In vitro dynamics and mechanisms of cefiderocol resistance development in wild-type, mutator and XDR Pseudomonas aeruginosa

Abstract

Objectives

To analyse the dynamics and mechanisms of stepwise resistance development to cefiderocol in Pseudomonas aeruginosa.

Methods

Cefiderocol resistance evolution was analysed in WT PAO1, PAOMS (mutS mutator derivate) and three XDR clinical isolates belonging to ST111, ST175 and ST235 clones. Strains were incubated in triplicate experiments for 24 h in iron-depleted CAMHB with 0.06–128 mg/L cefiderocol. Tubes from the highest antibiotic concentration showing growth were reinoculated into fresh medium containing concentrations up to 128 mg/L for 7 consecutive days. Two colonies per strain and experiment were characterized by determining the susceptibility profiles and WGS.

Results

Evolution of resistance was significantly enhanced in PAOMS, but was variable for the XDR strains, including levels similar to PAOMS (ST235), similar to PAO1 (ST175) or even below PAO1 (ST111). WGS revealed 2–5 mutations for PAO1 lineages and 35–58 for PAOMS. The number of mutations in the XDR clinical strains ranged from 2 to 4 except for one of the ST235 experiments in which a mutL lineage was selected, thus increasing the number of mutations. The most frequently mutated genes were piuC, fptA and pirR, related to iron uptake. Additionally, an L320P AmpC mutation was selected in multiple lineages and cloning confirmed its major impact on cefiderocol (but not ceftolozane/tazobactam or ceftazidime/avibactam) resistance. Mutations in CpxS and PBP3 were also documented.

Conclusions

This work deciphers the potential resistance mechanisms that may emerge upon the introduction of cefiderocol into clinical practice, and highlights that the risk of resistance development might be strain-specific even for XDR high-risk clones.

Introduction

Pseudomonas aeruginosa, a ubiquitous opportunistic pathogen considered one of the paradigms of antimicrobial resistance, is among of the main causes of hospital-acquired and chronic infections associated with high morbidity and mortality.^1,2 Indeed, P. aeruginosa infections are estimated to be associated with over 300 000 annual deaths and is at the top of the WHO priority list for the need for research and development of new antibiotics.^3,4 This growing threat results from the extraordinary capacity of this pathogen for developing resistance through chromosomal mutations and from the increasing prevalence of transferable resistance determinants, particularly those encoding carbapenemases or ESBLs.^1,5P. aeruginosa has a non-clonal epidemic population structure, composed of a limited number of widespread clones, which are selected from a background of a large quantity of rare and unrelated genotypes that are recombining at high frequency.⁶ Indeed, several surveys have provided evidence of the existence of MDR/XDR global clones, denominated high-risk clones, disseminated in hospitals worldwide; ST235, ST111 and ST175 are among those most widespread.^6,7 Epidemic clones are also occasionally documented in chronic infections, but the main resistance threat in this setting is the frequent emergence of mutator (hypermutable) variants that catalyse the evolution of antimicrobial resistance.²

The recent introduction of novel β-lactam/β-lactamase inhibitors combinations (such as ceftolozane/tazobactam, ceftazidime/avibactam or imipenem/relebactam) has helped to mitigate, to some extent, the problem of MDR/XDR P. aeruginosa.⁸ However, these novel antibiotics, stable against classical β-lactam resistance mechanisms (such as the overexpression of the β-lactamase AmpC or efflux pumps), are not exempt from resistance development, evidenced right upon their introduction into clinical practice.^9,10 Indeed, emerging resistance mechanisms to these novel agents were already predicted by previous in vitro evolution experiments and frequently included gain-of-function mutations leading to the modification of the catalytic centres of the intrinsic AmpC or acquired OXA-2/OXA-10 β-lactamases.^11–13 The use of mutator strains and MDR/XDR high-risk clone isolates in these evolution experiments was crucial for deciphering those resistance mechanisms, particularly when it required two or more simultaneous mutations or the presence of pre-existent first-step resistance mutations or horizontally acquired resistance determinants.^11–14

Cefiderocol is a new siderophore cephalosporin that has been recently approved for the treatment of respiratory and complicated urinary tract infections including those by carbapenem-resistant MDR/XDR P. aeruginosa.^15,16 Similarly to cefepime, it has a pyrrolidinium group on the C-3 side chain, which improves the antibacterial activity and stability against β-lactamases and, similarly to ceftazidime, it has a carboxypropanoxyimino group on the C-7 side chain, which improves transport across the outer membrane. However, cefiderocol adds to these classical antipseudomonal agents a catechol group on the end of the C-3 side chain, which confers siderophore activity. Thus, following chelation of iron, cefiderocol can be transported to the periplasmic space through ferric iron transport systems located on the outer membrane of Gram-negative bacteria and, once within the periplasmic space, cefiderocol dissociates from the iron and binds to PBPs, inhibiting peptidoglycan cell wall synthesis.¹⁷ For these reasons, cefiderocol displays a high stability against classical antimicrobial resistance mechanisms, showing potent antipseudomonal activity even against strains resistant to all other novel β-lactams.¹⁸ Therefore, this molecule represents a highly interesting model for exploring new antibiotic resistance pathways with outstanding clinical relevance potential. With this aim, here we performed in vitro evolution experiments coupled with WGS, using WT, mutator and MDR/XDR high-risk-clone P. aeruginosa isolates.

