Perspective The Role of in the Identification, Prediction, and Prevention of Biological Threats

W. Florian Fricke, David A. Rasko, Jacques Ravel* Institute for Sciences (IGS), University of Maryland School of Medicine, Baltimore, Maryland, United States of America

Since the publication in 1995 of the first Biodefense Funding for of programs, the genome sequences of over complete genome sequence of a free-living Genomic Research 90,000 influenza were rapidly organism, the bacterium Haemophilus influ- generated and are now deposited in enzae [1], more than 1,000 of Since the anthrax letter attacks of 2001, GenBank (http://www.ncbi.nlm.nih.gov/ species from all three domains of life— when letters containing anthrax spores genomes/FLU/aboutdatabase.html). Be- , , and Eukarya—have were mailed to several news media offices cause of the availability of large been completed and a staggering 4,300 and two Democratic senators in the capacity and the large amount of informa- are in progress (not including an even United States, killing five people and tion, the response to the 2009 H1N1 larger number of viral genome projects) infecting 17 others, funding agencies in influenza pandemic was rapid and efficient (GOLD, Genomes Online Database v. the US and other countries have priori- (Box 2): Genomics information was gener- 2.0; http://www.genomesonline.org/gold. tized research projects on organisms that ated within days and validated diagnostic cgi, as of August 2009). Whole-genome might potentially challenge our security tools were approved within weeks [5,6]. A shotgun sequencing remains the standard and economy should they be used as global response was made possible through in biomedical, biotechnological, environ- biological weapons. This has resulted in tremendous research efforts enabled by mental, agricultural, and evolution- large amounts of funding dedicated to so- genomic research. ary genomics (http://genomesonline.org/ called ‘‘biodefense’’ research, totaling close gold_statistics.htm#aname). While next- to $50 billion between 2001 and 2009 [4]. Access to and Documentation generation sequencing technology is Genomics has benefited greatly from this of Sequence Data changing the field, this approach will influx of research dollars and as a result, continue to be used and lead to a representatives of most major animal, plant, Open access to genomics resources (i.e., previously unimaginable number of ge- and human have been sequenced raw sequence data and associated publi- nome sequences, providing opportunities (http://www.pathogenportal.org/). Support- cations) is an essential component of the that could not have been thought of a few ed by federal funds from the National nation preparedness to biological threats years ago. These opportunities include Institutes of Health (NIH), the National (biopreparedness), whether intentionally studying genomes in real-time to under- Institute of Allergy and Infectious Diseases delivered or not. Although some consider stand the evolution of known pathogens (NIAID), and the US Department of De- open-source genomic resources a threat to and predict the emergence of new infec- fense, research programs, such as the Micro- security [7] because they make publicly tious agents (Box 1). With the introduction bial Sequencing Centers and the Bioinfor- available information that could facilitate of next-generation sequencing platforms, matics Resource Centers (http://www3. the construction of dangerous infectious cost has decreased dramatically, resulting niaid.nih.gov/topics/pathogenGenomics/ agents, we strongly disagree with this point in genomics no longer being an indepen- PDF/genomicsinitiatives.htm), have been of view. Rather, we and others [8] believe dent discipline, but becoming a tool established that carry out genomics re- that it is an enabling tool more useful to routinely used in laboratories around the search on pathogenic organisms and have those in charge of our public health and world to address scientific questions. This spearheaded a new phase of the genomics biosecurity than to those with ill inten- global sequencing effort has been focusing revolution. Similar programs were started tions. Genomic sequence data can provide primarily on pathogenic organisms, which in Europe, such as those at the Wellcome a starting point for the development of today are still the subject of the majority of Trust Sanger Institute in the United new vaccines, drugs, and diagnostic tests genome projects [2]. Sequencing two to Kingdom, and the multinational European [9], hence improving public health capa- five strains of the same has, in effort, The Network of Excellence Euro- bilities and increasing our bioprepared- recent years, afforded us not only a better PathoGenomics (http://www.noe-epg. ness. Access to the organisms from which understanding of evolution, virulence, and uni-wuerzburg.de/epg_general.htm). As the sequences are derived should be biology in general [3], but, taken to the an example of the success of these types restricted, not their genome sequences. next level (hundreds or thousands of strains) it will enable even more accurate Citation: Fricke WF, Rasko DA, Ravel J (2009) The Role of Genomics in the Identification, Prediction, and diagnostics to support epidemiological Prevention of Biological Threats. PLoS Biol 7(10): e1000217. doi:10.1371/journal.pbio.1000217 studies, food safety improvements, public Published October 26, 2009 health protection, and forensics investiga- Copyright: ß 2009 Fricke et al. This is an open-access article distributed under the terms of the Creative tions, among others. Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Competing Interests: The authors have declared that no competing interests exist. The Perspective section provides experts with a forum to comment on topical or controversial issues * E-mail: [email protected] of broad interest. This article is part of the ‘‘Genomics of Emerging Infectious Disease’’ PLoS Journal collection (http:// ploscollections.org/emerginginfectiousdisease/).

