Multiplexed Identification of Blood-Borne Bacterial Pathogens by Use of a Novel 16S rRNA Gene PCR-Ligase Detection Reaction-Capillary Electrophoresis Assay

Abstract

We have developed a novel high-throughput PCR-ligase detection reaction-capillary electrophoresis (PCR-LDR-CE) assay for the multiplexed identification of 20 blood-borne pathogens (Staphylococcus epidermidis, Staphylococcus aureus, Bacillus cereus, Enterococcus faecalis, Enterococcus faecium, Listeria monocytogenes, Streptococcus pneumoniae, Streptococcus pyogenes, Streptococcus agalactiae, Escherichia coli, Klebsiella pneumoniae, Haemophilus influenzae, Pseudomonas aeruginosa, Acinetobacter baumannii, Neisseria meningitidis, Bacteroides fragilis, Bacillus anthracis, Yersinia pestis, Francisella tularensis, and Brucella abortus), the last four of which are biothreat agents. The method relies on the amplification of two regions within the bacterial 16S rRNA gene, using universal PCR primers and querying the identity of specific single-nucleotide polymorphisms within the amplified regions in a subsequent LDR. The ligation products vary in color and size and are separated by CE. Each organism generates a specific pattern of ligation products, which can be used to distinguish the pathogens using an automated software program we developed for that purpose. The assay has been verified on 315 clinical isolates and demonstrated a detection sensitivity of 98%. Additionally, 484 seeded blood cultures were tested, with a detection sensitivity of 97.7%. The ability to identify geographically variant strains of the organisms was determined by testing 132 isolates obtained from across the United States. In summary, the PCR-LDR-CE assay can successfully identify, in a multiplexed fashion, a panel of 20 blood-borne pathogens with high sensitivity and specificity.

Identifying bacteria in the bloodstream is of the utmost importance in reducing morbidity and mortality in bacteremic patients. Correct identification of the causative agent and its antimicrobial susceptibility can guide treatment options and help determine appropriate and successful therapy (4, 5, 14, 15, 46, 53). The current gold standard for detecting bacteria in blood in clinical microbiology laboratories is the automated blood culture system (43, 61). These commercially available systems generally involve the cultivation of blood in broth culture bottles that are continuously monitored for the metabolites produced by the growing bacteria. Once the blood culture turns positive, routine microbiological testing involves Gram staining for early differentiation, followed by plating on appropriate selective and nonselective culture media, as well as additional biochemical tests for identification. Complete identification and susceptibility data are rarely available in less than 24 to 72 h after the blood culture becomes positive.

Molecular diagnostic methods based on the detection of bacterial nucleic acids from blood culture hold the promise of rapid detection and identification of the etiologic agent. Several different methods have been described, including either broad-range PCR (24, 35, 41, 62) or multiplex PCR (47, 48), real-time PCR (8, 18, 32, 33, 36), sequential PCR using sequence-specific probes (58), restriction fragment length polymorphism profile analysis (10), PCR-single-strand conformation polymorphism analysis (54), direct detection of bacteria from blood cultures using fluorescence in situ hybridization probes (29, 44, 56), and amplification of 16S and 23S ribosomal sequences for hybridization to a microarray (1, 12, 60). Relatively few assays are capable of detecting multiple pathogens simultaneously. They include the Hyplex Bloodscreen assay (57), the genotype blood culture DNA strip assay (17), PCR-mass spectrometry (7), resequencing microarrays (37, 38), and a DNA hybridization-based assay developed for the Luminex LabMAP system (16). Molecular assays that have been developed for the detection of biothreat agents rely on the use of real-time PCR or microarrays and were not designed for high multiplexing capacity (9, 11, 45, 49-51, 55).

Clinical microbiology laboratories today are compelled to outsource (to public health laboratories) isolates with any possibility of being a biothreat agent. This delays, and may negatively impact, the diagnosis, control, and handling of a bioterror attack. Additionally, as the anthrax “scare” of 2001 exhibited, public health laboratories may be overwhelmed by the extreme number of specimens submitted. For routine biothreat surveillance, the ability to screen for and detect multiple agents rapidly in a single reaction and with minimal sample processing is critical (11).

The ideal diagnostic assay would provide the rapid identification of any pathogen, whether a biothreat agent or a common bacterial pathogen. This would allow a physician to assess the likelihood of a bioweapon infection when faced with a patient who exhibits suspicious symptoms that are often indistinguishable from those caused by a common pathogen and to simultaneously obtain the identification of a nonbioterror etiologic agent.

