The Role of fosA in Challenges with Fosfomycin Susceptibility Testing of Multispecies Klebsiella pneumoniae Carbapenemase-Producing Clinical Isolates

With multidrug-resistant (MDR) Enterobacterales on the rise, a nontoxic antimicrobial agent with a unique mechanism of action such as fosfomycin seems attractive. However, establishing accurate fosfomycin susceptibility testing for non-Escherichia coli isolates in a clinical microbiology laboratory remains problematic. We evaluated fosfomycin susceptibility by multiple methods with 96 KPC-producing clinical isolates of multiple strains and species collected at a single center between 2008 and 2016.

multidrug-resistant (MDR) Gram-negative pathogens. Fosfomycin, which was originally discovered in 1969, has been shown to have in vitro activity against CPE (3). In the United States, the oral formulation is available for the treatment of uncomplicated urinary tract infections due to susceptible strains of Escherichia coli and Enterococcus faecalis. Outside the United States, the intravenous (i.v.) formulation is approved and available for the management of systemic infections (4). Zavante Pharmaceuticals received fast-track designation from the FDA in 2015 for i.v. fosfomycin and has completed phase III clinical trials for the U.S. market (5).
Fosfomycin is a bactericidal antibiotic that binds to the cysteine residue of UDP-Nacetylglucosamine enolpyruvyl transferase (MurA) and inhibits peptidoglycan biosynthesis (6). Fosfomycin has activity against a range of bacterial pathogens, including highly drug-resistant Enterobacteriaceae (6). Fosfomycin resistance in Enterobacterales has been primarily driven by mutations in the glpT and uhpT genes, preventing active transport of fosfomycin into the cell (7). These mutations are, however, thought to be associated with a fitness cost in E. coli and are thus unstable (8,9). The other major mechanism of resistance is hydrolysis of the drug via diverse Fos enzymes: FosA (FosA2, FosA3, FosA4, FosA5, FosA6, and FosA7), FosB, and FosX are metalloenzymes, whereas FosC is a serine enzyme (10). FosA was originally discovered on a transposon, Tn2921, in a Serratia marcescens plasmid and catalyzes the addition of glutathione to fosfomycin, rendering the drug ineffective (11). Transmissible fosA is of most concern in Enterobacterales, and plasmid-mediated fosA3 has been increasingly identified in E. coli in Europe (12). A recent evaluation of Fos enzymes in non-E. coli Enterobacterales demonstrated that different fosA variants are chromosomally located in a speciesspecific manner (11). Lastly, MurA target site alteration can also confer fosfomycin resistance. Amino acid substitutions in MurA, most notably Asp369Asn and Leu370lle, have been responsible for fosfomycin resistance (13).
Susceptibility testing of fosfomycin for non-E. coli Enterobacterales is difficult for clinical microbiology labs (14)(15)(16). Both the CLSI and EUCAST specifically recommend against the use of broth microdilution methods, which likely impacts the inaccuracies with most automated susceptibility testing platforms for E. coli or Klebsiella pneumoniae (17,18). Agar dilution is considered the reference method and is endorsed by the EUCAST; however, it is difficult to execute routinely in a clinical microbiology laboratory. Kirby-Bauer disk diffusion (DD) and Etests are more attractive options, as they can be performed easily in a clinical laboratory, but colonies often grow within the zones of inhibition, making interpretation difficult (19). Attempting to change the zone cutoff to better align with agar dilution has not proved successful with non-E. coli Enterobacterales (15). With agar dilution as the only accurate method for non-E. coli Enterobacterales, we aimed to characterize some of the molecular mechanisms (by whole-genome sequencing [WGS] for a subset of isolates) that may be contributing to the inaccuracies with these diffusion methods, using a diverse set of clinical, carbapenemase-producing strains and agar dilution-based reference phenotyping.

MATERIALS AND METHODS
Ninety-six retrospective samples of Klebsiella pneumoniae carbapenemase (KPC)-producing Gammaproteobacteria isolates were selected from those collected at the University of Virginia Health System since August 2008. Isolates were chosen to represent diverse species and strains for which Illumina sequence data were available. Species identification had been performed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) (Vitek-MS, Vitek-2; bioMérieux) and all isolates were bla KPC PCR positive as previously described (20).
