Comparison of Agar Dilution to Broth Microdilution for Testing In Vitro Activity of Cefiderocol against Gram-Negative Bacilli

Cefiderocol (CFDC) is a siderophore cephalosporin with activity against Gram-negative bacterial species that are resistant to carbapenems and other drugs. The MICs of CFDC were determined for 610 Gram-negative bacilli, including 302 multinational Enterobacterales isolates with characterized mechanisms of beta-lactam resistance, 180 clinical isolates from the Mayo Clinic and Mayo Clinic Laboratories not characterized for specific resistance mechanisms, and 128 isolates with CFDC MICs of ≥8 μg/ml obtained from International Health Management Associates, Inc.

used as QC strains in each trial. The MIC was reported as the lowest concentration with no visible growth, a single colony, or a faint haze due to the inoculum (27).
Determination of ion concentration in MHA. The fluid phase of hydrated MHA was prepared according to the method described by Hawkey et al. (28). Melted 40-ml aliquots were frozen at Ϫ80°C and then thawed in a water bath at 80°C. The process was repeated, and MHA was centrifuged at 47,000 ϫ g for 10 min to pellet insoluble components of the medium. The clear supernatant was decanted, and the amount of iron was determined using an iron test kit according to the manufacturer's instructions (Visocolor HE Iron; Macherey-Nagel, Germany). Testing was performed twice.
Data analysis. The MICs required to inhibit 50 and 90% of organisms (MIC 50 and MIC 90 , respectively) were calculated. Essential agreement (EA) was assessed by calculating the percentage of isolates with MICs within 1 doubling dilution of that determined by BMD with ID-CAMHB (ID-BMD). Categorical agreement (CA) was assessed by calculating the percentage of isolates tested by AD that yielded the same categories as ID-BMD. A percentage of Ն90% was considered acceptable for EA and CA. Categorical results that were not congruent were categorized as follows: minor error (mE), major error (ME), and very major error (VME). Acceptable percentages of errors were Յ1.5% for VME, Յ3% for ME, and Յ7% for combined mE and ME (29). CLSI, the FDA, and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) have different breakpoints for CFDC determined by ID-BMD (Table 1); no breakpoints are defined for AD (16,26,30). Here, MICs of A. baumannii and S. maltophilia were interpreted according to investigational CLSI breakpoints (since there are no FDA or EUCAST breakpoints), and MICs of P.

RESULTS
The in vitro activity of CFDC as assessed by the different methods is summarized in Table 2, with analyses of each bacterial group by CLSI, FDA, and EUCAST breakpoints. For the 610 isolates tested, regardless of species, MICs obtained by BMD with standard CAMHB were, as expected, higher than those obtained by ID-BMD (Table S1), so we focused on comparison between AD and ID-BMD. Scattergrams showing MICs for ID-BMD and AD for P. aeruginosa, S. maltophilia, B. cepacia complex, A. baumannii and Enterobacterales are shown in Fig. 2; cumulative percentages of isolates of all groups and species inhibited at each CFDC concentration tested are shown in Fig. S1; and QC MICs for each experiment are shown in Table S2.
By AD, 83, 48, and 70% of P. aeruginosa isolates were susceptible to CFDC using investigational CLSI, FDA, and EUCAST breakpoints, respectively; by ID-BMD, 97, 89, and 97%, respectively, were susceptible to CFDC. When AD was compared to ID-BMD, there was 38% EA, with 86, 52, and 74% CA when investigational CLSI, FDA, and EUCAST breakpoints were applied, respectively. All susceptible isolates were isolates from the Mayo Clinic or Mayo Clinic Laboratories which were not specifically characterized genetically.
All 71 S. maltophilia isolates tested (from the Mayo Clinic and Mayo Clinic Laboratories) had MICs of Յ4 g/ml (investigational CLSI susceptible breakpoint), with MIC 90 values of 1 and 0.25 g/ml for AD and ID-BMD, respectively. There was 100% CA and 30% EA for AD versus ID-BMD.
B. cepacia complex isolates had MIC 50 and MIC 90 values of 0.06 and 1 g/ml by AD, and of 0.03 and 1 g/ml by ID-BMD, with 77% EA for AD versus ID-BMD. CA and error rates could not be calculated due to the absence of breakpoints for this species complex.
A. baumannii isolates were 45% (44/97) and 57% (55/97) susceptible to CFDC by AD and ID-BMD, respectively, according to the investigational CLSI breakpoints; EA and CA of AD versus ID-BMD were 32 and 76%, respectively, with 6% ME and 16% mE. No FDA and EUCAST breakpoints have been established for this species.
At Յ2 g/ml (FDA and EUCAST susceptible breakpoints), CFDC inhibited 53% and 67% of Enterobacterales isolates when MICs were determined with AD and ID-BMD, respectively (Table 2). MIC 50 and MIC 90 values determined by AD were 2 and 32 g/ml, and those determined by ID-BMD were 2 and 8 g/ml. Comparing AD versus ID-BMD, there was 62% EA and 61% CA, with 25% mE and 13% ME, applying FDA breakpoints (Table 3).
EA rates for Enterobacterales isolates harboring bla CTX-M , bla NDM , and bla KPC were 87, 67, and 63%, respectively, when AD was compared to ID-BMD; an acceptability criterion of Ն90% was not met in any group. CA rates applying investigational CLSI and FDA breakpoints were 93 and 85%, respectively, for bla CTX-M -carrying isolates, 53 and 56%, respectively, for bla NDM -carrying isolates, and 63 and 61%, respectively, for bla KPCcarrying isolates.
The amount of iron determined in the MHA lot used in this study was Ͼ0.2 g/ml.