The ligase detection reaction (LDR) was originally developed for discriminating single-base mutations or polymorphisms and is described elsewhere (2, 3, 59). LDR is ideal for multiplexing when combined with PCR, and PCR-LDR has been successfully applied to the multiplexed detection of mutations and single-nucleotide polymorphisms (SNPs) in cancer genes (19-21, 23, 30, 31). Here, we describe a 96-well high-throughput PCR-LDR-capillary electrophoresis (CE) assay to detect and distinguish 20 blood-borne bacterial pathogens, including 4 biothreat agents. A schematic representation of the assay is shown in Fig. 1. Briefly, universal PCR primers amplify two distinct regions of the 16S rRNA gene. A subsequent LDR identifies SNPs at specific positions within the two PCR amplicons. Each organism is then identified based on the distinct pattern of ligation products produced in the reaction. The CE data can be displayed as a virtual gel image or analyzed by a software program for automated identification of the pathogen. The assay has been validated using local clinical isolates, national strains, and seeded blood cultures. A schematic representation of the studies undertaken in this report is shown in Fig. 2.

FIG. 1.

Schematic of the PCR-LDR-CE assay for the identification of bacterial pathogens. Universal PCR primer pairs are designed to amplify two distinct regions of the 16S rRNA gene (only one amplicon is illustrated for clarity). Each PCR primer contains between one and three degenerate positions to accommodate minor sequence variation in the 16S rRNA gene at the primer binding sites. Within each PCR amplicon, LDR primer pairs are designed to identify SNPs at multiple locations to allow distinction between different bacteria. At any given SNP, the allele-specific LDR primers are designed to ligate to a locus-specific common primer. The LDR primers bear either a Vic or a Ned fluorescent label, and the common primers bear tails of different lengths. Ligation of the appropriate LDR primers and common primers results in fluorescently labeled products of different lengths that are then separated using CE.

FIG. 2.

Schematic representation of the multiplexed PCR-LDR-CE assay and the types of samples used to validate the assay.

MATERIALS AND METHODS

NucPrep DNA purification kits for bacterial DNA extraction, AmpliTaq Gold for PCR amplification, the LIZ-500 DNA size standard, and Hi-Di formamide for CE were obtained from Applied Biosystems (Foster City, CA). Lysozyme, lysostaphin, and activated charcoal were obtained from Sigma Chemicals (St. Louis, MO). Proteinase K was obtained from QIAGEN (Valencia, CA), and T4 polynucleotide kinase was supplied by New England Biolabs (Ipswich, MA).

Bacterial isolates.

The isolates used in this study were recovered from blood culture and other clinical specimens at the clinical microbiology laboratory at Weill Cornell Medical Center (WCMC). National strains were obtained from Georgia, Ohio, Minnesota, Florida, Arizona, Oregon, Missouri, California, and Iowa (see Table SA1 in the supplemental material for the sources of these isolates). Genomic DNA was obtained from the American Type Culture Collection (ATCC) for the following strains: Staphylococcus aureus (ATCC 700699D), Staphylococcus epidermidis (ATCC 12228D), Bacillus cereus (ATCC 14579D and ATCC 10987D), Escherichia coli (ATCC BAA-460D), Enterococcus faecalis (ATCC 700802D), Listeria monocytogenes (ATCC 19115D), Streptococcus pneumoniae (ATCC BAA-334D and ATCC 6308D), Pseudomonas aeruginosa (ATCC 17933D), Neisseria meningitidis (ATCC 53415D and ATCC 53414D), and Bacteroides fragilis (ATCC 25285D). Bacterial genomic DNA for the four biothreat agents, Bacillus anthracis, Yersinia pestis, Francisella tularensis, and Brucella abortus, was kindly provided by Kimothy Smith at the Lawrence Livermore National Laboratory.

Primer design and synthesis.

PCR and LDR primers were designed using Oligo 6.0 software (Molecular Biology Insights, Cascade, CO). PCR primers were designed to universally amplify two distinct regions of the 16S rRNA gene with melting temperatures in the range of 65°C. Universal tail sequences were appended to the 5′ ends of forward and reverse PCR primers to prevent the formation of primer dimers and to allow an additional universal amplification if required. LDR primers were designed to distinguish SNPs within each PCR amplicon. The allele-specific discriminating primers had melting temperatures of 70°C and were labeled at the 5′ end with either a VIC or a NED fluorescent label. The locus-specific common primers were phosphorylated at the 5′ end and blocked at the 3′ end with a C3 spacer and had melting temperatures in the range of 70 to 75°C. Filler nucleotides were appended to both discriminating and common primers (at their 5′ and 3′ ends, respectively) to modulate the lengths of the expected ligation products. Both PCR and LDR primers were synthesized with degenerate bases where required due to sequence variation between the 16S rRNA genes of different bacteria. PCR primers were obtained from Integrated DNA Technologies (Coralville, IA), and LDR primers were obtained from Applied Biosystems (Foster City, CA).

Preparation of local and national clinical isolates.

