High-level ceftazidime/avibactam resistance in Escherichia coli conferred by the novel plasmid-mediated β-lactamase CMY-185 variant

Abstract

Objectives

To characterize a bla_CMY variant associated with ceftazidime/avibactam resistance from a serially collected Escherichia coli isolate.

Methods

A patient with an intra-abdominal infection due to recurrent E. coli was treated with ceftazidime/avibactam. On Day 48 of ceftazidime/avibactam therapy, E. coli with a ceftazidime/avibactam MIC of >256 mg/L was identified from abdominal drainage. Illumina and Oxford Nanopore Technologies WGS was performed on serial isolates to identify potential resistance mechanisms. Site-directed mutants of CMY β-lactamase were constructed to identify amino acid residues responsible for ceftazidime/avibactam resistance.

Results

WGS revealed that all three isolates were E. coli ST410. The ceftazidime/avibactam-resistant strain uniquely acquired a novel CMY β-lactamase gene, herein called bla_CMY-185, harboured on an IncI-γ/K1 conjugative plasmid. The CMY-185 enzyme possessed four amino acid substitutions relative to CMY-2, including A114E, Q120K, V211S and N346Y, and conferred high-level ceftazidime/avibactam resistance with an MIC of 32 mg/L. Single CMY-2 mutants did not confer reduced ceftazidime/avibactam susceptibility. However, double and triple mutants containing N346Y previously associated with ceftazidime/avibactam resistance in other AmpC enzymes, conferred ceftazidime/avibactam MICs ranging between 4 and 32 mg/L as well as reduced susceptibility to the newly developed cephalosporin, cefiderocol. Molecular modelling suggested that the N346Y substitution confers the reduction of avibactam inhibition due to steric hindrance between the side chain of Y346 and the sulphate group of avibactam.

Conclusions

We identified ceftazidime/avibactam resistance in E. coli associated with a novel CMY variant. Unlike other AmpC enzymes, CMY-185 appears to require an additional substitution on top of N346Y to confer ceftazidime/avibactam resistance.

Introduction

Avibactam is a class A, C and D β-lactamase inhibitor that can improve β-lactam activity against several Gram-negative organisms.¹ The potent inhibitory profile of avibactam has made ceftazidime/avibactam one of the preferred therapies for infections caused by Klebsiella pneumoniae carbapenemase (KPC)-producing Enterobacterales.² Ceftazidime/avibactam is also highly active against non-carbapenemase-producing carbapenem-resistant Enterobacterales clinical isolates including Klebsiella aerogenes, Enterobacter cloacae and others that produce AmpC or ESBLs.³

Nevertheless, there are certain AmpC variants that have reduced avibactam inhibition. Through a series of experimental passaging studies, a combination of ceftazidime or aztreonam with avibactam has been shown to select for a variety of amino acid hotspot mutations in derepressed chromosomal AmpC (cAmpC), including N346Y variants found in Citrobacter freundii and Enterobacter cloacae.^4,5 The functional role of the N346Y substitution has been characterized in recombinant, isogenic E. coli strains producing cAmpC of E. cloacae, PDC-5 (cAmpC of Pseudomonas aeruginosa) and DHA-1 (cAmpC of Morganella morganii) with each respective variant conferring increased ceftazidime/avibactam MICs as a result of reduced avibactam activity.⁶ A structural analysis of the AmpC complex with avibactam demonstrated that cAmpC N346 was one of eight conserved residues that specifically interact with the sulphate group of avibactam.⁷ Furthermore, a cAmpC N346T substitution would be predicted to create steric hindrance with the sulphate group and thus affect avibactam binding affinity across multiple cAmpC homologues.⁷

CMY-2 is the most common plasmid-associated AmpC (pAmpC) β-lactamase produced by Escherichia coli and other Enterobacterales species.⁸ The bla_CMY-2 gene is often detected in association with the mobile ISEcp1 element, which is likely responsible for its wide transmission across multiple Enterobacterales species.^9,10 A CMY variant, CMY-172, with K290_V291_A292del, N346I compared with CMY-2, has been reported to confer high-level ceftazidime/avibactam resistance in K. pneumoniae clinical isolates co-producing KPC-2 and CTX-M-65 in China.¹¹

Here, we report serial ST410 E. coli clinical isolates where the final ST410 serial isolate acquired an IncI-γ/K1 plasmid that encodes a novel CMY-2 variant, designated as CMY-185. This novel CMY-2 variant has a four amino acid substitution pattern that is associated with ceftazidime/avibactam resistance as well as cefiderocol resistance. We confirmed the ceftazidime/avibactam resistance correlation with this novel CMY-185 mutant through site-directed mutagenesis of CMY-2 to determine the contribution of single versus combination mutations observed within CMY-185.