Materials and methods

Strains

WT reference strain PAO1 and its hypermutable mutS-deficient derivative PAOMS were used.¹⁹ Additionally, three previously characterized XDR clinical strains belonging the high-risk clones ST111 [isolate NAV01-012: oprD (nt₉₉₁InsAC), parS (L137P), mexY (G530S), dacB (nt₆₆₄InsGGCCT), gyrA (T83I), bla_CARB-2], ST175 [isolate 101-E5; oprD (Q142*), ampR (G154R), mexZ (G195D), gyrA (T83I, D87N), parC (S87W)] and ST235 [isolate 109-F7: oprD (nt₁₂₀₅InsC), gyrA(T83I), mexZ (V48A), pmrB (V344M), parC (S87L), bla_OXA-2] were used.¹³ The three isolates were determined to be resistant to all classical antipseudomonal β-lactams tested (ceftazidime, cefepime, piperacillin/tazobactam, imipenem and meropenem), as well as ciprofloxacin and tobramycin. Moreover, they have been previously used for in vitro evolution experiments of ceftolozane/tazobactam and imipenem/relebactam resistance.¹³ PAO1 transposon mutants from a previously described library were used to assess the impact of specific genes on cefiderocol susceptibility.²⁰

In vitro resistance evolution experiments

Each strain was incubated in triplicate experiments for 24 h at 180 rpm in 10 mL iron-depleted CAMHB with 0.06, 0.12, 0.25, 0.5, 1, 2, 4, 8, 16, 32, 64 or 128 mg/L cefiderocol. Tubes from the highest antibiotic concentration showing growth were reinoculated (1:1000 dilution) into fresh medium containing concentrations up to 128 mg/L for 7 consecutive days. Two colonies per strain and experiment were then randomly selected, purified and stored frozen at −80°C for further study.

Susceptibility testing

MICs of ceftazidime, ceftazidime/avibactam, ceftolozane/tazobactam, cefepime, imipenem and imipenem/relebactam (0.25–128 mg/L) were determined by broth microdilution. MICs of cefiderocol (0.03–128 mg/L) were determined using iron-depleted Mueller–Hinton broth following EUCAST guidelines. EUCAST 2022 clinical breakpoints were used for interpretation (www.eucast.org).

WGS

Total DNA was isolated using a commercial capture system (High Pure PCR Template Preparation Kit, Roche Diagnostics) and indexed paired-end libraries were generated using the Illumina DNA Prep library preparation kit (Illumina Inc., USA) and then sequenced on an Illumina MiSeq^® platform. The reads for each isolate were mapped against the genome of the P. aeruginosa reference strain PAO1 (RefSeq accession number NC_002516.2) using Bowtie 2 software, version 2.2.6 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml).²¹ Pileups and raw files of the mapped reads were obtained using SAMtools, version 0.1.16 (https://sourceforge.net/projects/samtools/files/samtools/)²² and Picard, version 1.140 (https://github.com/broadinstitute/picard). Read alignments surrounding all putative indels were realigned using the Genome Analysis Toolkit (GATK), version 3.4-46 (https://www.broadinstitute.org/gatk/).²³ The list of SNPs was compiled from the raw files that met the following criteria: a quality score of >50, a root mean square (RMS) mapping quality of >25 and a coverage depth of >3. Indels were extracted from the total pileup files by use of the following criteria: a quality score of >250, an RMS mapping quality of >25 and a coverage depth of >3. SNPs and indels for each isolate were annotated by using SnpEff software version 4.3 (http://snpeff.sourceforge.net/index.html).²⁴ Finally, large chromosomal deletions were analysed with SeqMonk version 1.47.2 (https://www.bioinformatics.babraham.ac.uk/projects/seqmonk/).