A total of 315 clinical isolates from blood and other clinical specimens that had been previously phenotypically identified using conventional microbiological techniques were grown overnight at 35°C on sheep blood agar, and the growth was checked for purity. Three to five colonies were suspended in 100 μl of Tris-EDTA buffer, pH 7.5. Samples were prepared in quadruplicate and transferred to a 96-well deep-well plate for extraction of DNA. National strains were prepared in analogous fashion. Negative controls (Tris-EDTA buffer) were incorporated in each 96-well plate.

Seeding of blood cultures.

Patients' BacT/ALERT (bioMérieux, Durham, NC) blood culture bottles that had not turned positive after more than 5 days of incubation were deemed negative and were incubated with isolates for the study. Clinical isolates were first inspected for purity on blood agar plates before one to four colonies were inoculated into 1 ml of soy-Trypticase broth. The entire 1 ml of inoculated broth was immediately injected into the above-mentioned blood culture bottles, and the bottles were incubated in the BacT/ALERT incubator for 4 h or until the culture turned positive, whichever was earlier. Quadruplicate 100-μl aliquots of the incubated seeded blood culture were transferred to a 96-well deep-well plate for extraction of DNA. Negative controls (Tris-EDTA buffer) were incorporated in each 96-well plate.

Extraction of bacterial DNA from clinical isolates and national strains.

DNA from clinical isolates, as well as national strains, was extracted on the ABI 6100 Nucleic Acid Prep Station with the NucPrep DNA isolation kit, with minor modifications. Extractions were carried out in one of two ways.

(i) Method I.

To a 96-well deep-well plate containing bacterial colonies suspended in 100 μl of Tris-EDTA buffer was added 10 μl of a 50-mg/ml stock solution of lysozyme and 10 μl of lysostaphin, and the plate was incubated at 37°C for 60 min. Twenty microliters of 20-mg/ml Proteinase K was then added, followed by incubation at 70°C for 30 min. This was followed by the addition of 500 μl of NucPrep DNA purification solution, and the samples were further processed according to the manufacturer's instructions. The bacterial DNA was eluted in a total volume of 200 μl.

(ii) Method II.

One hundred microliters of 5 M guanidinium HCl was added to each well of a 96-well deep-well plate containing bacterial colonies suspended in 100 μl of Tris-EDTA buffer, and the plate was vortexed briefly. The plate was then sonicated in an ultrasonicator bath for 15 min to lyse the bacteria. This was followed by the addition of 1 ml of NucPrep DNA purification solution and brief vortexing. The samples were further processed according to the manufacturer's instructions. The bacterial DNA was eluted in a total volume of 200 μl.

Extraction of bacterial DNA from seeded blood cultures.

One-hundred microliters of a 12-mg/ml suspension of activated charcoal in 5 M guanidinium HCl was added to each well of a 96-well deep-well plate containing the 100 μl of seeded blood cultures. The plate was sonicated in an ultrasonicator bath for 15 min to lyse the bacteria. The resulting solution was filtered through a Tissue Prefilter tray II and collected in a 96-well microtiter plate. This plate was centrifuged at 1,800 × g for 5 min, and the supernatant was transferred to a fresh 96-well deep-well plate. One milliliter of ABI NucPrep DNA purification solution was added to the wells, and the plate was briefly vortexed and then processed according to the manufacturer's instructions.

PCR-LDR-CE assay.

PCR amplifications were carried out in 50 μl of 10 mM Tris-HCl, pH 8.3, containing 50 mM KCl, 2.5 mM MgCl₂, 200 μM of each deoxynucleoside triphosphate, 15 pmol of each PCR primer (16sUniB2-PCR1BFN, CGCTGCCAACTACCGCACATCACTGAGACACGGYCCARACTCCTAC; 16sUniB2-PCR2RN, CGCTGCCAACTACCGCACATCBATMTCTRCGCATTTCACYGCTAC; 16sUniB2-PCR3FN, CGCTGCCAACTACCGCACATCCAAACAGGATTAGATACCCTGGTAGTC; and 16sUniB2-PCR4RN, CGCTGCCAACTACCGCACATCAYTTGACGTCRTCCCCRCCTTC [underlining refers to the universal tail]), 5 μl of template DNA, and 1.25 units of AmpliTaq Gold. Samples were thermocycled using the following parameters: 10 min at 95°C, followed by 35 cycles (95°C for 15 s, 60°C for 1 min, and 72°C for 1 min) and a final extension at 72°C for 7 min, followed by 99.9°C for 30 min to destroy the polymerase, before being held indefinitely at 4°C.

Two separate LDR primer mixtures were prepared, one for each amplicon, containing 500 fmol/μl of each of the appropriate discriminating and common LDR primers. An aliquot of each primer mixture was separately kinased prior to its use in LDRs in 40 μl of 50 mM Tris-HCl, pH 7.5, containing 10 mM MgCl₂, 1 mM ATP, 10 mM dithiothreitol, 25 μg/ml bovine serum albumin, 10 μl of the LDR primer mixture, and 10 units of T4 polynucleotide kinase, and they were then incubated at 37°C for 60 min, followed by a 20-min incubation at 80°C to destroy the kinase enzyme.

