PDF
Citation
Print
Introduction
The ever-increasing number of antibiotic-resistant bacteria poses a worldwide public health threat [1,2]. Rapid antimicrobial susceptibility testings (RASTs) provide the tools for choosing appropriate antibiotics for prompt and effective treatment of bloodstream infections, thereby decreasing morbidity and mortality [3-5]. There are several commercial phenotypic antimicrobial susceptibility testings (ASTs) for identifying pathogens in blood, such as Vitek 2 (bioMérieux Inc., Marcy l’Etoile, France), the MicroScan system (Siemens Healthcare, Sacramento, CA, USA), and the Phoenix system (BD Diagnostic Systems, Sparks, MD, USA) [6-8]. However, these AST methods rely on the growth of blood cultures, which takes more than a day, thus delaying the treatment of patients with the most effective antibiotics [9].
We evaluated an automated, direct & RAST (dRAST) system called QMAC-dRAST (QuantaMatrix Inc., Seoul, Korea). This system uses micropatterned plastic microchips with a nutrient agarose well containing patient blood samples atop a layer of dried antibiotic at different concentrations [10]. The system analyzes bacterial microcolony growth over time directly in the wells with different antibiotics that diffuse into the agarose layer with the sample [6,10,11]. The bacterial growth is measured by time-lapse microscopic imaging in 6 hours [6,10,11]. Several studies have evaluated the clinical performance of QMAC-dRAST [1215], however, performance data on QMAC-dRAST in AmpC β-lactamase-producing Enterobacterales are still insufficient.
AmpC β-lactamases are enzymes that inactivate cephalosporin antibiotics by cleaving their β-lactam ring. The genes for these enzymes, which are found on the chromosomes of many Enterobacterales, are induced by β-lactam antibiotics, leading to resistance to broad-spectrum cephalosporins such as cefotaxime, ceftazidime, and ceftriaxone [16]. Enterobacterales with the ability to make inducible AmpC β-lactamases, such as Enterobacter spp., Citrobacter spp., Serratia spp., and Morganella morganii are major causes of bloodstream infections. However, resistance to antimicrobial agents has increased in these species and the development of new drugs is insufficient to ensure adequate treatments [17,18]. In a landmark study by Chow et al.[19], 19% (6 of 31) of the Enterobacterales strains isolated from patients with bacteremia were resistant to broad-spectrum cephalosporins. According to a multicenter study in Korea, the proportion of AmpC derepressed strains of Enterobacter cloacae, Citrobacter freundii, and Serratia marcescens was 27.5%47.3%, which is much higher than has been reported in European studies (11%-34%) [20,21]. Therefore, it is necessary to evaluate the effectiveness of the QMAC-dRAST system for AmpC β-lactamase-producing Enterobacterales.
In our preliminary research for the performance of QMAC-dRAST, M. morganii and S. marcescens showed many discrepant results in AST compared to Vitek 2 for the following antimicrobial agents: aztreonam, cefepime, cefotaxime, ceftazidime, imipenem, piperacillin/tazobactam. Therefore, we updated the system partially for these bacterial species-antibiotics combinations. Then, we compared multiple AST results of QMAC-dRAST for several AmpC β-lactamase-producing strains with results from the Vitek 2 system. We improved and evaluated the performance of QMAC-dRAST for AmpC β-lactamase-producing Enterobacterales in the clinical laboratory.
Materials and methods
Sample collection
This study was conducted at Severance Hospital, Yonsei University College of Medicine, South Korea, from April 2018 to March 2019. The performance of QMAC-dRAST was assessed using 80 Enterobacterales isolates: 20 M. morganii, 20 S. marcescens, 10 Klebsiella aerogenes, 10 E. cloacae, and 20 C. freundii. All isolates were collected from blood samples of patients hospitalized at Severance Hospital from January 2016 to June 2018 and identified by the Vitek 2 identification system and MALDI TOF M/S (Bruker Daltonics Inc., Billerica, MA, USA).
Bacterial spiking protocol
All 80 isolates, which had been inoculated in 15% glycerol stock and stored at -70℃, were cultured on blood agar. For the QMAC-dRAST method, each bacterial colony was dispersed in 1.0 mL of 0.9% saline and adjusted to approximately 1.5 × 108 CFU/mL (0.5 McFarland suspension). Mixtures were serially diluted to a final concentration of approximately 1.0 × 103 CFU/mL. A 1.0-mL sample of the final dilution was inoculated into a Bact/Alert FA blood culture bottle containing 5.0 mL of sheep blood and was incubated in the Bact/Alert 3D system (bioMérieux Inc., Marcy l’Etoile, France). After the Bact/Alert 3D system detected the blood culture bottle as positive, a 350-μL sample from the bottle was transferred to a glass tube and tested directly in the QMAC-dRAST system.
In the first test, the old interpretation algorithm of AST results (AST algorithm) was applied. Bacterial growth within the agarose matrix only was measured in the old AST algorithm. In the final test, the AST algorithm for M. morganii and S. marcescens in aztreonam, cefepime, cefotaxime ceftazidime, imipenem, piperacillin/tazobactam was changed to observe bacterial growth not only within the agarose matrix but also outside the agarose matrix.
