Rapid Identification of Bacterial Pathogens in Positive Blood Culture Bottles by Use of a Broad-Based PCR Assay Coupled with High-Resolution Melt Analysis

Abstract

We evaluated a broad-based PCR assay coupled with high-resolution melt analysis for rapid bacterial identification in patients with bacterial sepsis. With a reference library of 60 clinically relevant bacterial species, 52 positive blood culture samples were tested. Our assay identified 46/52 samples at the species level, with 100% concordance to culture findings.

Bacterial sepsis is a significant cause of morbidity and mortality, with an estimated 20 million cases worldwide each year and up to 215,000 registered deaths per year in the United States alone (2, 8, 11). Early diagnosis of the pathogens causing bacteremia and sepsis is particularly challenging in clinical settings due to the lack of rapid and specific diagnostic assays. As a consequence, this often leads to the conservative approach of prescribing broad-spectrum empirical antibiotics. Previous studies reported lower mortality rates for patients given appropriate targeted antibiotics versus those treated with a broad-spectrum therapy (9). Thus, for patients with suspected bacterial sepsis, especially the immunocompromised, early identification of the causative bacterial agent is crucial for directing appropriate antimicrobial therapy and improving patient care and associated outcomes (4).

Detection of the etiologic agent is currently accomplished by a multistep, time-intensive process involving procurement of blood cultures and incubation of the bottles, followed by Gram staining and subculturing of the bottles when positive, and finally detailed biochemical analyses of the recovered organisms. The complete process typically requires 48 h (and up to 72 h for fastidious organisms), during which time clinicians receive incremental information regarding culture positivity, microscopic description, speciation, and finally antimicrobial susceptibility results (7). Further, false-negative results based on conventional methods may occur in up to 30% of sepsis cases, especially for fastidious organisms or in samples obtained after antimicrobial therapy has been started (1, 3, 12). Molecular techniques to detect microbial nucleic acids have been adapted to identify a wide range of bacterial pathogens found in sterile body fluids in efforts to circumvent the technical shortcomings of current culture-based methods. Unlike culture methods, such molecular assays do not require bacterial growth for detection and identification. While methods such as peptide nucleic acid-fluorescence in situ hybridization (PNA-FISH) and real-time nucleic acid amplification methods are currently available for rapid identification of organisms directly from a positive blood culture bottle, they are pathogen-specific assays and are costly to perform.

Nucleic acid amplification methods that target the bacteria-specific 16S rRNA gene can offer many advantages; the highly conserved sequences of the gene allow broad-based detection of almost any bacterial species, while the hypervariable sequences can be exploited for species-level identification (12). We have previously combined 16S rRNA PCR with high-resolution melt analysis (HRMA) for broad-range species identification. HRMA is not only simple, fast, and easy to integrate with PCR, but also it has the capacity to discriminate single-nucleotide differences between closely related species (10). Our objective was to evaluate the performance of a broad-based PCR-HRMA assay in identifying common bacterial pathogens in positive blood culture bottles.

A mixture of 60 clinical and reference strains of bacterial species representing a broad range of microbial pathogens, including the 18 most frequent causes of bloodstream infections among our patient population, were used as reference organisms for analysis (Table 1) (5). All organisms were subcultured onto Trypticase soy agar with 5% sheep blood and incubated at 37°C for 24 h. Ten to 15 colonies of each bacterial organism were inoculated into 200 μl of molecular-grade water prior to DNA extraction. Negative blood culture bottles were used as standard negative controls.

TABLE 1.

