Structural Basis of PER-1-Mediated Cefiderocol Resistance and Synergistic Inhibition of PER-1 by Cefiderocol in Combination with Avibactam or Durlobactam in Acinetobacter baumannii

ABSTRACT

Cefiderocol is a novel siderophore cephalosporin that displays activity against Gram-negative bacteria. To establish cefiderocol susceptibility levels of Acinetobacter baumannii strains from China, we performed susceptibility testing and genomic analyses on 131 clinical isolates. Cefiderocol shows high activity against the strains. The production of PER-1 is the key mechanism of cefiderocol resistance. In silico studies predicted that avibactam and durlobactam could inhibit cefiderocol hydrolysis by PER-1, which was confirmed by determining cefiderocol MICs in combination with inhibitors.

INTRODUCTION

Acinetobacter baumannii is one of the most important nosocomial pathogens associated with high mortality rates (1). According to China Antimicrobial Surveillance Network (CHINET) data, the rates of resistance of Acinetobacter spp. to most antibiotics, except for colistin and tigecycline, in 2021 were close to or higher than 50%. With such unexpectedly high resistance rates for A. baumannii in China, new antimicrobials are urgently needed.

Cefiderocol is a novel siderophore-conjugated cephalosporin that shows potent in vitro activities against a broad array of Gram-negative bacteria, such as Enterobacterales, Pseudomonas aeruginosa, and A. baumannii, including isolates that are resistant to carbapenems (2–4). Although surveillance programs indicated that Gram-negative bacteria have high cefiderocol susceptibility rates, a small number of cefiderocol-resistant strains were found in previous studies (5, 6). Reported mechanisms of cefiderocol resistance mainly include the presence of β-lactamase genes and mutations in iron transporter systems. Research by Poirel et al. showed that PER-like β-lactamases and NDM β-lactamases contribute to decreased susceptibility to cefiderocol in A. baumannii (7). In addition, a deficiency of the iron transporter PiuA in P. aeruginosa or both CirA and Fiu in Escherichia coli caused 16-fold changes in cefiderocol resistance (8).

To establish the cefiderocol susceptibility levels of A. baumannii strains collected in China and to investigate cefiderocol resistance mechanisms, we collected 131 well-characterized A. baumannii strains from three hospitals in China (9) and determined their susceptibility to cefiderocol, tigecycline, and colistin. The susceptibility data for tigecycline and cefiderocol were interpreted using breakpoints for Enterobacterales recommended by EUCAST (for tigecycline: susceptible, MIC of ≤0.5 μg/mL; resistant, MIC of >0.5 μg/mL; for cefiderocol: sensitive, MIC of ≤2 μg/mL; resistant, MIC of >2 μg/mL) and those for colistin using the breakpoints for Acinetobacter spp. in the EUCAST breakpoint tables v12.0 (sensitive, MIC of ≤2 μg/mL; resistant, MIC of > 2 μg/mL) (10). The growth of 97.71% of the A. baumannii strains was suppressed at cefiderocol concentrations of ≤2 μg/mL, while 3 strains displayed resistance to cefiderocol. The MIC₅₀ and MIC₉₀ were 0.5 μg/mL and 1 μg/mL, respectively. The colistin and tigecycline susceptibility rates were 93.89% (123/131 strains) and 15.27% (20/131 strains), respectively (see Table S2 in the supplemental material). Cefiderocol had superior in vitro activity, compared to colistin or tigecycline, against the tested isolates.

Whole-genome sequencing (WGS) data obtained by Illumina HiSeq sequencing were analyzed to determine the molecular basis for cefiderocol resistance. BacAnt (http://bacant.net) (11) was used to identify antimicrobial resistance genes (ARGs). The WGS data showed that all resistant strains were identified as sequence type 2 (ST2) according to the Pasteur multilocus sequence typing (MLST) typing scheme and contained the β-lactamase genes bla_PER-1, bla_OXA-23, bla_OXA-66, and bla_ADC-25 (Table 1). Previous studies showed that, compared to strains without PER production, PER-producing strains showed a higher cefiderocol MIC distribution range (12). To confirm the role of PER-1 in cefiderocol resistance, we expressed bla_PER-1 in A. baumannii ATCC 17978 and E. coli DH5α using the plasmid pYMAb2-Hyg^r-PER-1 (13, 14). The cefiderocol MICs for PER-1-producing ATCC 17978 and E. coli DH5α increased >32-fold; this is consistent with previous research expressing PER-like β-lactamases in E. coli TOP10 and A. baumannii CIP70.10 (7). We then knocked out the bla_PER-1 gene in strain XH740 through allelic replacement (15, 16). The cefiderocol MIC for XH740ΔPER-1 decreased to a level that is defined as antibiotic susceptibility (0.25 μg/mL). When we introduced pYMAb2-Hyg^r-PER-1 into the knockout strain, the cefiderocol MIC was restored to 32 μg/mL (Table 2). In conclusion, we demonstrated, through bla_PER-1 knockout and plasmid complementation experiments, that the presence of bla_PER-1 causes cefiderocol resistance as the sole factor in clinical strains.