AST using the Vitek 2 system and the broth microdilution (BMD)
The performance of QMAC-dRAST was evaluated in comparison with that of Vitek 2. We conducted ASTs using QMAC-dRAST in parallel with Vitek 2 using 14 kinds of antibiotics. Discrepant AST results showing a different susceptibility category (resistant, susceptible, or intermediate) between two systems were adjudicated by BMD tests as the reference method to determine the correct results. The Vitek 2 ASTs were performed according to the manufacturer’s guidelines and BMD tests were carried out according to the Clinical and Laboratory Standards Institute (CLSI) guidelines [22].
Quality control
The three CLSI standard strains (Escherichia coli ATCC 25922, Pseudomonas aeruginosa ATCC 27853, and Klebsiella pneumonia ATCC 700603) were used for quality control for the QMAC-dRAST system and BMD tests. For the Vitek 2 system, E. coli ATCC 25922 and P. aeruginosa ATCC 27853 were used.
Data analysis
Based on the AST results of Vitek 2, isolates were categorized as resistant (R), susceptible (S), or intermediate (I) according to the manufacturer’s instructions. Using Vitek 2 as the comparator, the concordance in results was determined using categorical agreement (CA, i.e., agreement of the results between the test method and the comparator) and essential agreement (EA, i.e., agreement within ±1 twofold dilution of the test method under evaluation with the comparator minimum inhibitory concentration determination). The discrepancy rates for the detection of antimicrobial susceptibility were classified as very major errors (VME, i.e., S with test method vs. R with comparator), major errors (ME, i.e., R with test method vs. S with comparator), and minor errors (mE, i.e., I with test method vs. R or S with comparator or vice versa) [23]. Data were presented as numbers with percentages for categorical variables.
Results
Before and after QMAC-dRAST algorithm optimization
Initially, ASTs using the QMAC-dRAST system were conducted for 20 isolates of M. morganii and 20 isolates of S. marcescens using six antibiotics: aztreonam, cefepime, cefotaxime, ceftazidime, imipenem, and piperacillin/tazobactam. In case of ME results of these isolate-antimicrobial agent combinations, the actual wells of the QMAC-dRAST chip and their microscopic images were inspected. Then the AST algorithm for the combinations was updated based on the findings of the inspection. After the AST algorithm was improved, we conducted the final test. Susceptibility tests with the six antibiotics were performed again for 20 M. morganii and 20 S. marcescens.
Using six antimicrobial agents described above, we compared AST results between QMAC-dRAST and Vitek 2 for 20 M. morganii and 20 S. marcescens isolates before and after optimization of the AST algorithm for the two bacterial species. For all isolates, the agreement rates before optimization were 55.4% of CA, 19.1% of VME, 38.3% of ME, and after optimization were 83.8% of CA, 5.4% of VME, 4.3% of ME. For M. morganii, agreement rates before optimization were 67.5% of CA, 25.8% of VME, 17.9% of ME, and after were 79.2% of CA, 0% of VME, 7.1% of ME. Agreement rates of S. marcescens before optimization were 42.5% of CA, 0% of VME, 55.6% of ME, and after were 88.3% of CA, 16.7% of VME, 1.9% of ME (Table 1).
The performance of the QMAC-dRAST system for all samples
To fully evaluate the performance of QMAC-dRAST for AmpC β-lactamase-producing Enterobacterales, we performed the final ASTs for all 80 isolates using both QMAC-dRAST and Vitek 2 with 14 antibiotics: amikacin, amoxicillin/clavulanate, ampicillin, aztreonam, cefazolin, cefepime, cefotaxime, ceftazidime, ciprofloxacin, ertapenem, gentamicin, imipenem, piperacillin/tazobactam, and trimethoprim/ sulfamethoxazole. Among 1,120 total AST results, 121 showed discrepancies between QMAC-dRAST and Vitek 2. Therefore, BMD tests were performed for these samples to resolve discrepancies. The CA between QMAC-dRAST and BMD tests was 40.5% (49/121), lower than that between Vitek 2 and BMD tests, which was 48.8% (59/121). The VME between QMAC-dRAST and BMD tests was 3.3% (4/121) and that between Vitek 2 and BMD tests was 16.5% (20/121). In contrast, the ME between QMAC-dRAST and BMD tests was 14.0% (17/121), higher than that between Vitek 2 and BMD tests at 1.7% (2/121) (Table 2).
The CAs between QMAC-dRAST and Vitek 2 for each bacterial species were all above 90%. Only S. marcescens and C. freundii showed VMEs, but M. morganii, K. aerogenes, and E. cloacae showed relatively high ME rates (Table 3). Among antimicrobial agents, imipenem and piperacillin/tazobactam showed relatively high VME rates (Table 4). The CA between QMAC-dRAST and Vitek 2 for all 1,120 AST results was 93.0%, and the discrepancy rates were 1.7% for VME, 2.3% for ME, and 4.9% for mE (Table 3).