Melt library of common infectious pathogens

Pathogen	Grouping code of analysis subsetb			Signature codec	Pathogen	Grouping code of analysis subsetb			Signature codec
	V1	V3	V6			V1	V3	V6
Acinetobactersp. ATCC 5459	e	a	a	eaa	Mycobacterium fortuitum	m	l	e	mle
Aerococcus viridians	h	e	n	hen	Mycoplasma pneumoniaea	b	m	i	bmi
Bacteroides fragilisa	g	f	c	gfc	Mycoplasma hominisa	j	c	k	jck
Bordetella pertussis	j	d	f	jdf	Neisseria meningitidisATCC 6250	a	d	l	adl
Bordetella parapertussis	l	i	c	lic	Neisseria gonorrhoeaea	g	i	l	gil
Campylobacter jejunii	c	j	l	cjl	Oligella urethralis	i	h	l	ihl
Cornyebacteriumsp.a	a	j	a	aja	Pasteurella multocida	a	e	l	ael
Chlamydia pneumoniaea	g	a	b	gab	Pseudomonas aeruginosaATCC 10145	f	c	l	fcl
Chlamydia trachomatisa	m	n	n	mnn	Propionibacterium acnes	f	j	l	fjl
Citrobacter freundiia	a	b	m	abm	Proteus mirabilisa	a	b	c	abc
Clostridium difficile	h	d	a	hda	Proteus vulgarisa	j	g	m	jgm
Clostridium perfringens	d	b	b	dbb	Salmonellasp. ATCC 31194	c	g	m	cgm
Enterobacter aerogenes	c	b	e	cbe	Serratia marscecensATCC 8101	a	d	a	ada
Enterobacter cloacaea	e	d	j	edj	Staphylococcus aureusATCC 25923	h	e	g	heg
Enterococcus casseliflavusa	e	c	c	ecc	Staphylococcus epidermidisATCC 12228	n	e	a	nea
Enterococcus faecium	j	c	d	jcd	Staphylococcus capraea	d	e	a	dea
Enterococcus faecalisATCC 29212	c	h	a	cha	Staphylococcus haemolyticusa	b	e	b	beb
Enterococcus gallinarum	j	k	l	jkl	Staphylococcus hominisa	b	a	b	bab
Enterococcus raffinosusa	a	c	d	acd	Staphylococcus lugdunensis	d	a	e	dae
Escherichia coliATCC 25927	a	g	d	agd	Staphylococcus sapropyticusa	b	b	b	bbb
Helicobacter pyloria	g	b	f	gbf	Streptococcus pneumoniae ATCC 49619	b	e	a	bea
Haemophilus influenzaeATCC 49247	a	e	b	aeb	Streptococcus pyogenesa	m	a	d	mad
Klebsiella pneumoniaea	e	i	d	eid	Streptococcus agalactiaeATCC 13813	m	c	c	mcc
Legionella pneumophilaATCC 33495	c	d	a	cda	Streptococcus mitisATCC 49456	d	a	c	dac
Listeria monocytogenesATCC 7648	g	c	f	gcf	Streptococcus mutansATCC 25175	d	b	i	dbi
Micrococcus luteusa	g	i	h	gih	Streptococcus salivariusATCC 13419	g	d	b	gdb
Morganella morganiia	j	h	j	jhj	Streptococcus sanguisATCC 10556	g	e	i	gei
Moraxella catarrhalis	d	h	a	dha	Streptococcus oralisATCC 10557	d	b	c	dbc
Mycobacterium kansasii	g	j	h	gjh	Streptococcus gordoniiATCC 10558	k	b	c	kbc
Mycobacterium gordonae	k	b	b	kbb	Treponema palliduma	c	d	l	cdl

^aClinical isolate.

^bDifference plots generated for each organism were grouped based on curve similarity within each analysis subset (V1, V3, and V6), and the corresponding unique letter code was assigned to each group with a distinct curve shape.

^cCombined grouping code letters of the three analysis subsets.

The Clinical Microbiology laboratory at the Johns Hopkins Hospital (JHH) uses the BacT/Alert 3D blood culture system (bioMerieux, Durham, NC). The bottles in use contain the standard aerobic and standard anaerobic media. None of the bottles tested contained charcoal. A convenience sample consisting of 52 positive blood culture bottles (28 aerobic and 24 anaerobic) collected from 36 unique patients presenting to the JHH from July 2009 to August 2009 with suspected bloodstream infections were obtained from the hospital microbiology laboratory after standard blood culture testing was completed. Samples were deidentified for research purposes. The study was approved by The Johns Hopkins University Institutional Review Board.

The extraction of DNA from culture colonies was performed using a Roche MagNA Pure instrument (Roche Diagnostics, Indianapolis, IN) according to the manufacturer's instructions. For the isolation of bacterial DNA from clinical blood culture bottles, a 500-μl aliquot drawn from the blood culture bottle was treated with 50 μl of lysis buffer (MagNA Pure LC Kit-I; Roche Diagnostics) and incubated at room temperature for 20 min. The specimen was centrifuged at 4,000 × g for 10 min, and the cell pellet was resuspended in 100 μl molecular-grade water. The specimen was then treated with 130 μl of lysis buffer (MagNA Pure LC Kit-III; Roche Diagnostics), 20 μl of lysing enzyme mixture (10 μl of 0.8 μg/μl lysozyme and 10 μl of 1.25 μg/μl lysostaphin; Sigma-Aldrich, St. Louis, MO) and incubated at 37°C for 20 min. The specimen was then treated with 20 μl of 1× proteinase K (MagNA Pure LC Kit-I; Roche Diagnostics) and incubated at 65°C for 10 min. Subsequently, the specimen was subjected to a brief freeze-thaw at −80°C for 10 min and 95°C for 5 min and then sonicated for 10 min (T9000; Brasonic, Shelton, CT). The processed samples then underwent a final DNA extraction using the Roche MagNA Pure instrument (Roche Diagnostics) in accordance with the manufacturer's instructions.

