Abstract

Introduction

Sepsis caused by drug-resistant Gram-negative bacteria (GNB) often results in serious clinical outcomes in patients [1]. Inappropriate initial antibiotic treatment occurs in one third of patients with severe sepsis due to GNB, which is associated with increased hospital mortality and length of stay [2]. In contrast, early administration of appropriate antibiotics improves survival of sepsis patients [3]. Effective antibiotic administration within the first hour of documented hypotension was related to increased survival of patients with septic shock; however, 50% of septic shock patients did not receive effective antimicrobial treatment within 6 hours of documented hypotension [4]. To improve diagnosis of causative organisms of sepsis, automated continuous-monitoring blood culture systems were developed and introduced into clinical microbiological laboratories during the 1990s [5]. These early systems and subsequent generations of automated blood culture systems remain key diagnostic tools in the diagnosis of sepsis [5]. However, after a blood culture becomes positive, conventional bacteriological procedures still require 2 to 3 days for isolation, identification, and antimicrobial susceptibility testing. In addition to the time-consuming procedures, the emergence and spread of drug-resistant GNB producing various β-lactamases, including carbapenemase and extended spectrum β-lactamases (ESBLs), has been a serious problem for treatment of sepsis [6]. Thus, there has been an unmet need for rapid and automated technology to identify bacterial species as well as detection of drug resistance genes.

Recently, several rapid molecular diagnosis assays for sepsis diagnosis have been introduced and evaluated [7]; including LightCycler SeptiFast Test [8], peptide nucleic acid fluorescence in situ hybridization (PNA-FISH) [9], and matrix-assisted laser desorption-ionization time of flight mass spectrometry (MALDI-TOF MS) [10], and a DNA-based microarray platform [Prove-it sepsis assay [11] and the Verigene Gram-Positive Blood Culture (BC-GP) assay [12]–[19]].

The Verigene Gram-Negative Blood Culture (BC-GN) Nucleic Acid Test (Nanosphere Inc., Northbrook, IL) is a sample-to-result automated microarray-based, multiplexed assay for species identification of GNB and detection of their drug resistance genes in positive blood culture bottles. The Verigene BC-GN assay is designed to directly detect species of GNB from positive blood culture bottles with 5 minutes of hands-on and 2 hours of run time per sample. Recently, the BC-GN assay has been approved by the U.S. Food and Drug Administration (FDA), and the limit of detection, sensitivity, and specificity of the assay is shown on the FDA site (http://www.fda.gov/). In this study, we describe the performance of the BC-GN assay using simulated and clinical samples of positive blood culture bottles.

Materials and Methods

Results

Bacterial identification of simulated samples

Of the 397 positive blood culture samples in the study, 295 were simulated samples prepared by inoculating an isolate of one of the Gram-negative bacterial species listed in Table 2. All the samples became culture-positive. Of the 295 samples, 289 generated a result of the BC-GN assay on the first attempt (98.1%) (data not shown). The remaining 6 generated “No Call ” results meaning technical errors. When retested, all the 6 samples generated results (data not shown). The concordance rate of the 295 simulated samples between the BC-GN assay and the MicroScan WalkAway system for bacterial identification were as follows: Acinetobacter spp., Citrobacter spp., Enterobacter spp., E. coli, K. oxytoca, Proteus spp., and P. aeruginosa: 100%; K. pneumoniae: 88.2%; and S. marcescens: 81.3% (left columns in Table 3).

Table 3

Identification of bacterial isolates in blood culture samples with the BC-GN assay.

