Experimental evolution of gallium resistance in Escherichia coli

Abstract Background and Objectives Metallic antimicrobial materials are of growing interest due to their potential to control pathogenic and multidrug-resistant bacteria. Yet we do not know if utilizing these materials can lead to genetic adaptations that produce even more dangerous bacterial varieties. Methodology Here we utilize experimental evolution to produce strains of Escherichia coli K-12 MG1655 resistant to, the iron analog, gallium nitrate (Ga(NO3)3). Whole genome sequencing was utilized to determine genomic changes associated with gallium resistance. Computational modeling was utilized to propose potential molecular mechanisms of resistance. Results By day 10 of evolution, increased gallium resistance was evident in populations cultured in medium containing a sublethal concentration of gallium. Furthermore, these populations showed increased resistance to ionic silver and iron (III), but not iron (II) and no increase in traditional antibiotic resistance compared with controls and the ancestral strain. In contrast, the control populations showed increased resistance to rifampicin relative to the gallium-resistant and ancestral population. Genomic analysis identified hard selective sweeps of mutations in several genes in the gallium (III)-resistant lines including: fecA (iron citrate outer membrane transporter), insl1 (IS30 tranposase) one intergenic mutations arsC →/→ yhiS; (arsenate reductase/pseudogene) and in one pseudogene yedN ←; (iapH/yopM family). Two additional significant intergenic polymorphisms were found at frequencies > 0.500 in fepD ←/→ entS (iron-enterobactin transporter subunit/enterobactin exporter, iron-regulated) and yfgF ←/→ yfgG (cyclic-di-GMP phosphodiesterase, anaerobic/uncharacterized protein). The control populations displayed mutations in the rpoB gene, a gene associated with rifampicin resistance. Conclusions This study corroborates recent results observed in experiments utilizing pathogenic Pseudomonas strains that also showed that Gram-negative bacteria can rapidly evolve resistance to an atom that mimics an essential micronutrient and shows the pleiotropic consequences associated with this adaptation. Lay summary We utilize experimental evolution to produce strains of Escherichia coli K-12 MG1655 resistant to, the iron analog, gallium nitrate (Ga(NO3)3). Whole genome sequencing was utilized to determine genomic changes associated with gallium resistance. Computational modeling was utilized to propose potential molecular mechanisms of resistance.


INTRODUCTION
The spread of resistance to traditional antibiotics has spurred the search for new antimicrobial substances [1][2][3]. Ionic and nanoparticle metals have been proposed, including silver, copper, excess iron (II, III) and gallium. Unfortunately, many of these studies were conducted by material scientists and engineers who had little understanding of the evolutionary dynamics of populations exposed to toxic materials [4,5]. These studies claimed the fact that metals impacted so many different aspects of bacterial physiology, that they represented steep obstacles to the evolution of resistance [4].
Recently several studies have examined the impact of the iron analog gallium upon the growth of pathogenic bacteria [6]. The mechanism targeted by this approach was to use gallium (Ga 3+ ) to saturate bacterial siderophores that normally take up ferric iron (Fe 3+ ). Studies on the impact of gallium suggest that it may impede growth without killing cells (bacteriostatic) or eventually kill cells via prolonged iron starvation (bactericidal) [6]. Gallium's (Ga 3+ ) capacity to quench bacterial siderophores is due to this element's similarity with ionic radius and binding properties of iron (III, Fe 3+ ) [6]. Gallium is a group IIIA metal with atomic number 31 and shares certain properties with Fe 3+ , where the octahedral ionic radius of Ga 3+ is 0.620 Å compared with 0.645 Å for Fe 3+ and the tetrahedral ionic radius is 0.47 Å for Ga 3+ compared with 0.49 Å for Fe 3+ [7].