Materials and methods

Strains and susceptibility testing

The E. coli isolates were identified from a surgical drain of a patient admitted to a hospital in Texas. All three E. coli isolates (Ec1 through Ec3) were available for further laboratory evaluation. Initial susceptibility testing was performed with Vitek2 or Etest (bioMérieux, Marcy L’Étoile, France) per routine clinical practice at the hospital laboratory. For the purposes of testing cefiderocol susceptibility, broth microdilution was performed in triplicate with iron-depleted Mueller–Hinton broth and MIC reported in the range of 0.03–32 mg/L. Resistance was determined using CLSI M100 standards (29th edition).¹² MIC testing of the recombinant strains was conducted by the broth microdilution method.

WGS and computational analyses

Genomic DNA (gDNA) of E. coli isolates Ec1, Ec2 and Ec3 was extracted using the QIAGEN DNeasy blood and tissue kit following manufacturer’s instruction. Library preparation for WGS was completed using the Illumina Nextera kit and sequenced with the Illumina NovaSeq 6000 platform using 150 bp paired-end reads. FastQC-v0.11.9 was used to check short-read quality. Genomic DNA from each of the three isolates was next subjected to Oxford Nanopore Technologies long-read sequencing. The SQK-RBK114.96 library prep was used with 100 ng of gDNA input total for each isolate and then subsequently sequenced on a MinION flow cell (R10.4.1) following manufacturer’s instructions. Fast5 files were base-called, and adapters trimmed using the Guppy-v6.4.2 GPU base-caller with the dna_r10.4.1_e8.2_400bps_sup.cfg config file. Long-read quality control (QC) was performed using NanoPlot-v1.33.0. QC long-reads were used as inputs for Flye-2.9.2 assemblies with the —nano-hq read input option intended for Q20+ reads. Subsequent long-read polishing occurred with medaka-v1.8.0 using the medaka_consensus script and the r1041_e82_400bps_sup_g615 model. Illumina paired-end reads were used to polish the medaka output using polypolish-v0.5.0 with the polypolish_insert_filter.py python script to filter a subset of short reads, which reduces erroneous error correction based on insert size. Lastly, we polished with the MaSuRCA-v4.1.0 polca.sh short-read polish script and subsequently inspected the final assembly using snippy-v4.6.0 (T. Seemann, Snippy GitHub: https://github.com/tseemann/snippy) in conjunction with a minimap2 long-read alignment to check for erroneous variants. Long-read assemblies were quality controlled using BUSCO-v5.4.7 with all three assemblies having C:100.0%[S:100.0%, D:0.0%], F:0.0%, M:0.0%, n:124 BUSCO scores using the bacteria_odb10 (2020-03-06) set.

Antimicrobial resistance genes were detected using AMRFinderplus-v3.11.14 with the NCBI Reference Gene database (database version: 2023-04-17.1).¹³ Plasmid replicon typing was performed with the MOB-Suite-v3.0.0 (database download: 2023-06-01) mob_typer function, which can also predict likely conjugation through identification of a relaxase and coupling protein encoding gene.¹⁴ The snippy-multi script (T. Seemann, Snippy GitHub: https://github.com/tseemann/snippy) was used to create a core SNP phylogeny of the study isolates using a previously sequenced ST410 isolate (MB9108) as reference (RefSeq #: GCF_024917735.1). Gubbins-v.3.2.1 was used to create a recombination-free, core genome inferred maximum-likelihood phylogeny.¹⁵ All ST410 isolates were aligned to an IncIγ bla_CMY-42-positive plasmid, pCMY42-035148 (RefSeq #: GCF_003268695.1), using bwa mem and Samtools-v1.14 to subsequently check breadth of IncIγ coverage to further substantiate mob_typer results. An alignment of ftsI encoding genes, i.e. PBP3, from SPAdes assemblies was performed using mafft-v7.505 with mutations called based on comparison with MG1655 reference (Accession #: NC_000913.3). Pairwise SNP distances removing recombination sites were determined using snp-sites-v2.5.1¹⁶ and snp-dists-v0.8.2, respectively, using the gubbins.filtered_polymorphic_sites.fasta file as an input file. Tree visualization and metadata annotation was performed using the ‘ggtree-3.3.1’ R package.¹⁷

