1.1.1 The need for rapid detection and genetic characterization of pathogens as demonstrated by the 2001 anthrax attacks

In the fall of 2001, at least four envelopes containing Bacillus anthracis spores were mailed through the United States postal system. They were addressed to the New York Post, Tom Brokaw at NBC, and the Washington, DC offices of senators Daschle and Leahy. There were 22 suspected cases identified, divided evenly between inhalational and cutaneous cases. Of those, there were five fatalities (summarized in Rasko et al.¹). The aftermath included a climate of fear and an unprecedented forensic challenge in attribution of the crime to the perpetrator; i.e. finding the source of the spores.

This unfortunate act of domestic bioterrorism taught us several things. First, it demonstrated the importance of rapid identification of an etiological agent as well as rapid identification of those who have been exposed to the agent. Case histories of the 11 inhalational anthrax patients clearly demonstrate that early administration of the appropriate antibiotic results in improved chances of survival² (see Plate I in colour section between pages 256 and 257).

Second, in the forensic investigation that ensued, it became quite clear that existing genotyping methods, such as multi-locus variable-number tandem repeat analysis (MLVA) or single nucleotide polymorphism (SNP) typing systems, were insufficient to identify differences between isolates, as were initial efforts using Sanger sequencing. Therefore, whole genome sequencing (WGS) and analysis of various colony morphotypes were conducted. Additional findings that resulted from the forensic investigation are: evidentiary samples in a case such as this one need not be sequenced to ‘completion’ or ‘finished’, or closed to a single contig; following microbiological isolation of morphological variants, sequencing each morphotype to 9–12 × average depth of coverage was sufficient¹; although several genome sequences were closed in the course of the investigation, no novel high-quality variations were identified in the closed genome sequences that were not present in the draft sequences; and, finally, it was noted that high-quality reference sequences are essential to render evidentiary draft genomes useful.¹

In the case of the anthrax letter attacks, WGS was used to characterize the pathogen for the purpose of attribution, rather than as part of a rapid response scenario. However, given the short window in which the appropriate antimicrobial therapy or other countermeasures must be administered to save the victims, what would have been the outcome had the agent involved been a genetically engineered form of B. anthracis (e.g. ciprofloxacin resistant), and had the engineered phenotype not been discovered in the course of routine microbiological testing? In a mock rapid response exercise, Chen et al. demonstrated recently that WGS could have been employed upfront and operationally relevant information could have been obtained in time to improve survival rates.³

These lessons learned from the anthrax attacks are significant, but there are some caveats. B. anthracis is a highly monomorphic and new species, and spore formation supports its genomic stability. For instance, genome sequences of several Ames isolates were compared in the anthrax investigation and it was noted, that despite 21 years of laboratory growth, no distinguishing mutations were found, either in the chromosome or on the two plasmids. The mutations present in the Porton Down strain were attributed to the plasmid curing process. It is not clear whether a similar picture would emerge if the organism in question had not been such a monomorphic species – if, for instance, it had been an organism with as plastic a genome as Shigella sp. or Escherichia coli.

A further issue relates to scalability of the WGS sequencing approach for forensic genetic investigations. The combination of microbiological studies and individual WGS was costly and time-consuming. Is this approach feasible in a rapid response scenario when saving lives depends on the outcome of these genetic studies? The idea of direct metagenome sequencing of nucleic acid materials extracted from clinical specimens (ClinSeq) has gained momentum, and it could be a potential time-saver in a rapid response scenario. However, there are certain limitations of metagenomic sequencing, including the effects of matrix, depth and breadth of sequence coverage required and the bioinformatic challenges associated with sifting through mounds of data to identify the causal agent and causal genetic variations, if any (some of these aspects are treated in later sections). Finally, there is also the challenge of linking the potential agent to the disease, i.e. fulfilling Koch’s postulates, which may not be feasible within the time frame of real events.

1.1.2 Applications of WGS in a public health event caused by E. coli

Whereas, in the anthrax investigation, genomic characterization of the agent was conducted after rather than during the incident, in another more recent example WGS played a critical role in characterizing the agent of an outbreak in real time. This outbreak, involving a more virulent strain of E. coli than usual, occurred in May–June 2011 in Germany. The source of the bacterium was traced to fenugreek sprouts, but there was also significant secondary transmission (human to human and human to food). There were over 3800 total cases, including a higher than usual proportion of adults and an unusually high number of hemolytic uremic syndrome (HUS) cases (reviewed in Beutin et al.⁴).