It is known that iron (as well as other micronutrient metals Mg, Mn, S and Zn) play key roles in the capacity for pathogenic bacteria to maintain their virulence. These metals often serve as enzymatic co-factors. Therefore, their intracellular concentration must be maintained and tightly regulated to maintain cell viability [8]. Iron is one of the most important micronutrients as it fulfills many biological roles [9]. Iron-containing proteins (heme-proteins, ironsulfur cluster proteins and di-iron and mononuclear enzymes) play roles in nitrogen fixation and metabolism and serve as electron carriers for respiration [10]. Despite its critical role in bacterial metabolism, acquiring iron is one of the greatest challenges for bacterial growth [8,11] and iron deficiency is one of the most common nutritional stresses [12]. Despite iron being an essential metal, it can also be extremely toxic under aerobic conditions [13].
In aerobic environments, iron predominantly occurs as ferric iron (Fe 3+ ), however, Fe(OH) 3 is poorly soluble in aqueous solution (as low as 10 À18 M at pH 7.0). Under anaerobic conditions, the equilibrium shifts to ferrous iron (Fe 2+ ) that is more readily bioavailable to microorganisms and availability is a key to pathogenesis for a variety of microbes, thus many innate immunity mechanisms utilize iron sequestration (e.g. serum albumin, calprotectin [14,15]). As iron is essential, microbes have evolved means to take it up from their extracellular surroundings; for example, siderophores. The siderophore, enterobactin, in Escherichia coli is synthesized via genes such as entC and fep genes (in strain K-12: A, B, C, D, E and G). These siderophores are tightly controlled by the global iron homeostasis regulator, Fur [8]. Under aerobic conditions, excess iron has been shown to induce oxidative damage, whereas both Fe 2+ and Fe 3+ can interact with hydrogen peroxide and superoxides, respectively, which generate highly reactive hydroxyl radicals leading to cell damage and eventual cell death [13]. This is in addition to a variety of other effects summarized in Table 1. When oxygen free radicals form in their cells, bacteria have genes to defend themselves from the accrued stressors. This is aided by many proteins including OxyR that responds to the presence of hydrogen peroxide, SoxS and SoxR that respond to redox active compounds and RpoS that responds to general oxidative stress [17][18][19]. It is therefore important that intracellular iron be regulated to prevent toxicity by this essential metal [20,21].
We have recently shown that E. coli can evolve resistance to excess iron (II and III  Ewunkem et al. in preparation]. Furthermore, an unexpected result of excess iron resistance in these strains was resistance to gallium which is an iron analog possessing the same nuclear radius as iron (Fe 3+ ). For this reason, gallium has been shown to exchange for iron in siderophores such as in Pseudomonas aeruginosa where the siderophore pyochelin transports gallium into the cell [23], it also represses pvdS, a transcriptional regulator of the siderophore pyoverdine [6]. Once inside the cell gallium cannot be used for essential iron catalyzed reactions and is therefore inhibits growth (bacteriostatic) due to mismetallation. In addition, gallium has been shown to perturb the regulation of iron acquisition pathways and can eventually cause cell death [6]. Until recently, experiments showing the evolution of gallium resistance were limited to Pseudomonas [24][25][26]. However, a recent study utilized the Keio collection which consists of 3985 single, non-essential gene knockouts in E. coli BW25113 and examined the effect on gallium resistance [27,28]. Each mutant was scored as to whether it increased or decreased gallium sensitivity on a scale of À0.6 to +0.6 normalized to colony size. The mutants were classified by their impact on cellular metabolism (transcription, translation, DNA metabolism, RNA metabolism, protein metabolism, protein folding and secretion, cell exterior functions, biosynthetic functions, degradation, energy, cellular processes, response to stimulus and other pathways). Of the 3945 mutations studied, they found a relatively equal distribution of mutations that raised or lowered gallium sensitivity (1761 raised and 1878 lowered). However, the vast majority of these effects on colony size were minor. These sorts of studies are limited in their capacity to predict how bacteria might actually evolve resistance to gallium or any stressing material, as the gene knockouts were of non-essential functions. Furthermore, knocking out a gene may produce unanticipated pleiotropic effects on fitness that are hard to interpret in any given environment.