Conjugation experiment of IncI-γ/K1, pMB7671_5 plasmid

A conjugation experiment was conducted using E. coli J53AziR as the recipient by the broth mating method using a 1:10 donor:recipient ratio. Recipient cells were selected for with 50 mg/mL sodium azide whereas the pMB7671_5 plasmid was selected for with 25 mg/mL ampicillin and confirmed by PCR for bla_CMY-2. Conjugation frequency was calculated based on the number of transconjugants divided by the number of donor cells (cfu/mL). Antimicrobial susceptibility for selected antibiotics was tested using the Kirby–Bauer disc diffusion method.

Construction of CMY mutants

The CMY-2 gene was cloned into vector pBC SK(−) using the primers CMY-For XbaI (5′-GCTCTAGACATATGATGAAAAAATCGTTATGCTG-3′) and CMY-Rev BamHI (5′-CGGGATCCTTATTGCAGCTTTTCAAGAATG-3′), which was then electroporated into E. coli TOP10 as previously described.¹⁸ Single, double and triple mutants of CMY-2 were generated from the CMY-2-encoding vector using Q5 CMY_A114E_For (5′-ACCTATACGGAAGGCGGCCTA-3′) and Q5 CMY_A114E_Rev (5′-GGCTAAGTGCAGCAGGCG-3′), Q5 CMY_Q120K_For (5′-CCTACCGCTGAAGATCCCCGA-3′) and Q5 CMY_Q120K_Rev (5′-CCGCCTGCCGTATAGGTG-3′), Q5 CMY_V211S_For (5′-GCCCGTACACAGTTCTCCGGGACAAC-3′) and Q5 CMY_V211S_Rev (5′-TTCCCTTCGCGATAGCCC-3′), and Q5 CMY_N346Y_For (5′-AAGCTATCCTTACCCTGTCCG-3′) and Q5 CMY_N346Y_Rev (5′-TTGTTTGCCAGCATCACG-3′) and electroporated into E. coli TOP10. All mutations were confirmed by Sanger sequencing. Figure 1 is a flowchart documenting the process by which CMY mutants were generated using CMY-2 as the baseline allele.

Figure 1.

Overview of site-directed mutagenic CMY-2 variants. Parentheses indicate the amino acid mutation relative to CMY-2. A total of 15 CMY-2 mutants were created using a PCR-based mutagenesis method and subsequently assessed using antimicrobial susceptibility testing.

Molecular modelling

A homology model of CMY-185 was generated using the CMY-136 structure [Protein Data Bank (PDB) accession number 6G9T] as the template using the Modeller version 10.2.¹⁹ Ceftazidime or avibactam bound to CMY-185 was modelled by superimposition of the crystal structures of the AmpC^Ent385 complex with ceftazidime (PDB accession number 6LC9) or avibactam (PDB accession number 6LC8).²⁰

Data availability

E. coli Ec1, Ec2 and Ec3 WGS data have been deposited in BioProject PRJNA924946. ST410 isolate WGS data from previous studies can be acquired in BioProject PRJNA836696 and BioProject PRJNA388450.^21,22

Ethics

The collection of prospective clinical isolates and deidentified data was approved by the MD Anderson Institutional Review Board (IRB No: PA13-0334).

Results and discussion

Clinical case

A male with a retroperitoneal liposarcoma was admitted for an abdominal tumour resection with reconstruction. One week postoperatively he developed fever, which was treated with piperacillin/tazobactam for 1 week, at which time he was taken to surgery for a washout with drain placement. Two weeks after surgery, surgical drains grew a piperacillin/tazobactam-resistant E. coli (baseline in Table 1), for which he was placed on ertapenem. After 2 weeks, the drain clogged and was repositioned; cultures obtained at that time grew carbapenem-resistant E. coli (Day 14 in Table 1). He was placed on ceftazidime/avibactam at this time with a planned 8 week treatment course. Near the end of therapy, the drain began bleeding, and he was transferred to the ICU. A drain culture obtained at that time grew a ceftazidime/avibactam-resistant, carbapenem-susceptible E. coli (Day 65 in Table 1).