LDRs were carried out in 20-μl reaction volumes in 20 mM Tris, pH 7.6, buffer containing 10 mM MgCl₂, 100 mM KCl, 1 mM NAD, 1 mM dithiothreitol, 4 μl of kinased LDR primer mix, and 0.0125 μM AK16D thermostable ligase (52). The reaction mixtures were subject to thermal cycling using the following parameters: 94°C for 2 min, followed by 20 cycles (94°C for 30 s and 64°C for 4 min), before being held indefinitely at 4°C.

A 0.5-μl aliquot of each LDR mixture was added to 9.2 μl of Hi-Di formamide and 0.3 μl of LIZ-500 DNA size standard. The samples were denatured by heating them to 95°C for 3 min and cooled rapidly to 4°C before being loaded onto the ABI 3730 DNA analyzer for CE.

Data analysis and automated software identification.

Fragment analysis data from the CE of ligation products was analyzed and sized using GeneMapper 3.5 software (Applied Biosystems, Foster City, CA). The fragment size, color, fluorescence intensity, and peak area data were exported as text files that were then used to generate a virtual two-dimensional (2D) gel image (using Gelrender, a software program developed in our laboratory) or analyzed for the automatic identification of pathogens. A software program, “Infectious Agent Identifier,” was developed in our laboratory to process the text files exported from GeneMapper software as input, automatically filter noise peaks, detect signal peaks, and identify the organism(s) present. Identification was taken as definitive when two or more of the quadruplicate samples gave the same result. The program was designed to identify mixtures of organisms and to flag uncertain results for manual review. In such cases, a visual examination of the electropherogram was used to determine the identity of the organism present.

Determination of the limit of detection.

The limit of detection was initially determined at the DNA level by testing quadruplicate 10-fold serial dilutions of genomic DNAs from S. aureus, E. faecalis, E. coli, and P. aeruginosa. The limit of detection of the assay in detecting bacteria from blood cultures was determined by using 5-day-old negative blood cultures spiked with either S. aureus or E. coli. Tenfold serial dilutions of the bacterial suspensions were prepared in negative blood culture broth, and the CFU/ml was estimated by counting the number of colonies on triplicate plates streaked with aliquots of each dilution and incubated overnight. Bacterial DNA was directly extracted from each dilution without further incubation. For S. aureus, the concentration range tested was 1.4 × 10⁰ to 1.4 × 10⁸ CFU/ml. For E. coli, the concentration range tested was 9.8 × 10⁰ to 9.8 × 10⁷ CFU/ml. DNA was extracted from quadruplicate 100-μl aliquots of each dilution and tested by the PCR-LDR-CE assay.

Sequencing of discordant samples.

DNA extracted from discordant samples was subjected to sequencing of the 16S rRNA gene using the same PCR primers designed for the PCR-LDR-CE assay. PCRs were carried out as described above except that a single amplicon was generated in each PCR by using only one pair of amplification primers. The PCR products were purified using QIAQuick PCR purification plates (QIAGEN, Valencia, CA) according to the manufacturer's instructions. The PCR products were inspected for purity by electrophoresis on 1% agarose gels, as well as by measuring the absorbance at 260 and 280 nm. The purified PCR products were adjusted to concentrations of 3 to 5 ng/μl and sequenced using the BigDye Terminator 1.1 cycle-sequencing kit (Applied Biosystems, Foster City, CA) according to the manufacturer's directions. Sequencing reaction products were purified using Centrisep-8 strips (Princeton Separations, Adelphia, NJ), dried, and resuspended in 10 μl of Hi-Di formamide before CE on the ABI 3730 DNA analyzer.

The sequence data obtained were searched against the GenBank database using BLAST (http://www.ncbi.nlm.nih.gov/BLAST). The highest-scoring matches for each organism were recorded.

Detection of biothreat agents in excess of common pathogens.

DNAs from the four biothreat agents, B. anthracis, Y. pestis, F. tularensis, and B. abortus, were PCR amplified with the universal PCR primers to produce each amplicon in separate PCRs, as described above. Similarly, S. epidermidis and S. aureus DNAs were also PCR amplified with the universal primers. For each biothreat agent, each amplicon was diluted in increasing amounts of the corresponding PCR amplicon from either S. epidermidis or S. aureus to provide ratios of PCR products of 1:1, 1:5, 1:20, and 1:100. Two-microliter aliquots of these PCR product mixtures were subjected to LDR and CE as described above.

RESULTS

Assay design.

