We investigated the occurrence and molecular features of all fosfomycin-resistant E. coli isolated from hospitalized patients during a 12-month period (August 2014–July 2015). We performed bacterial identification by using VITEK 2 (BioMérieux, Marcy l’Etoile, France) and tested antibiotic susceptibility by using the disk diffusion method in accordance with 2016 Comité de l’Antibiogramme de la Société Française de Microbiologie/European Committee on Antimicrobial Susceptibility Testing guidelines (http://www.sfm-microbiologie.org/UserFiles/files/casfm/CASFM2016_V1_0_FEVRIER.pdf). We screened for fosfomycin resistance by using a 200-µg disk with a diameter cutoff of <13 mm. We determined MIC by using the Etest method with Muller-Hinton agar containing 25 mg/L G6P.

Figure

Figure. Inhibition of FosA-mediated fosfomycin resistance by phosphonoformiate. A modified Kirby-Bauer disk diffusion susceptibility assay was performed. In brief, a Mueller-Hinton agar plate was streaked with a 0.5 McFarland suspension of the...

Among 1,354 E. coli isolates tested, 12 (0.9%) showed confirmed resistance (MIC >128 mg/L). We explored the mechanism of fosfomycin resistance by growing these isolates for 48 hours at 35°C in M9 minimal medium agar supplemented with either G3P or G6P at 0.2% as a sole carbon source. Lack of growth showed impaired fosfomycin transport (12). Of the 12 isolates, 7 were double auxotrophic mutants with G6P and G3P (mean MIC 384 mg/L), 3 were auxotrophic only for G3P (mean MIC 597 mg/L), and 2 were capable of using both substrates and exhibited high-level resistance (MIC >1,024 mg/L). Paradoxically, single and double auxotroph strains had lower mean MICs than the 2 nonauxotroph strains. Because transport deficit could not account for the observed phenotype, we screened by PCR and sequenced genes coding enzymatic glutathione S-transferase variants fosA, fosA2, fosA3, fosA4, fosA5, and fosC2 (Table 1). In parallel, we screened inhibition of glutathione S-transferase activity by using FosA inhibitor phosphonoformiate (Figure) as described by Nakamura et al. (12). Results of these 2 tests were in agreement, with each detecting an ESBL-producing strain with enzymatic activity encoded by the fosA3 gene (MIC >1,024 mg/L). Neither strain had been previously reported in France.

We determined the prevalence of enzymatic resistance to fosfomycin in ESBL-producing strains isolated since 2012 by using the same 2 tests. Surprisingly, among 23 strains resistant to fosfomycin with no epidemiologic link, 7 additional FosA3­-producing and 1 FosA5-producing strains were detected, each with MIC >1,024 mg/L. Overall 83% of fosfomycin-resistant ESBL-producing E. coli with MIC >1,024 mg/L were FosA-positive. Auxotrophic tests showed that in addition of FosA production, fosfomycin transport was impaired in 6 strains. Chronologically, 2 strains were isolated in 2012, three in 2013, three in 2014, and two in 2015, meaning that FosA3-producing strains were present in France 6 years after the first isolation in Japan. The FosA5-producing strain was isolated from a clinical sample in France simultaneously with the original description of the strain in China (13). The sequencing of fosA5 showed 99% identity (with 96% coverage) with pHKU1, earlier described as an fosKP96-carrying IncN plasmid in Klebsiella pneumoniae (5).

Our sequencing of CTX-M genes showed that FosA3-producing strains were associated with CTX-M-15 (n = 5), CTX-M-55 (n = 3), and CTX-M-2 (n = 1), whereas the 1 FosA5-producing strain expressed CTX-M-14. We also conducted multilocus sequence typing and plasmid incompatibility group typing (14) (Table 2). These results show unambiguously that 8 strains of E. coli of different sequence types hosted 5 plasmid types that could be distinguished by their CTX-M variant and plasmid-incompatibility group types.

Nine out of 10 isolates yielded transconjugants in E. coli C600 (E. coli K12 derivative) or transformants in TOP10 (DH10B derivative) E. coli. All 9 of these isolates expressed high-level resistance to fosfomycin (MIC >1,024 mg/L), confirming that the observed resistance of the parent strain was indeed attributable to the fosA gene.

Although the prevalence of plasmid-mediated fosA3 genes in human clinical E. coli isolates has remained low in France since 2012, these genes are observed across numerous clones, sequence types, and molecular determinants and are always associated with ESBL CTX-M enzymes, suggesting multiple propagation events. Our results are consistent with FosA3-producing clinical strains previously isolated in Asia, which also co-express CTX-M enzymes. However, the CTX-M variant distribution between the strain in France and the strain in Asia are different, with CTX-M-15 having high prevalence in our collection. Medical records examination did not show a history of international travel in our patient population, and such a variety of fosfomycin-resistant E. coli lineages probably were not imported or transmitted. The broad use of oral fosfomycin has provided the opportunity to select for FosA producers. With the spread of CTX-M urinary tract infections in the community, the use of fosfomycin is likely to select for CTX-M–FosA co-producers and could lead to an increase of treatment failures with ESBL-producing organisms. Conversely, treatment of ESBL producers with fosfomycin should only be undertaken after testing for susceptibility because these ESBL producers can be linked to the same genetic determinant. Moreover, the indiscriminate use of the oral formulation in the community is jeopardizing the usefulness of this antimicrobial agent. While the world is bracing for an epidemic of infectious diseases bearing plasmid-mediated colistin resistance (15), a vast and ubiquitous reservoir for conjugative transmissible resistance to fosfomycin exists and can preclude its efficacy against extremely drug-resistant bacteria if the guidelines for the indiscriminate use of fosfomycin–trometamol are not urgently revised to safeguard this potent and well-tolerated agent. Because antimicrobial treatment of cystitis typically is motivated by concern for patient’s comfort, withholding treatment or the promotion of pivmecillinam as a first-line antimicrobial drug should seriously be considered.