fitness cost is obtained from competitions between susceptible and resistant strains at different antibiotic concentrations and a plot of the selection coefficients as a function of antibiotic concentrations where the intercept on the y axis (antibiotic concentration = 0) defines the fitness cost (Fig. 2). Depending on the resistance mechanisms and growth conditions, the fitness cost for planktonic growth varied from 1% for the trimethoprim (bgl::dfr)-resistant mutant to 18% for the rifampicin (rpoB S531L)-resistant mutant, and that during biofilm growth fitness cost varied from 2.8% for the fosfomycin (uhpT STOP 5 aa (the fifth amino acid was changed to a stop codon, TTA))-resistant mutant up to 37.4% for the rifampicin (rpoB S531L)-resistant mutant (Table 1). When comparing the six different resistant mutants, the fitness costs were slightly higher for biofilm than for planktonic growth for three of the mutants (Fig. 2B to D and Table 1). Thus, for rpsL K42R (streptomycin resistance), rpoB S531L (rifampicin resistance), and bgl::dfr (trimethoprim resistance), the fitness cost was 2- to 4-fold higher in biofilms. However, the increase in the fitness cost for growth in the biofilm was highly significant (P < 0.0001 by a two-tailed t test) for streptomycin (rpsL K42R) and rifampicin (rpoB S531L) resistance. For rpsL K42N (streptomycin resistance) and fosfomycin (uhpT STOP 5 aa), the costs were similar for both biofilm and planktonic growth (Fig. 2B and E and Table 1). For the nitrofurantoin (ΔnfsAB)-resistant mutant, no fitness cost could be estimated during biofilm growth, whereas during planktonic growth, it was 1.4% (Fig. 2F and Table 1). In conclusion, the fitness costs were similar (less than 2-fold in most cases) when comparing growth in biofilms and planktonic culture.

To better understand the impact of subinhibitory antibiotic concentrations on the selection of antibiotic resistance, the minimal selective concentration (MSC) was determined during both biofilm and planktonic growth. As described above, the MSC is obtained from competitions between susceptible and resistant strains at different antibiotic concentrations and from a plot of the selection coefficients as a function of the antibiotic concentration where the intercept on the x axis defines the MSC (Fig. 2). The difference in MSCs for the streptomycin-resistant mutants depended on the amino acid substituted in the rpsL gene, with a 2-fold increase (P = 0.05 by a two-tailed t test; t = 4.747; df = 5.116) in the MSC during biofilm versus planktonic growth for rpsL K42N and a 7-fold increase (P = 0.0003 by a two-tailed t test; t = 8.596; df = 5.192) for rpsL K42R (Fig. 2A and B and Table 1). For fosfomycin (uhpT STOP 5 aa) the increase in the MSC was 3-fold (P = 0.0085 by a two-tailed t test; t = 5.080; df = 3.739) for biofilm compared to planktonic growth, and for the rifampicin (rpoB S531L)- and trimethoprim (bgl::dfr)-resistant mutants, the MSCs were similar or the same independent of growth (Fig. 2C to E and Table 1), whereas for nitrofurantoin (ΔnfsAB), the MSC values could not be determined in the biofilm since these resistances showed no measurable fitness cost, and therefore, no intercept on the x axis was obtained (Fig. 2F and Table 1). To allow a comparison of how much lower the MSCs were than the MICs, we determined the MIC values during planktonic growth by broth microdilution and during biofilm growth by minimal biofilm inhibitory concentrations (MBICs) after 24 h of biofilm formation (Fig. 3). As expected, the biofilm MBIC values were significantly higher for all resistant mutants than MIC for planktonic growth, varying between a 2-fold-higher value for the streptomycin-resistant mutants to a >512-fold-higher value for trimethoprim (Table 2). The large difference between MBIC and MIC values leads to a drastic difference between MBIC/MSC and MIC/MSC ratios for rifampicin (>256 versus 22) and trimethoprim (50 versus >25,000) (Table 1). In conclusion, these results show that for all antibiotics, the resistant mutants were selected at subinhibitory concentrations during planktonic as well as biofilm growth.

