Modifiable Risk Factors for the Emergence of Ceftolozane-tazobactam Resistance Free

Abstract

Background

Ceftolozane-tazobactam (TOL-TAZ) affords broad coverage against Pseudomonas aeruginosa. Regrettably, TOL-TAZ resistance has been reported. We sought to identify modifiable risk factors that may reduce the emergence of TOL-TAZ resistance.

Methods

Twenty-eight consecutive patients infected with carbapenem-resistant P. aeruginosa isolates susceptible to TOL-TAZ, treated with ≥72 hours of TOL-TAZ , and with P. aeruginosa isolates available both before and after TOL-TAZ exposure between January 2018 and December 2019 in Baltimore, Maryland, were included. Cases were defined as patients with at least a 4-fold increase in P. aeruginosa TOL-TAZ MICs after exposure to TOL-TAZ. Independent risk factors for the emergence of TOL-TAZ resistance comparing cases and controls were investigated using logistic regression. Whole genome sequencing of paired isolates was used to identify mechanisms of resistance that emerged during TOL-TAZ therapy.

Results

Fourteen patients (50%) had P. aeruginosa isolates which developed at least a 4-fold increase in TOL-TAZ MICs(ie, cases). Cases were more likely to have inadequate source control (29% vs 0%, P = .04) and were less likely to receive TOL-TAZ as an extended 3-hour infusion (0% vs 29%; P = .04). Eighty-six percent of index isolates susceptible to ceftazidime-avibactam (CAZ-AVI) had subsequent P. aeruginosa isolates with high-level resistance to CAZ-AVI, after TOL-TAZ exposure and without any CAZ-AVI exposure. Common mutations identified in TOL-TAZ resistant isolates involved AmpC, a known binding site for both ceftolozane and ceftazidime, and DNA polymerase.

Conclusions

Due to our small sample size, our results remain exploratory but forewarn of the potential emergence of TOL-TAZ resistance during therapy and suggest extending TOL-TAZ infusions may be protective. Larger studies are needed to investigate this association.

Pseudomonas aeruginosa infections are associated with significant morbidity and mortality in hospitalized patients—particularly those with underlying medical conditions such as cystic fibrosis, malignancies, severe burns, or indwelling foreign material [1, 2]. Rates of both P. aeruginosa infections and carbapenem-resistant P. aeruginosa are rising [3]. Reduced effectiveness of β-lactam antibiotics by P. aeruginosa generally evolves because of an interplay of several mechanisms including alterations in the structure or function of OprD porins, hyperproduction or structural modification to AmpC β-lactamases (Pseudomonas-derived cephalosporinases [PDCs]), upregulation of efflux pumps (eg, MexAB-OprM, MexCD-OprJ), and mutations in penicillin-binding proteins, and—although uncommon in the United States—carbapenemase production [4, 5].

Recently, there have been several β-lactam-β-lactamase inhibitor (βL-βLI) combination agents developed with activity against carbapenem-resistant P. aeruginosa; one being ceftolozane-tazobactam (TOL-TAZ) [6]. Preclinical investigations indicate TOL-TAZ activity against approximately 85%–95% of US and Canadian carbapenem-nonsusceptible P. aeruginosa isolates [7, 8]. However, soon after the clinical introduction of TOL-TAZ, reports of resistance during therapy emerged [9–11]. Several of the chromosomal mutations identified reside within or adjacent to the AmpC omega loop (R164 to D179), a conserved region of the oxyminocephalosporin binding site [9–12]. Significant gaps in knowledge remain as we continue to develop novel agents to treat multidrug-resistant (MDR) P. aeruginosa infections including: (1) a clear understanding of bacterial chromosomal mutations and any notable acquired resistance genes that contribute to the emergence of clinically relevant TOL-TAZ resistance, (2) the frequency of and mutations leading to cross-resistance between TOL-TAZ and other novel βL-βLIs, and (3) an understanding of modifiable risk factors that may slow or even prevent the emergence of TOL-TAZ resistance so it continues to be a viable treatment option. Herein, we sought to address these questions using a cohort of patients infected with carbapenem-resistant P. aeruginosa with clinical isolates available before and after TOL-TAZ exposure.