Therefore, in this study we utilize experimental evolution on E. coli K-12 MG1655 to ask whether and how does E. coli evolve resistance to excess gallium. In addition, since we have already shown that this strain can evolve excess iron (II, III) resistance, we wanted to know whether gallium resistant E. coli would also have a correlated response that confers iron (II, III) resistance. Furthermore, we asked whether gallium resistance confers resistance to metals that may not have similar bacteriostatic/bactericidal effects to gallium such as silver (Table 1). We also examined the genomic changes, that allow E. coli to survive in environments containing excess gallium and compared those with the knockout results of Gugala et al. [28]. Finally, we modeled the structural changes in one of the proteins altered by selection to propose a mechanism by which this mutation conferred gallium resistance.

Evolution experiment
Experimental evolution is defined as the study of populations over defined and reproducible conditions over multiple generations under laboratory or natural conditions [29]. It has been used to study adaptation in a variety of organisms and conditions, including the study of antimicrobial resistance via antibiotics (ampicillin, beta lactams, chloramphenicol and others), as well as metals [30][31][32][33][34]. Other methods have been used, such as stepwise selection to isolate metal resistant mutants in E. coli [35]. However, the underlying principle of artificial selection is the same as that employed in experimental evolution. However, our method allows the study of adaptation to increase antimicrobial metal concentration over time, which does not result from the step-wise selection protocol.
Escherichia coli K-12 MG1655 (ATCC #47076) was chosen for this study due to the paucity of known metal resistant loci in this bacterium. This is also the strain we have used in our previous studies evaluating the evolution of silver and iron resistance [21,22,33,34]. Its chromosome has 4 641 652 nucleotides (GenBank: NC_000913.3) and contains no plasmids. We have sequenced this strain and polymorphisms not present in the published reference genome are given in in previous studies [33,34].
In our experiments, E. coli was cultured in Davis Minimal Broth (DMB; Difco, Sparks, MD) with 1 g of dextrose per liter (Fisher Scientific, Fair Lawn, NJ) carbon source. The medium is enriched with thiamine hydrochloride with 10 ml in the final volume of 10 ml, for a concentration of 0.3 mM. This is kept in 50-ml Erlenmeyer flasks and placed in a shaking incubator at 115 rpm at 37 C. Each replicate was founded from a unique colony on an agar plate isolated via serial dilution from the ancestral stock culture. These samples were placed in separate flasks and allowed to grow for 24 hours before being transferred to fresh media. Five replicates of the controls (C 1 -C 5 ) which are grown in standard DMB medium without the addition of gallium and five replicates of the Ga(NO 3 ) 3 selected (Ga 3+ _1-Ga 3+ _5) were founded and propagated by subculturing 0.1 ml into 9.9 ml of fresh sterile DMB daily. The population density increased from $10 7 to $10 9 cells per ml over the course of each daily transfer cycle. Replication of selection treatments is important because it allows for the observation of parallelism, that is, adaptations that are general responses to the selection regime.
A minimum inhibitory concentration (MIC) assay utilizing the ancestral E. coli K-12 MG1655 strain allowed the determination of the sublethal concentration of each gallium nitrate to be utilized for selection purposes. This was determined to be 100 mg/l of Ga(NO 3 ) 3 during 10 days of culture. The gallium selected and control populations were frozen at À80 C for future analysis. To determine the potential impact of the nitrate ion (NO À 3 ) on the growth of the controls and gallium nitrate selected populations, a MIC assay was conducted on both populations in sodium nitrate (NaNO 3 ). The reduction in 24-hour growth as measured by optical density at 625 nm from 6.25 to 500 mg/l was 0.012 for the controls and 0.014 for the gallium-selected populations. From these results, we determined that resistance to the nitrate ion would not be a significant factor in the experiment.