Table 1.

Susceptibility testing results of serial ST410 E. coli isolates

Antimicrobials	MICs (mg/L)
	E. coli EC1 (Baseline)	E. coli EC2 (Day 14)	E. coli EC3 (Day 65)
Aztreonam (ATM)	2 (S)	4 (S)	32 (R)
Piperacillin/tazobactam (TZP)	≥128 (R)	≥128 (R)	≥128 (R)
Cefotaxime (CTX)	4 (R)	8 (R)	≥64 (R)
Ceftriaxone (CRO)	4 (R)	8 (R)	≥64 (R)
Ceftazidime (CAZ)	4 (S)	4 (S)	≥64 (R)
Ceftazidime/avibactam (CZA)	0.75 (S)b	1.5 (S)b	>256 (R)b
Cefepime (FEP)	≤1 (S)	8 (SDD)	16 (R)
Ertapenem (ETP)	≤0.5 (S)	4 (R)	0.5 (S)a
Imipenem (IMI)	≤0.25 (S)	2 (I)a	2 (I)a
Meropenem (MEM)	≤0.25 (S)	2 (I)a	1 (S)
Cefiderocol (FDC)c	0.5 (S)	0.25 (S)	8 (R)

All testing was performed using Vitek2 unless otherwise indicated. R, resistant; S, susceptible; SDD, susceptible dose dependent.

Testing completed with Etest (bioMérieux).

Testing completed with gradient diffusion strip (Liofilchem).

Testing completed with the iron-depleted broth microdilution protocol (see Methods).

Phenotypic and genotypic resistance of E. coli clinical isolates

The baseline E. coli isolate Ec1 was resistant to ceftriaxone and aztreonam, but susceptible to cefepime and carbapenems (Table 1). E. coli isolate Ec2, which was collected after 2 weeks of ertapenem therapy, was additionally resistant to ertapenem, and intermediate to imipenem and meropenem, whereas it remained susceptible to ceftazidime/avibactam. E. coli isolate Ec3 was collected after a prolonged course of ceftazidime/avibactam and was highly resistant to this agent with an MIC of >256 mg/L. Additionally, Ec3 developed resistance to cefiderocol, a siderophore cephalosporin recently approved for clinical use. Co-resistance to ceftazidime/avibactam and cefiderocol has been reported to occur in AmpC with genetic changes in the R2 region through increased hydrolytic efficiency of ceftazidime and cefiderocol.^20,23,24 It was also resistant to other tested cephalosporins but susceptible to all carbapenems including ertapenem (Table 1).

We determined that all three isolates belonged to ST410 based on in silico typing using WGS data. For the purpose of determining genetic relatedness of these serial isolates in the context of ST410 we have previously sequenced,^21,22 we created a maximum likelihood phylogenetic tree inferred from a recombination-free, core genome alignment of 15 ST410 isolates collected from 10 patients (Figure 2a). After accounting for recombination regions, there were 831 polymorphic sites with a median pairwise SNP distance of 210 (IQR = 52.5). Notably, Ec1, Ec2 and Ec3 had <10 pairwise SNP distances. The prominent genomic difference detected between Ec3 and its antecedent isolates (Ec1 and Ec2) was the acquisition of a 34 280 bp IncI-γ/K1 plasmid (i.e. pMB7671_5; GenBank Accession #: CP127853.1) that harbours the class C β-lactamase pAmpC gene bla_CMY-185 (Accession #: OQ297612.1) as highlighted in Figure 2(b). The pMB7671_5 plasmid carries a mating pair formation-encoding gene (i.e. trbC), relaxase-encoding gene (i.e. nikB) and an oriT site (14 854 to 14 946 bp), which are plasmid features predictive of conjugation proficiency.¹⁴ The conjugation frequency of MB7671 with the recipient E. coli J53AziR strain was 1.1 × 10⁻⁷ transconjugants per donor cell. We further confirmed acquisition of bla_CMY-185 through the change in susceptibility profiles of the recipient strain, with the transconjugant strain notably developing ceftazidime/avibactam resistance (Figure 2b).

Figure 2.