All strains used in this study are listed in Table S1 in the supplemental material. These strains are all derived from Escherichia coli CFT073 (DA47111), a uropathogenic E. coli (UPEC) strain previously isolated from urine and blood samples of a patient with acute pyelonephritis (38). The use of chromosomal copies of an orange (dTomato) or a yellow (SYFP2) fluorescent protein (50) allowed measurements of cell-to-cell ratios during the competition experiments; therefore, all competing antibiotic-resistant strains were constructed in two isogenic strains, DA56709 (SYFP2) and DA56711 (dTomato). All constructed strains were whole-genome sequenced to confirm the absence of any additional unwanted mutations. Brain heart infusion (BHI) broth (Oxoid Limited, UK) or lysogeny broth (LB) with no salt was used for liquid cultures, and LB agar (LA) (Sigma-Aldrich, USA) was used for growth on plates. Strains were grown at 37°C unless otherwise noted, with planktonic growth in plastic tubes with shaking (180 rpm) and static biofilm growth in plastic 96-well plates with lids.

All mutant strains were constructed with the λ red recombineering system using the pSIM5-cam (chloramphenicol) vector (DA50218). The strains were grown overnight with 12.5 mg/L chloramphenicol at 30°C, diluted 1:200 in no-salt LB complemented with 12.5 mg/L chloramphenicol, and grown with shaking (150 rpm) at 30°C to an optical density at 600 nm (OD₆₀₀) of 0.20. When the cultures reached the target optical density, the cultures were transferred to a shaking 42°C water bath, inducing the expression of the temperature-controlled λ red genes. After 30 min, the cultures were placed on ice and washed three times with 10% glycerol. After the final centrifugation step, the cell pellets were resuspended in glycerol and mixed with DNA in Eppendorf tubes. Electroporation was performed in 50 μL of cell-DNA mix in 1-mm-gap electroporation cuvettes with a GenePulser Xcell system (Bio-Rad) at 1.8 kV, 2 μF, and 200 Ω. The transformants were recovered in no-salt LB with 12.5 mg/L chloramphenicol overnight at 30°C and spread onto plates containing the relevant antibiotics. All primers used are listed in Table S2 in the supplemental material.

To grow the biofilm, an in-house-developed biofilm growth system, FlexiPeg, was used (37). The FlexiPegs were printed using high-temperature resin (High Temp; Formlabs) at U-PRINT, Uppsala University’s three-dimensional (3D) printing facility at the Disciplinary Domain of Medicine and Pharmacy, using Formlabs form 3 (low-force stereolithography) 3D printers. The equipment is designed as a lid with removable pegs that fit over a 96-well flat-bottom plate (Thermo Scientific). By inoculating the wells, a biofilm can form on the FlexiPegs, where the lid allows the pegs to be moved between plates for cycling and washing purposes. Inoculated FlexiPegs were grown statically at 37°C in plastic containers with lids.

For biofilm extraction, the FlexiPegs were first washed by immersing them in 250 μL 1× phosphate-buffered saline (PBS) three times for 1 min each, with PBS replaced between dips. The design of the lid allows the FlexiPegs to be easily moved between the inoculated 96-well plate and a new plate with wells filled with PBS. When washed, the lid is placed into a holder above a rack specially designed to fit 24 glass tubes. The holder is positioned so that the FlexiPegs can be pushed through the lid from above, down into the glass tubes filled with 600 μL 1× PBS. To harvest the biofilm from the FlexiPegs, the tubes were vortexed at high speed for 2 min.

The introduction of either a yellow (SYFP2 [CH2037]) or an orange (dTomato [CH6016]) fluorescent gene in the otherwise isogenic resistant mutant strains and the corresponding susceptible wild type allowed ratio determinations during the competition assays (Fig. 1). Cultures of the wild-type strain grown overnight in BHI medium were mixed 1:1 (unless otherwise noted) with a resistant mutant carrying the other of the two fluorescent markers and diluted 10,000-fold, and 200 μL was then transferred to each well, with a final concentration of 2 × 10⁴ to 6 × 10⁴ cells/well. The biofilm was then allowed to establish attachment on the FlexiPegs for 3 h at 37°C without the antibiotics, followed by medium changes 4, 5, 6, 7, 8, 9, and 10 h after the inoculum with added antibiotics until a total of 12 h of growth from inoculation. Between each plate shift, the FlexiPegs were submerged in 250 μL fresh BHI broth. For each antibiotic concentration, competition was performed with three unique biological replicates, with a dye swap, resulting in six replicates per data point. All FlexiPegs were then harvested as described above, except for one set of three FlexiPegs that was harvested after 3 h, before the onset of selection pressure. For the harvested biofilms, the ratios between the strains were determined using the MACSQuant VYB device (Miltenyi Biotec), counting 10⁵ events per sample. The samples were prepared for analysis by diluting 20 μL of the culture in 200 μL of 1× PBS in 96-well plates (Thermo Scientific).