PLoS Biology | www.plosbiology.org 1 October 2009 | Volume 7 | Issue 10 | e1000217 Author Summary vocabulary should be developed to de- scribe these isolates and the genes they contain. Such efforts have already started, In all likelihood, it is only a matter of time before our public health system will for example through the open-access face a major biological threat, whether intentionally dispersed or originating from journal Standards in Genome Sciences a known or newly emerging infectious disease. It is necessary not only to increase (SIGS) (http://standardsingenomics.org/ our reactive ‘‘biodefense,’’ but also to be proactive and increase our preparedness. To achieve this goal, it is essential that the scientific and public index.php/sigen), but the dedicated re- health communities fully embrace the genomic revolution, and that novel sources are not adequate and highlight the bioinformatic and computing tools necessary to make great strides in our lack of understanding of the importance of understanding of these novel and emerging threats be developed. Genomics has metadata in genomics. Initiatives such as graduated from a specialized field of science to a research tool that soon will be those of the Genomics Standards Consor- routine in research laboratories and clinical settings. Because the technology is tium have made great strides [11,12], but becoming more affordable, genomics can and should be used proactively to still need widespread implementation build our preparedness and responsiveness to biological threats. All pieces, from the ever-expanding genomic com- including major continued funding, advances in next-generation sequencing munity. Open access to the genomic DNA technologies, infrastructures, and open access to data and that has been sequenced or the culture metadata, are being set in place for genomics to play a central role in our public from which the DNA was extracted and to health system. the associated metadata is key to success- ful genome sequencing projects, whether on single or several hundred genomes or Now that genomics technologies are species. Access to relevant, well-curated metagenomes. Well-documented genome broadly available, there is the potential culture collections [10] and DNA prepa- sequence data will form a key growing for commercial interests to hamper the rations suitable for sequencing may be- resource for biodefense and other re- release of genomic data in the public come a bottleneck in the future when search fields. domain. Thus it is important that federally sequencing resources are no longer limit- funded large-scale genome sequencing ing. More importantly, the impact of large Emerging New Bioinformatics efforts have enforceable rapid release genomic sequence datasets from clinical Resources policies. This accessibility could afford isolates will be limited without key clinical further opportunities to capitalize on metadata that characterize these isolates, As we enter a new era of modern investments in genome sequencing by such as patients’ medical information, genomics, the ever-expanding sequence providing the necessary resources to bio- date of isolation, and the number of datasets are becoming more challenging to preparedness. culture passages in the laboratory. Open analyze. Future analysts will require powerful Whereas genome projects aimed at access to large numbers of sequences and new bioinformatics tools in conjunction with sequencing one, two, or three isolates of associated metadata allows for powerful new computer systems engineered with a pathogen seemed adequate a few years comparative genomic analyses and thus genomic analysis in mind. Open-source ago, it is now possible to sequence rapidly provides major insights into the charac- new bioinformatics software tools are being hundreds of individual genomes for each teristics of a pathogen. Standardized developed that exploit Web-based services and the increasing computing power provid- ed by academic and commercial ‘‘cloud Box 1. Hot Spots for the Emergence of Infectious Disease computing networks’’ (large computing re- Can we define ‘‘hot spots’’ of microbial populations where new infectious sources provided as a service over the diseases are more likely to evolve? Human contact with new types of infectious Internet). For example, ‘‘Science Clouds’’ agents precedes the emergence of infectious diseases. Infectious agents can be (http://workspace.globus.org/clouds/) allow new in the sense of not having previously infected humans or new in the sense members of the scientific community to lease that a combination of preexisting genetic factors (for example, mobile elements cloud computing resources free of charge. or regulatory elements) have reassembled to give rise to an infectious agent with To leverage these capabilities, novel cloud- a substantially altered genome. The Ebola , which first emerged by infecting optimized bioinformatics tools are being