Evaluation of two rapid molecular test systems to establish an algorithm for fast identification of bacterial pathogens from positive blood cultures

Fast identification of pathogens directly from positive blood cultures is of highest importance to supply an adequate therapy of bloodstream infections (BSI). There are several platforms providing molecular-based identification, detection of antimicrobial resistance genes, or even a full antimicrobial susceptibility testing (AST). Two of such test systems allowing rapid diagnostics were assessed in this study: The Biofire FilmArray® and the Genmark ePlex®, both fully automated test system with a minimum of hands-on time. Overall 137 BSI episodes were included in our study and compared to conventional culture–based reference methods. The FilmArray® is using one catridge including a panel for the most common bacterial and fungal BSI pathogens as well as selected resistance markers. The ePlex® offers three different cartridges for detection of Gram-positives, Gram-negatives, and fungi resulting in a broader panel including also rare pathogens, putative contaminants, and more genetic resistance markers. The FilmArray® and ePlex® were evaluated for all 137 BSI episodes with FilmArray® detecting 119 and ePlex® detecting 128 of these. For targets on the respective panel of the system, the FilmArray® generated a sensitivity of 98.9% with 100% specificity on Gram-positive isolates. The ePlex® system generated a sensitivity of 94.7% and a specificity of 90.7% on Gram-positive isolates. In each case, the two systems performed with 100% sensitivity and specificity for the detection of Gram-negative specimens covered by each panel. In summary, both evaluated test systems showed a satisfying overall performance for fast pathogen identification and are beneficial tools for accelerating blood culture diagnostics of sepsis patients. Electronic supplementary material The online version of this article (10.1007/s10096-020-03828-5) contains supplementary material, which is available to authorized users.


Introduction
As bacteremia and sepsis are still leading causes of morbidity and mortality [1,2], fast diagnosis of the causative organism and appropriate antimicrobial therapy are essential for rapid treatment decisions [3]. Worldwide, the incidence of sepsis is about 19 million per year [4] with a lethality of 20-30% [5,6]. This emphasizes the need for rapid identification (ID) and detection of resistance genes. Conventional microbiological, agar-based methods used for blood culture diagnostics include culturedependent identification by MALDI-TOF and AST by phenotypic methods. As these techniques are time-consuming, automated molecular ID methods with lower hands-on and turnaround time were developed allowing for identification and the genotypic detection of resistance genes [7][8][9][10][11][12]. Therefore, using quick molecular-based ID tests directly from positive blood culture bottles can help with providing an adequate therapy in a timely manner by de-escalating broad-spectrum antimicrobial therapy or escalating insufficient treatment, respectively [13][14][15].
To speed up the identification and AST of Gram-negative bacteria in positive blood cultures, the Accelerate Pheno® system (Accelerate® Diagnostics, USA) was evaluated for fast ID and AST compared to culture-based diagnostics [12]. The system is implemented in our routine laboratory Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10096-020-03828-5) contains supplementary material, which is available to authorized users. accelerating reports for ID and AST especially for high-risk patients having BSI with Gram-negative bacteria. To establish a workflow for faster identification of all pathogens causing bloodstream infections, two different blood culture identification panels were compared in this study.
The BioFire FilmArray® Blood Culture Identification Panel (Biomerieux, Nürtingen, Germany) is a CE-marked multiplex PCR system, which allows identification of 24 targets at once including Gram-positives, Gram-negatives, and fungi, as well as three antibiotic resistance genes. DNA extraction, multiplex-PCR, and detection are fully automated with 2 min of hands-on time. A results report is generated approximately after 75 min [16].
The ePlex® system (Genmark Diagnostics, Carlsbad, USA) is endowed with three variable cartridges for pathogen identification. Deploying this test allows the detection of Gram-positives, Gram-negatives, and fungi. Besides the coverage of a large variety of organisms (55 bacterial/fungal targets) the test is designed to detect ten antibiotic resistance genes in addition. Nucleic acid extraction, amplification, and detection are completed in about 90 min using the microfluidic eSensor technology and require only 2 min of hands-on time.
The aim of the present prospective study was the comparative analysis of two molecular based Blood Culture Identification panels in comparison to conventional culturedependent ID methods in terms of (i) correct detection and identification of the pathogen, (ii) correct prediction of resistance genes and genotypes, and (iii) suitability for implementation in a diagnostic routine workflow.

