Cross-feeding modulates the rate and mechanism of antibiotic resistance evolution in a model microbial community of Escherichia coli and Salmonella enterica

Abstract

With antibiotic resistance rates on the rise, it is critical to understand how microbial species interactions influence the evolution of resistance. In obligate mutualisms, the survival of any one species (regardless of its intrinsic resistance) is contingent on the resistance of its cross-feeding partners. This sets the community antibiotic sensitivity at that of the 'weakest link' species. In this study, we tested the hypothesis that weakest link dynamics in an obligate cross-feeding relationship would limit the extent and mechanisms of antibiotic resistance evolution. We experimentally evolved an obligate co-culture and monoculture controls along gradients of two different antibiotics. We measured the rate at which each treatment increased antibiotic resistance, and sequenced terminal populations to question whether mutations differed between mono- and co-cultures. In both rifampicin and ampicillin treatments, we observed that resistance evolved more slowly in obligate co-cultures of E. coli and S. enterica than in monocultures. While we observed similar mechanisms of resistance arising under rifampicin selection, under ampicillin selection different resistance mechanisms arose in co-cultures and monocultures. In particular, mutations in an essential cell division protein, ftsI, arose in S. enterica only in co-culture. A simple mathematical model demonstrated that reliance on a partner is sufficient to slow the rate of adaptation, and can change the distribution of adaptive mutations that are acquired. Our results demonstrate that cooperative metabolic interactions can be an important modulator of resistance evolution in microbial communities.

Fig 1. Schematic of experimental setup and expectations.

A. Monocultures of E. coli (blue) and S. enterica (yellow), as well as cross-feeding co-cultures (green), were distributed in six replicate populations along an antibiotic gradient in 96-well plates. The antibiotics tested included rifampicin and ampicillin, and the concentration of antibiotic increased twofold at each well. 96-well plates were incubated with shaking at 30°C for 48 hours, then cells were transferred to fresh medium and antibiotic in a new plate. The transfer regime was to transfer 1μL of culture to a fresh well containing the same concentration of antibiotic at which it had previously grown, and 1μL to a fresh well containing one concentration step higher antibiotic. At each passage, the OD600 of the plate was measured, as well as species-specific fluorescence (CFP for E. coli, YFP for S. enterica). B-C. Hypothesis on why time under selection may be sufficient to explain MIC differences between monocultures and co-cultures. B. In monoculture, each species is under selection at every time step, thus selecting for increasing resistance with each passage. C. In obligate co-culture, only the more antibiotic-sensitive species is under selection at a given time, and effective co-culture resistance requires an increase in MIC in both species. This leads to the slower rise in resistance in the co-culture.

Fig 2

Resistance evolves more slowly in co-culture-evolved populations vs. monoculture-evolved populations. Six replicate populations each of monocultures and co-cultures were evolved along a rifampicin gradient (A-B) or an ampicillin gradient (C-D); gradients were identical for monocultures and co-cultures. Population MICs for each species (E. coli A, C; S. enterica B, D) were measured each transfer and the resulting MICs plotted. Statistical analysis was performed using a mixed-effects model with a randomized slope for each replicate within a culture type. The fitted slopes for each treatment are indicated by the dashed lines. P-values are for the interaction term between passage and culture type. Error bars represent the standard deviation of MIC among the six replicates.

Fig 3

Resistance-associated mutations in rifampicin-resistant evolved populations A-B. Lists of mutations which arose in E. coli (A) or S. enterica (B) monocultures and co-cultures. The number in each box represents the number of independent replicates in which the putative resistance mutation was observed. Additional mutations that occurred in only one population, and details on the nature of these mutations, may be found in S1 Table. C-D. Rifampicin MICs for isolates with wild-type vs. mutant rpoB genes in E. coli (C) and S. enterica (D). Isolates were obtained from passage 20 populations by streaking onto selective medium and picking isolated colonies. Each data point represents the average MIC for three isolates obtained from a single population. For each species- culture type combination, there are six populations total, and the statistical comparisons represent MIC comparisons between populations with wild type vs. mutant alleles. Note that there are only five S. enterica monoculture populations due to possible contamination in one population. MIC₉₀ values for isolates were defined as the lowest concentration of antibiotic which decreased growth by greater than 90% by 48 hours at 30°C. Additionally, monoculture lines evolved more mutations than co-culture lines. Mutations in mdoG and mdoH were observed in both monocultures of E. coli and S. enterica under rifampicin selection, but at much lower frequencies in co-culture (S1 Table, S2 Fig). All but one of these mutations were insertions or deletions of 1-6bp causing frameshift mutations (S1 Table). Both mdoG and mdoH likely influence cell membrane permeability [31,32]. The mutations were not associated with any changes in monoculture or co-culture growth rates (S3 Fig). Taken together, the pattern of rifampicin resistance mutations suggests that populations were moving along the same evolutionary trajectory in monoculture and co-culture, but progressed further along that trajectory in monocultures.

Fig 4

Resistance- associated mutations in ampicillin-resistant evolved populations A-B. Lists of mutations which arose in E. coli (A) or S. enterica (B) monocultures and co-cultures. The number in each box represents the number of independent replicates in which the putative resistance mutation was observed. Additional mutations that occurred in only one population, and details on the nature of these mutations, may be found in S1 Table. C. Image of co-culture populations containing ftsI mutations on Petri plates with LB pH = 7 (left) and pH = 4.7 (right). Blue colonies are E. coli which metabolize X-gal to a blue color; white colonies are S. enterica. No white colonies were observed on LB at pH = 7. D. MICs of isolates from libraries containing ftsI mutations in pH = 4.7 medium. Isolates were obtained from passage 10 populations by streaking onto selective medium and picking isolated colonies. Each data point represents the average MIC of three isolates from a single population. For each species- culture type combination, there are six populations total, and the statistical comparisons represent MIC comparisons between populations with wild type vs. mutant alleles MIC₉₀ values for isolates were defined as the lowest concentration of antibiotic which decreased growth by greater than 90% by 48 hours at 30°C. Each point represents the average MIC of three isolates from a single population. P = 0.0603, Mann-Whitney U.

Fig 5. Simulation model of the evolution of antibiotic resistance in single-species vs. multi-species obligately dependent communities.

A. A simulated evolution experiment with 35 replicate populations shows that increasing the number of interdependent species in a consortium results in a slower increase in the average MIC over time. Error bars represent the standard deviation among replicates. B. Effect of lower mutation rate on resistance evolution in model. Lower mutation rates result in bigger differences in the relative tolerance of one-species versus two species systems. C. Distributions from which MIC-altering mutations were randomly pulled. The x-axis is relative to the maximum MIC change possible in one mutation. The “null” model used a uniform distribution. The “neg exp” used a truncated negative exponential distribution with rate parameter 2.3, which kept 90% of random numbers less than the max MIC change of 1.0. D. Mean MIC increase conferred per mutation as a function of the mutation distribution and the number of species.