	Simulated samplesb (n=295)				Clinical samplesc (n=102)			Total (n=397)
Bacterial speciesa	Inocultaed isolates	Correctly identified isolatesd (%)	Not detected isolatese (%)	Incorrectly identified isolatesf (%)	Identified speciesg	Correctly identified isolatesd (%)	Not detected isolatese (%)	Concordance rateh (95% CIi)
Acinetobacter spp.	37	37 (100%)	0	0	2	2 (100%)	0	100% (91.0–100)
Citrobacter spp.	6	6 (100%)	0	0	1	1 (100%)	0	100% (59.0–100)
Enterobacter spp.	18	18 (100%)	0	0	12	11 (91.7%)	1 (8.3%)	96.7% (82.8–99.9)
E. coli	36	36 (100%)	0	0	51	51 (100%)	0	100% (95.8–100)
K. oxytoca	6	6 (100%)	0	0	7	7 (100%)	0	100% (75.3–100)
K. pneumoniae	51	45 (88.2%)	5 (9.8%)	1j (2.0%)	21	17 (81.0%)	4 (19%)	86.1% (75.9–93.1)
Proteus spp.	9	9 (100%)	0	0	2	2 (100%)	0	100% (91.0–100)
P. aeruginosa	116	116 (100%)	0	0	9	8 (88.9%)	1 (11.1%)	99.2% (95.6–99.9)
S. marcescens	16	13 (81.3%)	3 (18.7%)	0	5	5 (100%)	0	85.7% (63.7–97.0)
Total	295	286 (96.9%)	8 (2.7%)	1 (0.3%)	110	104 (94.5%)	6 (5.5%)	96.3% (94.0–97.9)

Open in a separate window

a Bacterial species inoculated into blood culture bottles or bacterial species identified in clinical samples of blood culture bottles.

b Blood culture samples into which bacterial isolates were inoculated.

c Clinical blood culture samples obatined from sepsis patients.

d Numbers of isolates which were correctly identified with the BC-GN assay.

e Numbers of isolates which were not detected with the BC-GN assay.

f Numbers of isolates which were incorrectly identified with theBC-GN assay.

g Numbers of bacterial species which were identified in clinical samples with the conventional methods. Some of the samples contained 2 or 3 bacterial species.

h Concordance rate between the BC-GN assay and the conventional methods for bacterial identification.

i CI indicates confidence interval

j A sample inoculated with K. pneumoniae was identified as Enterobacter spp. with the assay.

Of 51 samples inoculated with stored K. pneumoniae isolates (left columns in Table 3), 45 were detected with the BC-GN assay, but the remaining 6 were not correctly detected. Of these 6 samples, 5 were reported as “Not detected”, and 1 was reported as Enterobacter spp. The MicroScan WalkAway system reported these 6 samples as K. pneumoniae. The 16S rRNA sequences of these samples showed more than 99.3% similarity to both K. pneumoniae and K. variicola. Of 16 samples inoculated with stored S. marcescens (left columns in Table 3), 3 were not detected with the BC-GN assay. The MicroScan WalkAway system reported these 3 samples as S. marcescens. The 16S rRNA sequences of these samples showed more than 99.5% similarity to both S. marcescens and Serratia nematodiphila. Their biochemical profile based on carbohydrates utilization corresponded to those of S. marcescens [23].

The BC-GN assay was tested on blood culture bottles inoculated simultaneously with 2 or 3 clinical isolates. The following combinations of bacteria species were tested: A. baumannii (harboring bla _OXA-23) and K. pneumoniae (bla _CTX-M and bla _NDM); A. baumannii (bla _OXA-23) and MRSA; C. freundii, K. pneumoniae (bla _KPC) and P. mirabilis; C. freundii (bla _KPC), E. faecalis and MSSA; E. cloacae (bla _IMP) and S. marcescens; E. faecalis and E. coli (bla _CTX-M); E. coli (bla _CTX-M) and P. aeruginosa (bla _VIM); P. aeruginosa (bla _VIM) and MRSE. All Gram-negative bacterial isolates were correctly detected by the BC-GN assay regardless of the presence of other bacterial species, including Gram-positive strains. The BC-GN assay also detected all drug resistance genes under these conditions (data not shown). The BC-GN assay was also tested on 2 blood culture bottles inoculated with M. morganii and S. maltophilia, respectively, which are Gram-negative pathogens but are not among the targets of the BC-GN assay. As expected, the BC-GN assay did not detect these species (data not shown).