Study design
The study was performed from November 2017 to January 2018 at a tertiary university hospital in Southern Germany in accordance with the local ethics committee (no. 667/2014BO1 and 139/2016BO2).
The first positive blood culture of each patient with a positive initial Gram-stain (Gram-positive or Gram-negative bacteria as well as fungi) was included in the study until 100 cultures with Gram-positive isolates were reached. Samples having more than one obvious organism according to the Gram-stain were not included in the study.
Positive blood cultures with Gram-positive (n = 98) or Gram-negative bacteria (n = 33) as well as fungi (n = 6) were analyzed by two different blood culture identification instruments: FilmArray® Blood Culture Identification panel (BioFire Diagnostics, Salt Lake City, UT, USA) and ePlex® Research Use Only (RUO) Blood Culture Identification panels (Genmark Diagnostics, Carlsbad, CA, USA). A schematic overview of the study design is shown in Fig. 1.

Conventional diagnostics
Laboratory operation hours are weekdays from 7:30 AM to 5:30 PM and weekends from 7:30 AM to 4 PM. Blood cultures are processed every morning starting at 7:30 AM and during the operating hours as soon as they are flagged positive.
Incubation of Blood culture bottles was performed in the Bactec FX blood culture instrument (BD Diagnostic Systems, Franklin Lakes, USA). The routine diagnostic workflow included subcultures of positive blood culture bottles on CNA-(Biomerieux, Nürtingen, Germany) and Brain Heart agar (Oxoid, Munich, Germany) in addition to Gram-stain. Identification to the species level was achieved by MALDI-TOF mass spectrometry (Microflex LT, Bruker Daltonics, Germany) and supplemented, if necessary, by the VITEK® 2 identification system (bioMérieux SA, France).
T h e a p p r o p r i a t e t e s t i n g c a r d o f V I T E K ® 2 (bioMérieux, SA, France) was used for Antimicrobial Susceptibility Testing (AST). All bacterial isolates were stored at − 80°C for further analysis. The guidelines set o f t h e E u r o p e a n C o m m i t t e e o n A n t i m i c r o b i a l Susceptibility Testing (EUCAST) (http://www.eucast.org/ fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_ tables/v_8.1_Breakpoint_Tables.pdf) were consulted for interpretation of susceptibility results.
If necessary, resistance genes bla VIM , bla IMP , bla KPC , bla OXA-48 , and bla NDM for carbapenemases were confirmed by molecular diagnostics (Light Mix Modular Realtime-PCR, Roche, Mannheim). In house-PCR for mecA [17] was performed on a T3 thermal cycler (Biometra, Göttingen) and analyzed by agarose gel electrophoresis directly from positive blood culture bottles during laboratory operating hours, but not during weekends. For these cases, the mecA-PCR was repeated from culture isolates. Enterococcus species were tested for the presence of vancomycin resistance genes vanA and vanB by in-house PCRs [18] using T3 thermal cycler (Biometra, Göttingen) followed by gel electrophoresis. Resistance of selected Enterobacterales against 3rd generation cephalosporins was further analyzed phenotypically for the presence of extended-spectrum beta lactamases (bla CTX-M ) using ESBL detection disk set (Mast Diagnostica GmbH, Reinfeld, Germany). 16S rRNA-PCR directly from positive blood culture bottles [17] using T3 thermal cycler (Biometra, Göttingen) was performed in case of lacking cultural growth.
Fungi were cultivated on yeast agar supplemented with gentamicin (in-house preparation). Single colonies were identified with MALDI-TOF mass spectrometry (Microflex LT, Bruker Daltonics, Germany) and BBL CHROMagar for Candida (BD Diagnostic Systems, Franklin Lakes, USA). Resistance profiles for fungi were not assessed for the study.