Cloning and characterization of AmpC variants

The selected L320P AmpC variant was cloned as previously described.¹¹ Briefly, triplicate PCR products obtained with upstream (AmpC-F-EcoRI, 5′-TCGAATTCACGACAAAGGACGCCAATCC-3′) and downstream (AmpC-R-HinDIII, TCAAGCTTTCAGCGCTTCAGCGGCACC) primers were digested with EcoRI or HinDIII, ligated to pUCP24, and transformed into Escherichia coli XL1-Blue made competent by CaCl₂. Transformants were selected in 5 mg/L gentamicin Xgal-IPTG LB agar plates. The cloned genes obtained from three independent experiments were fully sequenced to ascertain the absence of mutations introduced during PCR amplification. Resulting plasmids were electroporated into an ampC knockout mutant of PAO1 (PAOΔC) and plated on 30 mg/L gentamicin LB agar plates. The obtained transformants were characterized through the determination in duplicated experiments of the MICs of ceftazidime, cefepime, ceftazidime/avibactam, ceftolozane/tazobactam, imipenem and cefiderocol, as described above. PAOΔC derivatives harbouring pUCP24 with cloned WT AmpC, as well as AmpC variants previously shown to be involved in ceftolozane/tazobactam resistance development (T96I, G183D or E247K) were tested for comparative purposes.⁹ Representations of the P. aeruginosa PAO1 AmpC β-lactamase structure were performed with the PyMOL Molecular Graphic System, v.2.3 (www.pymol.com).²⁵

Results

Dynamics of in vitro resistance development to cefiderocol, and cross resistance with other β-lactams, in WT, mutator and XDR P. aeruginosa

Figure 1 shows the results of stepwise resistance development to cefiderocol in WT PAO1, its mutS deficient derivative (PAOMS) and three XDR isolates belonging to ST175, ST111 and ST235 high-risk clones. Evolution of resistance was, as expected, significantly enhanced in PAOMS (reaching 128 mg/L by Day 6) compared with PAO1 (4 mg/L by Day 7) (Figure 1a). However, a huge variation in resistance development was documented for the XDR clinical strains, including levels similar to PAOMS (ST235), similar to PAO1 (ST175) or even well below PAO1 (ST111) (Figure 1b).

Figure 1.

Dynamics of stepwise resistance development to cefiderocol in iron-depleted CAMHB cultures from WT PAO1 and its mutS-deficient hypermutable derivative PAOMS (a) and ST111, ST175 and ST235 P. aeruginosa high-risk clones (b). Median values and ranges for three experiments are shown.

MICs of cefiderocol and other relevant antipseudomonal β-lactams for two isolated colonies from each of the three independent experiments per strain are shown in Table 1. Cefiderocol MICs for PAO1 mutants evolved in the presence of cefiderocol varied considerably depending on the lineage, ranging from 2 to 64 mg/L. MICs from PAOMS evolved mutants were more uniform and higher (range 64–256 mg/L). MICs for evolved mutants of the XDR clinical isolate were also strain and lineage dependent: 0.12–4 mg/L for ST111, 4–64 mg/L for ST175 and 32–128 mg/L for ST235. Interestingly, cross resistance with other β-lactams was overall very limited. In general, ceftazidime and particularly cefepime MICs were increased, but ceftazidime/avibactam and ceftolozane/tazobactam susceptibility remained in most of the cases, even for those mutants showing high-level cefiderocol resistance. Moreover, imipenem and imipenem/relebactam MICs were frequently reduced with respect to those of the parent strains.

Table 1.

Susceptibility profiles and WGS of mutants obtained after 7 days of exposure to stepwise concentrations of cefiderocol for WT PAO1, its mutS-deficient derivative PAOMS and XDR clinical isolates from ST175, ST111 and ST235 clones

