A genome wide analysis of Escherichia coli responses to fosfomycin using TraDIS-Xpress reveals novel roles for phosphonate and phosphate transport systems

Fosfomycin is an antibiotic which has seen a revival in use due to its unique mechanism of action and resulting efficacy against isolates resistant to many other antibiotics. Mechanisms of resistance have been elucidated and loss of function mutations within the genes encoding the sugar importers, GlpT and UhpT are commonly selected for by fosfomycin exposure in E. coli. There has however not been a genome wide analysis of the basis for fosfomycin sensitivity reported to date. Here we used ‘TraDIS-Xpress’ a high-density transposon mutagenesis approach to assay the role of all genes in E. coli in fosfomycin sensitivity. The data confirmed known mechanisms of action and resistance as well as identifying a set of novel loci involved in fosfomycin sensitivity. The assay was able to identify sub domains within genes of importance and also revealed essential genes with roles in fosfomycin sensitivity based on expression changes. Novel genes identified included those involved in glucose metabolism, the phosphonate import and breakdown system, phnC-M and the phosphate importer, pstSACB. The impact of these genes in fosfomycin sensitivity was validated by measuring the susceptibility of defined inactivation mutants. This work reveals a wider set of genes contribute to fosfomycin sensitivity including core sugar metabolism genes and two transport systems previously unrecognised as having a role in fosfomycin sensitivity. The work also suggests new routes by which drugs with a phosphonate moiety may be transported across the inner membrane of Gram-negative bacteria. Importance The emergence and spread of antibiotic resistant bacteria had resulted in increased use of alternative drugs which retain efficacy against isolates resistant to other classes of drugs. One example is fosfomycin; an old drug which has found greatly increased use in recent years. We studied the mechanisms of fosfomycin resistance by applying a genome wide screen based on comparing the fitness of a massive library of transposon mutants in the presence of fosfomycin. This approach identified the previously known mechanisms of resistance but also identified a number of new pathways which contribute to fosfomycin sensitivity including two importer systems. This information advances our knowledge about an increasingly important antibiotic and identifies new potential routes to resistance.


Abstract 22
Fosfomycin is an antibiotic which has seen a revival in use due to its unique mechanism of 23 action and resulting efficacy against isolates resistant to many other antibiotics. Mechanisms 24 of resistance have been elucidated and loss of function mutations within the genes encoding 25 the sugar importers, GlpT and UhpT are commonly selected for by fosfomycin exposure in 26 E. coli. There has however not been a genome wide analysis of the basis for fosfomycin 27 sensitivity reported to date. Here we used 'TraDIS-Xpress' a high-density transposon 28 mutagenesis approach to assay the role of all genes in E. coli in fosfomycin sensitivity. The 29 data confirmed known mechanisms of action and resistance as well as identifying a set of 30 novel loci involved in fosfomycin sensitivity. The assay was able to identify sub domains 31 within genes of importance and also revealed essential genes with roles in fosfomycin 32 sensitivity based on expression changes. Novel genes identified included those involved in 33 glucose metabolism, the phosphonate import and breakdown system, phnC-M and the 34 phosphate importer, pstSACB. The impact of these genes in fosfomycin sensitivity was 35 validated by measuring the susceptibility of defined inactivation mutants. This work reveals a 36 wider set of genes contribute to fosfomycin sensitivity including core sugar metabolism 37 genes and two transport systems previously unrecognised as having a role in fosfomycin 38 sensitivity. The work also suggests new routes by which drugs with a phosphonate moiety 39 may be transported across the inner membrane of Gram-negative bacteria. 40

Importance 41
The emergence and spread of antibiotic resistant bacteria had resulted in increased use of 42 alternative drugs which retain efficacy against isolates resistant to other classes of drugs. 43 One example is fosfomycin; an old drug which has found greatly increased use in recent 44 years. We studied the mechanisms of fosfomycin resistance by applying a genome wide 45 screen based on comparing the fitness of a massive library of transposon mutants in the 46 presence of fosfomycin. This approach identified the previously known mechanisms of 47

