In this manuscript, we describe the identification of highly pathogenic bacteria using an assay coupling biothreat group-specific PCR with electrospray ionization mass spectrometry (PCR/ESI-MS) run on an Ibis PLEX-ID high-throughput platform. The biothreat cluster assay identifies most of the potential bioterrorism-relevant microorganisms including Bacillus anthracis, Francisella tularensis, Yersinia pestis, Burkholderia mallei and pseudomallei, Brucella species, and Coxiella burnetii. DNA from 45 different reference materials with different formulations and different concentrations were chosen and sent to a service screening laboratory that uses the PCR/ESI-MS platform to provide a microbial identification service. The standard reference materials were produced out of a repository built up in the framework of the EU funded project “Establishment of Quality Assurances for Detection of Highly Pathogenic Bacteria of Potential Bioterrorism Risk” (EQADeBa). All samples were correctly identified at least to the genus level.
Citation: Jacob D, Sauer U, Housley R, Washington C, Sannes-Lowery K, Ecker DJ, et al. (2012) Rapid and High-Throughput Detection of Highly Pathogenic Bacteria by Ibis PLEX-ID Technology. PLoS ONE 7(6): e39928. doi:10.1371/journal.pone.0039928
Editor: Baochuan Lin, Naval Research Laboratory, United States of America
Received: December 1, 2011; Accepted: June 4, 2012; Published: June 29, 2012
Copyright: © 2012 Jacob et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This publication has been produced with the support of the European Commission's Executive Agency for Health and Consumers (EAHC). Its content is the sole responsibility of Robert Koch-Institut, Centre for Biological Security, and can in no way be taken to reflect the views of the EAHC or any other body of the European Union. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: Authors RH, CW, KSL, DJE and RS are employed by Abbott, or its affiliate, Athogen, where the assay described in the manuscript was developed and samples were tested. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials.
Highly pathogenic bacteria can often cause infectious zoonotic diseases and even large outbreaks in the human and animal population. A number of these agents have the potential for, or have been used, in bioterrorist attacks –. The intentional release of infectious agents can result in serious public health consequences as shown by the anthrax episodes in the USA in 2001. Rapid and reliable laboratory detection and confirmation strategies for potential bioterrorism agents contribute to reducing the health risk and improving emergency response and public health control initiatives .
In case of deliberately or naturally occurring pathogens, the agent must be identified as rapidly as possible to allow measures to be taken to prevent spread and to ensure proper treatment of casualties. In addition, unknown materials accompanied by declarations threatening individual persons, organizations, or events must be analyzed rapidly. In both these scenarios, the assay used must be able to identify a broad panel of potential threat microorganisms, possibly in a background matrix that is contaminated with non-pathogenic bacteria or viruses in order to exclude or confirm the presence of a biological threat.
A number of bacterial pathogens were classified by CDC (Centers for Disease Control and Prevention) as category A and B comprising the highest concern for use in bioterrorist attacks , . The categories include bacterial agents such as Bacillus anthracis, Francisella tularensis, Yersinia pestis, Burkholderia mallei and pseudomallei, Brucella species, and Coxiella burnetii. These microorganisms also cause naturally occurring diseases in animals and humans. In most European countries, the natural prevalence of these pathogens is low, although, from time to time some of the pathogens, e.g. F. tularensis, Brucella spp., or C. burnetii cause outbreaks in animal and human populations. Low disease prevalence and low diagnostic demands are probably the main reasons why existing diagnostic and detection assays are almost not commercially available for this group of zoonotic diseases. Existing diagnostic and detection assays for highly pathogenic bacteria are primarily based on classical microbiological and on molecular and immunological methods, mainly developed as in-house assays. Some multiplex assays have been designed to simplify and shorten the period of identification for pathogens responsible for infectious diseases. However, a highly sensitive yet broadly inclusive approach for the identification of all bacteria under suspicion is not yet available. In addition, due to the lack of reference samples for clinical as well as environmental samples, it is often difficult to validate and determine the accuracy of broad-spectrum assays.