Phenotypic assays: 24-hour growth Measurements of 24-hour growth in excess metal (Ga 3+ , Fe 2+ , Fe 3+ and Ag + ) and traditional antibiotics for the ancestral, gallium-selected and control populations were determined at 10 days of evolution. These values were compared with the five samples of the E. coli K-12 MG1655 ancestor which was grown over night in standard DMB medium. The range used was 0-1000 mg/l for the Ga 3+ ; 0-1750 mg/l for the Fe 2+ and Fe 3+ assays, 0-100 mg/l for Ag + and finally 0-500 mg/l for the antibiotics (chloramphenicol, rifampicin, sulfanilamide and tetracycline). Previous studies of E. coli K-12 MG1655 have shown that this strain has no resistance to these antibiotics [36]. Statistical analysis of the effect of selection regime and concentration (and their interaction) for all 24-hour growth data was performed via General Linear Model utilizing SPSS version 23 (SPSS Inc, Armonk, NY, USA). All graphs in this paper were made via SigmaPlot version 14. Finally, the phenotypic data from these studies will be submitted into DRYAD (https://datadryad. org/) upon acceptance of this manuscript for publication.

Genomic analysis
DNA was extracted from each population after 10 days of culture using the EZNA Bacterial DNA extraction kit (Omega Bio-tek Õ ) as per manufacturer instructions. DNA concentrations were normalized using the QuantiFluor Õ dsDNA system. Genomic libraries were prepared using the Illumina Nextera XT kit and samples were sequenced using the Illumina MiSeq sequencing platform. The depth of coverage of the sequencing runs ranged from $20Â to $80Â, with most exceeding 40Â coverage. The SRA accession number for sequencing data from this study are PRJNA521749 (Ga 3+ -resistant and controls). Sequence alignment and variant calling from the samples was achieved by use of the breseq 0.30.0 pipeline [37]. The breseq pipeline uses three types of evidence to predict mutations, read alignments, missing coverage and new junctions, and any reads that indicate a difference between the sample and the reference genome that cannot be resolved to describe precise genetic changes are listed as 'unassigned' [37]. These unassigned reads are not described or interpreted here.
The position at which the FecA mutations occurred were then identified in the published structure of FecA in complex with ironcitrate (PDB1PO3) [38] and visualized using PYMOL [39]. This was done in order to predict the functional consequences of the identified mutations.

RESULTS
Phenotypic changes in presence of metals Figure 1a-d shows 24-hour growth across increasing concentrations of gallium, iron (II), iron (III) and silver nitrate. The Ga 3+selected populations showed superior growth relative to the controls and ancestral strain across concentration for gallium, iron (III) and silver nitrate. The controls and ancestral populations showed superior growth relative to the Ga 3+ -selected populations in iron (II). For all substances tested, the Ga 3+ -selected, controls and ancestral population decreased their 24-hour growth with increased concentration of the metal tested. There were two significant interactions between the population and concentration variables, in gallium and iron (II) resistance comparisons between the Ga 3+ -selected and ancestral population. The F-statistics and P-values for all comparisons are given in Table 2.
The 24-hour growth of the Ga 3+ -selected, control and ancestral populations from 0 to 250 mg/l in rifampicin, tetracycline, chloramphenicol and sulfanilamide are shown in Figure 2a-d. The control populations showed superior growth relative to the Ga 3+ -selected and ancestral population in increasing concentrations of rifampicin. There was no significant difference between the selected, control and ancestral populations in chloramphenicol and tetracycline (growth ceased at 6.2 mg/l). There was no difference between Ga 3+ -selected, control and ancestral populations from 6 to 75 mg/l of sulfanilamide. However, from 100 to 250 mg/l, both the Ga 3+ -selected and control populations showed equivalent growth which was statistically significantly greater than that of the ancestral population. Table 3 shows selective sweeps in the Ga 3+ -resistant populations at day 10. There were eight hard selective sweeps (0.000 ancestor-1.000 in the