Introduction 52
The increasing prevalence of bacteria which are resistant to clinically important antibiotics 53 has led to searches for alternative options to treat problematic infections (1). There has been 54 limited progress in the development of new antibiotics and one strategy has been to revive 55 older drugs which may be effective but are not common in clinical practice (2). One example 56 is fosfomycin which has seen a sharp increase in clinical use in recent years. Fosfomycin 57 has a unique mode of action where it targets the initial stages of peptidoglycan biosynthesis 58 by acting as a phosphoenolypyruvate analogue and inhibiting MurA (3). This means that 59 fosfomycin retains activity against strains producing beta-lactamases as it targets an earlier 60 stage in peptidoglycan biosynthesis. This is attractive given the high prevalence of 61 production of beta-lactamase enzymes of many families in important pathogens. Fosfomycin 62 is a phosphonic-acid molecule produced in nature by Streptomyces species and is 63 commonly used for treatment of complicated urinary tract infections and increasingly for 64 more serious systemic infections. In Enterobacteriaceae, fosfomycin enters the cell by acting 65 as a mimic for two nutrient importer systems; GlpT and UhpT (4). 66 Resistance to fosfomycin has been shown to be relatively easy to select in vitro and resistant 67 mutants often show loss of function of GlpT or UhpT, or have mutations in adenylcyclase 68 (cyaA), or the phosphotransferase (ptsI) both of which control intracellular levels of cyclic-69 AMP, which in turn regulates expression of glpT and uhpT (5-9). In addition, alterations 70 within the drug target MurA (particularly those altering a Cys115 residue near the active site) 71 can decrease susceptibility by reducing its affinity for fosfomycin (10-12). Over-expression of 72

Susceptibility of BW25113 to fosfomycin 90
Baseline susceptibility to BW25113 was measured and the MIC determined to be 4µg/ml, 91 the transposon mutant library was then inoculated into fresh media (~10 7 mutants were 92 added to each reaction) containing multiples of the MIC and allowed to grow overnight.  (Table 1). These 104 included the majority of known, chromosomal mechanisms of resistance with strong signals 105 identified for murA; the target for fosfomycin, glpT and cyaA. This validates the specificity of 106 TraDIS-Xpress in identifying genes involved in fosfomycin susceptibility. The TraDIS-Xpress 107 method also proved able to assay essential genes. For example, murA is an essential gene 108 but mutants with inserts which are positioned upstream and in the same orientation as murA 109 were highly enriched in the presence of IPTG (Figure 1). These mutants will over-express 110 murA which will help the target saturate the activity of fosfomycin. The high density of the 111 library also allows very high resolution, for example cyaA was identified as a significant 112 target but enrichment of inserts was restricted to this gene's regulatory domain ( Figure 1). 113 Mutations within this domain have previously been reported as being important to determine 114 fosfomycin sensitivity (but not the rest of the protein) and TraDIS-Xpress was able to clearly 115 identify the sub-domain within the gene responsible for fosfomycin. Interestingly, no signal 116 was identified at the uhpT locus which has also previously been implicated in fosfomycin 117 import although is not expressed in the test conditions so would not be under selective 118

pressure. 119
Identification of new mechanisms of fosfomycin resistance 120 As well as loci known to be involved in determining fosfomycin sensitivity several new loci 121 were identified by Tradis-Xpress. These included several genes involved in sugar 122 metabolism beyond those already noted as contributing to fosfomycin susceptibility (table 1). 123 Significant patterns of inserts were seen at cra, crp, cyaA, galU, glpK, glpX and treC. 124 Together these genes are all involved in control of available glucose within the cell which will 125 in turn influence expression of the glucose importers that are known routes of entry for 126 fosfomycin. Mutants that over-expressed leuO, a pleiotropic regulator with known roles in 127 control of stress responses were strongly selected by fosfomycin (figure 1). As were mutants 128 with inserts within the phosphonate and phosphate uptake systems (figure 2). The phnC-P 129 phosponate uptake and degradation system is a large operon composed of an ABC importer 130 (phnCDE), regulator (phnF), a multisubunit phosphonate degradation complex (phnG-L) and 131 some ancillary components (phnM-P). The system has been reported to be phase variable 132 due to the presence of either 2 or 3 copies of a repeat element within phnE and the 133 downstream genes (PhnF-L) are thought to be cryptic in K-12. Despite this, there was a 134 massive enrichment of inserts throughout this system upon exposure to fosfomycin ( figure  135 2), sequencing of the parent strain of the library, and mapping of from reads obtained after 136 fosfomycin exposure confirmed our strain carried 2/3 inserts and was, therefore supposedly 137 cryptic. However, the TraDIS-Xpress data suggested a large fitness advantage from 138 inactivation of most of this system. Additionally, another ABC importer (PstSACB) was also 139 implicated in fosfomycin sensitivity (figure 2) with inactivation of this system favouring growth 140 in the presence of fosfomycin. Both mutL and tatD, involved in DNA repair processes were 141 protected under all fosfomycin exposure conditions. 142 None of the genes important in all conditions have been associated with resistance to other 143 drugs, i.e. there was no signal observed for porins or multidrug efflux pumps or their 144 regulators suggesting selection of cross-resistance to other agents by fosfomycin may be 145 limited. 146