DISCUSSION
Iron is an essential nutrient for both humans and pathogens. During infection in mammalian hosts, the innate immune system limits iron availability by hijacking iron to deprive pathogens of this essential nutrient (31). Therefore, low availability of iron in vivo during infection can be likened to iron depletion in broth in vitro. Depletion of iron in BMD media for CFDC susceptibility testing has been previously demonstrated to recapitulate in vivo activity of CFDC (32), and therefore, this is the method approved by CLSI (26,33). Our results, like those of others (25,34), support the use of iron-depleted media on the basis of the higher MICs obtained when CFDC was tested with standard CAMHB (Ͼ0.03 g/ml iron) than with iron-depleted CAMHB (Յ0.03 g/ml iron; Table S1) (35).
To the best of our knowledge, there are no published studies evaluating AD testing of CFDC. We showed low EA values (not meeting a 90% acceptance level) for all species tested, in addition to poor CA (except for S. maltophilia isolates, although all study isolates had low MICs), and high rates of mE and ME, with some VME when results of AD were compared to those of ID-BMD. Although the medium used for AD itself has been reported to have iron-chelating properties (36) and lower free-iron concentrations than CAMHB (37), the identified discordance between AD and ID-BMD could possibly have been related to the amount of iron in the MHA lot used for the duration of the study, as it was at least 6 times more than the maximum amount allowed in BMD media (0.03 g/ml).
MICs may be influenced by cation concentration of the culture medium, as demonstrated by Washington et al., who assessed activity of aminoglycosides against P. aeruginosa isolates using 14 lots of MHA (38). Girardello et al. demonstrated variability in polymyxin B MICs determined by AD in comparison to BMD using four MHA brands (37). MICs of tigecycline determined by Etest were 2 to 8 times higher with different MHA commercial brands and appeared to depend on the concentration of manganese (39). Thus, the amount of iron present in the specific MHA medium used for AD here may have affected CFDC activity.
It is noteworthy that all 71 S. maltophilia isolates tested were susceptible to CFDC, with MIC 90 values of 1 and 0.25 g/ml obtained by AD and ID-BMD, respectively. MIC 90 values  reported in previous surveillance studies ranged from 0.25 to 0.5 g/ml for ID-BMD (13,25). The activity of CFDC against S. maltophilia may be significant, given that this species is intrinsically resistant to many broad-spectrum antimicrobial agents, including carbapenems, as a result of production of inducible chromosomal metallo-and serine-␤-lactamases (L1 and L2) (40). The S. maltophilia and B. cepacia complex findings suggest that CFDC deserves further study for these often challenging-to-treat bacterial species. We evaluated discrepancies between categorization by MIC breakpoints defined by the FDA, CLSI, and EUCAST-most were observed among AD results. The investigational CLSI breakpoints are for research use as well as compassionate use of the agent when there is no other therapy available. In 2019, the FDA set breakpoints for CFDC that are more conservative than those of the CLSI (Table 1); FDA breakpoints should be used by laboratories until CLSI reevaluates its investigational breakpoints based on outcomes from more recent clinical trials. A possible revision is expected in 2021 (9). In May 2020, EUCAST established clinical breakpoints for Enterobacterales and P. aeruginosa (16).
In this study, we applied FDA breakpoints for all Enterobacterales, although the recommendation is technically only for some Enterobacteriaceae, including E. coli, K.  pneumoniae, E. cloacae complex, and P. mirabilis. For Enterobacterales, the susceptibility percentages obtained with ID-BMD were 65, 53, and 53% when investigational CLSI, FDA, and EUCAST breakpoints, respectively, were applied (Table 2). FDA and EUCAST have the same susceptible breakpoints for Enterobacterales. For P. aeruginosa, 96 and 89% of study isolates would be considered susceptible using investigational CLSI and FDA breakpoints, respectively. Discrepancies between MIC breakpoints also affected CA and error rates (Table 3). CA for P. aeruginosa when AD was compared to ID-BMD was 86, 52, and 74% when analyzed by investigational CLSI, FDA, and EUCAST breakpoints, respectively; there were 16 mEs when FDA breakpoints were applied and 8 when investigational CLSI breakpoints were applied.
Limitations of this study must be considered. AD using iron-depleted medium was not assessed. The collection of isolates studied was not representative of isolates in general clinical practice, nor was this a "surveillance" study. Instead, this study was enriched with a subset of drug-resistant GNB and included 83 CFDC-resistant isolates representing various species to challenge the breakpoints for method comparison purposes. A large number of A. baumannii isolates were specifically included on the basis of their being CFDC resistant. The main aim of this study was to compare the performance of AD to ID-BMD for MIC determination; therefore, we desired to test a range of susceptible to resistant CFDC isolates. Another limitation was the lack of genetic data regarding mechanisms of resistance for the nonfermenting GNB studied. Also, only a single lot of MHA was evaluated (BD Difco); differences in the cation composition of the MHA can generate categorical errors in susceptibility testing. On the other hand, inclusion of MDR GNB isolates from different countries representing a variety of MICs across the susceptible-resistant spectrum and mechanisms of resistance can be considered a strength of our work.
Overall, CFDC showed low EA rates and high error rates with AD in comparison to ID-BMD. The activity of CFDC against S. maltophilia alongside B. cepacia complex is encouraging. Based on the findings of this study, AD should not be used for in vitro susceptibility testing of CFDC using the described method.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.5 MB.