humans 1976 in Zaire [21], is an example of the former, whereas the acquisition of developed, such as the genome sequence antimicrobial resistance by Acinetobacter baumannii [22] is an example of the read mapper CloudBurst [13]. In addition, latter. In both cases, a change in the selective pressure on an infectious agent novel resources are currently under devel- allows its emergence from a specific setting. This selective pressure may be, for opment to increase the availability of open- example, the new niche that the human host provides to the pathogen or the source bioinformatics tools for cloud com- antimicrobial selection on a pathogen. Since both events rely on preexisting puting (http://www.nsf.gov/awardsearch/ genetic resources and not on the de novo evolution of virulence factors, the showAward.do?AwardNumber=0949201; potential of a setting to serve as a hot spot or reservoir for an emerging infectious http://www.nsf.gov/awardsearch/showAward. disease is theoretically predictable from the examination of the total metagen- do?AwardNumber=0844494). These emerging ome. In this scenario, traditional microbiological approaches that focus on single tools make access to the Worldwide Web the isolates of bacteria or viruses are limited in their predictive power since they lack a view of the complete genetic landscape. The potential infectious disease agent only requirement to join the genomic revolution could, however, arise from an environment that only contains pieces of a and achieve large scale bioinformatics analyses ‘‘virulence puzzle,’’ i.e., individual virulence factors encoded within the genomes that could not be possible on local servers. As a of different organisms (the metagenomic ‘‘gene soup’’). These pieces would have consequence, it is conceivable that in the future to be assembled in one species for the new pathogen to emerge as an infectious genomic research will increasingly move away agent. from the large sequencing centers toward a more decentralized organization. Decentralized

PLoS Biology | www.plosbiology.org 2 October 2009 | Volume 7 | Issue 10 | e1000217 Box 2. Pandemic H1N1 2009 Influenza: A Recent Example of the the first use of microbial genomics as an Impact of Genomics on Biopreparedness essential tool in a forensic investigation. In the course of the investigation, scien- Genomics can be readily applied to follow outbreaks of infectious diseases. This is tists had to establish culture repositories clearly illustrated during the severe acute respiratory syndrome (SARS) outbreak fromstrainsusedinresearchintheUS in 2002–2003 and the emergence and worldwide spread of the pandemic H1N1 and build databases of genome sequences 2009 influenza virus this year. In both cases, genomics played a key role in the of all B. anthracis isolates. This work took immediate response to the outbreak. Initially, very little was known about the several years and delayed the investiga- virus responsible for the SARS outbreak. Pangenomic virus microarrays identified tion significantly. A lesson to be learned it as a coronavirus [23]; however, it was only through detailed sequencing that the from this investigation should therefore specific genotype of this virus could be determined [24]. Comparative sequence be that there is a need for comprehensive analysis identified the SARS virus as distinct from other coronaviruses in terms of databases of unique DNA fingerprints of its encoded proteins responsible for antigen presentation. This finding ultimately stocks of potentially threatening patho- lead to development of diagnostics [25] and potential therapeutics [26]. This example of a sequencing approach as a rapid response to a virus outbreak gens. In the event that another bioterror demonstrates that genomics can be a useful and important, if not essential, attack were to take place such genomic epidemiological tool. In the ongoing H1N1 influenza outbreak, the National databases would be key in quickly Center for Biotechnology Information (NCBI) established the Influenza Virus establishing the source of the biological Resource (a database and tool for flu sequence analysis, annotation, and material. submission to GenBank; http://www.ncbi.nlm.nih.gov/genomes/FLU/SwineFlu. The concept of population genomics also html), containing 462 complete viral genome sequences from worldwide viral applies to epidemiological studies of out- samples (as of September, 2009). Some of the genomic data was completed, breaks of infectious diseases such as those compared, and released to the public within two weeks of isolation of the DNA. caused by food-borne or zoonotic patho- The rapid generation of genome sequence data is providing a paradigm shift in gens, such as Salmonella spp. Traditionally, the analysis of infectious disease outbreaks, from more classical methods of epidemiologists and pathologists have used isolation to the rapid molecular examination of the pathogen in