Three conserved primers (V1-F, V1-R, V3-F, V3-R, V6-F, and V6-R) flanking three hypervariable regions (V1, V3, and V6) within the 16S rRNA gene were designed and used for PCR amplification as previously described (25). Extracted DNA from each bacterial organism or clinical sample was subjected to three PCR amplifications, each targeting the three hypervariable regions. The PCR setup and cycling conditions were as previously described (25).

Each post-PCR sample amplicon was subjected to HRMA on the LightScanner instrument (Idaho Technology). Melting temperatures ranged from 60°C to 95°C. Fluorescence data acquisition was obtained for every 0.1°C increase in temperature. Each PCR sample was examined in triplicate for HRMA and analyzed using the Light Scanner software version 2.0 (Idaho Technology). Melting curves generated from HRMA were analyzed as previously described (10). Briefly, derivative plots were generated to assess the number of melting peaks. Analysis subsets (V1, V3, and V6) were defined by the primer sets used. “Autogrouping” was performed on the difference plots to group all positive samples with similar curve shapes within the same analysis subset. A unique letter code was assigned for each group identified, starting with the letter a and progressing alphabetically. A combination of each letter from each of the variable regions was accumulated to provide a signature code for each organism. A reference library based on the signature codes of 60 reference organisms was established accordingly for use in pathogen identification (Table 1).

Fifty-two positive blood culture bottles were tested using our PCR-HRMA assay, and melting curve data generated from each of the samples were compared to our reference library for pathogen identification. Fifty of 52 (96.2%) positive blood culture samples contained a pathogen included in our library. Our assay results were concordant with original blood culture findings in 46/52 (88.5%) samples, with detailed species identification (Table 2). In 4/6 discordant samples, HRMA generated multiple melting peaks in their derivative plots, suggesting the presence of multiple pathogens. Further microbiologic analysis of these blood culture samples (reculturing and sequencing) confirmed polymicrobial growth in all four samples (Table 2). The other 2/6 discordant samples were not identifiable by our HRMA analysis. These samples were positive for Fusobacterium varium and coagulase-negative staphylococci (CoNS) by conventional blood culture methods. Both samples generated a single melting curve in HRMA but were not associated with any specific pathogens in our reference library. The unidentified CoNS sample was recultured and sequenced for definitive discrepant analysis, and the isolate was confirmed as Staphylococcus xylosus. Total PCR-HRMA assay time was approximately 3 h (2 h for DNA extraction, 1 h for PCR amplification, and 5 min for HRMA).

TABLE 2.

Identification of 52 blood culture bottle samples by PCR-HRMA and conventional blood culture methods

Conventional culture identification (no. of isolates)	Grouping code of analysis subsets			Signature code	PCR-HRMA identification (no. of isolates)	Analysis result for discordant samples (based on reculturing/sequencing)
	V1	V3	V6
Concordant samples
S. aureus (10)	h	e	g	heg	S. aureus (10)
CoNS (9)	n	e	a	nea	S. epidermidis (4)
	b	e	b	beb	S. haemolyticus (2)
	d	a	e	dae	S. lugdunensis (1)
	b	a	b	bab	S. hominis (1)
S. agalactiae (5)	m	c	c	mcc	S. agalactiae (5)
MRSAa (3)	h	e	g	heg	S. aureus (3)
P. aeruginosa (3)	f	c	l	fcl	P. aeruginosa (3)
H. influenzae (3)	a	e	b	aeb	H. influenzae (3)
E. coli (2)	a	g	d	agd	E. coli (2)
K. pneumoniae (2)	e	i	d	eid	K. pneumoniae (2)
M. morganii (2)	j	h	j	jhj	M. morganii (2)
M. luteus (2)	g	i	h	gih	M. luteus (2)
S. epidermidis (1)	n	e	a	nea	S. epidermidis (1)
E. faecalis (1)	c	h	a	cha	E. faecalis (1)
E. aerogenes (1)	c	b	e	cbe	E. aerogenes (1)
Enterobacter sp. (1)	c	b	e	cbe	E. aerogenes (1)
S. pyogenes (1)	m	a	d	mad	S. pyogenes (1)
Discordant samples
F. varium (1) (unidentified)b						Staphylococcus xylosus
CoNS (1) (unidentified)b						Mixed β Streptococcus species
S. pyogenes (2) (polymicrobial)c						α Streptococcus species, Enterobacteriaceae species
Viridans streptococci (2) (polymicrobial)c						α Streptococcus species, Enterobacteriaceae species

^aMRSA, methicillin-resistant S. aureus.