Ga 3+ -resistant populations); one hard sweep found in (Ga 3+ _1, Ga 3+ _2 and Ga 3+ _5), Ga 3+ _3 displayed two hard sweeps and Ga 3+ _4 showed four hard sweeps. Of the hard sweeps, all populations showed a mutation in fecA (UniProtKB-P77279). Ga 3+ _1, Ga 3+ _2, Ga 3+ _3 and Ga 3+ _5 showed the same mutation, and Ga 3+ _4 had a different mutation in this gene. Ga 3+ _3 showed a hard selective sweep in insl1 (UniProtKB-P0CF91), Ga 3+ _4 in yedN (UniProtKB-A0A023BD4). There was one hard sweep in an intergenic region (qmcA /! fetA). The qmcA (UniProtKB-P0AA53) gene encodes a prohibitin homology domain-anchored (PHB) domain-anchored putative protease and fetA (UniProtKB-P77279) encodes an iron exporter, ATP binding subunit, ABC transporter FETAB subunit; peroxidase resistance protein. There were seven significant polymorphisms (0.351-0.39) distributed across the five populations. Ga 3+ _1 and Ga 3+ _3 show an identical mutation in fecE (dicitrate transport ATP binding protein, UniProtKB-P15031). Ga 3+ _4 shows a mutation in fecB (binds citrate-dependent Fe 3+ ; part of the binding protein-dependent transport system for uptake of citrate-dependent Fe 3+ , UniProtKB-P15028). Ga 3+ _3 displays a mutation in ydfE (Qin prophage, pseudogene; phage or prophage related, UniProtKB-Q7138). The remaining significant polymorphisms are intergenic, Ga3+_1 (fepD /! entS; fepD UniProtKB-23876 is part of the binding protein-dependent transport system for ferric enterobactin. Probably responsible for the translocation of the substrate across the membrane; entS UniProtKB-24077 has enterobactin transmembrane transporter activity, also involved in cellular response to DNA damage stimulus, and response to antibiotic). Ga 3+ _3 has an intergenic mutation [yfgF /! yfgG; yfgF UniProtKB-P77172 is a phosphodiesterase (PDE) that catalyzes the hydrolysis of cyclic-di-GMP (c-di-GMP) to 5 0 -pGpG; yfgG UniProtKB-P64545 is an uncharacterized protein]. Finally, six lowlevel polymorphisms were observed. Ga 3+ _1 displayed a mutation in prlF UniProtKB-15373, antitoxin component of a type II toxinantitoxin (TA) system, Ga 3+ _2 had a mutation in nhaB UniProtKB-P0AFA7, a Na + /H + antiporter that extrudes sodium in exchange for external protons and Ga 3+ _5 had one in rpoS UniProtKB-P13445, RNA polymerase sigma factor S. The remaining low-level polymorphisms were intergenic with Ga 3+ _1 showing ypjC / ileY; ypjC UniProtKB-76613 predicted protein and ileY tRNA isoleucine; agaI !/! yraH, agaI UniProtKB-P42912, putative deaminase, galactosamine-6 phosphate isomerase; yraH UniProtKB-P42913, putative fimbrial-like adhesin protein; dcuB (/( dcuR, dcuB, UniProtKB-P0ABN9, anaerobic C4dicarboxylate transporter DcuB and DcuR are members of the two-component regulatory system DcuB/DcuR, involved in the C4-dicarboxylate-stimulated regulation of the genes encoding the anaerobic fumarate respiratory system. Table 4 and 5 report   3 200-1000 mg/l F= 110.64, P = 0.0001 F = 18.09, P = 0.0001 F = 7.18, P = 0.0001 Iron (III) Fe 2 (SO4) 3 6. 2 All polymorphisms classified by breseq in the Ga 3+ -selected populations and the controls are found in Supplementary Tables 1a-c and 2a-c. None of the suspected mutations associated with Ga 3+ resistance were found in the controls, or have ever been observed in the control populations of our previous studies [21,22,33,34].

DISCUSSION
These studies have clearly shown that E. coli K-12 MG1655 can rapidly evolve resistance to gallium. We predicted that this would be true based upon our previous work that demonstrated that this strain could evolve resistance to excess iron (II, III) and that in each case, increased resistance to gallium resulted [21,22]. Our result is in direct opposition to previously published claims concerning gallium quenching as an evolutionarily stable antimicrobial treatment [24]. That study examined experimental evolution of gallium resistance and found no evidence of anti-gallium resistance over the course of 12 days for P. aeruginosa. On the face of it that result is counter intuitive, given that a 2013 study did find evidence of anti-gallium mutations in P. aeruginosa [26] and a recent 2018 study corroborated the 2013 result [25]. Thus, it seems clear that Gram-negative bacteria do have the capacity to evolve resistance to gallium quenching.