An interesting approach for a complete identification of a large number of microorganisms is the Ibis PLEX-ID technology offered by Abbott –. This technology is based on polymerase chain reaction/electrospray ionization-mass spectrometry (PCR/ESI-MS). A number of primer pairs are used to amplify nucleic acid markers from a wide variety of organisms, followed by an analysis of PCR amplicons using mass spectrometry. The analysis of precise masses of these amplicons allows the calculation of the base compositions of the PCR products; these base compositions are then compared to a database for identification of the pathogen or pathogens. This platform provides a broad infectious disease detection capability and has been used for the detection of bacteria and viruses from a variety of different sources , –. In the accompanying manuscript, Sampath et al.  describe an approach for the identification of most relevant biothreat agent groups, the so called “biothreat cluster assay”, for the PLEX-ID that uses 36 multiplexed primer pairs for detection of biothreat agents. The biothreat cluster assay was validated by analysis of a broad collection of biothreat organisms and near neighbors prepared by spiking biothreat nucleic acids into nucleic acids extracted from filtered environmental air. The biothreat assay detects 28 NIAID (National Institute of Allergy and Infectious Diseases) priority pathogens  and 18 HHS/USDA (United States Department of Health and Human Services/United States Department of Agriculture) select agents .
In the framework of the European Union (EU) funded project “Establishment of Quality Assurances for Detection of Highly Pathogenic Bacteria” (EQADeBa) we assembled a repository of the bacterial pathogens mentioned above which we then used to produce standardized reference material. The EQADeBa project provides the infrastructure and the design, the organization, and the management of an external quality assurance (EQA) scheme for potential bioterrorism agents (European Commission, EAHC – Agreement n° - 2007 204). The project includes designing and implementing practical proficiency tests for the detection of bacterial agents of CDC category A and B. Standardized reference materials have been prepared in different formulations, sometimes as mixed cultures and in different concentrations for these exercises. We used material from two proficiency tests, from which 45 different samples were chosen after DNA-preparation for analysis using the biothreat cluster assay and the PLEX-ID technology. All samples were correctly identified as described below.
Materials and Methods
The bacterial strains used are listed in Table 1 and Table 2. Strains from the EQADeBa repository were plated on blood agar plates with the exception of Francisella sp. which was plated on cysteine heart agar (QUELAB, Quebec, Canada). The identity of all used strains was verified by appropriate microbiological and molecular methods. Strains were cultivated under appropriate biosafety level (BSL) 2 or 3 standards. After sterility testing, inactivated suspension or DNA produced under BSL3 conditions were handled under BSL2 conditions. Inactivated cells from C. burnetii were kindly provided by Georg Baljer (Justus-Liebig-Universität, Gießen, Germany).
15 DNA samples were directly prepared from bacterial cultures and used as reference material (Table 1). Total DNA was purified according to a modified protocol for Gram-negative bacteria using the QIAGEN DNeasy Blood and Tissue kit (Hilden, Germany). Rather than harvest cells from broth media, bacteria from agar plate were inoculated directly into 180 µl ATL buffer. After adding 20 µl proteinase K, we incubated 90 minutes at 56°C followed by adding 200 µl AL buffer for 10 minutes at 70°C. To avoid cross contamination we mixed all samples only by pipetting and added extra centrifugation steps between heat incubation steps. After at least one week of sterility testing, DNA concentrations were measured using the nanodrop 8000 (peqlab Biotechnologie GmbH, Erlangen, Germany). Dilutions of DNA were carried out to 106 copies/µl using AE buffer. The mean values of three DNAs of each bacterium were used to estimate the genome equivalents. Each bacterial DNA was analyzed by real-time PCR to ensure that no cross-contamination had occurred.
Inactivation of Bacterial Samples
To prepare reference materials, each bacterial species was inoculated into PBS to a density of approx. OD600 = 1.0 directly from colonies of agar culture media. To estimate the colony forming units (cfu) of each suspension, dilutions were plated onto blood agar plates (or cysteine heart agar plates for F. tularensis). Each sample was tested in duplicate. Bacterial suspensions were inactivated by heating at 60°C for at least 22 h. The exception was B. anthracis which was treated by 1% peracetic acid for 30 min with subsequent washing. All inactivated suspensions were checked for sterility by one tenth of each volume over 14 days. After inactivation, cell numbers in each suspension were determined using a counting chamber and real-time PCR. Suspensions of all inactivated samples were analyzed for cross contamination with other target bacteria by real-time PCR and by immunological methods (direct immune fluorescence and ELISA).
