In Vitro Activity of Cefiderocol, a Siderophore Cephalosporin, against Multidrug-Resistant Gram-Negative Bacteria

Cefiderocol is a parenteral siderophore cephalosporin with a catechol-containing 3′ substituent. We evaluated its MICs against Gram-negative bacteria, using iron-depleted Mueller-Hinton broth. The panel comprised 305 isolates of Enterobacterales, 111 of Pseudomonas aeruginosa, and 99 of Acinetobacter baumannii, all selected for carbapenem resistance and multidrug resistance to other agents. At 2 and 4 μg/ml, cefiderocol inhibited 78.

avibactam is active against most Enterobacterales isolates with KPC and OXA-48-like enzymes, while meropenem-vaborbactam and imipenem-relebactam inhibit those with KPC enzymes only (7,8). However, none of these combinations are active against Enterobacterales isolates with MBLs, nor against the vast majority of carbapenemaseproducing isolates of Pseudomonas aeruginosa and Acinetobacter baumannii (9,10).
Cefiderocol is a novel parenteral siderophore cephalosporin approved in the United State for the treatment for complicated urinary tract infections (cUTI), and in the European Union (EU) for the treatment of infections due to aerobic Gram-negative bacteria in adults with limited treatment options. It has a catechol moiety attached via its 3-position side chain, allowing uptake into Gram-negative bacteria via the ferric iron transporter system (11)(12)(13). We evaluated its activity against multidrug-resistant clinical isolates of Gram-negative bacteria from the United Kingdom, prioritizing those with carbapenem resistance (Table 1).

RESULTS
Overall MIC distributions. The isolate panel was deliberately loaded with highly resistant organisms (Table 2), as reflected in the fact that no comparator agent was active against Ͼ90% of isolates and only colistin achieved Ͼ80% activity across all groups; tigecycline achieved activity versus Ͼ90% of Enterobacterales isolates based on the FDA 2-g/ml breakpoint but only 42.0% based on the EUCAST 0.5-g/ml value, which, strictly, is only applicable to Escherichia coli and Citrobacter koseri. Among comparator ␤-lactams, ceftazidime-avibactam was the most active combination against Enterobacterales, inhibiting 41.6% of isolates at its breakpoint of 8 g/ml for ceftazidime plus 4 g/ml for avibactam, essentially comprising almost all those without MBLs. Aztreonam was the most active ␤-lactam against the P. aeruginosa panel, inhibiting 46.8% of isolates-mostly MBL producers-at its 16-g/ml breakpoint. Resistance to established ␤-lactams, including ceftazidime-avibactam, was nearly universal in the A. baumannii panel.
Cefiderocol inhibited 78.7% of Enterobacterales isolates at 2 g/ml and 92.1% at 4 g/ml; corresponding proportions for the P. aeruginosa collection were 81.1 and 86.5%, respectively, although only 63.3% were inhibited at the FDA's 1-g/ml breakpoint. The proportions of the Acinetobacter collection inhibited at 2 and 4 g/ml were 80.8 and 88.9%, respectively; neither CLSI nor the FDA has breakpoints for this genus.
Irrespective of species, the MIC distribution for Enterobacterales isolates with NDM enzymes was extended and elevated compared with those for isolates with other MBLs. This behavior was independent of aztreonam resistance, indicating that higher cefiderocol values, where seen, were not contingent upon coresident ESBL or AmpC enzymes (Table 4). P. aeruginosa. At 2 g/ml, cefiderocol inhibited 81.1% of P. aeruginosa isolates tested, with rates between 90.0 and 93.3% for all groups, except for those with IMP (80.0%), PER (66.7%), and NDM (45.5%) ␤-lactamases (Table 3). At 4 g/ml, cefiderocol inhibited 86.5% of P. aeruginosa isolates tested, with rates between 90.0 and 100% for all resistance mechanism groups, except for those with IMP (80.0%), PER (73.3%), or NDM enzymes (72.7%). Cefiderocol MICs of Ն128 g/ml were recorded for two isolates: one with an NDM carbapenemase and one with a VIM carbapenemase. As with Enterobacterales, cefiderocol MICs for P. aeruginosa isolates with NDM carbapenemases were elevated compared with those for isolates with other mechanisms, although it should be cautioned that (i) the behavior seemed less marked and (ii) only 11 P. aeruginosa isolates with NDM carbapenemases were tested-far fewer than the 61 Enterobacterales isolates.
A. baumannii. At 2 g/ml, cefiderocol inhibited 80.8% of the test panel of A. baumannii isolates, with rates of Ն85% for all groups, except for those with NDM carbapenemases (50.0%); at 4 g/ml, it inhibited 88.9% (Table 2), with rates of Ն87% for all groups, except for those with NDM carbapenemases (80.0%) ( Table 3). Never-  Table 3 indicates that 10% of A. baumannii isolates with OXA-58 carbapenemases were resistant to cefiderocol at 128 g/ml, this represents only a single isolate, meaning that significance is limited.
Cefiderocol combined with ␤-lactamase inhibitors. Because cefiderocol MICs for isolates with NDM carbapenemases were generally higher than those for isolates with other carbapenemases, we sought to explore whether these isolates could inactivate the compound. To do this, we redetermined MICs of cefiderocol together with dipicolinic acid (100 g/ml), as an inhibitor of MBLs, and avibactam (4 g/ml), as an inhibitor of any coproduced serine ␤-lactamases, and with a combination of both these inhibitors. A panel of 40 Enterobacterales isolates with NDM enzymes was used together with small control groups representing other carbapenemases or mechanisms (Table 5).
Avibactam alone had predictably little effect on the MICs of cefiderocol for the isolates with NDM enzymes: only 1/40 NDM-positive Enterobacterales isolates showed a Ͼ2-fold reduction in cefiderocol MIC. Dipicolinic acid, in contrast, achieved 4-to 32-fold reductions in MIC for 25/40 bla NDM -positive isolates, with combination MICs ranging from Յ0.03 to 16 g/ml and a mode of 1 g/ml, compared with 4 g/ml for cefiderocol alone. There was a greater effect when both inhibitors were tested together in combination with cefiderocol, with 4-to 64-fold MIC reductions for 36/40 bla NDMpositive isolates. Using this triple combination, all 40 NDM-positive isolates were inhibited by cefiderocol at Յ2 g/ml compared with 22.5% for cefiderocol alone.
The control groups were small, reducing the strength of conclusions for individual enzyme types: nonetheless, MICs of cefiderocol for isolates with VIM enzymes generally Cefiderocol against MDR Gram-Negative Bacteria Antimicrobial Agents and Chemotherapy showed 2-to 4-fold reductions when dipicolinic acid was added alone or together with avibactam, whereas no reductions were seen with only avibactam added. Conversely, the MICs of cefiderocol for isolates with KPC or OXA-like carbapenemases, or with combinations of ESBL and porin loss, mostly were reduced by the addition of avibactam alone or combined with dipicolinic acid, whereas dipicolinic acid alone had minimal effect.