Etest was performed using glucose-6-phosphate (G6P)-supplemented fosfomycin Etest strips (bioMérieux) according to the manufacturer's instructions. DD was performed with a 200-g fosfomycin disk with 50 g of G6P (Becton and Dickinson, Franklin Lakes, NJ) on Mueller-Hinton agar as per CLSI guidelines (21). Agar dilution was performed using G6P-supplemented (25 g/ml) Mueller-Hinton agar with fosfomycin concentrations ranging from 0.5 to 1,024 g/ml per CLSI methods (22). A 0.5 McFarland inoculum for each isolate was placed in triplicate on the agar, placed in an incubator at 37°C for 16 to 20 h, and then interpreted. Plates were prepared without isolates at each concentration to serve as the control.
Per the package insert for interpretation of Etest, the crossing point of the ellipse was used to identify the MIC where colonies within the zone of the ellipse were accounted for if 5 colonies were present within 3 mm of the strip within the zone (bioMérieux). DD diameters were measured as the shortest distance between 2 separate colonies (17,21). For agar dilution, the median of the three interpreted MICs was recorded as the result. All fosfomycin susceptibilities were interpreted according to CLSI breakpoints for E. coli urinary isolates, as there are no breakpoints available for non-E. coli Enterobacterales (17).
With agar dilution as the reference method, essential agreement for Etest was defined as MIC variation within 1 dilution. Categorical agreement was defined as matching susceptible/intermediate/ resistant interpretation criteria for the two respective tests, as per CLSI guidelines for E. coli urinary isolates. Falsely susceptible results were deemed to be very major errors and falsely resistant results to be major errors. All other disagreements were deemed minor errors. Chi-square and Fisher's exact tests were used to compare rates of nonsusceptibility and categorical agreement. The Mann-Whitney test was utilized for statistical analysis of MIC distributions and DD zone diameter changes.
Disk potentiation testing with sodium phosphonoformate (PPF) was performed as described by Ito et al. to specifically evaluate the activity of fosA enzymes and the impact on susceptibility testing with the disk diffusion method (11). Cultures of each isolate were plated on Mueller-Hinton agar with 1 mg of a 50-mg/ml sodium phosphonoformate (Sigma-Aldrich) solution added to a 200-g fosfomycin disk supplemented with 50 g of G6P. The plates were incubated overnight at 37°C and the inhibition zone was recorded. These inhibition zones were then compared the DD inhibition zones of fosfomycin without supplementation of PPF. For each isolate tested, a blank disk with PPF was also placed on the agar plate to serve as a negative control.
Molecular mechanisms of resistance to fosfomycin were investigated in a subset of isolates previously whole-genome sequenced by Ilumina Sequencing (HiSeq 2000) as previously described (20). The quality-filtered short reads were de novo assembled using SPAdes v3.11 (23), and the contigs were screened for fos resistance genes using NCBI's AMRFiner (identity Ն 0.9; coverage Ն 0.5) (24). For isolates where AMRFinder failed to detect fos genes, we screened the contigs using a custom HMM model built from distinct fosA protein sequences published by Ito et al., with an E value threshold of 1eϪ20 (11,25).

RESULTS
Fosfomycin susceptibility across species. Ninety-six bla KPC -positive isolates across 12 species were included in the study (see Table 1 for species breakdown). Eighty-eight of the 96 isolates had undergone whole-genome sequencing (WGS). The MIC 50 across all isolates was 8 g/ml and the MIC 90 was 128 g/ml by agar dilution. Using the 2019 CLSI breakpoints (Յ64 g/ml indicates susceptibility), 84 of 96 isolates (88%) were susceptible, 11 of 96 (11%) were resistant, and 1 of 96 (1%) was classified as intermediate. The MIC distributions by agar dilution are shown in Fig. 1.
Diffusion method performance. Sixty of 96 isolates (63%) were susceptible by Etest and 65/96 (68%) were susceptible by DD. Categorical agreement of Etest with agar dilution occurred in 69/96 isolates (72%), with 1 very major error, 16 major errors, and 10 minor errors. Essential agreement occurred in 55 of 96 isolates (57%) overall and in 4 of 5 (80%) E. coli isolates. Categorical agreement of DD with agar dilution occurred in 72/96 isolates (75%), with 2 very major errors, 14 major errors, and 8 minor errors. Of note, when testing the non-E. coli species, colonies within the zone were frequently present, making interpretation challenging but adhered to package insert and CLSI guidance for Etest and DD, respectively (21). Figure 2 shows a comparison of DD and Etest discordance in categorical agreement with agar dilution.