Preparation of “Inactivated Bacteria Samples”
Six different matrices were chosen to spike with the different inactivated bacteria (Table 2): 1) PBS as a substitute for isolate samples (Iso), 2) inactivated mouse hybridoma cells (Ag8) with a concentration of 2×106 /ml as a surrogate for the tissue samples (Tiss), 3) autoclaved and sterile filtrated natural water from river “Spree” (Env-1 through Env-5), 4) autoclaved water from lake “Plötzensee” as matrix for the environmental samples (Env-6 through Env-10), 5) a reduced-fat, pasteurized and homogenized milk served as a substitute for the food samples (Food), and 6) fetal bovine serum from Invitrogen/Gibco (lot# 41A1268K) was used as surrogate for the clinical samples (Clin). Each bacterial suspension was diluted 1∶10 with the corresponding matrix. Each set of vials for the environmental samples Env-6 to Env-10 and for the food as well as for the clinical samples contained two different target bacteria with two different concentrations. Additionally, each matrix was included as a negative control (Table 2).
Preparation of DNA from “Inactivated Bacteria Samples” and PLEX-ID Analyses
DNA was isolated from “Inactivated Bacteria samples” using the QIAGEN DNeasy Blood and Tissue kit according to the protocol for Gram-negative bacteria. A total of 45 samples, 30 samples with approx. 100 µl DNA out of the “Inactivated Bacteria samples” and the 15 “DNA samples” were sent for analyses to AthoGen , a service screening laboratory that uses the PCR/ESI-MS platform to provide a microbial identification service. PCR was performed in a 50 µL reaction volume containing 5 µL nucleic acid extract in a reaction mix as previously described . One-step real time (RT)-PCR was performed in wells that targeted viral detection. Reverse transcription PCR cycling conditions were used for both RT-PCR and PCR reactions according to previously described methods , –. Samples and reaction components were aliquoted into PLEX-ID biothreat cluster assay (Cat. No. 03N35-63) plates using a Janus Automated Workstation. One 96 well plate allows to analyze six samples in parallel, means 16 wells per sample. Thermocycling was carried out on an Eppendorf Mastercycler. After thermocycling, plates were stored at –40°C until the samples could be analyzed by ESI/MS. After PCR amplification, 30 µL aliquots of each PCR reaction were desalted and analyzed by mass spectrometry as previously described , .
Data Analysis and Results Reporting
Data analysis and results reporting was performed as described in Sampath et. al. . Briefly, a customized reporting rule set was designed that allowed rapid and accurate detection of the biothreat targets. The biothreat assay report provides data obtained using 36 primer pairs. Each of the threat clusters is treated independently and the results are reported for each cluster separately. Mixed detections of two or more threats or a threat with an unrelated near neighbor in another group are also reported. The system is also capable of reporting organism/strain level matches based on the genomic sequence data in the PLEX-ID database. This is provided as a research utility tool in a separate analysis workstation. Additional details for each of the matches, including the detected base compositions, Q-scores and levels, are available using this report and are described in the accompanying manuscript.
The PLEX-ID biothreat cluster assay (Ibis Biosciences, Carlsbad, CA) described in the accompanying manuscript by Sampath et al.  provides the capacity to analyze samples for targets as B. anthracis, Y. pestis, F. tularensis, Vibrio cholerae, C. burnetii, as well as Rickettsia, and Orthopoxvirus or Filovirus. The 36 primer pairs used in the assay are multiplexed such that 16 wells on a 96-well plate are required per sample; six samples can be analyzed in parallel on a single plate. The PLEX-ID assay system correctly identified, with one exception, all of the 45 unknown samples targeting highly pathogenic bacteria, using DNA template from these organisms (Table 1, DNA-12). In one case, Brucella abortus was recognized as Brucella melitensis. At the same time, samples containing B. melitensis were analyzed correctly (Table 1: DNA-8; Table 2: Food-2, Food-5).