METHODS

Study Population

This study occurred at The Johns Hopkins Hospital (JHH), a 1162-bed academic medical center in Baltimore, Maryland. We identified 48 unique patients with carbapenem-resistant P. aeruginosa isolates treated with at least 72 hours of TOL-TAZ between January 2018 and December 2019. TOL-TAZ use at JHH requires approval by the Antimicrobial Stewardship Program and is limited to patients infected with P. aeruginosa isolates exhibiting resistant antimicrobial susceptibility testing (AST) results to piperacillin-tazobactam, cefepime, and meropenem. Furthermore, TOL-TAZ is only approved at JHH for administration to patients after susceptibility is demonstrated (ie, TOL-TAZ minimal inhibitory concentration [MIC] ≤4/4 mcg/mL using the Clinical and Laboratory Standards Institute [CLSI] interpretive criteria) [13]. Of the 48 patients, 28 had P. aeruginosa isolates available both (1) within 7 days before TOL-TAZ exposure (ie, index isolate) and (2) either during TOL-TAZ therapy or within 30 days of completion of TOL-TAZ therapy (ie, subsequent isolate). The subsequent isolate did not have to be from the same site of infection and was generally obtained because of an inadequate clinical response.

We conducted a case-control study including the 28 patients to identify modifiable risk factors associated with the development of TOL-TAZ resistance. Cases were defined as patients infected with P. aeruginosa with at least a 4-fold increase in P. aeruginosa TOL-TAZ MICs after exposure to TOL-TAZ. Controls were defined as patients infected with P. aeruginosa that did not have an increase in TOL-TAZ MICs, using the accepted standard of error in reporting AST results of ±1 doubling-dilution of the MIC [14], also after exposure to TOL-TAZ.

Data on the 28 patients collected via manual health record review included demographics, underlying medical conditions, source of infection and source control measures, microbiological data and AST data, and antibiotic treatment data. Source control was defined as the removal of infected hardware or drainage of infected fluid collections within the first 7 days of antibiotic therapy. Specific antibiotic treatment data collected included the following: (1) dosage of TOL-TAZ administered (1.5 grams every 8 hours vs 3 grams every 8 hours) or equivalent after adjustment for creatinine clearance calculated using the Cockcroft-Gault equation (Table 1), (2) duration of TOL-TAZ infusion (over 1 hour vs over 3 hours), (3) use of > 48 hours of combination therapy with an aminoglycoside or polymyxin, and (4) duration of TOL-TAZ exposure between the index and subsequent isolate.

Microbiological Data

Bacterial genus and species were identified using MALDI-TOF (Bruker Daltonics Inc., Billerica, MA, USA) and AST results were generated by the BD Phoenix Automated System (BD Diagnostics, Sparks, MD, USA), per routine JHH Medical Microbiology Laboratory protocol. The TOL-TAZ and imipenem-relebactam ETEST® (bioMérieux, Marcy-l’Étoile, France) were used to determine TOL-TAZ and imipenem-relebactam MICs, respectively. The ceftazidime-avibactam (CAZ-AVI) HardyDisks (Hardy Diagnostics, Santa Maria, CA, USA) were used to determine CAZ-AVI MICs. Isolates were stored at −80°C in glycerol until further testing was performed.

Thirty-two isolates from 16 patients were available for additional laboratory testing. Isolates were subcultured twice from frozen stock to tryptic soy agar with 5% sheep blood. MICs were confirmed by broth microdilution (BMD), in triplicate, using Food and Drug Administration (FDA)-cleared Sensititre GN7F gram-negative panels, which include TOL-TAZ and CAZ-AVI (ThermoFisher Scientific, Indianapolis, IN, USA); MICs reported by BMD were used for subsequent analysis.