Validation of targets. 147
To test the predictions made by TraDIS-Xpress we selected a set of nine genes and took 148 both the corresponding mutants from the KEIO collection for each and tested their sensitivity 149 to fosfomycin by growth on agar containing different concentrations of the drug (Figure 3). 150 BW25113 was inhibited by the MIC of the drug as expected, as were both leuO mutants 151 tested; this was as expected as the data predicted leuO over-expression was important for 152 survival. For all the other mutants one or both mutants tested showed significantly improved 153 growth compared to the parent strain. This was particularly evident for mutants in three 154 separate parts of the phn operon or the two pts mutants tested (Figure 3). Given the 155 indicated role for the phosphonate uptake system we also tested whether addition of an 156 exogenous phosphonate would impact fosfomycin sensitivity. We used etidronate, a small 157 bis-phosphonate and found in checkerboard assays with etidronate and fosfomycin a 158 consistent rescue of growth to 1-2 dilutions above the MIC of fosfomycin by addition of 400 159 mg/L of etidronate. Traditional TraDIS experiments have not been able to assay essential genes which are often 177 the targets for antibiotics, TraDIS-Xpress was able to clearly show a key role for MurA in 178 fosfomycin sensitivity with very strong enrichment of mutants that were positioned upstream 179 of this gene and in the same orientationthese will over-express murA helping to saturate 180 fosfomycin and allow escape of its inhibition of peptidoglycan biosynthesis. 181 In addition to genes known to be involved in fosfomycin resistance a wider set of new loci 182 were identified as contribution to sensitivity to the drug (table 1) The phosphonate uptake and metabolism system is made up of a 14 gene operon 195 containing an ABC transporter, phosphonate lyase complex (able to break the P-C 196 phosphonate bond) and regulatory genes. Fosfomycin contains a phosphonate moiety and 197 so is a plausible substrate for this system. The TraDIS-Xpress data showed a very strong 198 signal for enrichment of mutants inactivating components of this system with selection for 199 inactivation of all genes at sub-MIC exposures and the ABC transporter and regulator above 200 the MIC (figure 2). This suggests inactivation of the ABC importer provides a strong fitness 201 advantage in the presence of fosfomycin suggesting this is another, unrecognised route of 202 import for the drug. This system has been suggested to be cryptic in K-12 with expression 203 being recovered after a phase variation mediated alteration of the copy number of a repeat 204 element within phnE. This is however downstream of the start of the operon with the first 205 components of the ABC transporter upstream of this site so their expression would not be 206 influenced by this site in any case. The TraDIS-Xpress data does strongly suggest a 207 biological benefit from inactivation of most genes in the operon at some fosfomycin 208 conditions, this challenges the notion that this system is cryptic in these conditions in K-12. 209 Mutants in three parts of the operon all demonstrated a growth benefit in the presence of 210 fosfomycin, supporting a role for inactivation of the system. Addition of the bis-phosphonate, 211 etidronate did result in limited (1-2 fold MIC increase) rescue of growth in the presence of 212 fosfomycinthis data is supportive of the phosphonate importer being an additional route of 213 entry into the cell for fosfomycin. 214 The potential for phosphonate moieties to mediate import of molecules into the cell is 215 intriguing and may have potential utility, a major challenge in therapy is getting drugs into 216 cells and new routes to modify molecules and promote their uptake are likely to enhance 217 efficacy of drugs. It has become clear in recent years that most drugs cross the inner 218 membrane by active import rather than passive diffusion (27) and identifying side groups 219 which may improve uptake by changing importer specificity is important. 220 Similarly, the phosphate uptake system, PstBCAS was also identified as being involved in 221 fosfomycin sensitivity with inactivation of the system proving beneficial in the presence of the 222 drug ( figure 2). These data suggest that multiple importers, including novel systems are 223 involved in fosfomycin sensitivity, testing of defined mutants in both the Phn and Pst systems 224 confirmed a phenotype with the mutants able to grow above the MIC of the drug (figure 3). 225 One major worry with some drugs is selection of cross-resistance to other agents, this is 226 often mediated by generic mechanisms of resistance including multidrug efflux and or porins 227 where expression changes can influence accumulation of many drugs (28). There was not a 228 strong signal for these pathways after fosfomycin exposure which suggest the major 229 mechanisms of fosfomycin resistance (figure4) are likely to be relatively specific and not 230 influence other classes of agent. This is potentially important as, whilst fosfomycin resistance 231 is not hard to select a key feature of this drug is its activity against strains resistant to other 232 drugs, in particular various beta-lactams. 233 Taken together, these data show a genome-wide analysis of genes involved in sensitivity to 234 fosfomycin identifies new loci which are important for determining sensitivity to the drug, 235 suggest important roles for two transport systems not previously implicated in fosfomycin 236 resistance as well as providing new information about potential routes for drug entry into 237