question. low-resolution methods such as pulsed-field gel electrophoresis (PFGE), multi-locus sequence typing (MLST), or multi-locus rapid genome sequencing and bioinformatic genomicists joined forces to characterize variable number tandem repeats analysis analysis of infectious agents will enable near-real- the unique genetic traits of the Bacillus (MLVA) to trace an individual isolate from time global surveillance, detection of new anthracis spores recovered from the enve- a patient back to a potentially infected food pathogens, new virulence factors, antimi- lopes, which were quickly identified as source or to isolates from other patients crobial resistance determinants, or engineered the B. anthracis Ames strain (DAAR et al., [14–17]. In 2006, for example, during an organisms. unpublished data). Sequencing the ge- outbreak of pathogenic nome of several single colonies obtained O157:H7 infections in 26 states of the Population Genomics Applied from the spores revealed that the entire US, which was caused by contaminated to Single Cultures chromosome and its associated spinach, isolates of the pathogen were were 100% identical to the genome recovered from cows and wild pigs (the Because the resources for affordable sequence of the ancestral B. anthracis zoonotic reservoirs), bags of spinach (the high-throughput sequencing, data pro- Ames strain that was stored for over 20 vehicle of transmission), and ill patients cessing, and analysis are available, the years in a military laboratory in Freder- (http://www.cdc.gov/mmwr/preview/ time is right to think about microbial ick, Maryland. The only genotypic dif- mmwrhtml/mm55d926a1.htm). One population genomics and large-scale mi- ferences were found in a small, pheno- of these isolates was designated as the crobial in the context of typically and genetically distinct portion reference for the outbreak based on biodefense research (Box 3). Traditional- of cells grown from the spores used in the conserved PFGE patterns. Genome ly, the concept of population genomics attacks. Genomic characterization of sequencing of several isolates from the has applied to variation within a species. these phenotypic variants revealed a same outbreak performed in our labo- However, a bacterial culture, even if number of unique genetic alterations that ratory, however, revealed genomic derived from a single clone, is composed together provided a characteristic DNA variations that questioned a direct of millions of cells that are not necessarily fingerprint of the spore population that evolutionary link between all out- identical at the genome sequence level, could be unequivocally matched to the break-associated isolates (Eppinger hence forming a population of genomes. spore sample used in the attacks. Using et al., unpublished data). Comparative Therefore we propose to apply the this fingerprint, a genetic assay was genomics followed by whole-genome concept of population genomics to mi- developed to screen a B. anthracis spore phylogenetic analyses based on single crobial cultures. The assemblage of repository, which identified the origin of nucleotide polymorphisms demonstrat- genotypes defines what is called a ‘‘cul- the spores as a single spore stock of B. ed that these isolates were indeed ture,’’ ‘‘culture stock,’’ or ‘‘reference anthracis Ames. This stock was stored at closely related to one another and only strain.’’ Population genomics addresses the US Army Medical Research Institute distantly related to other E. coli the genomic diversity within these assem- for Infectious Diseases in Fort Detrick, O157:H7 isolates, hence linking all blages and has significant implications for Maryland, narrowing the pool of suspects isolates to the same outbreak, some- many fields of research but, most impor- to a manageable number (those who had thing that was not possible using PFGE tantly, for pathogen evolution, diagnos- access to the spore stock) for the investi- patterns. In this case, phylogenetic tics, epidemiology, and microbial foren- gative team. The police investigation that analyses suggest that several highly sics. For example, following the anthrax followed identified a potential suspect as related genotypes were at the source mail attacks of 2001, microbiologists and the custodian of the spore stock. This was of the outbreak, thus challenging the