Eleven of the 70 isolates (16%) were nonsusceptible to fosfomycin. Isolates carrying the fosA gene were associated with a higher MIC distribution than those without the gene (P Յ 0.00001) but did not differ in rates of nonsusceptibility (P ϭ 0.26). These results are shown in Table 2.
FosA inhibition effect on susceptibility testing by diffusion method. Disk potentiation testing with PPF was performed on all 96 isolates. Categorical agreement with agar dilution was found in 72/96 (75%) of isolates prior to the addition of PPF and in 81/96 (84%) isolates after the addition of PPF (P ϭ 0.11). In the 88 isolates with WGS data available, rates of categorical agreement with and without PPF were compared. In isolates that were negative for the fosA gene, categorical agreement was found in 18/18 (100%) and 17/18 (94%) of isolates before and after the addition of PPF, respectively (P ϭ 1). In isolates that carried the fosA gene, categorical agreement was found in 49/70 (70%) and 59/70 (84%) before and after the addition of PPF, respectively (P ϭ 0.04). When specifically isolating all nonsusceptible isolates by DD, categorical agreement was found in 9/31 (29%) isolates and 20/31 (65%) before and after the addition of PPF, respectively (P ϭ 0.005). Results are shown in Tables 3 and 4. The presence of PPF not only increased the zone size but also decreased the presence of colonies within the zone in DD testing for the fosA-positive isolates (Fig. 2). The control disk with PPF alone indicated no antibacterial activity, with no isolates demonstrating a zone of inhibition.

DISCUSSION
We demonstrate that fosfomycin susceptibility testing by routinely used laboratory diffusion-based methods (Etest and DD) largely overcalls resistance compared to agar dilution as a gold standard (Table 1). Fosfomycin susceptibility appears to be influenced by the presence of fosA among these KPC-producing Enterobacterales isolates. This is highly relevant to the clinical microbiologist who is frequently fielding requests for fosfomycin susceptibility testing for non-E. coli Enterobacterales. A prior study by Kaase et al. tested 107 carbapenem-nonsusceptible Enterobacteriaceae isolates, of which 80 produced various carbapenemases (KPC, VIM, NDM, and OXA-48), and they found 81% of isolates to be susceptible to fosfomycin, with a MIC of Յ64 g/ml by agar dilution. This study also found similar issues of discordance with diffusion testing methods (15), as was also seen in a study by Hirsch et al. (14).
Fosfomycin has been promoted as a useful, safe medication for the treatment of  (2,3,7,26,27), but susceptibility testing by agar dilution is practically difficult for a clinical microbiology lab. The CLSI cutoffs for E. coli were applied for Etest and DD interpretation for all species tested in this experiment, which requires accounting for scattered colonies within the ellipse per the package insert or zone per the CLSI (17). Fosfomycin susceptibility testing for E. coli was reviewed by the CLSI in 2018. At that time the recommendation that the susceptibility cutoffs apply only to E. coli was strengthened, and since scattered colonies are rare within the zone for this species, the practice of measuring the zone from the innermost colonies was upheld (17). This differs from the new EUCAST guidelines for fosfomycin and E. coli, which suggest ignoring scattered colonies within diffusion-based inhibition zones and utilizing agar dilution approaches for non-E.coli Enterobacterales (28). This decision for the former was based on the findings that subcolonies of E. coli within inhibition zones are rare (Ͻ1% of isolates) and largely less fit, with channel or transporter mutations (8,9). We postulate based on our findings that the colonies within the inhibition zone seen more frequently with some non-E. coli species may be driven by the presence of a chromosomal fosA gene rather than channel or transporter mutations. Thus, the advice to ignore the colonies in the zone in E. coli may not apply to non-E. coli Enterobacterales, and clinical microbiologists should proceed with caution. The discordance among commercially available DD and Etest with agar dilution we observed was largely due to bacterial colonies that grew within the zone of inhibition. Based on the change in zone size with the addition of a FosA inhibitor as well as the work of others demonstrating the activity of chromosomally expressed FosA, it may unwise to ignore subcolonies, as the clinical implications of this finding remain unknown (29).
