Environmental and genetic determinants of plasmid mobility in pathogenic Escherichia coli

Integrating genomics with growth-based phenotyping reveals divergent roles of antibiotics and incompatibility in plasmid mobility.


This PDF file includes:
Fig. S1. Conditions during the conjugation period effectively decouple conjugation from growth dynamics. Fig. S2. The log-linear correlation between T 0 and τ is maintained with respect to strain and choice of OD threshold.  Table S1. All strains used in this study. Table S2. Plasmid composition of pathogenic E. coli isolates. Table S3. MICs and IC 50 s for plasmid donors in this study. Model development and assumptions Reference (49) Fig. S2. The log-linear correlation between T 0 and τ is maintained with respect to strain and choice of OD threshold. (a) With a constant OD 600 threshold of 0.03, the standard curves for ESBL193, ESBL168, GN02545, GN02766, and GN05696 display only minor variations in the relationship between 0 and τ. For example, data and curve for ESBL193 are shown in red, with other strains' data alternatively colored. Serially diluted transconjugants were grown under conditions recreating post-conjugation outgrowth to establish the relationship between 0 and τ. Error bars represent +/-SD from triplicate measurements. (b) Standard curves are for ESBL193 with changing OD 600 thresholds of 0.01, 0.03, 0.05, 0.1, 0.2, and 0.3 (bottom to top). Serially diluted transconjugants were grown under conditions recreating post-conjugation outgrowth to establish the relationship between 0 and τ. As OD 600 threshold increases, τ also increases and shifts the curve. The relationship between 0 and τ degrades as cell growth slows near the end of exponential phase. Error bars represent +/-SD from triplicate measurements. Average values from triplicate OD measurements are displayed +/-SE and fit with a two-parameter log-logistic model to estimate IC 50 . Five antibiotics were tested at concentrations ranging from 0 to 64 µg/mL. The results from carbenicillin are not displayed as every donor displayed the same high-level beta-lactam resistance. (b) Strength of antibiotic resistance does not correlate with the effect of antibiotics on τ. IC 50 s determined for pathogen donors in Figure 3 show no correlation with τ normalized to no antibiotic controls (Δτ) across four antibiotics of varied mechanism of action: chloramphenicol (Cm), erythromycin (Ery), kanamycin (Kan), and norfloxacin (Nor). Donor-recipient strain interactions for GN02766. GN02766 and DA28102 displayed no significant changes in population density while mixed over two hours under no growth conjugation conditions, relative to separated cultures. This suggests the changes in 0 are not from donor-recipient interactions that cause growth or death. Error bars represent +/-SE from triplicate measurements. (c) Promotion of conjugation in GN02766 is specific to macrolide and chloramphenicol mechanisms of action and not solvent dependent. The effect of two additional bacteriostatic antibiotics on conjugation in GN02766 was assessed via the time to threshold method. Antibiotics were dosed as before ( Figure 3), with 1x concentrations for azithromycin (Az) and sulfamethoxazole (Sm) being 2 μg/mL and 16 μg/mL, respectively. Displayed Δτ are averages of triplicate measurements +/-SE, and normalized by subtracting the no antibiotic control τ. Promotion of conjugation is indicated by Δτ<0, while Δτ>0 indicates inhibition. Both macrolides, Ery and Az, displayed nearly identical conjugation promotion (p < 0.0001, Tukey HSD). Cm, which shares a similar mechanism of action with macrolides, repeated its promotive trend but returned insignificant for this experiment. Sm had no significant effect, indicating conjugation promotion is mechanism-specific, not bacteriostatic-specific. No antibiotic was dosed with EtOH solvent to control for solvent effects.

Fig. S5. Identification of plasmid features lost in strain GN02766
. Plasmid sequences were aligned to p168-1 via BLASTn and BRIG as the most closely related plasmid to p2766-1 from a no antibiotic modulation strain. Nucleotide identity ≥70% is indicated by a band colored according to Inc group, with darker shading corresponding to higher sequence match. Blank regions indicate <70% nucleotide identity. Inner GC plots, size map, and outer coding sequences (CDS) are for p168-1. The outermost IncFIB/FII/Col156 plasmid corresponds to p2766-1. Among plasmids displaying no antibiotic modulation of conjugation, no major commonalities among the majority were lost in p2766-1. Macrolide and chloramphenicol promotion of conjugation may then be acquired instead.

Fig. S6
. Sequencing read coverage plot for p2766-1. The 5x tnpA repeats spanning the ~1-25kb region of p2766-1 could be assembly error due to poor read coverage. However, we see no indication of poor coverage across the plasmid, which suggests that the tnpA repeats are genuine. The increase in coverage seen from ~90-100kb is due to a concentration of short reads and not collapsed repeats. This coverage plot was generated through the SMRTLink HGAP4 pipeline with parameters: minMatch = 12; bestn = 10; minPctIdentity = 70.0.

Fig. S7. Macrolide promotion of conjugation is transferrable. (a)
To measure plasmid-specific effects, 2766T, the DA28102 transconjugant produced by GN02766, was subsequently used as a donor with the fAYC002 recipient. Antibiotics were dosed as before during conjugation (Figure 3). Following conjugation, mixtures were diluted 50x instead of 150x to allow for low conjugation efficiency. Carb100 and Kan50 were used for selective outgrowth for fAYC002 transconjugants. Displayed Δτ are averages of triplicate measurements +/-SE, and normalized by subtracting the no antibiotic control τ. Promotion of conjugation is indicated by Δτ<0, while Δτ>0 indicates inhibition. Ery once again displayed significant conjugation promotion at both 0.5x and 1x IC 50 concentrations (p < 0.0001 and p < 0.05 respectively, Tukey HSD). (b) Macrolide promotion is specific to 2766T, not DA28102 transconjugants in general. In line with ESBL94, 94T also displayed no antibiotic modulation of conjugation when used as a donor with fAYC002 recipients. Carb100 and Kan50 were used for selective outgrowth for fAYC002 transconjugants. Displayed Δτ are averages of triplicate measurements +/-SE, and normalized by subtracting the no antibiotic control τ. Promotion of conjugation is indicated by Δτ<0, while Δτ>0 indicates inhibition.

Model development and assumptions
Our model of transconjugant outgrowth consists of three ordinary differential equations that account for growth and conjugation dynamics of the donor ( , 1), recipient ( , 2), and transconjugant ( , 3) populations. There are several parameters, including: , death rates for each population, , transconjugant growth rate, , culture carrying capacity, and η, rates for -and -conjugation.
In describing the dynamics of these three cell populations, we make several simplifying assumptions. The first being that, experimentally, we use antibiotic concentrations so high that degradation has a negligible effect on , , and . Furthermore, parental cell growth rates are insignificant relative to death rates under strong antibiotic selection. Finally, we assume that η does not change with the strain donating the plasmid (either parental donor or its transconjugant).

Parameter values
Parameters values and initial population sizes were chosen to mimic time to threshold experimental results. was found to be ~10 9 cells/mL, which was set to 1 for simplicity and served as the basis for other constants in the model. Both and were set to 0.5 to ensure significant parental death within a few hours, as found in preliminary tests. Doubling growth is expected during exponential phase, therefore = 1 with low relative = 10 -4 . Conjugation rate constants (e.g. η = 0.1) were based on experimental donor and recipient cell densities (~10 7 cells/mL) and the average conjugation efficiency (~10 -12 ) found for E. coli in preliminary tests. It then follows that: = 0.01, = 0.01 -, and was varied about 10 -5 for sensitivity analysis.