Discussion
AmpC β-lactamase, extended-spectrum β-lactamase, and carbapenemase are major reasons for antimicrobial resistance [24]. A study in Korea evaluated 732 patients with infections due to Enterobacter spp., S. marcescens, C. freundii, or M. morganii which have chromosomally encoded AmpC β-lactamases. Among 732 patients, 14 (1.9%) patients had bacteria that demonstrated antimicrobial resistance during antimicrobial therapy [25]. Another study included 46 patients who were initially infected with cephalosporin-susceptible Enterobacter spp. that became resistant to the antibiotic. Compared to 113 matched control patients who were infected with persistently susceptible isolates of the same organism, the first group of patients showed higher mortality, had a longer hospital stay, and had higher hospital costs [26].
Conventional AST systems such as Vitek 2, Phoenix, or Microscan are now widely used in the clinical setting, but they require 8–20 hours to produce results, excluding the time for isolation and growth of the bacteria [27]. In the QMAC-dRAST system, minimum inhibitory concentration (MIC) is determined by observing microcolony formation from a single bacterium without subculturing bacteria from blood culture bottles [10]. A QMAC-dRAST chip consists of 96 test wells. Each test well has a micropatterned chamber for an agarose matrix and a sub-well containing dried antibiotics. A sample of bacteria mixed with agarose is inoculated into the chamber. For multiplex ASTs, different concentrations of several antibiotics are dried in each well of the chip. The antibiotics are rehydrated by adding cation-adjusted Mueller-Hinton broth (CAMHB, BD Biosciences, CA, USA) and diffuse into the agarose matrix [10,11]. In the QMAC-dRAST instrument, bacterial microcolonies forming in the agarose matrix are imaged every hour and quantified by in-house software for automated image analysis to determine bacterial growth at each antibiotic concentration. This growth information is used to determine the MIC and the bacterial antibiotic susceptibility can be reported [10].
The QMAC-dRAST system uses bright-field microscopy. Bacteria located in the upper region of the agarose matrix can migrate outside the agarose matrix and grow in the CAMHB. This bacterial growth blocks the illumination and reduces the brightness of the image (Fig. 1). However, the old AST algorithm measured bacterial growth only within the agarose matrix as originally designed, regardless of bacterial growth outside the agarose matrix. For instance, if some wells with different concentrations of the same antibiotic showed similar bacterial growth each other in the agarose matrix but different bacterial growth outside the agarose matrix, the brightness of images of each well would be different from each other. However, in the old AST algorithm, bacterial growth is considered to have occurred in all these wells. As a result, high ME (38.3%) for M. morganii and S. marcescens was observed in the first test before AST algorithm improvement (Table 1). To address the discrepancy, we modified the old AST algorithm so that the new AST algorithm did not ignore the bacterial growth outside the agarose matrix. In the new AST algorithm, the brightness of the images is calculated numerically by its own standard. It was considered that actual bacterial growth had occurred only in the wells with reduced brightness of the images (Fig. 1). This new AST algorithm applied only to M. morganii and S. marcescens when testing aztreonam, cefepime, cefotaxime, ceftazidime, imipenem, and piperacillin/tazobactam. Otherwise, the old AST algorithm is still applied.
As we expected, ME rates of QMAC-dRAST for M. morganii and S. marcescens decreased after AST algorithm optimization (Table 1). However, the VME rate of QMAC-dRAST for M. morganii was changed from 25.8% to zero (Table 1). Therefore, we reviewed our data and found that only 2 out of 20 M. morganii isolates yielded VMEs (4 VMEs each) in the first test. In 4 out of the 8 isolate-antimicrobial agent combinations, susceptibility categories of Vitek 2 or BMD results were changed in the final test compared to the first test. In the other 4 combinations, susceptibility categories of QMAC-dRAST results were changed. MIC values of those four QMAC-dRAST results were increased in the final test, which was unexpected considering the mechanism of the new AST algorithm. However, we could not figure out the cause of these conflicting results and further studies are required.
Except for VME (1.7%), the overall performance of the QMAC-dRAST with AmpC β-lactamase-producing Enterobacterales met the criteria for AST in the FDA guidelines (e.g. CA ≥ 90%, VME ≤ 1.5%, ME ≤ 3%) [23] (Table 3). However, our study has some limitations. Firstly, BMD tests were performed only for samples that showed discrepancies between QMAC-dRAST and Vitek 2. In further studies, it is desirable to conduct BMD tests in all samples. Secondly, more bacterial species and samples may be required to validate our findings. Finally, further modifications based on the characteristics of each of the bacteria are still needed to improve the performance of the QMAC-dRAST.
In conclusion, we updated the AST algorithm used in the commercially available in-house software of QMAC-dRAST; the AST algorithm has been improved by considering the brightness of the images of each well of the QMAC-dRAST chip. The performance of the QMAC-dRAST automated system can likely be improved by further modifications of the AST algorithm and this system will show a more acceptable level of agreement for AmpC β-lactamase-producing Enterobacterales.
XML Download