Straina				MIC (mg/L)b					Genotypec
		IPM	IMR	FEP	CAZ	CZA	C/T	FDC
PAO1		1	0.25	1	1	1	0.25	0.06–0.12	—
1	m1	1	0.25	1	1	1	0.25	2–4	pilT(nt570Δ11), piuC(aa133Δ1)
	m2	1	0.12–0.25	1	1	1	0.25	2–4	pilT(nt570Δ11), piuC(aa133Δ1)
2	m1	1	0.25	2	1	1	0.5	4	pilZ(nt18InsT), cpxS(I279T), fptA(nt1679Δ1)
	m2	1	0.25	2	1	1	0.5	4	pilZ(nt18InsT), cpxS(I279T), fptA(nt1679Δ1)
3	m1	1	0.25	2	2	2	1	16	pilZ(nt218Δ4), cpxS(L80P), fptA(nt505Δ1), PA4701(L403R)
	m2	1	0.25	2	2	2	1	64	pilZ(nt218Δ4), cpxS(L80P), PA3421(L256Q), fptA(nt505Δ1), PA4701(L403R)
PAOMS		1	0.25	2	8–16	1	0.25	0.12–0.25	ΔmutS
1	m1	1	0.5	>64	32–64	4	2	128	speB(D276G), PA0468(V175A), nirM(V45A), impA(A512T, D280G), PA0714(nt73InsG), gacS(G418S), xphA(P18L), mexY(G287S), PA2214(R21Q), PA2286(nt1474InsC), nuoM(F132L), lolC(N263S), nagZ(Q42R), dacB(S390N), PA3143(nt285Δ1), cpxS(L16P), nbdA(G189D), PA3313(L257P), hsbR(nt1527InsG), PA3532(Y336C), yedE(V27I), PA3732(K56R), ampC(L320P), fptA(L591P), uvrA(I511T), ftsI(A421V), dppA4(G65S), piuC(P82L), radA(T109A), PA4736(nt34InsC), livM(L132P), pilN(L6P), mcpK(G23D), PA5158(A103V, A252T)
	m2	1	0.25–0.5	>64	64	4	2	256	speB(D276G), PA0468(V175A), nirM(V45A), impA(A512T, D280G), PA0714(nt73InsG), gacS(G418S), xphA(P18L), mexY(G287S), PA2214(R21Q), PA2286(nt1474InsC), nuoM(F132L), lolC(N263S), nagZ(Q42R), dacB(S390N), PA3143(nt285Δ1), cpxS(L16P), nbdA(G189D), PA3313(L257P), hsbR(nt1527InsG), PA3532(Y336C), yedE(V27I), PA3732(K56R), ampC(L320P), fptA(L591P), uvrA(I511T), ftsI(A421V), dppA4(G65S), piuC(P82L), radA(T109A), PA4736(nt34InsC), livM(L132P), pilN(L6P), mcpK(G23D), PA5158(A103V, A252T)
2	m1	1–2	0.5	64	16	8	4	64	PA0123(R131H), PA0205(G121D), PA0239(T176A), PA0334(W110*), laoC(nt1383InsC), pgk(A223V), fumC2(G352D), aruC(D226G), itaE(nt862InsG), gacS(A227V), tolA(S12G), fgtA(A208V), fleQ(R138C), dgt(H344R), PA1389(F252L), PA1791(nt46InsC), cysI(W145R), PA1897(A9V), PA1918(F417S, nt82InsC), PA1990(P58L), asnB(R99H), PA2201(A90T), PA2298(D246G), vgrG(S593N), pvdP(Q541R), PA2431(R724C), muxC(T143A), PA2547(A16V), PA2650(A206T), eco(Q67R), PA2773(V17A), pelE(W102R), xcpS(L189P), PA3140(nt180Δ1), cpxS(L285P), PA3324(V378A), PA3504(nt1064InsG), PA3614(G380D), PA3886(F154L), PA4058(P106L), PA4071(nt245InsGG), bfiS(N746S), fptA(Q105*), PA4371(E303K), mexW(I635T), ftsI(L506P), piuC(L188P), pilB(W280*), fimU(W77*), PA4718(A146V), retS(T772M), PA4866(N40S), phaC(K11R), tatC(F80L), thiI(T288A), dgcA(T276A), PA5412(E252K), PA5532(D55G)
	m2	1–2	0.25	64	16–32	8	8	64	PA0123(R131H), spuA(V208A), PA0334(W110*), pgk(A223V), fumC2(G352D), aruC(D226G), itaE(nt862InsG), gacS(A227V), tolA(S12G), fgtA(A208V), dgt(H344R), PA1389(F252L), PA1791(nt46InsC), cysI(W145R), PA1897(A9V), PA2201(A90T), PA2262(R293W), PA2298(D246G), vgrG(S593N), pvdP(Q541R), PA2431(R724C), muxC(T143A), PA2547(A16V), PA2650(A206T), PA2773(V17A), pelE(W102R), xcpS(L189P), PA3140(nt180Δ1), cpxS(L285P), PA3324(V378A), PA3504(nt1064InsG), PA3614(G380D), PA3994(nt316InsG), PA4058(P106L), PA4071(nt245InsGG), bfiS(N746S), fptA(Q105*), purU(G162S), PA4371(E303K), ftsI(L506P), piuC(L188P), fimU(W77*), phaC(K11R), tatC(F80L), dgcA(T276A), PA5462(S30P), PA5532(D55G)