The assay was designed to be able to identify and distinguish a panel of 20 organisms. They were the gram-positive bacteria S. epidermidis, S. aureus, B. cereus, E. faecalis, Enterococcus faecium, L. monocytogenes, S. pneumoniae, Streptococcus pyogenes, and Streptococcus agalactiae; the gram-negative bacteria E. coli, Klebsiella pneumoniae, Haemophilus influenzae, P. aeruginosa, Acinetobacter baumannii, and N. meningitidis; the anaerobic B. fragilis; and the four biothreat organisms, B. anthracis, Y. pestis, F. tularensis, and B. abortus. The sequences of the 16S rRNA genes for all of the organisms were aligned using the MultAlin software program (13). Two sets of degenerate PCR primers were designed to universally amplify two distinct regions of the 16S rRNA gene, with each amplicon ranging between 375 and 415 base pairs. Within each amplicon, we selected SNPs that were common to groups of organisms, as well as SNPs that were unique to each given organism.

The SNP positions were selected to provide a hierarchical readout. For example, in each amplicon, an SNP that would allow distinction between gram-positive and gram-negative bacteria was chosen. A second SNP within amplicon 1 was common to and helped identify the gram-positive staphylococci, bacilli, and enterococci and L. monocytogenes. Another SNP distinguished the staphylococci and bacilli from the remaining organisms. Subsequent SNPs distinguished the staphylococci from the bacilli, and finally, S. epidermidis from S. aureus. A total of nine SNPs were queried by the assay across both amplicons with sufficient redundancy that the dropout of an SNP in any one amplicon did not preclude the identification of an organism.

LDR primer pairs were designed to identify these SNPs so that ligation products carried either a VIC or a NED fluorescent label and differed in length to enable separation using CE (see Table SA2 in the supplemental material for the locations and identities of the SNPs queried in this assay and Table SA3 in the supplemental material for LDR primer sequences). Figure 3 shows the distinct ligation products and the pattern generated by each organism on the assay panel.

FIG. 3.

DNA was extracted from clinical isolates or obtained from the ATCC. One-hundred nanograms of extracted DNA was subjected to 35 rounds of multiplexed PCR, and two 1-μl aliquots were subjected to 20 cycles of LDR for each amplicon. Ligation products were separated by CE on an ABI 3730 DNA analyzer. The ligation products contained either a VIC (green) or a NED (yellow) fluorescent label and were sized using a LIZ-labeled internal standard (the approximate length is indicated on the y axes). The CE data are displayed as a reconstructed gel image generated by a software program developed in our laboratory. The LDR products in amplicons 1 and 2 varied between 41 and 77 bases.

We evaluated two methods for DNA extraction, one using enzymatic treatment for bacterial lysis and the other using sonication for the lysis. The two methods yielded similar amounts of bacterial DNA. The sonication method, being faster and easier to implement, was used for extracting DNA from most clinical isolates and all national strains.

Validation with clinical isolates.

The assay was initially validated with 315 organisms isolated from blood or other clinical sources at the WCMC clinical microbiology laboratory. The results of this study are shown in Table 1. Six of the 315 isolates did not amplify, presumably due to failure of the nucleic acid extraction or the presence of inhibitors in the extracted DNA. The assay sensitivity for identifying pure cultures of clinical isolates was 98%. Of the remaining 309 isolates, the assay correctly identified 304. Five isolates of coagulase-negative staphylococci (CoNS) were incorrectly identified as S. aureus. Blast analysis of the 16S sequencing of these isolates identified three of them as Staphylococcus haemolyticus and one as Staphylococcus cohnii. The fifth isolate could not be determined to species level but had the highest consensus with an uncultured staphylococcus species from the GenBank database.

TABLE 1.

Validation of the 16S rRNA gene PCR-LDR-CE assay on clinical isolates from WCMC

Isolate	No. of isolates
	Tested	Correct	Discordant	Not detected
CoNS	47	42	5a
S. aureus	41	41
E. faecalis	36	34		2
E. faecium	38b	37		1
S. pneumoniae	14	14
S. pyogenes	22	22
S. agalactiae	22	21		1
E. coli	33	32		1
K. pneumoniae	15	14		1
H. influenzae	1	1
P. aeruginosa	19	19
A. baumannii	5	5
B. fragilis	22	22
Total	315	304	5	6

^aAll five were incorrectly identified as S. aureus. Sequencing identified three isolates as S. haemolyticus and one as S. cohnii. One isolate could not be identified to the species level but had the highest consensus with an uncultured staphylococcus species.

^bTwo of the 38 isolates were incorrectly identified as E. faecalis by conventional microbiological testing and were verified as E. faecium by sequencing a portion of the 16S rRNA gene.