PLoS Biology | www.plosbiology.org 3 October 2009 | Volume 7 | Issue 10 | e1000217 Box 3. Simple Genomics, Population Genomics, and Challenges for the Future Metagenomics The field of biodefense has thoroughly It is now technically possible and scientifically desirable to combine sequencing embraced genomics and made it a projects on single genomes, genome populations, and metagenomes to study keystone for developing better identifica- genome evolution. Single-genome projects provide the greatest resolution for tion technologies, diagnostic tools, and identifying genetic factors responsible for specific virulence phenotypes and vaccines and improving our understand- provide answers to many important questions, such as: What is the minimal gene ing of pathogen virulence and evolution. set in a pathogen required to cause a specific disease phenotype? What does the Enabling technologies and bioinfor- genetic context of virulence or antibiotic resistance factors tell us about their matics tools have shifted genomics from evolutionary origin or the mobility between different microbial species or even a separate research discipline to a tool so genera? Population-level genome sequencing projects provide us with informa- powerful that it can provide novel tion about the pangenomic gene pool and the potential of a species to evolve insights that were not imaginable a few into a novel pathogen. Are certain bacterial species or strains more likely than years ago, including for example redefin- others to evolve pathogenic traits? What distinguishes a commensal from a ing the notion of strains or cultures in the pathogenic isolate? What provides the trigger or ability to convert a commensal context of biopreparedness or microbial or opportunistic strain into a pathogen? What role does horizontal gene transfer play in species evolution? Is an infection always caused by an individual isolate or forensics. Challenges remain, though, might infection be caused by a combination of individuals in a population that all mostly in the form of large amounts of have different attenuated infectious potentials? Metagenomics projects sample data that are being generated, and will the genetic reservoir (the set of genes carried by all members of a community) continue to be generated in the future, within a specific environment or sample. This ‘‘gene soup’’ reflects the maximum and are becoming difficult to manage. genetic potential accessible to individual isolates by horizontal gene transfer. The need for better bioinformatic algo- rithms, access to faster computing capa- bilities, larger or novel and more efficient utility of assigning a single reference ciated microbial communities (e.g., Vib- data storage devices, and better training strain to a specific outbreak. Instead, rio cholerae, the etiologic agent of cholera) in genomics are all in critical demand, collecting and sequencing tens or but potentially also by slight shifts in the andwillberequiredtofullyembracethe hundreds of isolates from each source proportions of different populations wi- genomic revolution. Our nation’s pre- or patient linked to an outbreak would thin the community that give an other- paredness for biological threats, whether provide a better basis for understand- wise harmless species or strain an un- they are deliberate or not, and our public ing the genomic diversity within the desirable advantage over others, a sim- health system would benefit greatly by outbreak population and would aid in ilar situation to what is observed in leveraging these capabilities into better defining the population dynamics of an bacterial vaginosis [20]. Probiotic die- real-time diagnostics (in the environment outbreak. tary supplements of live microorganisms as well as at the bedside), vaccines, a deliver beneficial bacteria that promote greater understanding of the evolution- A New Concept: Contrabiotics an healthy state of the targeted micro- ary process that makes a friendly microbe become a pathogen (Box 3) (hence to Insufficient attention has been paid to biota. In a completely hypothetical the (i.e., the consor- possibility, the opposite would also be better predict what microbial foes will be tium of microbes that inhabit the human plausible, where the healthy microbiota facing us in the near future), and better body) as it relates to our efforts to (skin, gut, or upper respiratory tract, forensics and epidemiological tools. The increase biopreparedness. New analyses among others) may be disturbed by time is right to be bold and capitalize on of the diversity and composition of the introducing large amounts of ‘‘contra- these enabling technological advances to human microbiome are making it in- biotics,’’ i.e., living nonpathogenic bac- sequence microbial species or complex creasingly clear that human health teria that would shift the microbiota microbial communities to the greatest depends on a delicate equilibrium be- away from a healthy state. A better level possible—that is, hundreds of ge- tween the microbial inhabitants and the understanding of the ecological princi- nomes per species or samples—but let us human host [18,19]. Severe effects on ples that shape the composition of our notforgetthatinformaticsandcomput- health could be caused not only by the microbiome might contribute to our ing resources are now becoming the introduction of true pathogens in the biopreparedness for such a threat to bottleneck to actually making major traditional sense into these human-asso- public health. progress in this field.

