Bacterial populations vary in their stress tolerance and population structure depending upon whether growth occurs in well-mixed or structured environments. We hypothesized that evolution in biofilms would generate greater genetic diversity than well-mixed environments and lead to different pathways of antibiotic resistance. We used experimental evolution and whole genome sequencing to test how the biofilm lifestyle influenced the rate, genetic mechanisms, and pleiotropic effects of resistance to ciprofloxacin in Acinetobacter baumannii populations. Both evolutionary dynamics and the identities of mutations differed between lifestyle. Planktonic populations experienced selective sweeps of mutations including the primary topoisomerase drug targets, whereas biofilm-adapted populations acquired mutations in regulators of efflux pumps. An overall trade-off between fitness and resistance level emerged, wherein biofilm-adapted clones were less resistant than planktonic but more fit in the absence of drug. However, biofilm populations developed collateral sensitivity to cephalosporins, demonstrating the clinical relevance of lifestyle on the evolution of resistance.

15Many opportunistic pathogens live in surface-attached communities called biofilms that 16 generate ecological structure and can increase stress tolerance. Theory suggests that 17 bacterial populations evolving in biofilms may harbor greater genetic diversity and 18 become resistant to antibiotics by different pathways than in well-mixed environments. 19We used experimental evolution and whole genome sequencing to test how the mode of 20 growth influences dynamics and mechanisms of antibiotic resistance in Acinetobacter 21 baumannii populations. Biofilm and planktonic populations were propagated under 22 conditions lacking antibiotics, under constant sub-inhibitory concentrations of 23 ciprofloxacin, or under steadily increasing concentrations of this drug. As predicted, both 24 the evolutionary dynamics and the identities of selected mutations differed between 25 treatments and lifestyle. Planktonic populations exposed to ciprofloxacin underwent 26 sequential selective sweeps of single mutations including the primary drug targets, gyrA 27 and parC. In contrast, biofilm-adapted populations diversified by multiple contending 28 mutations in regulators of efflux pumps. Mutants isolated from both lifestyles exhibited 29 a trade-off between fitness and resistance level, wherein biofilm-adapted clones were less 30 resistant but more fit in the absence of drug. Further, biofilm-adapted populations evolved 31 collateral sensitivity to cephalosporins whereas the planktonic populations displayed 32 cross-resistance with several classes of antibiotics. This study demonstrates that growth 33 in biofilms, arguably the predominant bacterial lifestyle, may substantially alter the 34 routes, dynamics, and consequences of the evolution of antibiotic resistance and is 35 therefore an important consideration when treating infections. 36

Different species exposed to a common stress may adapt by mutations in shared pathways or in unique systems, depending on how past environments have molded their genomes. Understanding how diverse bacterial pathogens evolve in response to an antimicrobial treatment is a pressing example of this problem, where discovery of molecular parallelism could lead to clinically useful predictions. Evolution experiments with pathogens in environments containing antibiotics, combined with periodic whole-population genome sequencing, can be used to identify many contending routes to antimicrobial resistance. We separately propagated two clinically relevant Gram-negative pathogens, Pseudomonas aeruginosa and Acinetobacter baumannii, in increasing concentrations of tobramycin in two different environments each: planktonic and biofilm. Independently of the pathogen, the populations adapted to tobramycin selection by parallel evolution of mutations in fusA1, encoding elongation factor G, and ptsP, encoding phosphoenolpyruvate phosphotransferase. As neither gene is a direct target of this aminoglycoside, mutations to either are unexpected and underreported causes of resistance. Additionally, both species acquired antibiotic resistance-associated mutations that were more prevalent in the biofilm lifestyle than in the planktonic lifestyle; these mutations were in electron transport chain components in A. baumannii and lipopolysaccharide biosynthesis enzymes in P. aeruginosa populations. Using existing databases, we discovered site-specific parallelism of fusA1 mutations that extends across bacterial phyla and clinical isolates. This study suggests that strong selective pressures, such as antibiotic treatment, may result in high levels of predictability in molecular targets of evolution, despite differences between organisms’ genetic backgrounds and environments. IMPORTANCE The rise of antimicrobial resistance is a leading medical threat, motivating efforts to forecast both its evolutionary dynamics and its genetic causes. Aminoglycosides are a major class of antibiotics that disrupt translation, but resistance may occur by a number of mechanisms. Here, we show the repeated evolution of resistance to the aminoglycoside tobramycin in both P. aeruginosa and A. baumannii via mutations in fusA1, encoding elongation factor G, and ptsP, encoding the nitrogen-specific phosphotransferase system. Laboratory evolution and whole-population genome sequencing were used to identify these targets, but mutations at identical amino acid positions were also found in published genomes of diverse bacterial species and clinical isolates. We also identified other resistance mechanisms associated with growth in biofilms that likely interfere with drug binding or uptake. Characterizing the evolution of multiple species in the presence of antibiotics can identify new, repeatable causes of resistance that may be predicted and counteracted by alternative treatment.