DISCUSSION
Cefiderocol is the first catechol ␤-lactam to be licensed. It combines efficient entry into Gram-negative bacteria with considerable ␤-lactamase stability. These factors support in vitro activity against otherwise extremely resistant Enterobacterales isolates and nonfermenters, at least under iron-deficient conditions. Thus, as found here and by others (12)(13)(14), cefiderocol inhibited the majority of carbapenem-resistant Gramnegative bacteria, tested irrespective of species, at 2 or 4 g/m. This activity encompassed many MBL-producing isolates of Enterobacterales, P. aeruginosa, and A. baumannii, as well as isolates of A. baumannii with OXA carbapenemases, whereas these groups typically are unequivocally resistant to new ␤-lactamase inhibitor combinations, including ceftazidime-avibactam (as confirmed here), meropenem-vaborbactam, and imipenem-relebactam (15)(16)(17).
These aspects are positive, but three interlinked areas of uncertainty remain: the "correct" breakpoints, the activity against bacteria with NDM carbapenemases, and the clinical efficacy against carbapenemase producers. The breakpoint issues have been recently and well summarized by Simner and Patel (18). Succinctly, a provisional CLSI value of susceptible (S) Յ 4 g/ml and resistant (R) Ͼ 4 g/ml was adopted in development and has been retained as a reference point here; however, EUCAST has subsequently advised breakpoints of S Ͻ 2 g/ml and R Ͼ 2 g/ml for all species, and the FDA has adopted values of S Ͻ 2 g/ml and R Ͼ 4 g/ml for Enterobacterales and S Յ 1 g/ml and R Ͼ 2 g/ml for P. aeruginosa, with the last of these values predicated on a cUTI trial (19) where there was only a single P. aeruginosa isolate with a MIC of Ͼ1 g/ml. It is not now clear whether CLSI will retain their provisional values or adopt the lower FDA values; a decision is anticipated in 2021 (18).
These breakpoint issues become particularly pertinent for bacteria with NDM carbapenemases, as the MICs of cefiderocol for these mostly were higher than those for isolates of the same species with other carbapenemases. In particular, cefiderocol MICs were Ͼ4 g/ml for 17/61 (27.9%) of Enterobacterales isolates with NDM carbapenemases compared with 2/62 (3.2%) of those with VIM and IMP MBLs (P Ͻ 0.001, chi-square test). This behavior was unrelated to aztreonam resistance or susceptibility, implying that the higher MICs reflected the NDM enzymes themselves and not coproduced ESBLs or AmpC enzymes. To explore this aspect further and to determine if NDM enzyme protected bacteria against cefiderocol, we undertook combination tests with ␤-lactamase inhibitors. The results supported the view that cefiderocol is not completely ␤-lactamase stable, as it was frequently, though weakly, potentiated by dipicolinic acid against the isolates with NDM (particularly) and VIM MBLs and by avibactam for isolates with KPC, OXA-48, ESBL, and AmpC enzymes. Additional synergy for Enterobacterales isolates with NDM enzymes when both dipicolinic acid and avibactam were added is surprising. It may be that avibactam inhibits coresident class A or D ␤-lactamases, but were this a significant factor, one would expect the MICs of unprotected cefiderocol to be higher for aztreonamresistant, NDM-positive Enterobacterales isolates than for their aztreonam-susceptible counterparts lacking ESBLs or AmpC enzymes, and this was not the case ( Table 4).
The significance of this slight ␤-lactamase lability is difficult to judge: imipenem remains clinically useful against P. aeruginosa, unless OprD is lost, despite lability to the organism's chromosomal AmpC enzyme (20); on the other hand, breakpoints for oxyimino cephalosporins against Enterobacterales have had to be lowered substantially because even modestly raised values have been associated with clinical failures when ESBLs are present (21).
Relating these MIC observations to clinical data is presently difficult. Recent phase III trials have shown cefiderocol to be effective and noninferior to imipenem-cilastatin for the treatment of cUTIs and to meropenem in nosocomial pneumonia caused by Gram-negative bacteria (19,22). However, these trials mostly recruited patients with broadly susceptible pathogens, and a further study, CREDIBLE-CR (23, 24), gave more equivocal results. This compared cefiderocol with "best available therapy"-comprising colistin or its combinations in 66% of cases-in multiple infection types involving carbapenem-resistant Gram-negative bacteria. Overall, cefiderocol achieved comparable clinical and microbiologic outcomes to its comparators, but disturbingly, there was a significant excess of deaths in the cefiderocol arm, many involving infections with Acinetobacter spp. Formal publication and analysis of these data are awaited, and the issue of outcomes relative to enzyme type as well as species and MIC will be of vital importance. More positively, several case reports have appeared describing the successful use of cefiderocol as compassionate therapy in infections involving difficult extremely resistant pathogens (25)(26)(27).
Ultimately, accumulating clinical experience will determine cefiderocol's utility against carbapenemase-producing Gram-negative bacteria. Given the drug's unusual mode of uptake, this utility may depend on the degree of iron starvation that applies at a particular infection site. What can fairly be said at this stage is that its MICs for many isolates that are resistant to other new agents are sufficiently low to permit a degree of guarded optimism.