Characterization of ST410 E. coli with putative bla_CMY-185 vector. (a) ST410 population structure with the bla_CMY-185-harbouring strain. A total of 15 ST410 isolates from 10 patients were identified through WGS from 2013 to 2019 locally. Ec1, Ec2 and Ec3 are indicated on the tree having <10 pairwise SNPs within this group. Ec3 is highlighted in red font with the novel bla_CMY-185 variant (purple) indicated in the column. All IncIγ-positive isolates had 100% breadth of coverage when aligned with pCMY42-035148 (RefSeq #: GCF_003268695.1). ftsI and ompC mutations, which are predicted to contribute to reduced susceptibility to β-lactams based on previous studies,^26,28 are indicated in the legend. (b) IncI-γ/K1 description harbouring bla_CMY-185 with Kirby–Bauer (KB) disc diffusion testing results from conjugation experiments. The 34 kbp plasmid has the ORF labelled based on function as highlighted in legend. The bla_CMY-185-encoding gene is labelled in red. Zone diameters for the recipient (J53) and the transconjugant clone (J53 + pMB7671_5) are in the accompanying table with susceptibility (S) and resistance (R) labelled based on the CLSI guidelines.¹² AMP, ampicillin; CAZ, ceftazidime; CAZ + APB, ceftazidime + 3-aminophenylboronic acid (300 mg); CZA, ceftazidime/avibactam; FOX, cefoxitin; MEM, meropenem.

The predicted CMY-185 β-lactamase differed from the most commonly detected pAmpC in E. coli, i.e. the CMY-2 enzyme, by four amino acid substitutions (A114E, Q120K, V211S, N346Y). The other three ST410 IncIγ-positive isolates in our population harboured bla_CMY-42, which encodes for an enzyme that only differs by one amino acid (V211S) relative to CMY-2 and showed moderate hydrolysis of ceftazidime compared with the CMY-2 enzyme.²⁵ All four IncIγ-positive ST410 isolates had 100% breadth of coverage to the pCMY42-035148 plasmid, which is a previously identified vector for bla_CMY-42 carriage (RefSeq #: GCF_003268695.1).

Interestingly, all bla_CMY-positive isolates had a mutation in the ftsI gene, which encodes PBP3, a key divisome target of cephalosporins and monobactams, where either a predicted ‘YRIN’ (i.e. YRIN N337N) or ‘K(P)YRI I336I’ duplication event occurred (Figure 1). Furthermore, Ec1, Ec2 and Ec3 had mutations in ftsI corresponding to three substitutions (Q227H + E349K + I532L). These PBP3 substitutions have been associated with reduced susceptibility to ceftazidime and cefepime,²⁶ to aztreonam/avibactam in the presence of bla_CTX-M-15,²⁷ and associated with the acquisition of carbapenemase genes in ST410 backgrounds.²⁸ An additional frameshift variation in the rseA gene (p.Ser111fs) was detected in Ec2 compared with Ec1, which may have contributed to the increase in ertapenem MIC (Table 1). Disrupting mutations of rseA, which encodes a sigma E anti-sigma factor, can reduce OmpF, OmpC and OmpA membrane content through the increase of DegP protease, which in turn can reduce carbapenem susceptibility.^29,30 All three isolates carried two copies of bla_TEM-1b adjacent to IS26 transposase genes on an IncFIA/IncFIB/IncFII/IncQ1 multireplicon plasmid. A full roster of all acquired antimicrobial resistance genes detected in Ec1–Ec3 is presented in Table 2. These results suggest that bla_CMY-185 may have developed through selective pressures that beget mutations in a bla_CMY-42 gene harboured on an IncI-γ/K1 conjugative plasmid.

Table 2.

Acquired antimicrobial resistance genes for serial ST410 isolates

Acquired resistance elements	E. coli Ec1	E. coli Ec2	E. coli Ec3
β-Lactams	bla TEM-1b	bla TEM-1b	bla TEM-1b, blaCMY-185
Aminoglycosides	aadA5 mph(A)strA/strB	aadA5, mph(A) strA/strB	aadA5, mph(A) strA/strB
Folate synthase inhibitors	dfrA17	dfrA17	dfrA17
Fluoroquinolones	qnrB7	qnrB7	qnrB7
Sulfonamides	sul1, sul2	sul1, sul2	sul1, sul2
Tetracyclines	tet(B), tet(D)	tet(B), tet(D)	tet(B), tet(D)