words)13 An important problem in evolution is identifying the genetic basis of how different species adapt 14 to similar environments. Understanding how various bacterial pathogens evolve in response to 15 antimicrobial treatment is a pressing example of this problem, where discovery of molecular 16 parallelism could lead to clinically useful predictions. Evolution experiments with pathogens in 17 environments containing antibiotics combined with periodic whole population genome 18 sequencing can be used to characterize the evolutionary dynamics of the pathways to 19 antimicrobial resistance. We separately propagated two clinically relevant Gram-negative 20 pathogens, Pseudomonas aeruginosa and Acinetobacter baumannii, in increasing 21 concentrations of tobramycin in two different environments each: planktonic and biofilm. 22Independent of the pathogen, populations adapted to tobramycin selection by parallel evolution 23 of mutations in fusA1, encoding elongation factor G, and ptsP, encoding phosphoenolpyruvate 24 phosphotransferase. As neither gene is a direct target of this aminoglycoside, both are relatively 25 novel and underreported causes of resistance. Additionally, both species acquired antibiotic-26 associated mutations that were more prevalent in the biofilm lifestyle than planktonic, in electron 27 transport chain components in A. baumannii and LPS biosynthesis enzymes in P. aeruginosa 28 populations. Using existing databases, we discovered both fusA1 and ptsP mutations to be 29 prevalent in antibiotic resistant clinical isolates. Additionally, we report site-specific parallelism of 30 fusA1 mutations that extend across several bacterial phyla. This study suggests that strong 31 selective pressures such as antibiotic treatment may result in high levels of predictability in 32 molecular targets of evolution despite differences between organisms' genetic background and 33 environment.The notion that evolution can be forecasted at the level of phenotype, gene, or even 36 amino acid is no longer a fantasy in the post-genomic era (Lässig et al., 2017). If we 37 acknowledge that most forecasting efforts rely on history to anticipate the future, the explosive 38 growth of whole-genome sequencing (WGS) sets the stage to resolve evolutionary phenomena 39 in action and suggest the next selected path. Among the best examples, bacterial populations 40 exposed to strong selection like antibiotics and analyzed by WGS are likely to identify gene 41 regions that produce resistance (Ahmed et al., 2018a; Cooper, 2018; Feng et al., 2016; Palmer 42 and Kishony, 2013). Repeated instances of the same antibiotic selection may enrich the same 43 types of mutations and ultimately enable some measure of predictability (Ibacache-Quiroga et 44 al., 2018; Wong et al., 2012). For instance, we can be confident that exposure of many bacteria 45 to high doses of fluoroquinolones like ciprofloxacin may select for substitutions in residues 83 or 46 87 of the drug target, DNA gyrase A (Fàbrega et al., 2009; Wong and Kassen, 2011).47 Furt...