Genomic DNA was extracted from the 32 isolates (from 16 patients) using the DNeasy Blood & Tissue Kit (Qiagen, Inc, Valencia, CA, USA). Whole genome sequencing (WGS) was conducted using Illumina MiSeq short-read sequencing (Illumina, San Diego, CA, USA). Approximately 100–500 ng of gDNA was used to prepare sequencing libraries using the Nextera Flex Kit. Each Illumina library was then sequenced using v3 2 × 300 reagents on an Illumina MiSeq.

Resistance Mechanism Identification

Sequenced isolates were evaluated using FASTQC v0.11.6 and MultiQC v1.6 [15, 16]. Trimmomatic v0.39 removed adapters and trimmed low-quality paired-end reads [17]. Trimmed and de-duplicated reads (FastUniq v1.1) were de novo assembled with SPAdes v3.12.0 and annotated with Prokka v1.13 [18–20]. Quast v4.6.3 confirmed assembly quality [20]. Genomic distances for cluster analysis were calculated with SourMash 2.0.0a [21]. MUMmer3 v3.23 was used for pairwise differential genome analysis [22]. Genes present in the index isolate but absent in the subsequent isolate, or vice versa, as well as genes with point mutations, were extracted for manual assessment. Gene annotations were determined with nucleotide BLAST v2.9.0+ against the reference genome P. aeruginosa PA01. Resistance genes were identified using ARESdb [23]. Isolate variant analysis was carried out with Snippy 4.6.0 against the same reference of PA01 using default parameters, before complements of variants for paired isolates were extracted using bcftools 1.10.2 [24, 25]. Intergenic and synonymous variants were removed. Bioinformatics analysis were performed by Ares Genetics.

Statistical Analysis

Baseline and treatment characteristics of cases and controls were compared using the Fisher exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. Due to the small sample size, a multivariable analysis was not performed. P-values of ≤ .05 were considered significant. All analyses were performed using Stata, version 15 (StataCorp, College Station, TX, USA).

RESULTS

General Characteristics of Patients

Of the 28 patients infected with P. aeruginosa who received at least 72 hours of TOL-TAZ therapy with P. aeruginosa isolates available before and after therapy, 14 (50%) had P. aeruginosa isolates that developed at least a 4-fold increase in the TOL-TAZ MIC (ie, cases). (Table 2). Baseline characteristics were similar between cases and controls. Sites of P. aeruginosa infection among the 28 patients included pneumonia (68%), bacteremia (18%), and intra-abdominal infections (14%).

Modifiable Risk Factors

Twelve (86%) cases and 14 (100%) controls received 3 grams IV of TOL-TAZ every 8 hours (P = .14); Table 2. No association was observed with the dosage of TOL-TAZ administered and the emergence of resistance. Similarly, no association was observed with the use of combination therapy (ie, addition of intravenous aminoglycoside or polymyxin to TOL-TAZ) and the prevention of resistance. No cases or controls received concomitant fluoroquinolone therapy or two carbapenem agents simultaneously [26, 27]. Interestingly, all patients who developed significant increases in TOL-TAZ MICs received the agent over 1 hour. In contrast, no patients who received the agent over 3 hours developed TOL-TAZ resistance (P = .04). Deep-seated sources of infection (eg, left ventricular-assist devices, vascular grafts), all of which were unable to be removed (ie, no source control) were more likely to result in TOL-TAZ resistance than when source control was achieved (P = .04). For every additional day TOL-TAZ was administered, a 3% increase in TOL-TAZ resistance was observed.

Changes in Antimicrobial Susceptibility Results

The median TOL-TAZ MIC prior to TOL-TAZ therapy for the 28 patients was 2 mcg/mL (interquartile range [IQR] 1–4), with no differences between cases or controls. The median TOL-TAZ MIC after TOL-TAZ therapy in the cases was 256 mcg/mL (IQR 16–256), and remained at 2 mcg/mL (IQR 2–4) for the controls. Reversion to meropenem susceptibility was not observed for any subsequent P. aeruginosa isolates. The median days of TOL-TAZ exposure between the index and subsequent isolate for cases was 15 days (IQR 8–22), compared to 9 days (IQR 6–14) for controls; P = .17).