3	m1	1	0.25	>64	128	8	2	128	PA0061(nt251Δ1), siaB(V55A), PA0326(E241K), thyA(A189V), PA0884(R122C), alpD(nt484InsG), PA0987(S211L), fliJ(I55V), PA1211(F202S), hxuI(L122P), bkdB(S349N), uvrY(T180I), PA2777(nt37Δ1), PA2913(L104F), gltS(V342A), PA3264(L184P), hasD(W157R), tli5(nt636InsC), PA3592(P41S), spdH(G81S), PA3739(Y403H), PA3756(L123P), PA4041(P348S), ispA(F104L), ampC(L320P), fptA(Q93*), PA4319(F199L), pyeM(G106S), piuC(G214D), pilC(nt482Δ1), PA4576(P585L), cdrA(nt1364InsG), PA4739(E114K), PA5209(G47S), PA5310(A103T), uvrD(F536L), PA5529(G400S)
	m2	1	0.25	>64	128	4	2	128	PA0061(nt251Δ1), siaB(V55A), PA0326(E241K), thyA(A189V), PA0884(R122C), alpD(nt484InsG), PA0987(S211L), fliJ(I55V), PA1211(F202S), hxuI(L122P), bkdB(S349N), uvrY(T180I), PA2777(nt37Δ1), PA2913(L104F), gltS(V342A), PA3264(L184P), hasD(W157R), tli5(nt636InsC), PA3592(P41S), spdH(G81S), PA3739(Y403H), PA3756(L123P), PA4041(P348S), ispA(F104L), ampC(L320P), fptA(Q93*), PA4319(F199L), pyeM(G106S), piuC(G214D), pilC(nt482Δ1), PA4576(P585L), cdrA(nt1364InsG), PA4739(E114K), PA5209(G47S), PA5310(A103T), uvrD(F536L), PA5529(G400S)
ST175		8–16	1	8–16	16	2	0.5	0.12	oprD(Q142*), ampR(G154R), mexZ(G195D), gyrA(T83I, D87N), parC(S87W)
1	m1	8	1	8–16	16	4	2	8–16	cpxS(T163P), fptA(Q261*), piuC(nt506Δ2)
	m2	8	1	16	16	4	2	8–16	cpxS(T163P), fptA(Q261*), piuC(nt506Δ2)
2	m1	8	1	32	32	4	1	4	fptA(Q261*), PA4338(A191V), ftsI(R551C), piuC(L188Q)
	m2	8	1	32	32	4	1	4	clpA(aa465Δ3), fptA(Q261*), piuC(L188Q)
3	m1	2	0.5	>64	64	4	1	32-64	PA3602(Q233L), ampC(L320P), ΔpchE
	m2	2	0.5	>64	64	4	1	32	PA3602(Q233L), ampC(L320P), ΔpchE
ST111		4	0.5	8–16	8–16	1	0.5	0.06	oprD(nt991InsAC), parS(L137P), mexY(G530S), dacB(nt664InsGGCCT), gyrA(T83I), blaCARB-2
1	m1	4	0.5	8	8–16	2	0.5	1	pirR(nt180Δ1), PA2529(nt664InsAGCAG), pchR(W45*)
	m2	4	0.5	8	8	2	1	1	pirR(nt180Δ1), pchR(W45*)
2	m1	4–8	0.5	8	8	2	0.5	4	pirA(nt1100InsC), pchR(A270V), piuC(nt6InsG)
	m2	8	0.5	8	8	2	0.5	4	pirA(nt1100InsC), pchR(A270V), piuC(nt6InsG)
3	m1	8	0.5	8–16	8–16	2	0.5	1–2	pcm(nt504InsC), fptA(nt191InsG), piuC(nt103InsC)
	m2	0.5	0.5	16	128	16	32–64	0.12–0.25	pcm(nt504InsC), ampC(P243Q), rpsE(T86P)
ST235		8	1	16	4-8	2	0.5	1	oprD(nt1205InsC), gyrA(T83I), mexZ(V48A), pmrB(V344M), parC(S87L), blaOXA-2
1	m1	2	1	64	16	4	0.5	128	PA0434(nt8InsA), pirR(nt395InsG), PA1297(nt395InsG), PA1766(T202A), nuoM(W223*), ampC(L320P), piuC(W26*), morA(nt270InsC), gcbA(Q141R), mutL(Q619*)
	m2	2	1	64	16	4	0.5–1	128	PA0434(nt8InsA), PA0547(R105Q), pirR(nt395InsG), PA1768(A51T), pfeA(A603T), PA2345(A365V), PA2701(E35G), gdhB(A942V), ampC(L320P), piuC(L119P), morA(nt270InsC), gcbA(Q141R), mutL(Q619*), PA5436(H456R), yjbB(E572K)
2	m1	8	1	16	64	32	16	64	ftsY(aa130Ins2), piuC(G148D)
	m2	8	1	16	64	32	16	32	ftsY(aa130Ins2), piuC(G148D)
3	m1	8–16	4	16	16	4	1	32	galU(D67E), piuC(Q141*)
	m2	8–16	4	16	16	8	1	32	galU(D67E), piuC(Q141*)