TABLE 1.

ARGs and susceptibility testing of cefiderocol-nonsusceptible A. baumannii strains

Strain	ST	ARGs	Cefiderocol MIC (μg/mL)
XH740	ST2	blaPER-1, blaADC-25, blaOXA-23, blaOXA-66, sul1, amvA, adeC, tet(B), ant(3″)-IIa, aph(6)-Id, aph(3″)-Ib, armA, msr(E), mph(E), abaF	16
XH801	ST2	blaPER-1, blaADC-25, blaOXA-23, blaOXA-66, sul1, amvA, adeC, tet(B), ant(3″)-IIa, aph(6)-Id, aph(3″)-Ib, armA, msr(E), mph(E), abaF	32
XH1820	ST2	blaPER-1, blaADC-25, blaOXA-23, blaOXA-66, sul1, amvA, adeC, tet(B), ant(3″)-IIa, aph(6)-Id, aph(3″)-Ib, armA, msr(E), mph(E), abaF	>32

TABLE 2.

In vitro activity of cefiderocol and cefiderocol plus β-lactamase inhibitors against A. baumannii and E. coli recombinant strains and reconstructed clinical isolates

Strain	MIC (μg/mL)a
	FDC	FDC-AVI	FDC-DUR	FDC/SUL	AVI	DUR	SUL
XH740	16	0.5	0.25	8/4	>32	>32	32
XH801	32	0.5	0.125	16/8	>32	>32	16
XH1820	>32	1	0.125	32/16	>32	>32	32
E. coli DH5α	≤0.03	≤0.03	≤0.03	≤0.03/0.015	16	0.25	32
E. coli DH5α::pYMAb2	≤0.03	≤0.03	≤0.03	≤0.03/0.015	16	0.25	32
E. coli DH5α::pYMAb2-PER-1	1	≤0.03	≤0.03	0.25/0.125	32	0.25	>32
ATCC 17978	0.06	0.06	0.06	0.06/0.03	>32	>32	2
ATCC 17978::pYMAb2	0.125	0.125	0.125	0.06/0.03	>32	>32	2
ATCC 17978::pYMAb2-PER-1	>32	0.5	0.25	16/8	>32	>32	32
XH740 ΔPER-1	0.25	0.125	0.125	0.25/0.125	>32	>32	8
XH740 ΔPER-1::pYMAb2	0.125	0.125	0.06	0.25/0.125	>32	>32	8
XH740 ΔPER-1::pYMAb2-PER-1	32	0.25	0.25	16/8	>32	>32	32

^aFDC, cefiderocol; AVI, avibactam; DUR, durlobactam; SUL, sulbactam. The concentration of avibactam was 4 μg/mL. The concentration of durlobactam for A. baumannii strains was 4 μg/mL, and that for E. coli strains was 0.06 μg/mL.

In addition, we performed WGS of XH1820 using the Oxford Nanopore Technologies platform (Oxford Nanopore Technologies, Oxford, UK). The map of the XH1820 chromosome and the locus of ARGs on the chromosome are shown in Fig. S1 in the supplemental material. Two copies of bla_PER-1 are located on a variant of the resistance island AbGRI5 that was identified in XH1056 (17). In a BLAST search, a similar sequence was found in strain XH859, which carried one copy of bla_PER-1 without the tandem repeat of the ISCR1-sul1-bla_PER-1-ISCR1 element (see Fig. S1C). ISCR1 appears to be important in the dissemination of bla_PER-1, as Xie et al. detected a circular intermediate comprising bla_PER-1-ISCR1-qacED1/sul1 (18). Additionally, ISCR1 contains a promoter region that could drive the expression of downstream genes (19, 20). The study by Lallement et al. indicated that ISCR1 significantly increases the expression level of downstream ARGs bla_CTX-M-9 and dfrA19 via promoters located in the oriIS region of ISCR1 (21). Given the similar sequence of the oriIS region of ISCR1, we speculate that ISCR1 might also increase the expression of bla_PER-1 to mediate cefiderocol resistance in strain XH1820. In addition, we mapped Illumina reads to the corresponding region of XH859 to determine the bla_PER-1 copy numbers in the other 2 cefiderocol-resistant strains. Strains XH740 and XH801 both carry 1 copy of bla_PER-1, which might explain why their cefiderocol resistance levels are lower than that of strain XH1820 (see Fig. S1D).