Fourteen (50%) patients had index P. aeruginosa isolates susceptible to CAZ-AVI (ie, CAZ-AVI MIC ≤8/4 mcg/mL using the CLSI interpretive criteria) [13]. Interestingly, 6 of 7 cases (86%) with index isolates susceptible to CAZ-AVI had subsequent P. aeruginosa isolates with a 4-fold or greater CAZ-AVI MIC after TOL-TAZ exposure—and without any exposure to CAZ-AVI. None of the controls with index isolates susceptible to CAZ-AVI had subsequent CAZ-AVI isolates with 4-fold or greater MICs (P < .01), underscoring the overlapping mechanisms of resistance between TOL-TAZ and CAZ-AVI.

Sixteen of 28 patients had both index and subsequent TOL-TAZ isolates available for imipenem-relebactam MIC testing. None of the 16 patients received imipenem-cilastatin-relebactam during the interval when the index and subsequent P. aeruginosa isolate was obtained. The FDA susceptibility breakpoint for imipenem-relebactam is against P. aeruginosa ≤ 2/4 mcg/mL (CLSI breakpoint not yet published) [28]. Three (19%) of the index isolates (all of which were TOL-TAZ susceptible) were susceptible to imipenem-relebactam. One of these 3 isolates became resistant to imipenem-relebactam after TOL-TAZ exposure (0.5 mcg/mL to 4 mcg/mL).

Mechanisms of Acquired Resistance

WGS was pursued on 32 P. aeruginosa isolates from 16 patients (Table 3). Phylogenetic trees were constructed to confirm relatedness between index and subsequent isolates and sequence types were also confirmed. Paired susceptible (index) and subsequent resistant strains were compared to identify mutations associated with TOL-TAZ resistance (Table 3). Isolates with high-level resistance had 2–6 mutations leading to amino acid changes. Most mutations occurred in the AmpC-AmpR region and DNA polymerase subunits gamma and tau. Three isolates had penicillin binding protein 3 (PBP3) R504C mutations prior to developing TOL-TAZ resistance and one isolate acquired a PBP3 E466K mutation that likely contributed to resistance. Metallo-β-lactamase genes were not identified in any TOL-TAZ resistant isolates.

DISCUSSION

In a cohort of 28 patients with invasive P. aeruginosa infections treated with TOL-TAZ, 50% of patients with P. aeruginosa isolates available before and after TOL-TAZ exposure had at least a 4-fold increase in TOL-TAZ MICs after TOL-TAZ exposure. Although resistance to antibiotics is inevitable—and particularly problematic for P. aeruginosa as multiple mechanisms of resistance are working in concert [4, 5], we sought to identify modifiable risk factors that can mitigate or even prevent the emergence of TOL-TAZ resistance. We found that infusion of TOL-TAZ over 3 hours, rather than the package insert recommendation of infusion over 1 hour [29], appeared to be protective against the emergence of resistance. Only 2 patients in our cohort received the 1.5 gram dosage of TOL-TAZ, precluding us from further investigating the role of TOL-TAZ dosage on subsequent resistance [30].

Prolonged β-lactam infusions (encompassing both extended infusion and continuous infusion) are increasingly being used as the default approach for β-lactam antibiotic administration [31]. In a meta-analysis of 632 patients with severe sepsis, prolonged infusion strategies were associated with a 7% lower mortality risk and 9% higher likelihood of clinical cure compared to standard (ie, 30–60 minute) β-lactam administration strategies [32]. An advantage of this approach is that it increases the likelihood of achieving target serum drug levels (ie, free time of the drug above its MIC) even for patients infected with bacteria with elevated MICs [33], and this has been demonstrated specifically for TOL-TAZ [34, 35].