IPM, imipenem; IMR, imipenem/relebactam; FEP, cefepime; CAZ, ceftazidime; CZA, ceftazidime/avibactam; C/T, ceftolozane/tazobactam; FDC, cefiderocol.

Data for two colonies (m1 and m2) from each of the three experiments per strain are shown.

MICs were determined in duplicate and both values obtained are shown when not identical.

Resistance genotype of the parental strain is indicated, and all acquired mutations detected in each of the mutants are listed.

Mechanisms of in vitro resistance development to cefiderocol in WT, mutator and XDR P. aeruginosa

In addition to the susceptibility profiles, Table 1 also includes the mutations acquired during the evolution in the presence of cefiderocol. WGS revealed 2–5 mutations for PAO1 lineages and 35–58 for PAOMS. The number of mutations in the XDR clinical strains ranged from 2 to 4, except for one of the ST235 experiments in which a mutL-deficient lineage was selected, thus increasing the number of mutations (10–15). The most frequently mutated genes were piuC (detected in 21 of 30 mutants) followed by fptA (detected in 15 of 30 mutants), both related to iron-uptake regulation. Other genes mutated in some strains related to iron-uptake regulation included pirR, pirA, pchE, pchR, pvdP, optJ and pfeA. Beyond iron-uptake genes, particularly noteworthy is the selection of a specific L320P AmpC mutation in ST235, ST175 and PAOMS lineages. Other noteworthy mutations include those in PBP3, detected in PAOMS derivatives and in one ST175 lineage. Likewise, cpxS mutations were detected in up to one-third of the mutants from different genomic backgrounds.

To assess the impact of the individual mutated genes on cefiderocol susceptibility, an available transposon mutant library was used.²⁰ Table 2 shows the cefiderocol MICs (and those of ceftazidime for comparative purposes) for selected transposon mutants. Cefiderocol MICs (unlike those of ceftazidime) were highly increased in piuA, piuC and pirR mutants, whereas the impact on MICs was modest (within one 2-fold dilutions) for other tested mutants, including the frequently mutated fptA.

Table 2.

Ceftazidime and cefiderocol MICs for selected mutants from the University of Washington transposon PAOUW library

PAOUW Transposon locus mutants	Gene name	Biological function involved	MIC (mg/L)a
			CAZ	FDC
WT	—	—	1	0.06–0.12
PA0434	—	siderophore transport	1	0.12
PA0547	—	regulation of transcription	1	0.5
PA0929	pirR	regulation of transcription	0.5–1	1
PA0931	pirA	siderophore transport	1	0.12–0.25
PA1297	—	cation/transmembrane transport	1	0.12
PA1768	—	unknown	1	0.12–0.25
PA1922	cirA/feuA	TonB-dependent outer membrane receptor	0.5–1	0.06
PA2345	—	oxidoreductase	0.25	0.06–0.12
PA2529	—	proteolysis	1	0.12–0.25
PA2620	clpA	protein unfolding	1-2	0.25–0.5
PA2648	nuoM	generation of precursor metabolites and energy	1	0.12
PA2701	—	transmembrane secretion	1	0.12–0.25
PA3068	gdhB	cellular amino acid metabolic process	1	0.12–0.25
PA3206	cpxS	phosphorelay sensor kinase activity	1	0.25
PA3421	—	unknown	1	0.12–0.25
PA3602	yerD	glutamate biosynthetic process	1	0.25
PA3624	pcm	protein repair	1	0.25
PA4221	fptA	siderophore transport	1	0.12–0.25
PA4226	pchE	pyochelin biosynthetic process	1	0.25
PA4227	pchR	regulation of transcription	1	0.25
PA4246	rpsE	translation	1	0.12
PA4338	—	transferase activity	1	0.25
PA4514	piuA	siderophore transport	1	2
PA4515	piuC	iron ion binding	1	2
PA4601	morA	motility regulator	1	0.12–0.25
PA4701	—	unknown	1	0.25
PA4843	gcbA	phosphorelay signal transduction	1	0.25
PA5436	—	ligase activity	1	0.25
PA5473	yjbB	sodium-dependent phosphate transport	1	0.25–0.5

CAZ, ceftazidime; FDC, cefiderocol.

MICs were determined in duplicate and both values obtained are shown when not identical.

Impact of the L320p AmpC mutation on cefiderocol resistance

The effect of the L320P (L294P of mature protein) mutation, specifically and independently selected by cefiderocol exposure in different strains and lineages, on AmpC structure is represented in Figure 2. The L320P mutation is located in the R2 loop of AmpC and is found to widen the active site, likely making the enzyme more accessible to β-lactam molecules carrying a bulky R2 side chain such as cefepime.²⁶ Thus, L320P AmpC variant was cloned, expressed in PAOΔC and tested in parallel with cloned WT AmpC, as well as AmpC variants previously shown to be involved in ceftolozane/tazobactam resistance development in vitro and in vivo (T96I, G183D or E247K). MICs of cefiderocol and other relevant β-lactams are shown in Table 3. The L320P AmpC mutation produced a 16-fold increase in cefiderocol MICs as compared with WT AmpC. This effect on cefiderocol resistance was similar or slightly higher than that documented for classical mutations determining ceftolozane/tazobactam resistance. Additionally, the L320P mutation had a much greater impact on cefepime MICs than did T96I, G183D or E247K mutations. However, remarkably, unlike those, the L320P mutation barely affected ceftolozane/tazobactam or ceftazidime/avibactam MICs (Table 3).