To investigate the modes of cefiderocol binding to PER-1 at the molecular level, we employed in silico structural modeling and conducted covalent docking using Schrödinger software (see the supplemental material). As shown in Fig. 1, the predicted covalent binding residue Ser70 is located at the bottom of the ligand binding pocket of PER-1, while the Ω loop, the R2 loop, and the hairpin loop linking the β7 and β8 sheets are situated at the lip of the pocket. We showed that cefiderocol binds to both the large cavity and the adjacent small shallow cavity that the trabecular Thr235 to Gln239 of the β7 sheet divide, as the siderophore moiety seems to occupy the small cavity. In contrast, β-lactamase inhibitors such as avibactam and durlobactam seem to bind only to the large cavity. Ser130, Trp105, Arg220, Thr237, and Lys240A are key interacting residues in our in silico analysis (see Fig.S2 to S5). Previous studies showed that most β-lactamases, except PER and NDM, are unable to effectively hydrolyze cefiderocol (7, 22, 23); the reason might be that the bulkier siderophore side chain impedes binding to most β-lactamases. Next, we purified the β-lactamases PER-1, NDM-1, and VIM-24 and performed enzymatic kinetic analyses (24) using cefiderocol and ceftazidime as the substrates (see Table S4). The catalytic efficiencies of PER-1 and NDM-1 for cefiderocol and ceftazidime were similar but significantly higher than that of VIM-24 according to k_cat/K_m values (Table 3).

FIG 1.

Binding mode analysis of cefiderocol, avibactam, and durlobactam bound to PER-1. (A to C) Nearby residues interacting with ligands. (D to F) Ribbon representation of ligand-bound PER-1. (G to I) Superposed binding surface of PER-1. Gray refers to the cefiderocol-bound state, green refers to the avibactam-bound state, and cyan refers to the durlobactam-bound state. The carbon of cefiderocol is colored gray, while those of avibactam and durlobactam are colored green and cyan, respectively.

TABLE 3.

Kinetic parameters of PER-1, NDM-1, and VIM-24 hydrolysis of cefiderocol

Enzyme	Parameters witha:
	Ceftazidime			Cefiderocol
	Km (μM)	Kcat (s−1)	Kcat/Km (s−1 μM−1)	Km (μM)	Kcat (s−1)	Kcat/Km (s−1 μM−1)
PER-1	ND	ND	0.680 ± 0.008	ND	ND	0.046 ± 0.000
NDM-1	85 ± 8	71.4 ± 3.8	0.844 ± 0.034	67 ± 2	4.5 ± 0.04	0.067 ± 0.001
VIM-24	ND	ND	0.059 ± 0.001	ND	ND	0.005 ± 0.000

^aThe K_m, k_cat, and k_cat/K_m values shown are the mean ± standard deviation of three independent experiments. ND, not determined. K_m and k_cat values for PER-1 and VIM-24 with ceftazidime and cefiderocol were not determined because the calculated K_m values were much higher than the measured substrate concentrations. k_cat/K_m values were approximately equal to the slope extracted through linear regression of the initial velocity (v)/enzyme concentration ([E]) versus substrate concentration ([S]) when ignoring the [S] in the denominator of the Michaelis-Menten equation, i.e., v/[E] = k_cat × [S]/(K_m + [S]).

To predict whether β-lactamase inhibitors can reduce PER-1-mediated enzymatic hydrolysis of cefiderocol, we carried out covalent docking with molecular mechanics/generalized Born surface area (MM/GBSA) scoring. Avibactam, durlobactam, and taniborbactam received more negative covalent docking scores than cefiderocol (see TableS5 and Fig. S6A). This is likely due to their small volume, which might make it easier for them to adopt a position suitable for the hydrolysis reaction. Consequently, avibactam, durlobactam, and taniborbactam might be more likely than cefiderocol to react with PER-1. In contrast, the inhibitory effect of sulbactam is likely weaker, as we observed a less negative ΔG (see Fig. S6B). To prove the results of our computational predictions, we added avibactam, durlobactam, or sulbactam to cefiderocol and performed susceptibility testing. Consistent with the predictions of the covalent docking analyses, a >32-fold reduction in cefiderocol MIC values in strains producing PER-1 was observed when avibactam was added, while durlobactam resulted in a ≥64-fold MIC reduction (Table 2). To evaluate the inhibitory action of β-lactamase inhibitors toward PER-1, competition assays were performed with nitrocefin as a reporter substrate. The K_i values of durlobactam were significantly lower than those of avibactam (0.0016 ± 0.0009 μM versus 0.48 ± 0.02 μM), and avibactam and sulbactam had similar K_i values toward PER-1 (see Table S3). A reduction of cefiderocol MIC values caused by avibactam was observed in previous studies (7), while durlobactam used in combination with cefiderocol is first reported here. Durlobactam is a new β-lactamase inhibitor with broad-spectrum activity against Ambler class A, C, and D serine β-lactamases (25). Cefiderocol combined with avibactam or durlobactam might be an effective therapeutic approach against A. baumannii strains producing PER-1.

Cefiderocol was highly effective against 131 clinical A. baumannii strains we collected in Hangzhou, China. We demonstrate that the production of the extended-spectrum β-lactamase PER-1 is a key factor mediating cefiderocol resistance, because it can hydrolyze cefiderocol, while the bla_PER-1 copy number positively correlates with the resistance level. Most importantly, PER-1-mediated hydrolysis of cefiderocol can be inhibited in the presence of avibactam or durlobactam.

Approval was obtained from the Ethics Committee of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (reference number 2022-446-01).

Data availability.

The complete sequence of A. baumannii strain XH1820 has been deposited in the GenBank nucleotide database under accession numbers CP096890, CP096891, and CP096892.