Strains used in this study are listed in Supplementary Table S1. Strains derived from MG1655 were grown in M9 minimal medium supplemented with glucose [0.2% (w/v)], liquid or agar plates [1.5% (w/v)]. Unless otherwise indicated, cultures were incubated at 37°C for 48 h. MG1655 ΔdnaQ was constructed according to Datsenko and Wanner (Datsenko and Wanner, 2000) and ΔdnaQ::Kan was amplified from strain JW0205 (Baba et al., 2006). When needed for strain selection and plasmid propagation, antibiotics were added at the following concentrations: ampicillin (AMP) at 100 μg/ml, chloramphenicol (CHL) at 50 μg/ml, and kanamycin (KAN) at 50 μg/ml. For cloning and propagation of pGE-derivatives, E. coli CC118 λ-pir was used (de Lorenzo and Timmis, 1994; Herring et al., 2003).

Plasmids used in this study are listed in Supplementary Table S2. Primers were synthetized by Sigma Genosys and are listed in Supplementary Table S3. Gene amplification was performed with Expand High Fidelity PCR System (Roche).

Construction of mutant strains of E. coli MG1655 was based on the expression of I-SceI endonuclease (Sniegowski et al., 1997). Selected genes were amplified from the evolved strain CIM8M using oligonucleotides listed in Supplementary Table S3 and cloned into pGE vector (KAN^R). Mutants construction was performed as described in Piñero-Lambea et al. (2015).

Complementation assays were carried out using vectors (Supplementary Table S2) from the ASKA collection for all selected strains (Kitagawa et al., 2005).

Prior to the evolutionary process, a wild-type E. coli MG1655 strain and its ΔdnaQ derivative were adapted to M9 minimal medium by serial passages into fresh media every 24 h during 7 days, equivalent to 56 generations, approximately. This previous adaptation was done to diminish the putative confounding effects of mutations not directly related with resistance to gentamicin. Bacterial evolution was performed in chemostats with M9 minimal medium with glucose as only carbon and energy source and increasing concentrations of gentamicin. For each evolutionary process, four chemostats containing 40 ml of culture medium supplemented with gentamicin at 0.03 μg/ml, equivalent to 0.125-fold their minimal inhibitory concentrations (MIC), were inoculated with 400 μl of an overnight culture of the ancestor. Chemostats were continuously fed with fresh culture medium with gentamicin at a dilution rate of 0.24 h⁻¹, and maintained at 37°C and aeration of 1.2 l/h. Gentamicin concentration in culture media was doubled every 72 h until it reached 256 μg/ml. Samples from every stage of the evolutionary process were taken and cryopreserved at −80°C, with glycerol at a final concentration of 20% (v/v). Samples from the last stage of the evolution process were plated on M9 minimal medium agar with glucose and gentamicin (256 μg/ml). After 72 h of incubation, 22 MG1655-derivatives and 38 MG1655 ΔdnaQ-derivatives were isolated and cryopreserved for further characterization. Selection of these colonies was based on colony size and morphology.

Antimicrobial susceptibility profiles were established by evaluating MICs of different antimicrobials in M9 minimal medium supplemented with glucose at 0.2% (w/v), through micro-broth dilution method. The evaluated compounds were amikacin (AMK), ampicillin (AMP), chloramphenicol (CHL), colistin (COL), ciprofloxacin (CIP), fosfomycin (FOS), gentamicin (GEN), kanamycin (KAN), tetracycline (TET), rifampicin (RIF), trimethoprim (TRM), and ethidium bromide (EtB).

For strains derived from wild-type MG1655, the mutant frequency was determined on plates of M9 minimal medium with glucose containing 40 μg/ml of nalidixic acid, while rifampicin at 50 μg/ml was used for ΔdnaQ-derivatives. For both ancestors, overnight cultures were plated on media with antibiotic. Viable cells were measured by plating 10³ cells from each overnight culture on M9 minimal medium supplemented with glucose and incubated under the same conditions. Mutant frequency was estimated by the ratio between colony-forming units (CFU) on rifampicin or nalidixic acid and total viable CFU. Frequency was evaluated in triplicates for each strain.

Bacterial susceptibility to hydrogen peroxide was measured on plates with M9 minimal medium agar with glucose, by measuring growth inhibition. Bacterial cells from overnight cultures were inoculated into top agar and exposed to 26 μmol of hydrogen peroxide. The halo of growth inhibition was measured in millimeters (mm) after 48 h. Susceptibility to hydrogen peroxide was evaluated in triplicates for each strain.