Staphylococci are among the most frequently isolated bacteria in blood cultures and are generally characterized as S. aureus or CoNS. At WCMC, CoNS are not routinely identified to the species level, except when Staphylococcus lugdunensis is suspected or when requested by the attending clinician. The primers in our assay were specifically designed to identify S. epidermidis and to distinguish it from S. aureus on the basis of two SNPs, one in each PCR amplicon. An examination of the 16S rRNA gene sequences of other CoNS indicated that these SNPs were not specific for S. aureus. The SNP selected in amplicon 1 is common to S. aureus, Staphylococcus gallinarum, S. haemolyticus, and Staphylococcus hyicus. The SNP queried in amplicon 2 is common to S. cohnii, Staphylococcus saphrolyticus, Staphylococcus xylosus, Staphylococcus equorum, S. haemolyticus, and Staphylococcus delphini. Others have also reported the limited discriminating power of genotyping identification based on 16S rRNA gene sequences for different staphylococcus species (25). This limitation is shared by the current assay but can be rectified, as detailed in the Discussion.

Two isolates that had been identified as E. faecalis by conventional microbiology in the clinical laboratory were identified as E. faecium by the PCR-LDR-CE assay. In this case, sequencing verified that the assay had correctly identified these as E. faecium.

Validation with national strains.

The specificity of the assay in detecting a variety of strains from distinct geographical locations was validated with a sample pool of “national strains” of the panel organisms. The results of this study are shown in Table 2. The assay correctly identified all 132 isolates.

TABLE 2.

Validation of the 16S rRNA gene PCR-LDR-CE assay on geographically variant national strains

Isolate	No. of isolates
	Tested	Correct
S. epidermidis	9	9
S. aureusa	20	20
B. cereus	5	5
E. faecalis	8	8
E. faeciumb	10	10
L. monocytogenes	5	5
S. pneumoniae	6	6
S. pyogenes	8	8
S. agalactiae	9	9
E. coli	10	10
K. pneumoniae	10	10
H. influenzae	7	7
P. aeruginosa	10	10
A. baumannii	9	9
B. fragilis	6	6
Total	132	132

^aNine of the 20 isolates tested were methicillin-resistant S. aureus.

^bSix of the 10 isolates tested were vancomycin-resistant enterococci.

Validation with seeded blood cultures.

The assay was validated with seeded blood cultures (seeded with all 132 national strains and an additional 352 clinical isolates) to determine the ability of the assay to identify organisms from blood culture broth. The results of this validation study can be seen in Table 3. The assay identified bacterial DNA in 473 of the 484 samples, for a sensitivity of 97.7%, while 11 of the samples did not amplify. While this was most likely due to failure of the nucleic acid extraction or the presence of inhibitors in the extracted DNA, it is also possible that organisms seeded into the blood culture were not viable, did not grow during the 4-hour incubation, and were therefore present in very low numbers in the aliquots tested. The assay correctly identified 469 of the 473 samples, and 4 isolates of CoNS were incorrectly identified as S. aureus. Blast analysis of the 16S sequencing of these isolates identified all four of these as S. haemolyticus.

TABLE 3.

Validation of the 16S rRNA gene PCR-LDR-CE assay on spiked blood cultures

Isolate	No. of isolates
	Tested	Correct	Discordant	Not detected
CoNS	47	40	4a	3
S. aureus	55	53		2
B. cereus	3	3
E. faecalis	52	50		2
E. faecium	47	46		1
L. monocytogenes	5	5
S. pneumoniae	8	8
S. pyogenes	23	23
S. agalactiae	40	39		1
E. coli	46	46
K. pneumoniae	48	47		1
H. influenzae	17	17
P. aeruginosa	43	43
A. baumannii	41	40		1
B. fragilis	9	9
Total	484	469	4	11

^aAll four were incorrectly identified as S. aureus. Sequencing identified all four isolates as S. haemolyticus.

Determination of the limit of detection.

The limit of detection was first determined at the DNA level using serial dilutions of reference DNA for S. aureus, E. faecalis, E. coli, and P. aeruginosa. The 16S rRNA gene PCR-LDR-CE assay was able to detect identifiable signals above background from 200 fg of starting DNA for S. aureus, 800 fg for E. coli, 500 fg for P. aerigunosa, and as little as 100 fg of E. faecalis DNA. Based on the average genome size of 2.7 to 2.8 Mb for S. aureus, this corresponds to a detection limit of approximately 120 whole-genome equivalents in the initial PCR (26). The corresponding numbers of whole-genome copies required in the initial PCR for E. faecalis, E. coli, and P. aerigunosa were 60, 240, and 130, respectively. These results are comparable to the limits of detection reported by other assays, such as LightCycler PCR assays (limit of detection, 1 pg for S. aureus, 10 pg for E. faecalis, 100 fg for E. coli, and 10 fg for P. aerigunosa) (58) and the Luminex LabMAP platform (10 pg of genomic DNA) (16).