In addition to gallium resistance, correlated resistance to excess iron (III, Fe 3+ ) and ionic silver (Ag + ) was observed in the Ga 3+ -resistant populations. The latter result can be questioned, in that the genomic analysis showed no selection in genes that have previously been associated with silver resistance. Although selection for iron (II) and iron (III) resistance in our previous after 10 days of evolution. There was no difference between the Ga3+-selected, controls and ancestral populations from 6 to 75 mg/l; however, the Ga3+-selected = controls and both were significantly > than the ancestral populations from 100 to 250 mg/l studies did confer a minor increase in silver resistance without evidence of selection in the genes cusS, ompR, that played a major role in conferring resistance [21,22,33,34]. Therefore, it is possible that gallium selection is conferring a minor increase in silver resistance by mechanisms, we do not as of yet understand. Finally, our Fe 2+ -and Fe 3+ -resistant populations showed cross resistance to iron (II) and iron (III). The Ga 3+ -resistant populations are resistant to iron (III), but not to iron (II). The Ga 3+ -selected populations also showed inferior 24-hour growth relative to the controls and the ancestors in the traditional antibiotic rifampicin. This can be linked to the selective sweeps in rpoB that were found in the control populations, but not in the Ga 3+ -resistant populations or ancestors. Mutations in rpoB are commonly observed in E. coli experimental evolution, including adaptation to media-containing minimal amounts of carbon source (glucose/lactose) and these are associated with increased resistance to rifampicin [40][41][42]. It seems that pleiotropic effects of such mutations may be a crucial means by which the evolution of antibiotic resistance proceeds in the absence of antibiotics [43,44].
There was no significant difference between the Ga 3+ -resistant, control and ancestral populations in chloramphenicol and tetracycline. The Ga 3+ -resistant, control and ancestral populations showed equivalent growth at lower concentrations of sulfanilamide (6-75 mg/l). However, both the Ga 3+ -resistant and control populations performed better at higher concentrations (100-250 mg/l) of sulfanilamide compared with the ancestor. As neither of these populations were exposed to sulfanilamide, it is likely that this improvement in growth is due to some other aspect of the environment shared by both populations (such as minimum sugar content). ATCC cultures this strain using Luria broth which has a higher sugar content than DMB (https://www.atcc.org/ products/all/47076.aspx#culturemethod).
The lack of correlated antibiotic resistance in the Ga 3+ -resistant populations is significant in that our iron-selected populations (Fe 2+ and Fe 3+ ) did show correlated resistance to antibiotics [21,22]. The difference between these strains in correlated responses to selection is best explained by the differences in the genomic foundations of their respective resistances. All the Ga 3+resistant populations showed a mutation in the fecA gene (ferric citrate outer membrane transporter, the G400S mutation was observed in all populations except Ga 3+ _4 which displayed N169K). This gene encodes the protein FecA is responsible for Fe 3+ transport in aerobic conditions [11]. We propose that these mutations reduce the function of FecA, thereby limiting Ga 3+ import in these strains. For example, we have shown the FecA mutations (A559T and G243C) observed in our iron (II)-resistant populations express 128-fold less fecA than the controls in absence of iron indicating a likely preemptive mechanism of defense [22].
Several of the other genomic variants showing strong selection involved in conferring gallium resistance are in genes and intergenic regions associated with iron metabolism (fecE; fecB; qmcA/fetA; fepD/entS). The remaining polymorphisms showing evidence of selection have no clear relationship to iron metabolism, thus some could be responses to excess NO 3À in the medium, or they could have yet undiscovered relationships to gallium resistance. It is also important to note the difference in our genomics results (via experimental evolution) and those obtained via knock out mutations in E. coli [9] in that none of our genes resulting from selection for gallium resistance were found in the previous study. This results from the fact that the Keio strains were produced by knockout of non-essential genes in E.
coli K-12 [27] and our selective sweeps occurred in genes essential to iron metabolism in E. coli K-12 (e.g. fecA, fecE and fecB). In addition, the knock out method cannot identify intergenic regions that contribute to gallium adaptation, as no intergenic regions are knocked out in these strains. Thus, at best, knock out methods can identify genes that could possibly play a role in adaptation, but this technique is not necessarily going to tell you which genes will play a role in adaptation.
