Fosfomycin, a phosphonic acid derivative discovered in 1969, has become the first-choice antibiotic for the ‘single-dose’ oral treatment of uncomplicated urinary tract infections (UTIs) (Falagas et al., 2016). It is a bactericidal antibiotic with a broad spectrum of activity that interferes with the first step of peptidoglycan synthesis in both Gram-positive and Gram-negative bacteria (Castaneda-Garcia et al., 2013; Falagas et al., 2016). As a phosphoenolpyruvate analog, fosfomycin inhibits the cytosolic UDP-N-acetylglucosamine enolpyruvyltransferase (also named MurA) by covalent binding to key residue C115 of the enzyme active site, preventing the formation of N-acetylmuramic acid (Kahan et al., 1974). This low-molecular-weight antibiotic enters into the bacterial cell via two transport uptake systems: the glycerol-3-phosphate permease (encoded by glpT) constitutively expressed and, the hexose phosphate uptake transporter (encoded by uhpT) inducible by extracellular glucose-6-phosphate (G6P) (Castaneda-Garcia et al., 2013). While transcription of glpT and uhpT is regulated by glpR and uhpABC, respectively, their expression also requires high levels of cyclic AMP (cAMP) combined with, as a complex, the cAMP receptor protein (CRP) (Castaneda-Garcia et al., 2013). cAMP levels depend on the activity of CyaA adenyl cyclase and are regulated by the phosphotransferase enzyme PtsI (Castaneda-Garcia et al., 2013).

Despite its widespread clinical use for many years in several countries, the prevalence of fosfomycin resistance is still low among E. coli clinical isolates, usually below 3% (4,5). Concerning multi-drug-resistant (MDR) isolates as ESBL-producing E. coli, levels of susceptibility to fosfomycin remain as high as 80% (Falagas et al., 2016, 2019; Aghamali et al., 2019). By contrast, the selection of fosfomycin-resistant mutants is much easier under in vitro conditions at high mutation frequencies (ca. 10^–8–10^–7) (Karageorgopoulos et al., 2012). This paradox is partially due to a significant resistance-associated fitness cost with decrease in vitro rate and attenuated virulence in vivo (Marchese et al., 2003; Nilsson et al., 2003; Pourbaix et al., 2017), and higher fosfomycin activity under urinary tract physiological conditions (i.e., urine acidification and anaerobiosis counterbalanced by negligible amounts of urinary G6P) that enhance expression of GlpT and UhpT (Martin-Gutierrez et al., 2018; Pourbaix et al., 2019).

Due to the unique mechanism of action of fosfomycin, there are no cross-resistances with other antibacterial agents (Falagas et al., 2016; Silver, 2017). However, three specific mechanisms of fosfomycin resistance were described in E. coli: impaired drug uptake, enzymatic drug inactivation and target modification (Cattoir and Guérin, 2018). Reduced drug uptake is the most frequent resistance mechanism for in in vitro mutants and clinical isolates. It results from chromosomal mutations that alter the function or expression of GlpT and/or UhpT transporters. These mutations (mutations, insertions, deletions) can arise either in structural genes (i.e., glpT and uhpT) or in genes coding for regulators (i.e., uhpA, cyaA, and ptsI) (Castaneda-Garcia et al., 2013; Silver, 2017). More recently, there is the emergence of plasmid-mediated metallo-dependent enzymes (including FosA, FosB, and FosX) that inactivate the drug, of which FosA3 is, by far, the most frequently variant in E. coli (Yang et al., 2019). Much more uncommon, fosfomycin resistance can be mediated by qualitative and/or quantitative modifications of MurA (Silver, 2017).

The aim of this study was to (1) investigate the genetic basis of fosfomycin resistance in E. coli mutants selected in vitro that had no mutations in genes previously reported to be involved in resistance (i.e., glpT, uhpT, uhpA, murA, cyaA, and ptsI), (2) demonstrate experimentally the role of novel mutations identified in four different genes, and (3) determine their prevalence among a collection fosfomycin-resistant E. coli clinical isolates recently collected in France.