In analyses of “DNA samples” and spiked “Inactivated Bacteria samples” containing Burkholderia, B. mallei (Table 1: DNA-13; Table 2: Clin-3, Clin-4) could not be discriminated from B. pseudomallei. Further, signatures for some isolates of B. pseudomallei (Table 1: DNA-10; Table 2: Env-7, Env-8) could not be differentiated from Burkholderia xenovorans (Table 1: DNA-10; Table 2: Env-7, Env-8). This might be a problem for risk assessment because B. pseudomallei is classified in risk group 3, whereas B. xenovorans is classified in risk group 1 and is not pathogenic for humans or animals. This limitation has been described by Sampath et. al. Interestingly, samples containing Burkholderia thailandensis (risk group 1), the nearest relative of B. pseudomallei, were clearly identified (Table 1: DNA-4; Table 2: Iso-4, Tiss-2, Env-4).
Different samples containing F. tularensis were always identified correctly at the species level (Table 1: DNA-2, DNA-11; Table 2: Iso-3, Tiss-1; Env-5, Clin-1, Clin-2), but there is no differentiation between the subspecies in the species summary reporting. Since strain differentiation would be of interest because subsp. tularensis, subsp. holarctica, and subsp. mediasiatica cause tularemia in humans, whereas subsp. novicida shows a low virulence and causes very rarely disease in immune-compromised individuals. In this case, a secondary data evaluation would be offered by the PLEX-ID platform using appropriate software that contains detailed subspecies information, and researchers should use this offer if a Francisella subspecies is identified. This approach was not applied in our study. A sample comprising a distinct Francisella species, Francisella philomiragia, was correctly identified (Table 1: DNA-7).
Several samples analyzed contained identical bacteria with different quantities. There was reasonable agreement between quantitative analysis of bacterial numbers obtained by the counting of bacteria in counting cell chambers and the “Genomes per well” calculated by PLEX-ID technology based on a calibrant in the assay wells for B. pseudomallei (Env-7 and Env-8), Y. pestis (Env-9 and Env-10), F. tularensis (Clin-1 and Clin-2), B. melitensis (Food-2 and Food-5), and B. mallei (Clin-3 and Clin-4). No significant difference in quantities of B. anthracis present in samples Food-1 and Food-3 was observed using the PLEX-ID technology although previous counting in the sample preparation process revealed a 10-fold difference.
The closely related species Y. pestis and Yersinia pseudotuberculosis as well as B. anthracis and other Bacillus cereus-group strains were correctly identified using the PLEX-ID biothreat cluster assay. Bacillus thuringiensis and B. cereus were not differentiated (Table 1: DNA-9, DNA-15), consistent with the description in Sampath et. al.
There were observed some anomalous results, yelding in false positive and false negative data or indications of potential cross-contaminations in some of the assays. In one negative control (Table 2: Iso-5) a low genome number of F. tularensis was detected, partly caused by the very high concentrations of the bacteria in some samples. The source of the contamination problem remains to be determined. The generation of the reference material and the DNA extraction process that were performed within the laboratory at the Robert Koch-Institut was shown not to be the cause of the contamination. The unexpected identification of Clostridium perfringens in sample Food-4 (negative control milk) should be clarified. The detection of Burkholderia vietnamensis in sample Env-6 comprising natural lake water could be explained by a cross reactivity or presence of naturally occurring DNA.
In addition to specific targets, C. burnetii DNA could be detected in the commercial milk matrix. The presence of this bacterium was confirmed by the External Quality Assurances Exercises (EQAEs) in the framework of the EQADeBa project.
In most of the DNA samples (Table 1) and several other samples (Table 2) traces of Escherichia coli DNA were detected. It is not possible to evaluate retrospectively, whether this contamination was due to the amplification or the measurement process.
In conclusion agents present in the samples can be identified with a high level of confidence.
The biothreat assay performed using the PLEX-ID technology can be used to detect a broad collection of biothreat organisms and near neighbors as described in the accompanying manuscript. This assay uses 36 pairs of multiplexed primers to amplify bacterial or viral DNA present in environmental or clinical samples. After analysis of the amplicons by mass spectrometry, base counts of the amplicons are determined. By comparison with the more than 850000 entries in the PLEX-ID database, biothreat agents present in the sample can be identified. There are other multiplex PCR or MALDI-TOF mass spectrometry based methods for analysis of infectious agents –; however, these methods are less broad than the PLEX-ID technology. The PLEX-ID approach provides the possibility for a simultaneous differentiation of multiple pathogens resulting in quantitative data given by “Genomes per well” and the probability of correct identification calculated by the “Q-Score”.