The ability of bacterial strains to use different organic compounds as sole energy and carbon source was evaluated using Biolog'S GN2 system, according to manufacturer's instructions. Biochemical characterization was performed using API 20E strips (Biomerieux), according to manufacturer'S instructions. Both systems were incubated at 37°C for 72 h.

Bacterial growth was evaluated in presence and absence of gentamicin (64 μg/ml). Ninety-six-well plates were filled with 200 μl of minimal medium and inoculated with 10³ cells per well. Growth curves were performed in a Tecan Infinite M200 spectrophotometer with incubation, at 37°C, with linear shaking, for 48 h. Growth rates were determined using GrowthRates program (Hall et al., 2014). Growth curves were evaluated in triplicates for each strain.

Whole genome sequencing of all selected strains and their ancestors was carried out using the pair-end protocol in an Illumina HiSeq2000 Sequencer. More than 1,000,000 sequences, between 35 - 82 bp, were obtained per sample, and their quality was analyzed through FASTAQC. Sequences were aligned with BWA (Li and Durbin, 2009), using “bwa aln” and “bwa sampe” commands. Genome sequence of MG1655 was obtained from Genebank Database (NCBI): https://www.ncbi.nlm.nih.gov/nuccore/U00096.3. Alignment files were transformed to BAM format, sorted and indexed through samtools (Li et al., 2009). snpEff (Cingolani et al., 2012) was used for the annotation of gene polymorphisms. For mutations in upstream and downstream regions, 200 bp were considered. Results were filtered and analyzed in an interactive manner using a visualizer developed by the Computational Genomic Service of the National Center for Biotechnology (CNB-CSIC) in Madrid, Spain. Polymorphisms were considered relevant when they reached a minimal quality of QC > 30 and a biological impact of “High” or “Moderate.” The sequences obtained in this study have been deposited in the NCBI-SRA database (https://www.ncbi.nlm.nih.gov/sra) under the accession number SRP128878.

To determine the evolutionary trajectories of the selected genes, samples from chemostats were taken at 0.5, 2.0, 4.0, and 8.0 μg/ml of gentamicin, and plated on M9 minimal medium agar with glucose. Two single colonies were selected from each chemostat. Genes of interest were amplified from the selected strains, sequenced, and analyzed in order to determine if gene mutations were present at the selected evolutionary steps.

Statistical analysis was carried out with StatPlus:mac LE (AnalystSoft Inc., www.analystsoft.com). For all analyses, the level of significance was set at 95% (p-value < 0.05).

To minimize the confounding effects of mutations related to medium adaptation, both MG1655 and its ΔdnaQ derivative were adapted to M9 minimal medium without antibiotic, during 56 generations prior to the evolutionary process. Evolution to gentamicin resistance was carried out in chemostats under increasing concentrations of this antibiotic. Four parallel cultures of MG1655 and MG1655 ΔdnaQ, with MIC of gentamicin of 0.25 μg/ml in both cases, were evolved until antibiotic concentration reached 256 μg/ml. In total, 60 E. coli strains resistant to gentamicin were isolated (Supplementary Table S4): 22 derived from MG1655 (denominated CIM) and 38 from MG1655 ΔdnaQ (denominated CIQ), all of them presenting gentamicin MIC ≥256 μg/ml. These strains presented different colony sizes and morphology, including the presence of small colony variants especially in MG1655-derivatives. Thus, the 60 E. coli strains selected in this study represent the different colony phenotypes observed in each sample. Clones were named according to the author's name (CI), their ancestor (M for MG1655-derivatives and Q for ΔdnaQ), the replicate from which they were isolated (N°) and the isolated colony (letter).