The Uhp hexose phosphate transport pathway and its regulation are well described in E. coli (Kadner, 1973; Kadner and Winkler, 1973; Kadner and Shattuck-Eidens, 1983; Weston and Kadner, 1987, 1988; Island et al., 1992; Island and Kadner, 1993; Wright et al., 2000; Verhamme et al., 2001, 2002). UhpT is a member of the Major Facilitator Superfamily (MFS) containing 12 transmembrane protein segments, and it is responsible for the accumulation of G6P into the bacterial cells. The UhpT system is tightly controlled by the UhpABC phosphorelay system UhpABC, which is necessary for high-level expression of uhpT. UhpC is also an MFS member that shares approximately 30% amino acid sequence identity with UhpT. UhpC is a membrane-bound protein that senses external G6P in the periplasm and interacts with UhpB, stimulating its kinase activity. UhpB is a membrane-bound histidine kinase (HK) in a two-component system that possesses eight predicted transmembrane helices and a C-terminal cytoplasmic domain containing the conserved sequence elements common to HK proteins (i.e., the H-box around the phosphorylated histidine, the N-box, and the G-box comprising the ATP-binding and phosphate transfer region) (Parkinson and Kofoid, 1992). Upon interaction with UhpC, UhpB autophosphorylates the conserved histidine residue (His313), with subsequent phosphorylation at Asp54 of its cognate response regulator UhpA. Phosphorylated UhpA increases the affinity for its specific DNA binding sites, hence promoting transcription of uphT.

Many mutants defective in the hexose phosphate transport were isolated between 1970s and 1990s, but shortcomings can be found in these old studies, such as the imprecise position of the mutation/deletion/insertion due to the poor annotation of the uhp region sequence, the absence of determination of fosfomycin MICs, and the ‘artificial nature’ of many in vitro mutants that were obtained by transposon insertion (Mu, Tn10), or a resistance cassette (Kadner, 1973; Kadner and Shattuck-Eidens, 1983; Weston and Kadner, 1987, 1988; Island et al., 1992). Also, mutations/insertions can have different impacts on fosfomycin susceptibility since some of them do not impair uhpT expression and others confer constitutive expression (Weston and Kadner, 1987; Island and Kadner, 1993). Deleted mutants with a kanamycin resistance cassette in uhpA, uhpB, or uhpC from the E. coli BW25113 parental strain only conferred a modest increase in fosfomycin MICs to 8, 8, and 4 mg/L (Castaneda-Garcia et al., 2009), respectively, which is different from our findings. Altogether, it suggests that ‘artificial’ insertional mutants do not represent systematically how bacteria develop fosfomycin resistance.

Unexpectedly, we found here novel mutations in uhpB and uhpC in mutants, which are not often detected in fosfomycin-resistant clinical isolates. Indeed, fosfomycin resistance in E. coli clinical isolates is usually due to chromosomal mutations in uhpT, uhpA, glpT, murA, cyaA, and ptsI genes (Nilsson et al., 2003; Oteo et al., 2009; Takahata et al., 2010; Li et al., 2015b; Tseng et al., 2015; Ohkoshi et al., 2017; Lucas et al., 2018; Seok et al., 2020), and little is known about the involvement of mutations in other genes, especially those in uhpB and uhpC that have been exceptionally reported (Castaneda-Garcia et al., 2013).

Recently, mutations in uhpB or uhpC were described in E. coli BW25133-derived laboratory mutants ΔcyaA, ΔglpT-cyaA, ΔglpT-ptsI, and ΔptsI-cyaA recovered in vitro after time-kill experiments with fosfomycin (Ballestero-Tellez et al., 2017) and in two E. coli clinical isolates (Martin-Gutierrez et al., 2018). All the mutants were resistant to high levels to fosfomycin (MICs > 1,024 mg/L) and possessed the following mutations one or two mutations in uhpB (48del, W181^∗, L255^∗, and Q262^∗) and uhpC (T27^∗, T72P and 541_548del) (Ballestero-Tellez et al., 2017). In the two clinical isolates, one uhpB mutation (D205A) and three uhpC mutations (Y18H, G282D, T435A) were found in the first while two uhpC mutations (I14M, Q17Y) were found in the second (Martin-Gutierrez et al., 2018). We found here two mutations at the exact same position (D205A in UhpB and T72 in UhpC) of these previous studies (Figure 3), which is in favor of their role in fosfomycin resistance. In our study, uhpB mutations were distributed all along the 500-amino-acid-long protein in either periplasmic (n = 3), transmembrane (n = 5), or cytoplasmic (n = 6) regions, including one in the autophosphorylation H-box (H313Y) and another in the conserved G-box (G469R) that part of the ATP-binding domain (Figure 3). Concerning uhpC mutations, they were more frequently detected within the transmembrane segments (8/15) than into the cytoplasm (n = 6) or periplasm (n = 1) portions of the 439-amino-acid protein, suggesting that it could impair external G6P sensing through the membrane (Figure 3). Among fosfomycin-resistant clinical isolates, five had a full deletion of the uhp region (uhpA-uhpB-uhpC-uhpT), as reported (Lucas et al., 2018).