We used 45 samples from External Quality Assurances Exercises (EQAEs) performed in the framework of the above-mentioned EU project on quality assurance of the diagnosis of highly pathogenic bacteria (EQADeBa) to evaluate the ability of the PLEX-ID to correctly identify pathogens within this biothreat cluster. In addition, we assessed the capability of the assay system to differentiate between the more highly pathogenic species from congeners that were either low or apathogenic.
In the current configuration of the instrument, DNA was extracted in a separate step. An integrated DNA extraction step could be helpful.
In general, the biothreat assay run on the PLEX-ID technology had an identification rate of 100% for the highly pathogenic target bacteria at the genus level for the 45 reference samples. With exception of Burkholderia and in one case Brucella, all other target bacteria could also been discriminated at the species level. In addition to the correct and rapid identification of the target bacterial DNA we have seen several unspecific reactions. Most striking was the contamination with E. coli DNA. In parallel cultures of the primary samples, E. coli was not found except sample DNA-1 which was prepared from E. coli cultures and was used as the source for the DNA-control samples in the first set of experiments (Table 1, data not shown). Thus, an intrinsic contamination during the amplification or measurement process of the sample set must be excluded. Most unexpected results were observed in the negative control samples, where for one negative control a low genome number of F. tularensis was detected (Table 2: Iso-5) and is assumed to be the result of cross-contamination.
The ability to differentiate between B. pseudomallei and B. xerovorans would be most useful as related to potential medical treatment decisions to be made or for the purposes of microbial forensic analysis. Mainly for forensic questions, but also for threat estimation the differentiation of B. mallei and pseudomallei as well as of F. tularensis subspecies would be of interest. In addition, the species F. tularensis contains subspecies tularensis, holarctica, and mediasiatica, pathogenic to humans, and subspecies novicida, not pathogenic to humans, which are also of interest to be identified. Currently these are distinguished by the assay and details of speciation are available only in a secondary report rather than the primary sample report.
The PLEX-ID Genomes per well data reflects the same general trend of quantification as the reference method; however, the magnitudes were quite different. The PLEX-ID assays are only semi-quantitative and Genomes per well reflects a relative measure of the abundance as compared to the internal standard . The normal range for reporting these levels is between 0.1X and 10X the levels of internal controls in the assay, which in the case of the PLEX-ID biothreat assay represents a working range of ∼10 GE/well to 1000 GE/well. Above 1000 GE/well, the target organism levels far exceed the calibrant, thus, the reported level only serves to indicate high target concentrations. This would explain that quantitative differences in some samples were not detected.
The PLEX-ID technology is a high-throughput technology that allows rapid analysis of the DNA content of environmental and clinical samples. Here we evaluated 45 samples of genomic DNA in various matrices. The high specificity and high sensitivity of the PLEX-ID biothreat assay reported in the accompanying manuscript by Sampath et al.  was confirmed by our analysis of these bacterial samples. An advantage of this assay is that it is semi-generic; a wide range of bacterial agents and their near-neighbours are identified without need for assumptions about the pathogens present. If more than one pathogen is present in a sample, all are identified simultaneously. The results are scored for a probability, which allows the interpretation of uncertain results in the context of further information. The PLEX-ID instrument is useful for the application in central service facilities with a high-throughput of samples.
We thank the European Commission's Executive Agency for Health and Consumers (EAHC) for the advisory support and the EQADeBa consortium for its cooperation in this project. We also thank A. Barduhn, C. Schulze, I. Klein, and R. Krüger for the excellent technical assistance.
This publication has been produced with the support of the European Commission's Executive Agency for Health and Consumers (EAHC). Its content is the sole responsibility of Robert Koch-Institut, Centre for Biological Security, and can in no way be taken to reflect the views of the EAHC or any other body of the European Union. The supporting agency has no influence on the outcomes of the experiments. This manuscript is linked to the manuscript by Sampath et al. published in the same issue .
Conceived and designed the experiments: DJ US RG. Performed the experiments: DJ US CW KSL. Analyzed the data: DJ US CW KSL RH RS DJE RG. Contributed reagents/materials/analysis tools: RG DJE RS. Wrote the paper: RG RS DJE.