Interestingly, the impact of prolonged-infusion strategies on the emergence of resistance has been relatively unexplored [36]. Dhaese and colleagues evaluated the role of intermittent versus continuous infusion piperacillin-tazobactam and meropenem on the emergence of gram-negative bacterial resistance in a retrospective cohort of 205 ICU patients [37]. Resistance developed in 12% of isolates and the duration of infusion was not associated with the emergence of resistance. Yusuf and colleagues evaluated the association between duration of drug infusion and the emergence of resistance to several commonly used gram-negative agents among 187 intensive care unit (ICU) patients infected with P. aeruginosa, and no association was identified [38]. As was observed in our cohort, combination therapy did not appear to protect against the emergence of β-lactam resistance in the study by Yusuf and colleagues [38]. It remains difficult to draw meaningful conclusions from limited, observational, single-center studies where small proportions of patients developed resistance. It is unknown if there are pharmacokinetic-pharmacodynamic index correlations that may be protective or even conductive to the development of resistance. Theoretically, prolonging drug infusion may reduce precipitous reductions in drug concentrations, especially in those with augmented renal clearance, limiting the opportunity for bacteria to develop resistance as may occur in the presence of suboptimal antibiotic levels [39, 40]. The impact of extended-infusion TOL-TAZ strategies on the prevention of acquired resistance warrants further investigation.

Similar to what we found in the paired P. aeruginosa isolates that exhibited high-level resistance to TOL-TAZ after TOL-TAZ exposure, in vitro studies indicate that several mutations can be implicated. Commonly reported mutations involve either the overexpression or structural modification of AmpC to reduce the activity of ceftolozane [11, 12, 41]. We identified 2 paired isolates with mutations in AmpC (G183D, E247K) and 1 in AmpR (D135G) associated with resistance. Haidar and colleagues described the emergence of TOL-TAZ resistance in three (14%) TOL-TAZ exposed P. aeruginosa isolates from Pittsburgh [10]. Molecular analysis was undertaken for 2 resistant and 3 TOL-TAZ susceptible isolates [10]. No mutations were identified in the AmpC-AmpR region for isolates that remained TOL-TAZ susceptible, although both TOL-TAZ resistant isolates developed mutations in the AmpC-AmpR genomic region [10]. Similarly, emergence of TOL-TAZ resistance during therapy was described in 5 (11%) P. aeruginosa isolates in a Spanish cohort [9]. Various amino acid substitutions and deletions in the AmpC repressor, AmpR, were found to be the primary mechanism of TOL-TAZ resistance in 4 isolates [9]; resistance to the fifth isolate was due to the emergence of a bla_OXA-14 gene [9].

Aside from the AmpC-AmpR mutants identified in our cohort, the most common mutations contributing to acquired TOL-TAZ resistance involved mutations in DNA polymerase subunits gamma and tau (dnaX gene). Further studies defining the role of DNA polymerase mutations in TOL-TAZ resistance are needed [41]. TOL-TAZ exhibits a high affinity to PBP3 [42]. We identified PBP3 mutations, the target of ceftolozane, in 3 sets of paired isolates (in both index and subsequent isolates) and in 1 isolate that developed resistance to TOL-TAZ. These findings are not surprising as PBP3 inhibition leads to global effects on transcription in P. aeruginosa including induction of the SOS response, subsequently impacting upregulation of error-prone DNA polymerases, furthering mutational derived TOL-TAZ resistance [43]. Our findings suggest that, although PBP3 mutations may contribute to increased TOL-TAZ MICs through downstream mechanisms, PBP3 mutations alone do not necessarily lead to frank TOL-TAZ resistance.

All but one P. aeruginosa isolate that developed resistance to TOL-TAZ in our cohort also developed resistance to CAZ-AVI. The high likelihood of cross-resistance between these agents is not surprising as the AmpC-AmpR region encompasses the omega loop, the substrate binding site for both ceftolozane and ceftazidime. This region serves as a “hot spot” for mutations that widen the binding site, effectively trapping the β-lactam component of TOL-TAZ and CAZ-AVI and facilitating hydrolysis [12]. The intricacies of P. aeruginosa resistance and the complex patient population prone to recurrent, invasive, and often deep-seated P. aeruginosa infections highlights the need to explore strategies to reinforce this error prone region of the bacterial genome as well as to further investigate nonantibiotic treatment options such as bacteriophage therapy [44].