A majority of clinical isolates harbored uhpB and uhpC mutations that were widely distributed over the protein sequences. It is likely that all these mutations impact fosfomycin susceptibility differently, as described for insertion mutants exhibiting variable Uhp phenotypes (Weston and Kadner, 1988; Island et al., 1992; Island and Kadner, 1993). We also demonstrated that deletions and mutations in uhpB and uhpC were responsible for an absence of induction by G6P of uhpT expression, as described in several ΔuhpA, ΔuhpB, and ΔuhpC laboratory mutants and one clinical isolate with a truncated UhpA (Weston and Kadner, 1988; Island et al., 1992; Island and Kadner, 1993; Wright et al., 2000; Lucas et al., 2018). This confirms the role of UhpB and UhpC as G6P-response regulators required for the induction of uhpT expression. In addition, it appears that the mutation in uhpC (leading to the substitution F384L) also alters in vitro bacterial growth rate in LB and NB (regardless the pH) but not in urine, suggesting that it may occur in vivo.

Besides uhpB and uhpC mutations, two novel mutations were also identified in the two-step in vitro mutants. The first mutation occurred in galU that codes for a 302-amino-acid-long protein named UTP-glucose-1-phosphate uridylyltransferase, which catalyzes synthesis of UDP-D-glucose from UTP and α-D-glucose 1-phosphate (Weissborn et al., 1994). It is a central precursor for synthesis of cell surface carbohydrates, colanic acid, trehalose, cellulose, capsule- and membrane-derived oligosaccharides, and also has a major role in galactose metabolism (Ebrecht et al., 2015). Then, the deletion of galU has many consequences on different carbon metabolic pathways: for instance, they are unable to ferment galactose and fail to incorporate glucose and galactose into bacterial cell membranes, resulting in the incomplete synthesis of lipopolysaccharides (Fukasawa et al., 1962; Sundararajan et al., 1962). Also, the absence of galU leads to a reduced level of TolC into the outer membrane (Sharma et al., 2009), which might be related to antibiotic susceptibility. Here, we identified a non-synonymous mutation (R282V) in the C-terminal region of GalU that is outside the enzyme active site formed by the key residues T20, R21, and K202 (Ebrecht et al., 2015). Then, it is difficult to explain the implication of R282V mutation into the fosfomycin MIC two-fold increase. Note that it seems that this mutation also impacts on bacterial fitness when grown in acidic pH or in urine, suggesting that it may be difficult to develop in vivo.

The second mutation appeared in lon coding for an ATP-dependent serine protease that plays a major role in protein quality control, degrading incorrect proteins, and has an important role into many biological processes in bacteria (Tsilibaris et al., 2006). It degrades abnormal and misfolded proteins, but has also regulatory proteins as substrates, such as MarA and SoxS (Griffith et al., 2004). Here, we identified a lon mutation giving rise to a premature stop codon (Q558^∗), and then a truncated protein, probably not functional. Indeed, with a length of 784 amino acids in E. coli, a large part of the C-terminal domain is lacking (Amerik et al., 1991). Therefore, we can assume that this truncated protein is inactive since the Ser679-Lys722 catalytic dyad is absent (Botos et al., 2004). Interestingly, it was demonstrated that mutations in lon were implicated in the development of multiple antibiotic resistance phenotype related to the efflux pump system AcrAB-TolC, and to the OmpF porin (Nicoloff et al., 2006, 2007; Duval et al., 2009; Nicoloff and Andersson, 2013; Bhaskarla et al., 2016). MarA, SoxS, and Rob, positively control the expression of acrAB, tolC, and micF, and micF regulatory RNA post-transcriptionally represses the translation of ompF mRNA (Li et al., 2015a). In a lon mutant, the accumulation of MarA and SoxS could enhance the micF-mediated inhibition of the OmpF production, that could impact on fosfomycin activity since OmpF can facilitate the spontaneous diffusion of the antibiotic across the outer membrane (Golla et al., 2019).

In conclusion, we demonstrated here experimentally the role of novel mutations in four genes implicated in fosfomycin resistance, and the prevalence of uhpB and uhpC mutations among fosfomycin-resistant E. coli clinical isolates.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

VC and FG conceptualized the study. VC, AP, MM, FC, VL, BF, and FG contributed to methodology. VC and FG provided the formal analysis and visualization. VC, AP, MM, FC, VL, BF, BF, and FG carried out the investigation. VC, BF, and FG were responsible for the resources. VC, BF, BF, and FG wrote the manuscript. All authors read and approved the manuscript.

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.