All isolates were characterized for their antibiotic susceptibility profiles (Supplementary Tables S5, S6). All strains decreased their sensibility to fosfomycin. Increased susceptibility to chloramphenicol was also observed in 68% of strains. Interestingly, only MG1655-derivatives presented an increased resistance to rifampicin. From all 60 strains, 4 MG1655-derivatives (CIM5H, CIM5N, CIM8C, CIM8M), and 4 ΔdnaQ-derivatives (CIQ1E, CIQ1G, CIQ2J, CIQ4J) were selected for further studies. These strains were selected accordingly to their antibiotic-resistance profile in order to represent the different phenotypes observed among all characterized strains from each replicate. All MG1655-derivatives showed an increase in gentamicin resistance (≥2,000-fold) as well as rifampicin and fosfomycin resistance (Tables 1, 2). In particular, CIM5H presented an increase in chloramphenicol (16-fold) and ethidium bromide (≥4-fold) susceptibility; CIM5N showed an increase in chloramphenicol (4-fold) and a decrease in trimethoprim susceptibility (4-fold); CIM8C presented a decrease in chloramphenicol susceptibility (4-fold); CIM8M presented a decrease in chloramphenicol susceptibility (≥32-fold) and an increase in trimethoprim susceptibility (4-fold). Regarding the MG1655 ΔdnaQ derivatives, all strains showed increased fosfomycin resistance and were selected according to the following parameters: CIQ1E was selected due to its increase in chloramphenicol and tetracycline susceptibility (8-fold); CIQ1G showed an increase in its susceptibility to chloramphenicol (16-fold), trimethoprim (8-fold) and rifampicin (≥4-fold), and a 16-fold increase in resistance to ethidium bromide; CIQ2J presented increased resistance to ampicillin (4-fold) and a decrease to chloramphenicol (4-fold); CIQ4J showed increased susceptibility to ciprofloxacin (2-fold) and rifampicin (≥4-fold).

Fitness cost of gentamicin resistance was observed for all evaluated strains according to growth curves, metabolic and enzymatic profiles. All isolates presented lower growth rates relative to their ancestor, showing a significant reduction of this parameter in all strains (p-value < 0.05, one-sample t-test) (Table 1). No significant differences in relative growth rates were observed between MG1655-derivatives and ΔdnaQ-derivatives (p-value < 0.05, two-samples t-test). All strains presented a reduced ability to use different organic compound as a sole energy and carbon source (Supplementary Table S7). Changes on their enzymatic activities were also observed (Supplementary Table S8). Interestingly, even though hypermutation has been associated with an increased deleterious burden (Giraud et al., 2001), dnaQ-derivatives did not displayed a particularly large reduction in growth rate or metabolic capacity compared to MG1655-derivatives with respect to their ancestors (77.7 vs. 72.8% reduction, p-value < 0.05, Fisher's exact test). On the contrary, even when mutations in electron transfer chain and exposure to aminoglycosides have been associated with the development of small colony variants (SCV) (McNamara and Proctor, 2000), only two (CIM8C and CIM8M) out of eight strains showed smaller colony sizes (data not shown). Of note, these strains were the only ones to maintain the normomutator phenotype after the evolutionary process, in line with previous suggestions that mutators could keep their evolutionary advantage through substituting secondary mutations that compensate costly phenotypes (Perron et al., 2010).

The effect of adaptation on the spontaneous mutant frequencies of MG1655 and MG1655 ΔdnaQ was analyzed. The MG1655 evolved strains CIM5H and CIM5N showed mutant frequencies two orders of magnitude higher than their ancestor MG1655 (Figure 1A). Genome sequence analysis of CIM5H and CIM5N revealed a H162Q substitution in dnaQ, affecting a functional residue, likely reducing its proofreading activity (Cisneros et al., 2009). Note that this change is unrelated with the ΔdnaQ derivative, ruling out the possibility of a cross-contamination from the MG1655 ΔdnaQ strain. On the other hand, the evolved ΔdnaQ derivative CIQ4J showed a frequency of spontaneous mutants one order of magnitude higher than that of its ancestor (Figure 1B). CIQ4J showed an I695S substitution in MutS, a key enzyme of the Mismatch Repair System (Junop et al., 2003). This is a semi-conserved residue located in a highly conserved region of the protein. The addition of a mutation on genes of the mismatch repair pathway (MMR) on a dnaQ background has shown to generate a slight increase in mutant frequency (Schaaper, 1993), thus this mutation on mutS may be involved in the activity of MutS and in the increase of mutant frequency in CIQ4J.

In order to identify the genetic modifications arisen during the evolutionary process related to gentamicin resistance, whole genome sequencing and bioinformatic analysis of all selected strains and their ancestors were performed. Agreeing with their ancestors' phenotype, ΔdnaQ-derived strains presented a higher number of mutations, being CIQ1G the strain with more genetic modifications (Table 3). After the evolutionary process, strains derived from MG1655 showed two different mutations profiles. While CIM8C and CIM8M retained the normomutator phenotype, showing 29 and 22 mutations, respectively, CIM5H and CIM5N developed a mutator phenotype, gaining 345 and 636 genetic modifications, respectively (Table 3).