Figure 2.

WT AmpC (4GZB) (a, b) and L320P AmpC variant (c, d). Surface representation (a, c) of each variant with conserved β-lactamase motifs SVSK (red), YSN (green) and KTG (blue), and residue 320 (orange) indicated. AmpC active site detail (b, d), residues responsible for β-lactamase activity are represented and labelled. Numbering of residues for mature protein is indicated in brackets.

Table 3.

MICs of imipenem, cefepime, ceftazidime, ceftazidime/avibactam, ceftolozane/tazobactam and cefiderocol for cloned AmpC variants expressed in an ampC-deficient mutant PAO1

Strain		MIC (mg/L)a
	IPM	IMR	FEP	CAZ	CZA	C/T	FDC
PAO1	1	0.25	1	1	0.5–1	0.25	0.12
PAOΔC	0.25	0.25	1	1	0.5	0.25	0.12
PAOΔC + pUCPampC-WT	0.5	0.25	8	16	1	0.5	0.25
PAOΔC + pUCPampC-L320P	0.25	0.25	64–128	64	1	1	4
PAOΔC + pUCPampC-G183D	0.25	0.12–0.25	4	32	8	16	2
PAOΔC + pUCPampC-E247K	0.25	0.25	4–8	>64	8	64	4
PAOΔC + pUCPampC-T96I	0.25	0.25	4	32	4–8	32	2

IPM, imipenem; IMR, imipenem/relebactam; FEP, cefepime; CAZ, ceftazidime; CZA, ceftazidime/avibactam; C/T, ceftolozane/tazobactam; FDC, cefiderocol.

MICs were determined in duplicate and both values obtained are shown when not identical.

Discussion

According to the WHO, MDR P. aeruginosa, along with Acinetobacter baumannii and Enterobacterales, is on the top of the list of pathogens for which the development of novel antibiotic treatments is critical.⁴ The recent development of novel β-lactam/β-lactamase inhibitor combinations (such as ceftolozane/tazobactam, ceftazidime/avibactam or imipenem/relebactam) has helped to mitigate, to some extent, the problem of MDR/XDR P. aeruginosa⁸ but they are not exempt from resistance development, evidenced right upon their introduction into clinical practice.^9,10 Cefiderocol, a novel siderophore cephalosporin, provides an additional step in the fight against P. aeruginosa antimicrobial resistance. Thanks to massive entrance through iron transport systems, cefiderocol shows antipseudomonal activity even against strains producing potent acquired β-lactamases conferring resistance to all available β-lactams such as the MBLs.^27,28 However, as with any newly introduced antibiotic, the potential for resistance development needs to be closely monitored. Indeed, here we show that high-level cefiderocol resistance may emerge from WT strains devoid of any acquired resistance mechanisms. Moreover, as expected, resistance evolution is dramatically enhanced in a mutator background. On the other hand, the evolution of cefiderocol resistance in XDR strains showing diverse resistance mechanisms to classical antipseudomonal β-lactams was found to be strain dependent. Despite cefiderocol resistance being readily selected, the positive finding of our work is the limited cross-resistance with other β-lactams, notably the carbapenems and the novel β-lactam/β-lactamase inhibitor combinations (such as ceftolozane/tazobactam, ceftazidime/avibactam). As could have been expected, this specificity was partially driven by the selection of mutations related to iron-uptake systems as key cefiderocol resistance mechanisms. Our results are in agreement with previous findings showing that mutations in TonB-dependent receptors such as piuA/piuC, pirA/pirR or fptA (pyochelin receptor) are frequently selected in the presence of siderophore β-lactams.^29–31 Our results point towards piuC as a particularly relevant cefiderocol resistance target, according to both frequency of selection across different strains and resistance level conferred. piuC encodes an iron-dependent oxygenase involved in the expression of the adjacent piuA (or its homologue piuD, depending on the strain) iron receptor. On the other hand, ftpA mutations, despite being frequent, do not seem to have a direct major impact on cefiderocol MICs, and thus selection might reflect adaptive mutations for growing in the presence of cefiderocol. Mutations in these genes (piuA/piuC, pirA/pirR or fptA) were not found to be associated with cefiderocol resistance among isolates from chronically infected cystic fibrosis patients not treated with this antibiotic, but genomic variations in iron-uptake genes were associated with increased MICs.³² Thus, emergence of such mutants during cefiderocol treatment should be closely monitored when used in clinical practice.