Biofire FilmArray® blood culture identification panel
The FilmArray® instrument was run according to manufacturer's instructions. In brief, the reagents in the pouch were adjusted with approximately 300 μl rehydration solution. Next, 100 μl medium from the positive blood culture bottle was mixed with sample buffer and loaded into the pouch, which requires only 2 min hands-on time. Once inserted in the FilmArray® instrument, nucleic acid extraction was performed followed by nested multiplex PCR giving results in about 1 h. The BioFire FilmArray® is using one reagent pouch for identification of Gram-positives, Gram-negatives, and fungi. The organisms covered by the assay are shown in Table S1. The software of the FilmArray® generates an automated report. One system was available during the study period to analyze one sample at a time.

Genmark ePlex® BC panels
The GenMark ePlex® system is using the eSensor technology deploying competitive DNA hybridization followed by electrochemical detection. Three different blood culture identification panels are available: one cartridge each for the identification of Gram-positives (BCID-GP), Gram-negatives (BCID-GN), and fungi (BCID-FP). The decision of which cartridge to use is based on the Gram-stain result. To run the assay in accordance with the manufacturer specifications, 50 μl of positive blood culture are transferred into the sample port of the respective cartridge followed by loading into the system. One tower with six test bays was available during the study period, offering the possibility to analyze six blood culture samples simultaneously. Result reports were generated  6]) was included in the study (in total n = 137). Samples having more than one obvious organism according to the Gram-stain (n = 1) and Gramstainings showing no organism (n = 8) were excluded from the study. Blood cultures positive for Gram-positive, Gram-negative organisms or fungi were analyzed with the FilmArray® and the ePlex® blood culture panels. On weekdays, PCR for Staphylococcus aureus and the mecA gene was performed directly from positive blood culture bottles, based on the Gram-stain result indicating the presence of staphylococci. Results obtained from the two evaluated test systems were compared to culture-(MALDI-TOF identification and Vitek-AST) or molecular-based (inhouse PCR for Sa442 and mecA or vanA/vanB respectively) reference method automatically after about one and half hours from starting the test. Organisms included in the panel are illustrated in Table S1. The cartridges for Gram-negative and Grampositive bacteria include a PAN target, that detects the presence of Gram-positives organisms in the Gram-negative (BCID-GN) cartridge and vice versa helping to identify polymicrobial mixed BSI that may have been missed by Gram-staining initially. In addition, a PAN-target for Candida is included on the bacterial detection cartridges.

Data analysis
Identification of blood culture isolates by MALDI-TOF was defined as reference method. In case of lacking growth, 16S rRNA-PCR was performed directly out of positive blood culture bottles. Dependent on the fast ID system tested, some species are only identified to the genus level. Exact genus ID-as identified by the reference method-was rated as correctly identified in case a fast ID system did not provide species level identification. If one of the two systems produced an invalid or failed run, it was repeated with a new identification cartridge. Due to operating hours of the laboratory during the study period, it was not possible to repeat every invalid or failed run generated by one of the two test systems. Therefore two samples (n = 2) were excluded from the study.
Detection of genotypic resistance markers by the test systems were compared to the results of in-house PCRs in relation to the phenotypic resistance obtained by Vitek 2 or agar diffusion (ESBL screening) respectively.
An identification was called 'true positive' for any organism covered by the FilmArray® or the ePlex® panel respectively and identified by the reference method as well as by FilmArray® or the ePlex®. 'True negative' was defined as any species not detected by the panels of the two systems and different from the reference method result. Organisms are called 'false positive' if identified by the FilmArray® or ePlex® system and not by the reference method. Any organism missed or misidentified by the two ID systems, but detected by the reference method, were marked as "false negative." The calculated specificity and the number of true negatives refer to positive blood cultures, which were negative for the respective target organism.