Identification of drug resistance genes

With respect to drug resistant genes, the BC-GN assay reported that, of the 295 simulated samples tested, 184 were positive for one of the resistance gene targets detected by the BC-GN assay and 18 were positive for two of the targets. These results agreed with those of PCR and/or DNA sequencing. As shown on left columns in Table 4, 42 bla _CTX-M (2 E. cloacae, 19 E. coli, 2 E. hormaechei, 18 K. pneumoniae, and 1 P. mirabilis isolates), 119 bla _IMP (1 A. calcoaceticus, 1 A. nosocomialis, 1 A. pittii, 9 E. cloacae, 4 K. pneumoniae, and 103 P. aeruginosa), 8 bla _KPC (2 C. freundii, 5 K. pneumoniae, and 1 S. marcescens), 16 bla _NDM (2 A. baumannii, 1 E. hormaechei, 1 E. coli, 11 K. pneumoniae, and 1 P. aeruginosa), 24 bla _OXA-23 (24 A. baumannii), 1 bla _OXA-24/40 (1 A. baumannii), 1 bla _OXA-48 (1 K. pneumoniae), 3 bla _OXA-58 (2 A. baumannii and 1 A. pittii), and 6 bla _VIM (6 P. aeruginosa) were detected by the BC-GN assay.

Table 4

Identification of resistance genes in blood culture samples with the BC-GN assay.

Resistance genesa	Simulated samplesb (n=295)			Clinical samplesc (n=102)			Total (n=397)
	Positive genes for PCR/DNAsequences	Correctly identified genesd (%)	Not detected genese	Positive genes for PCR/DNAsequences	Correctly identified genesd (%)	Not detected genese	Concordance ratef (95% CIg)
bla CTX-M	42	42 (100%)	0	12	12 (100%)	0	100% (93.3–100)
bla IMP	119	119 (100%)	0	0	0	0	100% (96.9–100)
bla KPC	8	8 (100%)	0	0	0	0	100% (63.1–100)
bla NDM	16	16 (100%)	0	0	0	0	100% (79.4–100)
bla OXA-23	24	24 (100%)	0	0	0	0	100% (85.8–100)
bla OXA-24/40	1	1 (100%)	0	0	0	0	100% (2.5–100)
bla OXA-48	1	1 (100%)	0	0	0	0	100% (2.5–100)
bla OXA-58	3	3 (100%)	0	1	1 (100%)	0	100% (39.8–100)
bla VIM	6	6 (100%)	0	0	0	0	100% (54.1–100)
Total	220	220 (100%)	0	13	13 (100%)	0	100% (98.4–100)

Open in a separate window

a Resistance genes which were harbored by bacterial species inoculated into blood culture bottles or isolated from sepsis patients.

b, c See footnotes b and c in Table 3.

d Numbers of genes which were correctly identified with the BC-GN assay.

e Numbers of genes which were not detected with the BC-GN assay.

f Concordance rate between the BC-GN assay and the conventional methods for resistance-genes identification.

g CI indicates confidence interval

Of the 102 clinical samples tested, 13 were positive for one of the 9 drug resistance genes. As shown on middle columns in Table 4, 12 bla _CTX-M (2 E. cloacae, 8 E. coli, and 2 K. pneumoniae), and 1 bla _OXA-58 (1 A. baumannii) were detected by the BC-GN assay. These results were completely agreed with those of our laboratory methods.

MICs of E. coli isolates from patients harboring or not harboring bla _CTX-M were determined (Table 5). The E. coli isolates harboring bla _CTX-M showed higher MIC₅₀ and MIC₉₀ of β-lactams tested than the isolates not harboring bla _CTX-M. MICs of IMP were ≤1 μg/ml in E. coli isolates harboring or not harboring bla _CTX-M. An isolate of Acinetobacter spp. harboring bla _OXA-58 was susceptible to all β-lactams tested (Table S1). MICs of all clinical isolates were shown in Table S1, 2.