A particularly relevant finding of this work is the selection of a specific mutation (L320P) in AmpC. This mutation had a major impact on cefiderocol MICs without significantly affecting those of ceftolozane/tazobactam or ceftazidime/avibactam. Indeed, previous work has focused on analysing the impact of ceftolozane/tazobactam and ceftazidime/avibactam resistance mutations on cefiderocol susceptibility,^27,33 but the possibility of the occurrence of cefiderocol resistance mutations not conferring cross-resistance to the novel β-lactam/β-lactamase inhibitor combinations has not been previously considered and has obvious major clinical implications: (i) ceftolozane/tazobactam and/or ceftazidime/avibactam susceptibility may not predict cefiderocol susceptibility; and (ii) ceftolozane/tazobactam and/or ceftazidime/avibactam might be an option as rescue therapy in some cases of cefiderocol resistance development. The L320P mutation is located in the R2 loop of AmpC and is found to widen the active site, likely making the enzyme more accessible to β-lactam molecules carrying a bulky R2 side chain.²⁶ Indeed, it should be noted that up to nine AmpC variants containing the L320P mutation have already been described (https://arpbigidisba.com/pseudomonas-aeruginosa-derived-cephalosporinase-pdc-database/) and that they have been previously associated with high-level ceftazidime and cefepime resistance in a certain (low) number of clinical strains much before the introduction of cefiderocol.³⁴ Moreover, similar mutations in the R2 loop of Enterobacter cloacae AmpC, associated with cefepime resistance and reduced susceptibility to cefiderocol, have also been recently documented.³⁵ Thus, monitoring of the emergence and prevalence of such AmpC variants upon cefiderocol introduction seems crucial.

Another relevant previously described mechanism of resistance development for ceftolozane/tazobactam and ceftazidime/avibactam is the selection of mutations in the catalytic centre of acquired OXA-2 or OXA-10 enzymes.^9,36,37 However, unlike what was previously documented for ceftolozane/tazobactam,¹³ an evolution of the spectrum of the OXA-2 enzyme produced by the used ST235 was not observed with cefiderocol in this work. These results are, to a certain extent, in agreement with recent results showing only a modest impact on cefiderocol susceptibility by OXA-2/OXA-10 mutations conferring ceftolozane/tazobactam and ceftazidime/avibactam resistance.²⁷ Thus, our data may indicate that the risk of selection of resistance to cefiderocol might be lower than for ceftolozane/tazobactam and ceftazidime/avibactam among OXA enzyme producers.

Likewise, another relevant previously described mechanism of resistance development for ceftolozane/tazobactam and ceftazidime/avibactam is the selection of mutations in the β-lactam recognition site of PBP3. Our results, consistently with previous findings in P. aeruginosa and other species,^27,33,38 indicate that this type of mutation may also play a role in cefiderocol resistance. Finally, another frequently mutated gene upon cefiderocol exposure was cpxS. Although a major increase in cefiderocol MICs for the corresponding transposon mutant was not documented, it is noteworthy that all mutants documented showed amino acid changes and not directly inactivating mutations such as frameshifts or stop codons. Thus, the possibility that the mutations detected might be gain-of-function mutations is currently under investigation in our laboratory. In any case, it is particularly noteworthy that CpxS has been recently shown to be a two-component sensor involved in cellular hysteresis response to β-lactam antibiotics, driven by the selection of the same mutation (T163P) detected in one of our cefiderocol-evolved lineages.³⁹

In summary, in this work we deciphered the potential resistance mechanisms that may emerge upon the introduction of the novel siderophore cephalosporin cefiderocol in clinical practice, including both specific and common mechanisms of resistance as compared with other novel β-lactams. Moreover, our results highlight that resistance development may occur even in WT strains and that the risk of resistance development might be strain-specific for XDR high-risk clones. However, larger further studies are required to decipher which specific markers of the strains are associated with lower or higher risk of resistance development, and the impact of the production of acquired ESBLs or carbapenemases still needs to be determined. Likewise, our study did not test the order of selection of the different resistance mutations, and neither did it evaluate their clinical relevance.

Funding

This work was supported by Plan Nacional de I + D + i 2013–2016 and Instituto de Salud Carlos III, Subdirección General de Redes y Centros de Investigación Cooperativa, Ministerio de Economía, Industria y Competitividad, Spanish Network for Research in Infectious Diseases (REIPI RD16/0016) and grant PI21/00017 co-financed by European Development Regional Fund ERDF “A way to achieve Europe”, Operative program Intelligent Growth 2014–2020.

Transparency declarations

A.O. has received speaker fees and research grant funding from Shionogi, MSD and Pfizer.

Data availability

Sequence files have been deposited in the European Nucleotide Archive under study number PRJEB58820.