Results and discussion
During the study period, 153 positive blood cultures gave a positive signal and were assessed for inclusion in our analysis. Overall, eight positive blood cultures were excluded from the study, due to a negative initial Gram-stain result. In total, 137 positive blood cultures were included in our study (Fig. 1). Among those 98 episodes were monomicrobial BSI with Gram-positive organisms, 33 were monomicrobial with Gram-negative organisms and 6 episodes were caused by yeast strains (Tab.1). In addition, in eight samples, more than one organism was detected by culture. One example was not included in the study as the presence of multiple organisms was already suspected by the Gram stain, the results of the remaining seven samples are summarized in Table S2.

Identification of Gram-positive BSI pathogens
In summary 21 different Gram-positive organisms (98 isolates in total) were determined by routine-diagnostic methods. Identification by MALDI-TOF was compared to identification results created by the BioFire FilmArray® and the GenMark ePlex® system.
Correct identification by the FilmArray® was achieved for 86/98 (87.8%) isolates on genus and species level, as defined in the system panel for the respective target organism. The system scored 12 samples negative (12.2%). These 11 samples contained organisms that are not included in the system panel, namely C. acnes (n = 5), Actinomyces meyeri (n = 1), Actinomyces species (n = 1), P. faecalis (n = 1), Lactobacillus species (n = 1), Corynebacterium species (n = 1) and Atopobium parvulum (n = 1). One Staphylococcus warneri isolate was not detected as Staphyloccus species by the system. ID failure of S. warneri has been reported before and was explained by a reduced sensitivity of the FilmArray® system for some CoNS including S. warneri [19], which is also mentioned in the package insert of the panel. The FilmArray® therefore identified 86/87 (98.9%) correctly taking only the organisms included in the panel into account. The results are shown in detail in Table 1. For the Gram-positives, the FilmArray® showed a sensitivity of 98.9% and a 100% specificity (Table 4) for organisms covered by the panel.
In contrast to the FilmArray® system, the ePlex® system aims at identifying enterococci and coagulase negative staphylococci to the species level. In total identification to the species level was achieved in 89/98 (90.8%) of the Gram-positive bacterial isolates. Two Staphyloccus species were misidentified by the system: S. hominis as well as S. warneri were identified as S. epidermidis. A larger sample size is required to assess if this assay is more prone to misidentification of CoNS. Furthermore, it is difficult to rule-out that one or more of these specimens contained multiple Staphylococus species. In addition, as all these three CoNS species belong to the S. epidermidis group as part of the normal skin flora with similar pathogenicity [20] it is not mandatory to identify these CoNS on the species level. Worth mentioning is the inclusion of S. lugdunensis in the Genmark ePlex® panel as it is a common pathogen with species specific virulence factors [21] associated with aggressive causes of infective endocarditis [22][23][24]. n/a n/a n/a n/a Not included in the FilmArray® and ePlex®panel N. polysaccharea (n = 1) n/a n/a n/a n/a Not included in the FilmArray® and ePlex® panel C. canimorsus (n = 1) a n/a n/a n/a In addition, two E. faecalis isolates were misidentified as E. faecium. The correct identity of these two E. faecalis isolates was verified by MALDI-TOF and confirmed by next generation sequencing (NGS) of the whole genome on an Illumina Nextseq platform followed by calculating the average nucleotide identity (ANI) [25]. The ePlex® cartridges used in this study were for Research Use Only and according to the package insert of the test assay in a subsequent clinical study evaluating the performance of the CE-IVD and U.S. FDA cleared assays, the ePlex® demonstrated high sensitivity and specificity for identification of Enterococcus faecalis. Enterococcus species cause approximately 10% of nosocomial blood stream infections [26] caused by a variety of different entries of the pathogen [27]. Treatment of Enterococcus related BSI is strongly dependent on correct species identification. Whereas E. faecalis is mostly susceptible to the smallspectrum antibiotic ampicillin, E. faecium is resistant [28].
Classifying CoNS and enterococci only to the genus level, correct identification could be achieved for 93/98 (94.9%) of the Gram-positive bacterial isolates. Based on the broader panel C. acnes (n = 5), Lactobacillus species (n = 1) and Corynebacterium species (n = 1) were detected in contrast to the FilmArray® system. Four bacterial isolates were not identified as they were not part of the panel: Actinomyces meyeri (n = 1), Actinomyces species (n = 1), P. faecalis (n = 1) and Atopobium parvulum (n = 1). Moreover one S. haemolyticus strain was not detected by the system. Regarding only organisms covered by the panel, the ePlex® correctly identified 93/ 94 (98.9%) strains to genus level and 89/94 (94.7%) to the species level. For two blood cultures, the ePlex® detected a second Gram-positive organism (Micrococcus in addition to S. aureus and accordingly Staphylococcus spp. in addition to E. faecium). None of these additional organisms could be confirmed by culture ( Table 1). The ePlex® panel allows in contrast to the FilmArray® identification of potential contaminants namely Corynebacterium spp., Micrococcus spp., and C. acnes usually resulting in no need for treatment unless there is direct evidence for infection by the organism. Fast identification of such contaminants can help in quick de-escalation of the ongoing therapy and prevent redundant administration of anti-infectious therapies [29][30][31].
The ePlex® system showed a sensitivity of 94.7% and a specificity of 90.7% for Gram-positive pathogens included in the panel ( Table 4).
Percentages of detected organisms covered by the respective blood culture identification panel as well as for all microorganism detected by the reference method are shown in Table 2.