Despite the effort in pre-adapting both backgrounds to the medium without antibiotic, it remains the possibility that some of the observed mutations were linked to adaptation to the environment. For that reason, we focused on mutational targets for which previous knowledge rendered reasonable to expect some involvement with gentamycin susceptibility. All eight evolved strains shared non-synonymous mutations on five common candidate genetic elements: fhuA and fusA genes and atpIBEFHAGDC, cyoABCDE, and potABCD operons (Table 4). In Table 5 we show that, indeed, most of these mutations are associated with increases in resistance. Isolate CIM8M was selected as model for further studies in order to evaluate the putative effect of the acquired mutations on gentamicin resistance, due to its high-level resistance to gentamicin (MIC of 512 μg/ml) and lower number of mutations on its genome. Beside its mutations in the five common genetic elements, strain CIM8M presented a SNP in csrA, affecting the fifth residue (T5I) and a H526Y substitution in RpoB, which is related to its resistance to rifampicin. CsrA is a small RNA-binding protein that acts as a global regulator in E. coli, playing a key role in central carbon metabolism (Wei et al., 2000). It is also involved in biofilm formation and flagellum biosynthesis (Wei et al., 2001; Jackson et al., 2002).

Related to the five common genetic elements, strain CIM8M presented mutations on atpG, cyoA, fhuA, fusA, and potA. In the case of atpG, cyoA, and fhuA, mutations generated frameshifts in residues 245, 68, and 197, respectively, while F593L and Q208L substitutions were identified in fusA and potA, respectively (Table 4).

fhuA encodes a ferrichrome transporter (Hantke and Braun, 1975) and its mutation in CIM8M, as well as in all evaluated strains, generated a frameshift on the amino acid sequence. In CIM8M the insertion of a nucleotide in position 197 generated a frameshift (Endriss et al., 2003). This mutation did not generate variations in the gentamicin susceptibility of the strain FhuA197, a MG1655-derivative with a nucleotide insertion at residue 197 of fhuA (constructed as indicated in Materials and Methods). Complementation of CIM8M with the wild-type fhuA gene reduced gentamicin MIC from 512 to 128 μg/ml (Table 5). No fitness cost was observed for FhuA197 (Table 2), nevertheless, the complementation indeed decreased the duplication time (Table 6), suggesting that in this genetic background it has a detrimental effect on bacterial fitness. Anyway, a positive effect of the expression of fhuA in multicopy on growth rate cannot be discarded.

fusA encodes the Elongation Factor G (EF-G) and mutations were mapped to domains IV or V of the protein (Table 4), which are essential for its activity (Savelsbergh et al., 2000). CIM8M complementation with wild-type fusA produced a slight but consistent decrease on the MIC of gentamicin from 512 to 256 μg/ml (Table 5), and a bacterial growth improvement in absence of gentamicin (Supplementary Figure S1). The substitution F593L in fusA (strain EFG593) generated an 8-fold decrease in bacterial susceptibility to gentamicin respect MG1655 (Table 5), without any fitness cost (Table 2). However, complementation with wild-type fusA increased growth rate in the CIM8M strain (Supplementary Figure S1), suggesting that even if FusA^F593L has no effect on fitness in a wild-type background, it is detrimental in the presence of other mutations.

We also identified alterations on electron transport chain, specifically on the atp and cyo operons. Even though electron transport chain has been mainly related to the aminoglycoside uptake due to the importance of membrane potential for this process (Damper and Epstein, 1981), it has been recently described that aminoglycosides accelerate respiration, potentiating their lethal effect on bacteria (Lobritz et al., 2015). CIM8M complementation with the wild-type atpG gene reduced gentamicin MIC (from 512 to 128 μg/ml) and duplication time in absence of gentamicin (Table 6). Complementation with the wild-type cyoA did not generate changes on its MIC of gentamicin, while a decrease in growth rate was observed (Table 6). Despite all efforts, it was not possible to generate single mutants on atpG and cyoA reproducing those on CIM8M. For that reason we resorted to use gene knockout strains from the KEIO collection (Baba et al., 2006), JW3711 (ΔatpG::Km) and JW0422 (ΔcyoA::Km), to evaluate the effect of these genotypes on gentamicin resistance. The cyoA deletion caused a 2-fold decrease in gentamicin susceptibility (Table 5). In contrast to previous studies where atpG deletion caused an increase in gentamicin susceptibility (Liu et al., 2010), we failed to observe any effect over resistance. On the other hand, complementation of CIM8M with the wild-type atpG gene reduced its resistance to gentamicin and decreased its duplication time (Tables 5, 6). Because aminoglycoside uptake is related to growth rate in absence of the antibiotic (Muir et al., 1984), this genetic modification could generate a decreased uptake. The difference in gentamicin susceptibility between atpG deletion in JW3711 and atpG complementation in CIM8M may be related to the characteristics of the mutation. It has been described that only certain genetic changes in atpG generate decreased aminoglycoside susceptibility (Mogre et al., 2014). While JW3711 has a complete deletion of the protein, CIM8M presents a frameshift in position 245.