This was a retrospective study and, the possibility of missing clinical data exists. However, as all patients were hospitalized for the entire duration of evaluation (ie, from the time the index isolate was obtained to the time the subsequent isolate was obtained), it is unlikely that critical clinical data were missing. As our analysis was limited to 28 patients, our results are exploratory and must be confirmed in larger cohorts. There were some differences observed with cases and controls, and we are unable to determine if they would have reached statistical significance had our patient population been larger. For example, we found that 29% of cases and 7% of controls required renal replacement therapy while receiving TOL-TAZ. Although this difference was not statistically significant, others have found renal replacement therapy to be an independent risk factor for the emergence of CAZ-AVI resistance [45, 46]. Inadequate and fluctuating drug exposures can result from renal replacement therapy, making this a plausible precipitant for the emergence of drug resistance [29]. The role of renal replacement therapy and subsequent TOL-TAZ resistance merits further consideration.

Our TOL-TAZ resistance—baseline and acquired resistance—appears substantially higher than other reports in the literature. For example, evaluating 159 carbapenem nonsusceptible isolates from The JHH Medical Microbiology Laboratory, 61% exhibited susceptibility to TOL-TAZ, prior to TOL-TAZ exposure (unpublished data), compared to other North American studies describing initial susceptibility of carbapenem nonsusceptible P. aeruginosa to TOL-TAZ in the range of 85–95% [7, 8]. We used a more stringent definition amongst the P. aeruginosa isolates included in our study, compared to most surveillance studies; in our study isolates tested for TOL-TAZ susceptibility had to be resistant to several commonly used anti-pseudomonal β-lactams (eg, piperacillin-tazobactam, cefepime, and meropenem). This likely contributed to our lower TOL-TAZ susceptibility estimates. Humphries and colleagues evaluated 105 P. aeruginosa isolates from the greater Los Angeles area also resistant to all these agents, and only 52% of isolates retained susceptibility to TOL-TAZ [47]. Future work investigating acquired TOL-TAZ resistance should include isolates with substantial geographic diversity.

As we only included patients with initial and subsequent P. aeruginosa growth after exposure to TOL-TAZ, we likely enriched the population at risk for TOL-TAZ resistance. More specifically, for the patients without subsequent P. aeruginosa recovery, although we will never know what percentage experienced the emergence of TOL-TAZ resistance, it is presumably less than the population who had P. aeruginosa growth after TOL-TAZ exposure because of the potentially successful “eradication” of pseudomonal infections in the former. Importantly, recovery of P. aeruginosa on subsequent isolates reflected colonization for some patients. We cannot translate emergence of resistance into clinical failure for patients in our cohort.

These limitations notwithstanding, our findings suggest that the emergence of TOL-TAZ resistance during therapy is not a rare event and extending the infusion of TOL-TAZ may be protective against the development of resistance. Larger interventional studies are needed to validate this association. In the meantime, appropriate source control and limiting durations of therapy to evidence-based durations, cornerstones of antibiotic stewardship, remain modifiable risk factors that can limit the emergence of resistance to all antibiotics—including TOL-TAZ.

Notes

Acknowledgments. This work was supported by the an American Lung Association Research Grant awarded to P. D. T.

Potential conflicts of interest. S. B. is an employee of Ares Genetics, and A. E. P. is the CEO of Ares Genetics. Ares Genetics performed bioinformatics analysis for this work. S. E. C. receives consulting fees from Novartis, Theravance, and Basliea, outside the submitted work. P. J. S. has received grants and personal fees from Accelerate Diagnostics, OpGen Inc, and BD Diagnostics; grants from bioMerieux, Inc., Affinity Biosensors, Check-Points Diagnostics, BV, and Hardy Diagnostics; and personal fees from Roche Diagnostics, GeneCapture, CosmosID, and Shionogi Inc. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References