The limit of detection from spiked blood cultures was also determined with serial dilutions of organisms spiked into 5-day-negative blood culture broth bottles. S. aureus could be detected at a dilution of 1.4 × 10⁵ CFU/ml. This corresponds to the equivalent of 350 CFU of S. aureus in the initial PCR based on the calculation that DNA was extracted from a 100-μl aliquot and eluted in a total volume of 200 μl of buffer, with a 5-μl aliquot of the extracted DNA used in a 50-μl PCR mixture. For E. coli spiked into blood culture, the lowest dilution that could be detected was 9.8 × 10³ CFU/ml, which corresponds to 24.5 CFU of E. coli in the initial PCR mixture.

Detection of biothreat agents in the presence of a common pathogen.

Signals for the correct biothreat agent could be detected from mixtures of PCR products of the biothreat agent and a common pathogen at a ratio of 1:100. Figure 4 shows examples of mixtures of B. anthracis and F. tularensis with S. epidermidis at a ratio of 1:100. As expected, the signals that are specific to the biothreat agent are of much lower intensity than those of S. epidermidis, reflecting the 1:100 ratio of starting PCR products.

FIG. 4.

Detection of biothreat agent DNA in an excess of S. epidermidis DNA. PCR products from either B. anthracis or F. tularensis were mixed in different ratios with PCR products from S. epidermidis and subjected to LDR, followed by CE. Fluorescence intensity is indicated on the y axes, and the number of bases is indicated on the x axes. Fragments less than 40 bases in length represent unligated primers. The blue arrows indicate LDR products specific to S. epidermidis, the red arrows indicate LDR products specific to either B. anthracis (A) or F. tularensis (B), and the black arrows indicate LDR products that arose from both S. epidermidis and the biothreat agent present.

DISCUSSION

This study describes a novel 16S rRNA gene PCR-LDR-CE assay for the identification of 20 blood-borne bacterial pathogens, including four biothreat agents. Our objective was to develop an assay that could rapidly identify multiple pathogens in routine clinical use, as well as allow high-throughput screening of samples to identify biothreat agents in case of a suspected bioterror event. The assay uses universal PCR primers to amplify two distinct regions of the bacterial 16S rRNA gene. SNPs within these PCR amplicons are then queried using LDR primer pairs that generate ligation products that are separated by CE. Each bacterial pathogen generates a specific 2D molecular barcode based on the color and length of the ligation products. A software program, Gelrender, was developed to display the CE data as a virtual gel image for ease of visualization. We also developed a software program, Infectious Agent Identifier, to analyze the CE profile for automatic scoring of the ligation products and identification of the pathogen.

The bacterial pathogens on the assay panel were chosen to include the organisms most frequently isolated from blood culture at WCMC, as well as organisms representative of various categories of bacteria. For example, S. epidermidis was selected as a representative of CoNS, L. monocytogenes was chosen as a representative of small gram-positive non-spore-forming aerobic bacilli, H. influenzae was selected as a representative gram-negative fastidious bacillus, and B. fragilis was selected to represent gram-negative anaerobes. B. cereus was chosen as a surrogate for B. anthracis, while the four biothreat organisms are a part of the National Institute of Allergy and Infectious Disease select agent list. The WCMC routinely performs over 46,000 blood cultures annually, approximately 10% of which are positive. The time to final identification and determination of antimicrobial susceptibility is generally 1 to 3 days after the blood culture turns positive. In contrast, the PCR-LDR-CE assay, including DNA extraction, PCR amplification, LDR, CE, and data analysis, can be completed in less than 8 h. The assay has been developed in a 96-well format and is amenable to automation using robotic liquid-handling platforms, allowing rapid identification in a high-throughput fashion.

Figure 3 shows the distinct 2D molecular barcode generated by the assay for each organism on the panel. The lowest-running VIC-labeled band in amplicon 1 distinguished the gram-positive bacteria from the gram-negative bacteria (except F. tularensis and the anaerobe B. fragilis), which were characterized by the lowest-running NED-labeled band. Likewise, a double set of VIC-labeled bands in amplicon 2 were distinctive of the gram-positive organisms. Each organism provided sufficient signals to distinguish it from all other organisms on the panel. A total of six SNPs were selected to identify B. fragilis, three in each amplicon.

The assay was initially verified with clinical isolates from blood and other sources that had been collected at WCMC. The numbers of samples for H. influenzae and L. monocytogenes, which are rare isolates at our institution, were limited. Definitive identification was obtained for 309 isolates, and all but five samples were correctly identified. The five discordant samples were isolates of CoNS that were misidentified as S. aureus. A number of isolates of the pathogens were also obtained from clinical laboratories around the United States to account for geographical variants among the organisms. A total of 132 isolates were identified correctly, suggesting that the assay design is robust enough to identify organisms from different geographical locations.