The fifth mutation identified in all resistant strains affected the potABCD operon, encoding the spermidine-preferential uptake system PotABCD (Igarashi and Kashiwagi, 1999). Polyamines are involved in several cellular processes such as energy metabolism, oxidative stress tolerance, biofilm formation, and iron transport, as well as in gene expression of an oligopeptide transporter OppA (Tkachenko and Nesterova, 2003; Yoshida et al., 2004; Karatan et al., 2005; Patel et al., 2006; Tkachenko et al., 2012). PotA^Q208L identified in CIM8M affected the ATPase subunit of the transporter (Igarashi and Kashiwagi, 1996), generating a 4-fold increase in gentamicin MIC (Table 5), without fitness cost (Table 2). It is worth noting that both transporters, OppA and PotABCD, have been described to be able to uptake some aminoglycosides into bacterial cell (Holtje, 1978; Acosta et al., 2000), suggesting that PotA^Q208L may generate an altered gentamicin intake. In CIM8M, complementation with wild-type potA reduced gentamicin MIC 4-fold (Table 5). Also, we observed a reduction in the cellular growth in absence of this antibiotic (Supplementary Figure S1).

Besides gentamicin resistance, CIM8M presented increases in both kanamycin and amikacin MICs (Supplementary Table S9), indicating that its mechanism of resistance is shared with other aminoglycosides.

Evolutionary trajectories of the common resistance determinants were determined by analyzing the time of apparition of mutations along the experiment. For this purpose, two single colonies were randomly selected from chemostat populations at different gentamicin concentrations. Only clones from populations obtained at 0.5, 2.0, 4.0, and 8.0 μg/ml gentamicin were selected, as isolates from higher gentamicin concentrations showed high-level gentamicin resistance. For instance, gentamycin MIC of isolates from 8.0 μg/ml ranged from 16 to >128 μg/ml (data not shown). We chose five loci (atpG, cyoA, fhuA, fusA, and potA) as representative of the five common genetic elements that were repeatedly targeted by selection to be subjected to PCR amplification and Sanger sequencing. Results show a sequential and cumulative appearance of mutations in both backgrounds, MG1655 and its hyper-mutator derivative (Figure 2; Supplementary Table S10). The first locus to present alterations was fusA (observed first, in both backgrounds, at 0.5 μg/ml). Secondly, we found mutations in cyoA and potA, observed earlier in the hypermutator than in the normomutator (0.5 vs. 4 μg/ml). Next, we detected mutations in fhuA, also appearing earlier in the hypermutator than in the normomutator (2 vs. 8 μg/ml). The last mutations to be added mapped onto atpG, being present in both backgrounds from a concentration of 8 μg/ml onwards.

The effect of gentamicin over oxidative stress has been widely studied and discussed during the past years. While several studies have established that these antibiotics generate increased concentration of reactive oxygen species (ROS) as a consequence of antibiotic-stress and that this oxidative stress contributes to the bactericidal effect of aminoglycosides (Zhao and Drlica, 2014), other studies indicate that bactericidal effect of aminoglycosides is unrelated to oxidative stress (Keren et al., 2013; Liu and Imlay, 2013). On the other hand, Ezraty et al. (2013) established that iron concentration is related to gentamicin susceptibility in a Fenton-independent manner, where iron depletion reduces the activity of respiratory Complex I and therefore reduces proton motive force generation (essential for aminoglycoside uptake). Consequently, iron depletion reduces aminoglycosides entry and bacterial susceptibility to them (Ezraty et al., 2013). In order to evaluate if gentamicin resistance is related to a reduction in oxidative stress, the susceptibility of gentamicin-resistant strains to hydrogen peroxide was evaluated. Hydrogen peroxide, as well as increased intracellular concentration of iron, induces oxidative stress due to Fenton's reaction, where ferrous iron is reduced by superoxide, generating hydroxyl radicals. Therefore, an increase in hydrogen peroxide or iron concentration may lead to oxidative stress. Unexpectedly, all resistant strains were more susceptible to hydrogen peroxide than their ancestors (p-value < 0.05, two-samples t-test) (Figure 3), suggesting that their mechanisms of resistance are not related to enhanced tolerance to ROS. Differences in the effect of aminoglycoside over oxidative stress may be related to differences in experimental approaches and cell growth status. Wang et al. recently established that several genes involved in oxidative stress protection are under RpoS regulation in late exponential and early stationary phase of growth, showing that aminoglycosides induce oxidative stress in this phase of growth and that RpoS is directly related to oxidative stress protection (Wang et al., 2014). Because these strains were evolved in continuous cultures, maintaining the exponential phase of growth, the effect of gentamicin on oxidative stress may have not been significant in this condition and, therefore, the mechanism of resistance of these strains does not involve decreased susceptibility to ROS.