MATERIALS AND METHODS
Bacteria. The test isolates (Table 1) comprised (i) 305 isolates of Enterobacterales, selected to represent diverse carbapenemase producers and isolates with carbapenem resistance via combinations of porin loss with AmpC or ESBL activity; (ii) 111 isolates of P. aeruginosa, selected to represent producers of MBLs and GES carbapenemases, along with isolates that produced VEB or PER ESBLs and were carbapenem resistant via OprD loss; and (iii) 99 isolates of A. baumannii with NDM MBLs or various OXA carbapenemases.
In selecting isolates from Public Health England's Antimicrobial Resistance and Healthcare Associated Infections (PHE-AMRHAI) Reference Unit collections for inclusion, we favored organisms with phenotypes for other ␤-lactams (cefiderocol had not previously been tested), typical of their mechanisms but including susceptible and resistant representatives for antibiotics that divide isolates within groups. Thus, for example, groups with metallo-␤-lactamases were selected to include both aztreonam-resistant organisms (inferred also to have ESBLs or AmpC enzymes) and those that were aztreonam susceptible (inferred to lack ESBLs or AmpC enzymes). We ensured that the isolates were from diverse hospitals and excluded multiple isolates from single patients. Almost all of the organisms were submitted by hospital laboratories in the United Kingdom between 2008 and 2018 for investigation of unusual resistance phenotypes and/or susceptibility testing for therapeutic guidance. Exceptions were 3 isolates referred to the PHE-AMRHAI Reference Unit from hospitals in the Republic of Ireland and 11 isolates of Pseudomonas aeruginosa with PER ESBLs, which were collected in Turkey in the early 1990s (28).
Carbapenemases and VEB and PER enzymes were identified by PCR of their encoding genes or by whole-genome sequencing (WGS). Carbapenem resistance due to porin loss combined with ESBL or AmpC activity was inferred from previous susceptibility results and the absence of carbapenemase, as confirmed by PCR or WGS. Species identification was by matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectroscopy.
We reviewed results against published breakpoints: EUCAST has values of S Յ 2 g/ml and R Ͼ 2 g/ml for Enterobacterales and P. aeruginosa; the FDA has values of S Յ 2 g/ml and R Ͼ 4 g/ml for Enterobacterales and S Յ 1 g/ml and R Ͼ 2 g/ml for P. aeruginosa; CSLI still has values under review, but previously, when cefiderocol was in trial, had investigational values of S Յ 4 g/ml and R Ͼ 8 g/ml. MICs of comparator antibiotics were interpreted using EUCAST guidelines where available, the excep-