TraDIS-Xpress library 240
We recently described the construction of a high density TraDIS-Xpress library in E. coli 241 BW25113 which was used in this work (29). The transposons (a mini-Tn5 transposon coding 242 for kanamycin resistance (aph(3')-Ia)) used contain an outward facing tac promoter 3' to the 243 kanamycin cassette which is inducible by IPTG allowing over-expression or repression of 244 genes (depending on insert orientation) as well as traditional inactivation. This allows the 245 roles of essential genes in a stress to be analysed based on expression changes; 246 traditionally these loci have been cryptic in TraDIS experiments as insertions within them are 247 lethal. 248

Fosfomycin exposure conditions and TraDIS-Xpress sequencing 249
The minimum inhibitory concentration (MIC) of fosfomycin against BW25113 was 250 determined using microbroth dilution in LB broth, the same growth medium that was used for Tnp-i5 oligonucleotides being used instead of i5 index primers, and 28 PCR cycles. The 261 resulting DNA was size selected to purify fragments between 300bp-500bp and sequenced 262 on a NEXTSeq 500 sequencing machine using a NextSeq 500/550 High Output v2 kit (75 263 cycles). 264

Bioinformatics 265
Results were analysed using the AlbaTraDIS tool (version 0.0.5) which we developed for 266 TraDIS-Xpress analysis and recently described (29). Briefly, sequence reads were mapped 267 against the BW25113 reference genome (CP009273), aligned and insertion plots created. 268 The patterns of inserts were compared between fosfomycin exposed and control conditions,

Validation experiments 281
A total of 18 mutants were selected from the KEIO library to validate predictions about 282 sensitivity to fosfomycin made from by TraDIS-Xpress (33). These included genes in the 283 phosphonate uptake and metabolism and phosphorous import systems identified as major 284 contributors to fosfomycin sensitivity as well as a set of randomly selected control genes not 285 expected to have any impact on fosfomycin sensitivity. Both replicate mutants present in the 286 KEIO collection for each gene were independently analysed. Mutants were tested for their 287 sensitivity to fosfomycin by MIC determination. All experiments were duplicated (giving at 288 least four datasets for each gene; two repeats from each of the two mutant alleles of each 289 gene). BW25113 was included in all experiments as a control. 290

Data access 291
All sequence data have been deposited with EBI under project accession number 292

Funding 296
The author(s) gratefully acknowledge the support of the Biotechnology and Biological

Disclosure declaration 302
No authors have any conflicts to declare. 303 The funders had no role in study design, data collection and analysis, decision to publish, or 304 preparation of the manuscript.   The bottom of each panel illustrates the genomic context and the panels above illustrate the 400 mapped reads, red bars indicate reads orientated left-to-right and blue bars the opposite. 401 The height of each bar reflects abundance of each insert. 402