The competition experiments during planktonic growth were performed using the fluorescent markers in the same way as described above, mixing a susceptible wild-type strain with a resistant mutant carrying different fluorescent tags. Cultures grown overnight in BHI broth were mixed 1:1 (unless otherwise noted) and diluted 1,000-fold in 1 mL medium containing a range of antibiotic concentrations below the MIC. For each antibiotic concentration, competition was done with 8 biological replicates, with a dye swap, resulting in 16 replicates per data point and experiment. The cultures were then grown under shaking conditions for 24 h at 37°C for 10 generations of growth. One microliter of each culture was then added to fresh BHI broth (1 mL) with the selected antibiotic concentration and grown for an additional 24 h. Cycling was performed two times, resulting in a total of 30 generations of growth.

After every 24 h of growth, the ratios between the competing strains were determined for each antibiotic concentration in all replicates, providing 16 independent ratio measurements per antibiotic concentration at 10, 20, and 30 generations. From all cultures, 2 μL was mixed with 200 μL 1× PBS in 96-well plates (Thermo Scientific) and measured using the MACSQuant VYB device (Miltenyi Biotec), counting 10⁵ events per sample.

To calculate the fitness cost and MSC values for the antibiotics, the selection coefficients were determined using the regression model s = ln[R_(t)/R₍₀₎]/(t), where R is the ratio of the resistant mutant to the susceptible wild type obtained by MACSQuant analysis. By plotting the s values as a function of the antibiotic concentration, the intercept on the y axis represents the fitness cost, and the intercept on the x axis equals the MSC value (12). The s values as a function of the antibiotic concentration for each biological replicate, including a dye swap, were plotted separately (16 biological replicates for planktonic growth and 6 for biofilm growth, with the exception of fosfomycin, which had only 4 biological replicates). The fitness costs and MSCs were calculated for each biological replicate with the standard errors of the means for each antibiotic and growth (biofilm and planktonic growth). For MSC calculations, biological replicates missing an MSC value due to the lack of a fitness cost were excluded from the calculation of the average MSC value and the error of the mean (in the trimethoprim biofilm experiment, one out of six replicates in biofilms showed no fitness cost and was excluded). The P value was calculated by a two-tailed t test with Welch’s corrections.

In biofilms, the minimal biofilm inhibitory concentration (MBIC) can be used to measure bacterial tolerance toward antibiotics. The MBIC measures the impact of antibiotic treatment on a preformed biofilm, in our case after 24 h. For the determination of MBICs, cultures of the susceptible wild type grown overnight were diluted, and 200 μL/well was inoculated with 6 × 10⁴ cells/well. The MBIC was determined after 24 h, meaning that the biofilms were allowed to establish for 24 h without selection pressure, after which antibiotics were added. As for the competition experiments, a range of antibiotic concentrations was tested. After 24 h, the FlexiPegs were moved to medium containing the antibiotic, grown for an additional 24 h, and then harvested. After dilution and plating of the harvested biofilm, the CFU per FlexiPeg were determined and plotted against the antibiotic concentration. The MBIC was determined as the antibiotic concentration where fewer than 200 CFU were detected in the biofilm extracted from each FlexiPeg.

MIC assays were performed by broth microdilution in round-bottom 96-well plates (Thermo Scientific) according to EUCAST guidelines. A few colonies were diluted in 1 mL 0.9% NaCl to a final OD₆₀₀ of a 0.5 McFarland standard. One hundred microliters of the bacterial suspension was diluted in 10 mL of medium, giving a concentration of 1 × 10⁸ CFU/mL. Fifty microliters of the suspension was added to wells prefilled with 50 μL of medium with decreasing concentrations of the relevant antibiotic. The final CFU per well were 0.5 × 10⁶ to 1 × 10⁶ CFU per mL. The MIC was determined after static growth at 37°C for 16 to 20 h by identifying the well with the lowest concentration of the antibiotic without visible growth. In the case of pinpointing or skipped wells, EUCAST guidelines were followed when reading the results.

PCR amplification of mutated regions was performed using DreamTaq green PCR master mix (2×) (Thermo Scientific) with primers binding upstream and downstream of the mutation (see Table S2 in the supplemental material). The GeneJet gel extraction kit (Thermo Scientific) was used for purification, and the product was sent to Eurofins Genomics Europe for sequencing.