Identification of Gram-negative BSI pathogens
We further assessed identification of Gram-negative organisms for the BioFire FilmArray® and the Genmark ePlex® Due to the lack of bacterial growth, one sample with Capnocytophaga canimorsus was not verified by MALDI-TOF, but by 16S rRNA PCR out of the positive blood culture bottle. Two Klebsiella variicola isolates were reported as Klebsiella pneumoniae by the FilmArray® and GenMark ePlex®, but were not assessed as misidentification, as MALDI-TOF based differentiation of these two species has only recently become available [32]. Klebsiella variicola was described as a new species genetically isolated from K. pneumoniae and can be phenotypically distinguished by the inability of adonitol fermentation. The pathogenicity potential and analysis of virulence factors of K. variicola are still under investigation and high-risk antibiotic resistance genes are already described for K. variicola [33][34][35]. Thus, identification of K. variicola on species level and dissociation from K. pneumoniae could be of relevance for a future panel update. In contrast to the FilmArray® panel, the ePlex® system can identify the Bacteroides fragilis, which is of clinical importance since the B. fragilis group belongs to the most prevalent anaerobic pathogens causing BSI [36].

Detection of polymicrobial BSI pathogens
In total, for seven polymicrobial BSI episode samples, both test systems were performed (summarized in Table S2). The FilmArray® and ePlex® detected all pathogens covered by the panels at least on genus level with only one exception: the FilmArray® missed detecting a Streptococcus constellatus isolate although covered by the panel, which was found in addition to Proteus mirabilis in one blood culture bottle. In Table 2 Resistance genes identified for  the initial Gram-stain, only Gram-negative rods were observed. Furthermore, culture-based diagnostic showed only a few colonies of Streptococcus on CNA agar plates, indicating a low burden of Gram-positives in this blood culture. For the ePlex® the cartridge for Gram-negatives was used for the analysis of that blood culture resulting in the detection of the Gram-positive PAN target indicating the precence of the Streptococcus constellatus isolate. The Gram-positive PAN target was also detected in a polymicrobial BSI episode containing S. epidermidis in addition to K. pneumoniae.
In another polymicrobial blood culture, Fusobacterium nucleatum and Dialister pneumosintes were identified by culture. In the initial Gram-stain, the Gram-negative rods appeared Gram-positive, whereupon the ePlex® cartridge for Gram-positives was used. Both species are not covered by the FilmArray® panel and were not detected accordingly. However, no PAN target for Gram-negatives was detected by the ePlex®.