In order to evaluate if the evolution to gentamicin resistance altered the susceptibility to other antibiotics, the MICs of nine antimicrobial compounds were determined (Supplementary Tables S5, S6). All selected strains developed fosfomycin resistance, but whole genome sequencing revealed that they did not present mutations on genes previously related to this phenotype (Nilsson et al., 2003; Takahata et al., 2010; Couce et al., 2012). Nevertheless, 6 and 4 out of 8 strains were unable to grow with glucose-6-phosphate and glycerol-3-phosphate, respectively, as sole energy and carbon source (Supplementary Table S7), suggesting that a defect in fosfomycin uptake may be involved in their resistance. On the other hand, to evaluate whether gentamicin resistance genetic determinants were related to this phenotype, the MIC of fosfomycin was determined for CIM8M and its derivatives complemented with the wild-type genes (Supplementary Table S11). Except for a weak reduction with fhuA, complementation with wild-type genes did not modified fosfomycin resistance. Further studies are required to elucidate the genetic basis of this indirectly acquired resistance.

Besides fosfomycin resistance, all studied strains derived from MG1655, but not from its ΔdnaQ derivative, developed rifampicin resistance due to a H526Y substitution in RpoB (Cavusoglu et al., 2006; Ma et al., 2006). The H526Y mutation is known to cause temperature-sensitive growth and decreased termination efficiency at Rho-dependent and independent transcriptional terminators in vivo and in vitro (Jin and Gross, 1989), increasing the transcription elongation rate and decreasing pausing in vitro (McDowell et al., 1994). Changes in rpoB have been related to changes in the expression of hundreds of genes and large fitness advantages (Rodríguez-Verdugo et al., 2016). The importance of this mutation, if any, on the adaptation to gentamicin remains to be studied. Its absence from the genomes of ΔdnaQ-derivatives merits also a brief consideration. A simple possibility is that the wild-type mutational spectrum is particularly prone to produce the CAC → TAC substitution that underlies mutation H526Y, although this was not found in previous studies (Couce et al., 2013). Another, more speculative possibility is that many of the neutral or slightly deleterious mutations that the mutator has accumulated become strongly deleterious due to the pleiotropic effects associated to changes in the RNA polymerase (Rodríguez-Verdugo et al., 2016), precluding its selection. Interestingly, the ΔdnaQ-derivative CIQ4J displays an increased rifampicin resistance, yet without any mutation in rpoB. This is an intriguing result that deserves further studies to stablish the molecular basis of this resistance mechanism.

On the contrary, all selected strains increased their susceptibility to chloramphenicol (Table 1). This finding agrees with previous reports that indicate that aminoglycoside resistance is related to an increased susceptibility to chloramphenicol (Imamovic and Sommer, 2013; Lázár et al., 2013; Lobritz et al., 2015). It has been proposed that this phenomenon could be related to alterations in electron transport chain and thus a reduction in the activity of efflux pumps that export chloramphenicol but not aminoglycosides, like AcrAB-TolC (Nishino and Yamaguchi, 2001). The effect of single mutations on previously selected genes related to gentamicin resistance over chloramphenicol susceptibility was evaluated through MIC of the single gene mutants (Supplementary Table S12). FhuA197 and EFG593 were the only strains that decreased 2-fold their MIC of chloramphenicol, while atpG deletion increased it 2-fold, suggesting that, at least in an individual manner, these mutations do not explain CIM8M increase in chloramphenicol susceptibility from 4 to ≤ 0.25 μg/ml. Liu et al. determined the contribution of gene deletions in E. coli over antibiotic sensibility, including chloramphenicol (Liu et al., 2010). However, no mutations on genes described by Lui et al. were detected in CIM8M. An effect of the other mutations (synonymous, intergenic, etc.) found in this and other strains on gene regulation cannot be discarded.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.