Ceftazidime/avibactam and cefiderocol resistance conferred by CMY-185

To determine the phenotype conferred by CMY-185, bla_CMY-185 and bla_CMY-2 were constitutively expressed in a cloning vector. CMY-185 conferred a 256-fold higher MIC of ceftazidime/avibactam compared with CMY-2, confirming the role of the four substitutions in the ceftazidime/avibactam resistance observed in Ec3 (Table 3). We were also able to confirm cefiderocol resistance associated with the CMY-185 gene, which conferred a 16-fold higher MIC of cefiderocol, compared with CMY-2. However, none of the substitutions in CMY-185 were located in the R2 region, and both increased hydrolysis of ceftazidime and impaired inhibition by avibactam appeared to contribute to its ceftazidime/avibactam resistance phenotype (see discussion below). Given that there is a 4-fold cefiderocol MIC increase in the clinical isolate Ec3 compared with the isogenic CMY-185 mutant, there are likely other factors such as the ftsI insertion mutations, which contribute to the cefiderocol phenotype as has been documented previously.^24,26

Table 3.

MICs of E. coli TOP10 strains carrying pBC SK(−) vectors that encode single, double, triple and quadruple (CMY-185) mutants of CMY-2

CMY-2 and its mutants	MICs (mg/L)
	CAZ	CZA	FDC	FEP	ATM	MEM
pBC SK(−) (control)	0.25	0.12	0.06	<0.03	0.12	<0.03
CMY-2	16	0.12	0.12	0.12	2	<0.03
Single mutants
A114E	8	0.25	0.12	0.25	4	<0.03
Q120K	>32	0.5	0.25	0.25	4	<0.03
V211S	>32	0.25	0.12	0.12	4	<0.03
N346Y	2	0.25	0.25	<0.03	0.25	<0.03
Double mutants
A114E Q120K	32	1	0.5	0.25	4	<0.03
A114E V211S	>32	0.25	0.25	0.12	16	<0.03
A114E N346Y	16	8	0.5	2	2	<0.03
Q120K V211S	>32	0.25	0.5	0.12	8	<0.03
Q120K N346Y	32	32	1	0.25	2	<0.03
V211S N346Y	32	4	0.5	0.5	8	<0.03
Triple mutants
A114E Q120K V211S	>32	2	0.5	0.25	8	<0.03
A114E Q120K N346Y	32	16	2	1	4	<0.03
A114E V211S N346Y	32	16	1	0.5	16	<0.03
Q120K V211S N346Y	>32	16	1	0.25	8	<0.03
Quadruple mutant (CMY-185)
A114E Q120K V211S N346Y	>32	32	2	0.5	16	<0.03

ATM, aztreonam; CAZ, ceftazidime; CZA, ceftazidime/avibactam; FDC, cefiderocol; FEP, cefepime; MEM, meropenem.

Impact of amino acid substitutions observed in CMY-185

Given the presence of four amino acid substitutions in CMY-185 relative to CMY-2, bla_CMY-2 mutants with single, double and triple amino acid substitutions contained in CMY-185 were constructed for every combination to decipher their roles in ceftazidime/avibactam resistance. None of the single amino acid mutant strains showed increased ceftazidime/avibactam MICs (Table 3). Rather, the N346Y mutant strain appeared to be impaired in its hydrolytic activity of ceftazidime and aztreonam, with a decrease in MICs by a factor of eight compared with the strain producing CMY-2. In contrast, all three double mutant strains containing the N346Y substitution showed ceftazidime/avibactam MICs that were 32- to 256-fold higher than that of the CMY-2-producing strain, without major differences in ceftazidime MICs. These observations suggest that secondary substitutions were required in addition to N346Y for the enzymes to confer ceftazidime/avibactam resistance, most likely due to impaired inhibition by avibactam. This trend was extended to triple mutations, where the three triple mutant strains containing N346Y showed a ceftazidime/avibactam MIC of 16 mg/L, a 128-fold increase compared with CMY-2. The triple mutant strain from which N346Y was absent (A114E, Q120K, V211S) had a ceftazidime/avibactam MIC of 2 mg/L, which was 16-fold higher than that of CMY-2. On balance, the quadruple mutant (CMY-185) had the highest MICs across the tested cephalosporins (ceftazidime, ceftazidime/avibactam, cefiderocol) except for cefepime. All the mutants remained susceptible to meropenem.