Our studies of experimental evolution of iron resistance showed that co-selection of metal and antibiotic resistance, and this can occur from linkage as well as pleiotropy. In the case of metals and antibiotics, pleiotropy is well known. For example, resistance mechanisms to metals such as reduction in cell wall permeability, substance alteration, efflux, alteration of cellular targets and sequestration are wide spread, and these traits also produce resistance to traditional antibiotics [45,46]. However, our Ga 3+ -resistant lines, do not display pleiotropy between gallium and antibiotic resistance whereas our iron (II, III) populations showed positive pleiotropy for five antibiotics: ampicillin, chloramphenicol, rifampicin, sulfanilamide and tetracycline [33,34]. Genomic analysis illustrated that four of the five Fe 2+ -selected replicates showed at least one or more of these populations identified hard sweeps in genes such as murC, cueR, tolC, yeaG and ptsP. Several of these genes have been shown to be associated with antibiotic resistance, such as in the case of the murC which plays a role in cell wall biosynthesis as well as repairing oxidative damage [47]. However, the gallium-resistant populations do not display any selective sweeps in genes with an established relationship to antibiotic resistance, whereas the controls did display mutations in rpoB, that are consistently associated with antibiotic resistance [40][41][42]. We believe that the most functionally relevant mutations observed in this study is in FecA. As previously stated, FecA, through the siderophore citrate, is an iron (III) transporter under aerobic conditions. FecA is a transmembrane b-barrel and its structures has previously been solved both in its apo form and in complex with iron citration [48]. We used the complexed structure (PDB1PO3) as a guide to predict functional outcomes of the observed mutations. Both residues identified in this study (G400S and N169K) were found directly in the ligand binding pocket (Figure 3). The G400 amino acid sits directly adjacent to iron-citrate and the introduction of a polar group through mutation to a serine could easily hinder binding of the siderophore to the pocket. The second residue N169 sits on the backside of a helix which is also in contact with the iron citrate. The mutation to a large and positively charged amino acid such as lysine could potentially displace this helix into the ligand-binding pocket, also occluding entry of the siderophore. These mechanisms remain to be evaluated. In summary, we have found that E. coli K-12 MG1655 can rapidly evolve resistance to gallium in opposition to claims of the evolutionary stability of gallium quenching. This trait also conferred resistance to ferric iron (iron III) and ionic silver, but no correlated responses to antibiotics were observed. The genomic results indicated that gallium resistance results from selection on mutations associated with iron metabolism. This is not surprising as gallium is an iron analog. We propose that these mutations are likely to be associated with the down regulation of genes involved in iron acquisition; as we observed in our iron (II)-and iron (III)resistant populations [33,34].
As in the evolution of silver, copper and iron resistance, a small number of genomic changes were required to produce the gallium-resistant phenotypes. All replicates showed at least one hard selective sweep (fecA). The parallelism of the mutations in fecA is a strong evidence that this gene played a major role in conferring gallium resistance. The Ga 3+ _4 replicate showed three additional hard sweeps, while the other populations had associated polymorphisms associated with iron metabolism not observed in the ancestor or control populations. We also suspect that these genomic changes must be associated with a profound difference in gene expression associated with iron and general metal metabolism. This was unmeasured in this study but is consistent with our previous results [33,34]. Finally, our studies have important applications from the view of informing attempts to utilize metals such as gallium to control pathogenic and multidrug resistant bacteria. Despite optimism concerning the sustainability of metallic antimicrobials [4,5,24], bacteria can also rapidly evolve resistance to these materials as well.

supplementary data
Supplementary data is available at EMPH online.