- 1. Pappas G, Panagopoulou P, Akritidis N (2009) Reclassifying bioterrorism risk: are we preparing for the proper pathogens? J Infect Public Health 2: 55–61.
- 2. Centers for Disease Control and Prevention website. Available: http://www.cdc.gov/Accessed 2012 Jun 7.
- 3. Branda JA, Ruoff K (2002) Bioterrorism. Clinical recognition and primary management. Am J Clin Pathol 117: 116–123.
- 4. Wagar EA, Mitchell MJ, Carroll KC, Beavis KG, Petti CA, et al. (2010) A review of sentinel laboratory performance: identification and notification of bioterrorism agents. Arch Pathol Lab Med 134: 1490–1503.
- 5. Horn JK (2003) Bacterial agents used for bioterrorism. Surg Infect 4: 281–287.
- 6. Hofstadler SA, Sampath R, Blyn LB, Eshoo MW, Hall TA, et al. (2005) TIGER: the universal biosensor. Inter J Mass Spectrom 242: 23–41.
- 7. Ecker DJ, Sampath R, Blyn LB, Eshoo MW, Ivy C, et al. (2005) Rapid identification and strain-typing of respiratory pathogens for epidemic surveillance. Proc Natl Acad Sci USA 102: 8012–8017.
- 8. Ecker DJ, Sampath R, Massire C, Blyn LB, Hall TA, et al. (2008) Ibis T5000: a universal biosensor approach for microbiology. Nat Rev Microbiol 6: 553–558.
- 9. Crowder CD, Matthews H, Schutzer S, Rounds MA, Luft BJ, et al. (2010) Genotypic variation and mixtures of lyme Borrelia in Ixodes ticks from North America and Europe. PLoS ONE 5: E10650 10651–10659.
- 10. Ecker DJ, Drader J, Gutierrez J, Gutierrez A, Hannis J, et al. (2006) The Ibis T5000 universal biosensor - an automated platform for pathogen identification and strain typing. JALA 11: 341–351.
- 11. Eshoo MW, Whitehouse CA, Zoll ST, Massire C, Pennella TT, et al. (2007) Direct broad-range detection of alphaviruses in mosquito extracts. Virology 368: 286–295.
- 12. Sampath R, Russell KL, Massire C, Eshoo MW, Harpin V, et al. (2007) Global surveillance of emerging Influenza virus genotypes by mass spectrometry. PLoS ONE 2: e489.
- 13. National Institute of Allergy and Infectious Diseases website. Available: http://www.niaid.nih.gov/Accessed 2012 Jun 7.
- 14. National Select Agent Registry website. Available: http://www.selectagents.gov/Accessed 2012 Jun 7.
- 15. AthoGen website. Available: http://www.athogen.com/Accessed 2012 Jun 7.
- 16. Janse I, Hamidjaja RA, Bok JM, van Rotterdam BJ (2010) Reliable detection of Bacillus anthracis, Francisella tularensis and Yersinia pestis by using multiplex qPCR including internal controls for nucleic acid extraction and amplification. BMC Microbiol 10: 314.
- 17. Lasch P, Beyer W, Nattermann H, Stämmler M, Siegbrecht E, et al. (2009) Identification of Bacillus anthracis by using matrix-assisted laser desorption ionization-time of flight mass spectrometry and artificial neural networks. Appl Environ Microbiol 75: 7229–7242.
- 18. Lasch P, Drevinek M, Nattermann H, Grunow R, Stämmler M, et al. (2010) Characterization of Yersinia using MALDI-TOF mass spectrometry and chemometrics. Analytical Chem 82: 8464–8475.
- 19. He J, Kraft AJ, Fan J, Van Dyke M, Wang L, et al. (2009) Simultaneous detection of CDC category “A” DNA and RNA bioterrorism agents by use of multiplex PCR & RT-PCR enzyme hybridization assays. Viruses 1: 441–459.
- 20. Deshpande A, Gans J, Graves SW, Green L, Taylor L, et al. (2010) A rapid multiplex assay for nucleic acid-based diagnostics. J Microbiol Methods 80: 155–163.
- 21. Sampath R, Mulholland N, Blyn L, Massire C, Whitehouse CA, et al. (2012) Comprehensive Biothreat Cluster Identification by PCR/Electrospray-Ionization Mass Spectrometry. PLoS ONE 7(6): e 36528.