Table 5

Drug susceptibility of β-lactams of E. coli harboring or not harboring.

			β-lactams
		Number of isolates	PIPC			CAZ			CTX			IMP
			MIC50	MIC90		MIC50	MIC90		MIC50	MIC90		MIC50	MIC90
			(μg/ml)	(μg/ml)		(μg/ml)	(μg/ml)		(μg/ml)	(μg/ml)		(μg/ml)	(μg/ml)
E. coli (bla CTX-M -positive)		8	>512	>512		8	>16		>32	>32		≤1	≤1
E. coli (bla CTX-M -negative)		43	≤8	>64		≤1	≤1		≤8	≤8		≤1	≤1

Open in a separate window

Discussion

We determined the overall concordance rate, 96.3%, between the BC-GN assay and standard laboratory methods (Table 3). All bacterial species tested except for K. pneumoniae and S. marcescens showed the concordance rate of over 95%. In addition to the good performance, the BC-GN assay can be completed within 2 h with less than 5 min of hands-on time, enabling same day analysis and starting on appropriate treatment. The concordance rate of the BC-GN assay was comparable to that of the BC-GP assay. However, the BC-GN assay will need to be evaluated using various blood culture systems, since the BC-GP assay have already evaluated as well [12]–[19].

The BC-GN assay did not accurately identify a total of 15 isolates from 9 simulated samples and 6 clinical samples (Table 3). Of the 15 isolates, 10, 3, 1 and 1 were reported as K. pneumoniae, S. marcescens, Enterobacter spp., P. aeruginosa by the MicroScan WalkAway system, respectively. Relatively higher rates of misidentification in K. pneumoniae and S. marcescens isolates were likely to be due to the probes used in the BC-GN assay. These probes will be improved in the further study. The concordance rates of Gram-positive bacteria were comparable in the BC-GP assay [12]–[19]. The lower accuracy results of K. pneumoniae were shown in both the spiked and clinical observations. The 16S rRNA sequence revealed that the 10 misidentified isolates shared over 99% similarity to both K. pneumoniae and K. variicola. Previous studies have reported that routine methods may identify K. variicola as K. pneumoniae [24]. Further studies are required to clarify the clinical significance of K. variicola in sepsis patients. As described above, three samples containing S. marcescens were not detected with the BC-GN assay, which were observed only in the simulated samples. It is unclear whether clinical samples containing S. marcescens were not detected accurately with the BC-GN assay, since the number of samples tested was low.

As for the detection of polymicrobial samples, the BC-GN assay performed well in the simulated samples, but not in the clinical samples. All isolates were not detected completely in 3 of 7 clinical samples containing polymicrobial organisms. Although fungal species like Candida albicans were not tested, these results can have significant clinical implications. Rapid molecular-based assays including the BC-GP assay and the FilmArray blood culture identification assay showed relatively low sensitivity in polymicrobial infections [14], [25].

We evaluated the performance of the BC-GN assay in detecting various drug resistance genes using simulated samples. The BC-GN assay correctly detected all the drug resistance genes tested, even when multiple drug resistance genes were present in a sample. In addition, MICs of E. coli isolates from clinical samples indicated that the presence of bla _CTX-M was related to drug susceptibility profiles with high MICs of PIPC, CAZ, and CTX, not IPM. These results suggest that the BC-GN assay is useful for earlier administration of appropriate antibiotics and de-escalation. Whereas, no genotypic-phenotypic correlation was found in the Acinetobacter spp. isolate harboring bla _OXA-58. In that case, the BC-GN assay may lead to inappropriate use of antibiotic, and de-escalation is needed after conventional drug susceptibility testing.