^bA single melt curve was generated for each analysis subset but was not associated with a pathogen within our reference organism library.

^cMultiple melt curve peaks indicated polymicrobial infection.

In this study, we developed a strategy combining traditional blood cultures with broad-based PCR and HRMA for rapid identification of bacterial pathogens. Discrete pathogen identification based on our reference panel of 60 bacterial organisms was obtained in 88.5% of the positive blood culture samples tested, but more importantly, 100% of these PCR-HRMA identifications were concordant with the culture results at the species level. Due to the ease of integrating HRMA with PCR, our approach offers a simple work flow with an assay time of 3 h without laborious post-PCR procedures (i.e., sequencing, microarray analysis, and mass spectrometry). However, the actual turnaround time to results when integrated into the workflow of a clinical laboratory may be longer. The use of sophisticated instruments (i.e., a SmartCycler for faster thermal cycling) for automated back-to-back testing may offer more real-time results than batched testing. Given that traditional identification methods based on biochemical analyses typically require a minimum of 26 h after initial culture positivity (6), our approach may provide results ∼23 h earlier than the standard methods.

One of the unique features of our assay platform is the potential range of identifiable pathogens. We have included in our reference panel not only the top 20 (or 95% of) organisms commonly recovered from blood cultures but also other clinically relevant organisms found in other body sites for future testing (5). The total number of organisms identifiable can be easily expanded (5). In our prior study, we demonstrated that identical signature codes were shared among various strains of the same species (10). Moreover, nonmatching signature codes generated from a positive amplification reaction can either represent the presence of polymicrobial infection if multiple peaks are observed in the melt curves or signify the presence of an uncommon, mutant, or emerging pathogen if a single dominant peak is seen. A similar strategy can also be adapted for fungal detection, and such work is under way. Of note, signature letter codes assigned to each organism may change if a different panel of reference organisms is selected (which explains the differences in the signature codes for the same organisms listed in Table 1 versus those listed in our previous publication [10]).

Our PCR-HRMA assay also demonstrated superior discriminating power over conventional methods for speciation. Among the 52 samples tested, conventional methods were only able to identify the organisms at the genus level in 11 samples (i.e., 10 CoNS and 1 Enterobacter). Our PCR-HRMA assay was able to further differentiate 9 of 10 CoNS isolates at the species level and 1 isolate of Enterobacter as Enterobacter aerogenes (Table 2). Although most CoNS are regarded as contaminants, the ability to discriminate CoNS species may help ascertain their clinical relevancy.

One of the current limitations of PCR-HRMA is its inability to resolve polymicrobial infections. Although the number of dominant peaks in the derivative melting curve may suggest the number of etiologic agents involved, the ability to identify the individual organisms is still lacking (10). Nonetheless, in our study, four discrepant samples with initial monomicrobial growth by culture were found to have multiple organisms by PCR-HRMA, which was confirmed by reculturing and sequencing. These findings suggest either subsequent contamination or a potential advantage of PCR-HRMA over conventional methods for determining the presence of polymicrobial infections.

Knowing the species of a pathogen at an earlier time point may offer significant clinical benefit. The information would be useful in de-escalating antibiotic coverage based on institutional susceptibility patterns to reduce emergence of drug-resistant pathogens, prompt an escalation in antibiotic use (for example, adding colistin for Acinetobacter or double-covering for Pseudomonas), or allow clinicians to disregard the report of a positive culture altogether in the case of identification of a common contaminant. Although PCR detection directly from whole blood without culturing would offer even more drastic time saving, the technical challenges in achieving adequate detection sensitivity still exist. Currently under way is an examination of an alternative approach to enrich the bacterial concentration in whole blood with brief culturing (2 to 3 h) prior to PCR-HRMA. Given that PCR-HRMA does not provide antibiotic susceptibility information, however, this method should not replace but rather offer an adjunctive role to standard culture methods.