Molecular modelling

To visualize the substitution sites, we generated a homology-model structure of CMY-185 (Figure 3a). The model structure of CMY-185 showed that the A114E substitution is located at the inner molecule assembled by the hydrophobic contact with W100, I104, L109, L117 and W138. The Q120K, V211S and N346Y substitutions are located at the molecular surface where the substrate is bound. The N346 residue is a significant residue for interacting with the cephalosporins or avibactam and makes hydrogen bonds with the carboxyl group at the 4-position of the cephalosporin ring and the amide group at the 2-position of avibactam (Figure 3d, e). Compain et al.⁶ reported that the N346Y substitutions in class C β-lactamases AmpC_cloacae, PDC-5 and DHA-1 bring about large decreases in carbamoylation efficacy of avibactam due to steric hindrance by the bulky side chain of tyrosine (Figure 3b, c). In contrast, the impact of the N346Y substitution on the MIC of ceftazidime depends on the enzyme. The N346Y mutants of AmpC_cloacae or PDC-5 showed moderate hydrolysis of ceftazidime, whereas the N346Y mutant of DHA-1 showed a decrease in the MIC of ceftazidime by a factor of four.⁶ The single N346Y mutant of CMY-2 showed a decrease in the MIC of ceftazidime by a factor of eight compared with a WT CMY-2 in this study. These results suggest that the reduction of the inhibition of avibactam appears to be a major cause of the reduced susceptibility to ceftazidime/avibactam by the N346Y substitution in CMY-2. However, the secondary substitutions were required in addition to the N346Y substitution to confer ceftazidime/avibactam resistance in the CMY-2 mutants. Given the concerning susceptibility patterns, the additional A114E, Q120K and V211S substitutions appeared to restore the resistance to ceftazidime diminished by the N346Y substitution and further conferred resistance to cefiderocol, which contains a bulkier R2 side chain. Further studies are needed to understand the molecular mechanism for ceftazidime/avibactam or cefiderocol resistance conferred by the combination of N346Y substitution with A114E, Q120K and V211S substitutions.

Figure 3.

Model structure of CMY-185. (a) Overall structure of CMY-185. The substitution residues are shown as CPK representations coloured in magenta. To clarify the substrate-binding site, the ceftazidime molecule is shown as a white ball-and-stick representation. (b, c) Close-up views of the model structures of the CMY-185 complex with ceftazidime (b) or avibactam (c). (d, e) Close-up views of the crystal structures of the AmpC^Ent385 complex with ceftazidime (d, PDB accession number 6LC9) or avibactam (e, PDB accession number 6LC8). Hydrogen bonds are represented as orange dashed lines.

Conclusion

Our case illustrates how CMY-2, the most common pAmpC β-lactamase, can undergo a small number of amino acid changes to confer high-level ceftazidime/avibactam resistance in particular E. coli backgrounds. Additionally, this study also highlights a unique enzymatic feature whereby secondary substitutions complemented the β-lactam degradation activity due to the primary substitution, which is compromised by trading off the escape from the inhibitor. Given the widespread and transferable nature of CMY and other pAmpC enzymes in E. coli and other Enterobacterales species, this finding raises concern for additional cases of resistance with increasing usage of ceftazidime/avibactam.

Funding

This work was supported by the National Institute of Allergy and Infectious Diseases at the National Institutes of Health (R01AI104895, R21AI151362 to Y.D.; R21AI175821 to W.R.M.; R21AI151536, P01AI152999 to S.A.S.); a training fellowship from the Gulf Coast Consortia, on the Texas Medical Center Training Program in Antimicrobial Resistance (TPAMR) (T32AI141349 to W.C.S. and S.L.E); A. Kawai is partly supported by a grant from the Takeda Science Foundation. Core grant CA016672(ATGC) and NIH 1S10OD024977-01 grant provide funding for the Advanced Technology Genomics Core (ATGC) sequencing facility at The University of Texas MD Anderson Cancer Center.

Transparency declarations

W.R.M. has received research support from Merck and royalties from UpToDate. Y.D. has received research support from Entasis and Shionogi, consulting fees from Gilead, Moderna, GSK, MeijiSeika Pharma, Shionogi and bioMérieux, and honoraria from Gilead, MSD and Shionogi. All other co-authors have none to declare.