Although bla _CTX-M is the most prevalent among genes of extended-spectrum β-lactamases (ESBLs)[26], [27], the BC-GN assay cannot detect other ESBLs genes like bla _TEM and bla _SHV, and there remains concern of bacteria harboring drug resistance genes which were not detected by the assay. More than 200 bla _TEM and more than 100 bla _SHV genes are currently known in the Lahey clinic database [28]. Thus, it remains unfeasible to detect these variants in a rapid and simple manner. The emergence of various β-lactam resistance genes, including carbapenemases and ESBLs, have been extensively reported [6]. The timely direct detection of resistance genes from positive blood cultures will be very useful in clinical situations, since multidrug-resistant bacteria and polymicrobial infections are important causes of failure in the treatment of sepsis. Because drug resistance genes will continue to evolve and spread, it is necessary to carefully monitor the spread of drug resistance genes.

Of the 10 most frequently isolated microorganisms in ICU-acquired bloodstream infection in Europe, 7 were Gram-negative pathogens targeted by the BC-GN assay [29]. Similar results were obtained from monomicrobial nosocomial bloodstream infections both in ICU and non-ICU wards in the USA [30]. The National Healthcare Safety Network at CDC reported that a significant proportion of central line-associated bloodstream infections are caused by drug-resistant Gram-negative pathogens in the USA [31]. Of the 6 major pathogens other than coagulase-negative staphylococci associated with nosocomial blood stream infections in Japan, 3 were Gram-negative pathogens, 2 were Gram-positive ones, and 1 was Candida spp. [32]. The BC-GN assay, which targets 9 Gram-negative bacterial species, will contribute to the diagnosis and management of patients with Gram-negative sepsis.

A number of new diagnostic methods have been developed for rapid detection of pathogens in positive blood cultures. A real-time PCR-based assay (LightCycler SeptiFast Test) can detect a number of Gram-negative and Gram-positive bacteria and fungi in a single assay, although it cannot detect drug resistance genes [8]. PNA-FISH can detect 3 Gram-negative pathogens, including E. coli, K. pneumoniae, and P. aeruginosa, but cannot detect drug resistance genes [9]. MALDI-TOF MS can detect bacterial isolates in blood culture, but cannot detect 2 or more bacterial species in polymicrobial samples or drug resistance gene products [33]. A DNA-based microarray platform, Probe-it sepsis assay, can detect various Gram-negative and Gram-positive pathogens in blood cultures. Nevertheless, it cannot detect drug resistance genes other than mecA [11].

To integrate the Verigene system into a laboratory workflow, Gram stain was needed after blood culture was positive to determine whether the BC-GN or BC-GP assay was used. The Verigene system cannot measure MICs of antibiotics, and thus should be used in combination with biochemical characteristics of bacteria obtained from conventional drug susceptibility testing.

This study has some limitations. Firstly, we could not fully evaluate the ability of the BC-GN assay to detect some drug resistance genes, such as bla _OXA-24/40 and bla _OXA-48. Secondly, to select samples containing organisms listed in the BC-GN panel, positive blood culture was stored for 3.1±1.4 days (maximum 5 days) before the BC-GN assay was done. The BC-GN assay should be done immediately after the culture becomes positive, although the storage of the culture in the study was unlikely to affect the assay results. A large-scale prospective clinical evaluation of the BC-GN assay is planned to determine the clinical impact of the assay including selection of antibiotics, length of stay, morbidity, and cost-effectiveness.

In conclusion, the Verigene system BC-GN assay accurately detects common Gram-negative bacterial isolates and their drug resistance genes from positive blood cultures in a rapid manner. The BC-GN assay demonstrated high sensitivity and specificity in the study, and it will generate significantly faster results compared to conventional methods in the medical settings. This new anitibioticroarray platform can be easily introduced into microbiological laboratories in various clinical settings. It will contribute to improved sepsis management as a result of earlier reporting of critical information leading to earlier administration of appropriate antibiotics and de-escalation.

Supporting Information

Acknowledgments

Funding Statement

This work was supported by the grant of National Center for Global Health and Medicine (24-207). East-West Diagnostics/Theranostics, LLC provided support in the form of salaries for author HT, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.

References