In our subset of E. coli isolates, categorical agreement was found in 5 of 5 (100%) isolates. All E. coli isolates were susceptible to fosfomycin by both DD and agar dilution, and no colonies were observed within the zones of inhibition. Although our numbers are small, this is consistent with other reports (8).
fosA (alleles 1 to 7) was identified in the majority of the clinical isolates in this study. FosA was present in all Klebsiella species isolates, with a large portion harboring the fosA6 allele. fosA6 was first reported in 2016, from an extended-spectrum-␤-lactamase-(ESBL)-producing, fosfomycin-resistant E. coli strain in Pennsylvania. It shared 96% identity with fosA5 and 79% identity with fosA3 but was located on a plasmid, unlike the chromosomally borne fosA in K. pneumoniae (30). It has been suggested that fosA6 was mobilized from the chromosome of K. pneumoniae to an E. coli plasmid (30). However, in our study, no E. coli harbored fosA6, but rather one isolate harbored fosA7, which has been described to be on the chromosome of Salmonella enterica (31). We postulate that this gene may have been acquired via plasmid transfer, with Salmonella enterica serving as the reservoir for this allele. No fosC was detected, as expected, as it is a gene found most commonly within Pseudomonas spp., which were not included in this study.
In this subset of isolates, the presence of fosA resulted in a trend toward higher MIC values than for isolates not harboring the gene. Despite the higher distribution of MIC values, there was no statistical difference in fosfomycin susceptibility when determined by the agar dilution method. However, the E. coli breakpoints used in this study are based on obtainable urinary concentrations. The impact of increased MIC distributions on fosfomycin susceptibility in nonurinary sources of infection is unclear, as MIC breakpoints will likely be lower.
Disk potentiation testing with PPF was performed to specifically evaluate the activity of fosA enzymes and the impact on susceptibility testing with DD. The addition of PPF significantly increased zone diameter size and, subsequently, improved categorical agreement of disk diffusion with agar dilution, particularly in fosA-positive isolates, in which most major errors were eliminated. These improvements were largely due to the elimination of subcolonies within the zone of inhibition, as illustrated for isolate CAV 1217 (Fig. 2). As expected, there was no statistical change in categorical agreement in isolates that did not harbor a fosA allele. The results of the disk potentiation testing indicate that fosA impacts fosfomycin activity and limits convenient diffusion-based susceptibility testing. Some isolates had substantial zone diameter increases (6 to 8 mm) after the addition of PPF, yet this did not alter the susceptibility interpretation. This is likely due to alternative mechanisms of fosfomycin resistance, such as transporter mutations or MurA mutations. An alternative explanation is that certain alleles of fosA may have the ability to overcome the inhibition of PPF that was added to the disk.
Lastly, our data suggest that the addition of PPF may have a synergistic effect with fosfomycin against fosA-positive organisms. This is corroborated by a recently published study that found significant MIC reductions and restored fosfomycin susceptibility in fosA-positive Gram-negative organisms (29). PPF is available as the antiviral foscarnet but would likely be unattractive as an adjuvant therapy due to toxicity.
Our study is limited by a small sample size of bla KPC -positive multidrug-resistant isolates collected from a single center. We are unable to make inferences on non-KPC-positive isolates, as they were not included in this study. However, the clinical utility of fosfomycin is primarily against MDR isolates, and our study highlights the difficulty in accurately providing fosfomycin antimicrobial susceptibility testing (AST) for these organisms. A further limitation is the lack of exploration of other molecular mechanisms of resistance, which were not evaluated in all isolates. Our study also lacks outcome data, and therefore, we can make no conclusions on the clinical implications of fosfomycin susceptibility testing results.
In conclusion, fosfomycin appears to have reliable in vitro activity against KPCproducing Gram-negative organisms by agar dilution. However, methods readily available in a clinical microbiology laboratory, Etest and DD, generate frequent major errors, with the presence of fosA impacting the interpretation of these diffusion-based methods. Caution is advised when interpreting and releasing AST results derived from diffusion-based methods for non-E. coli Enterobacterales. Regardless, further research is needed to establish correlations between antimicrobial susceptibility testing, fosA presence, and clinical outcomes.
tance at University of Oxford in partnership with Public Health England (PHE). This work is supported by NIHR Oxford Biomedical Research Centre.