Detection of genotypic resistance markers
In total, 71 Staphylococcus species were identified in the study. Five Staphylococcus isolates could not be verified for the presence of mecA by in-house PCR as the isolates were not stored by the reason of routine laboratory workflow. Nevertheless, the susceptibility results of these five Staphylococcus isolates correlate with the detection results for mecA generated by the FilmArray® and the ePlex®. For the residual 66 Staphylococcus isolates in-house mecA-PCR results were available. All S. aureus isolates (n = 18) were methicillin-susceptible and negative for mecA as reported by the FilmArray®, the ePlex® and the in-house PCR (Table 3). Moreover, no mecC was detected by the ePlex® system. In the 48 BSI episodes with CoNS, 15 isolates were methicillin- Table 3 Resistance   [37]. However, due to increasing bacterial resistance, rapid detection of genes encoding carbapenemases is a desirable characteristic of molecular-based blood culture identification panels. In this regard the carbapenemase resistance genes covered by the ePlex® panel for Gram-negatives can be highlighted as the most common genes [38]. The ePlex® detected the presence of bla CTX-M for three study isolates (E. coli (n = 2) and K. pneumoniae (n = 1)), phenotypically confirmed by the presence of an ESBL phenotype in the cultured strains.
A summary of the overall performance of the two test systems is illustrated in Table 4. Concerning the detection of Gram-positive bacteria included in the panels of the FilmArray® and the ePlex®, the best performance with slightly higher sensitivity, specificity, PPV and NPV values was shown by the FilmArray® (Table 4). However, due to the broader identification panel of the ePlex®, putative contaminants or frequent anaerobes can be identified, resulting in on overall higher number of identified Gram-positive organisms (93/98) as compared to the FilmArray® with identification of 86 of 93 specimens. Regarding the identification of Gram-negative organisms and yeast strains, no major differences were found between the two systems.
In conclusion and in view of our study aims, both blood culture identification systems showed good results for fast pathogen recognition directly from positive blood cultures as well as for resistance gene detection. The broad coverage three panel approach of the ePlex® system implicate a drawback as it necessitates a Gramstaining prior to decide which cartridge to use for identification. However, the intended use of the other system also requires a Gram-stain as part of their approved instructions for use. Both systems require only short hands-on time, can easily be implemented in a routine microbiological diagnostic workflow and cause similar costs. The Accelerate Pheno® system, which is already implemented for high risk patients in our blood culture workflow [12], allows accelerating reports for ID and especially full AST for hard to predict resistance profiles for Gram-negatives. The panels of the FilmArray® and ePlex® systems with their ability for detecting common antimicrobial resistance genes are suitable for fast identification and rough genotypic resistance characterization of Grampositive organisms, specifically for Staphylococcus species demarcating S. aureus and MRSAs and Enterococcus species defining VREs. Specific performance characteristics of the assays, costbenefit ratio, laboratory operating hours, manning and state of knowledge of the laboratory personnel, patient population as well as effect on patient care from individual identification panels affect the advantages of rapid PCR-based blood culture diagnostics implicating the need of a considered selection and implementation of a rapid molecular ID system [39].
Acknowledgments Both companies were given the opportunity to comment on the manuscript. They had no role in the study design, experimental procedure, and data analysis.
We thank Matthias Willmann for critical discussion of the manuscript and the team of diagnostic technicians for their excellent technical assistance. Funding information Open Access funding provided by Projekt DEAL. The study was funded in part by the Institute of Medical Microbiology and Hygiene, Medical Faculty, University of Tübingen, Germany. During the study the FilmArray® (Biomerieux, Nürtingen, Germany) and ePlex® (GenMark Diagnostics, Carlsbad, USA) systems were provided by the companies. In addition, the reagents for the ePlex® system were provided, while for the FilmArray® reagent a discount was received during the study period.

Compliance with ethical standards
Conflict of interest SP received a speakers' honorarium from Biomerieux (Nürtingen, Germany) and is a consultant for IDbyDNA (San Francisco, USA).
Ethical approval The study was approved by the local ethics committee (no. 667/2014BO1 and 139/2016BO2).

Informed consent Not applicable to this study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.