References 1. Fleischmann RD, Adams MD, White O, 4. Franco C (2008) Billions for biodefense: Federal 7. Aldhous P (2001) Biologists urged to address risk Clayton RA, Kirkness EF, et al. (1995) Whole- agency biodefense funding, FY2008-FY2009. of data aiding bioweapon design. Nature 414: genome random sequencing and assembly of Biosecur Bioterror 6: 131–146. 237–238. Rd. Science 269: 5. Rowe T, Abernathy RA, Hu-Primmer J, 8. Read TD, Parkhill J (2002) Restricting genome 496–512. Thompson WW, Lu X, et al. (1999) Detection data won’t stop . Nature 417: 379. 2. Guzman E, Romeu A, Garcia-Vallve S (2008) of antibody to avian influenza A (H5N1) 9. Bambini S, Rappuoli R (2009) The use of Completely sequenced genomes of pathogenic virus in human serum by using a combina- genomics in microbial vaccine development. bacteria: A review. Enferm Infecc Microbiol Clin tion of serologic assays. J Clin Microbiol 37: Drug Discov Today 14: 252–260. 26: 88–98. 937–943. 10. Tindall BJ, Garrity GM (2008) Proposals to clarify 3. Binnewies TT, Motro Y, Hallin PF, Lund O, 6. Maurer-Stroh S, Ma J, Lee RT, Sirota FL, how type strains are deposited and made available to Dunn D, et al. (2006) Ten years of bacterial Eisenhaber F (2009) Mapping the sequence the scientific community for the purpose of systematic genome sequencing: Comparative-genomics- of the 2009 H1N1 influenza A virus research. Int J Syst Evol Microbiol 58: 1987–1990. based discoveries. Funct Integr Genomics 6: neuraminidase relative to drug and antibody 11. Garrity GM, Field D, Kyrpides N, Hirschman L, 165–185. binding sites. Biol Direct 4: 18. Sansone SA, et al. (2008) Toward a standards-

PLoS Biology | www.plosbiology.org 4 October 2009 | Volume 7 | Issue 10 | e1000217 compliant genomic and metagenomic publication 17. Boxrud D, Pederson-Gulrud K, Wotton J, natural history of Ebola virus in Africa. Microbes record. OMICS 12: 157–160. Medus C, Lyszkowicz E, et al. (2007) Compar- Infect 7: 1005–1014. 12. Field D, Garrity GM, Sansone SA, Sterk P, ison of multiple-locus variable-number tandem 22. Peleg AY, Seifert H, Paterson DL (2008) Gray T, et al. (2008) Meeting report: The fifth repeat analysis, pulsed-field gel electrophoresis, Acinetobacter baumannii: Emergence of a successful Genomic Standards Consortium (GSC) work- and phage typing for subtype analysis of Salmonella pathogen. Clin Microbiol Rev 21: 538–582. shop. OMICS 12: 109–113. enterica serotype Enteritidis. J Clin Microbiol 45: 23. Wang D, Urisman A, Liu YT, Springer M, 13. Schatz MC (2009) CloudBurst: Highly sensitive 536–543. Ksiazek TG, et al. (2003) Viral discovery and read mapping with MapReduce. Bioinformatics 18. Gao Z, Tseng CH, Strober BE, Pei Z, Blaser MJ sequence recovery using DNA microarrays. PLoS 25: 1363–1369. (2008) Substantial alterations of the cutaneous Biol 1: e2. doi:10.1371/journal.pbio.0000002. 14. Gerner-Smidt P, Hise K, Kincaid J, Hunter S, bacterial biota in psoriatic lesions. PLoS One 3: 24. Marra MA, Jones SJ, Astell CR, Holt RA, Rolando S, et al. (2006) PulseNet USA: A five- e2719. Brooks-Wilson A, et al. (2003) The genome year update. Foodborne Pathog Dis 3: 9–19. 19. Turnbaugh PJ, Ley RE, Mahowald MA, sequence of the SARS-associated coronavirus. 15. Urwin R, Maiden MC (2003) Multi-locus se- Magrini V, Mardis ER, et al. (2006) An obesity- Science 300: 1399–1404. quence typing: A tool for global epidemiology. associated gut microbiome with increased capac- 25. Zhu M (2004) SARS immunity and vaccination. Trends Microbiol 11: 479–487. ity for energy harvest. Nature 444: 1027–1031. Cell Mol Immunol 1: 193–198. 16. Keim P, Price LB, Klevytska AM, Smith KL, 20. Srinivasan S, Fredricks DN (2008) The human 26. Haagmans BL, Osterhaus AD (2006) Coronavi- Schupp JM, et al. (2000) Multiple-locus variable- vaginal bacterial biota and bacterial vaginosis. ruses and their therapy. Antiviral Res 71: number tandem repeat analysis reveals genetic Interdiscip Perspect Infect Dis 2008: 750479. relationships within . J Bacteriol 21. Pourrut X, Kumulungui B, Wittmann T, 397–403. 182: 2928–2936. Moussavou G, Delicat A, et al. (2005) The

PLoS Biology | www.plosbiology.org 5 October 2009 | Volume 7 | Issue 10 | e1000217