Since the assay has been designed for detecting blood-borne pathogens, the technique was also validated with seeded blood cultures. Sodium polyanetholesulfonate (SPS) is a component of blood culture medium and a known inhibitor of PCR. Others have successfully used benzyl alcohol treatment steps (22) or alkaline lysis (39) methods to remove SPS and other PCR inhibitors from blood cultures during DNA extraction. However, none of these methods are easily transferable to a 96-well high-throughput format. A modified protocol for DNA extraction, involving sonication of the bacteria in the presence of charcoal and guanidinium HCl, was both compatible with a 96-well high-throughput platform and successful in removing SPS from the blood culture medium. The assay was successful in correctly identifying bacteria seeded into blood cultures, again with the exception of four isolates of CoNS that were misidentified as S. aureus. In this case, all four isolates were identified as S. haemolyticus by sequencing of the 16S rRNA gene.

In the current study, we encountered a total of nine samples (five clinical isolates in the initial validation set and four additional isolates seeded into blood cultures) in which CoNS were mistyped as S. aureus. This was due to shared SNPs rather than nonspecific LDRs. One approach to improve the ability of the assay to distinguish S. aureus from CoNS involves querying additional SNPs. An alternative method is the addition of PCR and LDR primers for the S. aureus coagulase gene into the assay to allow unambiguous distinction of S. aureus from CoNS species. The latter approach has been validated with over 1,000 clinical blood cultures in a subsequent study (unpublished data). As seen in Fig. 3, B. cereus and B. anthracis cannot be distinguished on the basis of ligation products formed in amplicon 1. The single-base difference between the two organisms seen in amplicon 2 may only reflect base pair variation between the multiple copies of the 16S ribosomal gene. To improve the ability of the assay to distinguish these two organisms, we have incorporated primers targeted to genes on the pathogenicity-conferring plasmids pXO1 and pXO2 of B. anthracis (data not shown).

The limit of detection ranged from 350 CFU/PCR for S. aureus to approximately 25 CFU/PCR for E. coli. We expect this analytical sensitivity to be sufficient for detecting bacteria from positive blood cultures. The assay sensitivity may also be increased by exploring other methods of DNA preparation that provide higher yields of bacterial DNA or by extracting bacterial RNA, which is present in higher copy numbers.

The assay presented in this study shows very high specificity for detection and identification, with pure clinical isolates, as well as with bacteria spiked into blood culture. Negative controls incorporated at all stages of the assay showed no false-positive identification. Broad-range 16S rRNA gene PCR-based assays, while sometimes more sensitive, are more likely to detect contaminating bacterial DNA from reagents used in PCR or in DNA extraction (27, 34). Taq polymerase is known to be contaminated with bacterial rRNA genes (6, 28). Contamination of commercial DNA extraction kit reagents has also been reported (40). Furthermore, false positives may arise from residual nucleic acid contaminants in blood culture (22), as well as bacterial DNA in the blood samples derived from healthy controls (42). The use of an orthogonal detection technology, LDR, in addition to PCR greatly reduces the chances of false positives and results in high specificity. Another advantage of using LDR coupled with PCR in this context is the high multiplexing capability of LDR. This makes it feasible to design additional primers to distinguish a larger panel of organisms without significantly affecting the sensitivity or specificity of the technique. The assay design is compatible with the universal microarray, which allows the detection of a large number of ligation products and avoids the issue of unpredictable mobility in CE.

The four biothreat agents tested in our assay are part of the select agent list of the National Institute of Allergy and Infectious Disease. The assay was designed to detect and identify these organisms, along with common pathogens, in the same test. Genomic DNAs from different strains of these organisms have only recently become readily available and will be tested in the near future for further validation. Nevertheless, we did test the ability of the assay to distinguish the biothreat agents from a large excess of normal pathogens, specifically S. epidermidis and S. aureus. In each instance, LDR products arising from the biothreat agent could be readily distinguished from those of the two staphylococci.

In conclusion, we have developed a novel high-throughput PCR-LDR-CE assay for the identification of 16 blood-borne pathogens and four biothreat agents. The assay has been successfully validated against 447 clinical isolates and 484 seeded blood cultures. The modular nature of the assay lends itself to the possibility of expanding the number of organisms that can be identified. The incorporation of additional PCR and LDR primers can improve the ability of the assay to unambiguously distinguish S. aureus from CoNS and B. anthracis from B. cereus. The assay can provide “same-day” identification in a high-throughput 96-well format, which represents a significant reduction in time compared to conventional microbiological methods for identification of bacteria from positive blood cultures. While some molecular identification methods provide faster identification, they are not designed to identify multiple pathogens simultaneously. Samples can be assayed with the same experimental protocol irrespective of the pathogen present and without requiring Gram staining or prior differentiation of the pathogen. This assay represents a significant improvement over conventional and molecular techniques due to its rapid high-throughput format and multiplexing capabilities.