This outbreak was notable in that it became the debut of so-called ‘open-source’ genomic analysis and served as a paradigm for future outbreaks/events. It was characterized by rapid, crowd-driven, round-the-clock analysis of Ion Torrent draft sequence by bioinformaticians worldwide and aggregation of the resulting data in a wiki. Notably, this rapid sequencing and analysis resulted in design of diagnostic primers 5 days after release of the draft genome.⁵

Independently, another team of researchers used third-generation PacBio sequence data to elucidate the reason for increased virulence. By running three sequencers in parallel for 5 h per isolate, these researchers rapidly achieved 75× average coverage per genome. All the bioinformatic analyses of the sequence data from various sources revealed some clues to the unusual virulence of the EAEC bacteria that is not normally associated with HUS syndrome and a possible model for the evolution of this pathogen itself.

In short, there were at least three genetic changes that occurred during the evolution of this pathogen: (1) acquisition by an enteroaggregative E. coli (EAggEC) strain of Stx2 genes (Shiga toxin genes found on a bacteriophage in enterohemorrhagic E. coli (EHEC)), (2) acquisition of a plasmid encoding a Type III aggregative adhesion fimbrial gene cluster (AAF/III) that has been postulated to enhance the virulence by aggregating the bacteria on the intestinal epithelium, and (3) acquisition of a plasmid that confers multiple antibiotic resistance. However, the key virulence factor is the ability of these bacteria to produce shiga toxin, and, notably, toxin production is enhanced in the presence of certain classes of antibiotics normally prescribed to combat bacterial diseases. Rasko et al. provided experimental evidence that exposure to ciprofloxacin did indeed result in increased expression of the shiga toxin gene by the German isolate.⁶ Thus, the German E. coli outbreak highlighted the power of the WGS approach to decipher relevant genetic characteristics of a pathogen from an outbreak scenario, and provided valuable diagnostic assays and possible treatment options.

A further instance in which time is limited and WGS data may be useful is when ruling out bioterrorism, such as in the recent case of a rapidly progressive, fatal, inhalational anthrax-like infection of a welder in Texas. In this case, the patient sought medical care just 2 h after onset of his illness. Within 10 h of his arrival, he exhibited signs of multi-organ system failure and he was started on antibiotic therapy. On day 3, B. cereus was identified from his cultures and ciprofloxacin was added to his regimen; nevertheless, he died later that day. PCR ruled out the possibility that the strain was a so-called ‘conventional’ strain of B. cereus. Given those PCR data and the rapid course of infection in the patient, his healthcare providers wondered whether the organism had acquired genes conferring increased virulence, and, if it had, whether it was a naturally emergent strain or genetically engineered. Therefore, to rule out the possibility that the pathogen had been manmade and to practice their institution’s emergency response preparedness, they undertook WGS using the Illumina GXII platform. Bioinformatic analysis indicated that the isolate was likely a natural strain of B. cereus with a pXO1-like plasmid and B. anthracis-like virulence factors,⁷ similar to the previous cases in welders involving a similar strain called G9241.^8,9 Although in this case WGS was very useful for ruling out terrorism, it begs the question as to whether immediate metagenomic sequencing of the patient’s tracheal aspirate and/or bronchoalveolar lavage would have identified B. cereus faster and resulted in earlier administration of appropriate antibiotics.

Early administration of appropriate antibiotic is also important in the context of routine lower respiratory tract infections. In fact, the Centers for Medicare and Medicaid Services (CMS) require that antibiotics be administered within 6 h of the onset of symptoms. However, despite the fact that most cases are treatable if the etiologic agent is known, an etiologic identification is made in fewer than 10% of cases.¹⁰ Could a rapid metagenomic sequencing approach potentially increase the number of cases in which a causative agent can be identified in lower respiratory tract infections? Can MGS provide the same valuable information for other types of infections, and possibly decrease the time taken to identify sources of neonatal illness, such as the recent fatal case of Cronobacter sakazakii infection in a ten-day-old infant?¹¹

1.1.3 Forensics and attribution

Genomics vaulted to prominence as a forensic tool during the investigation of the 2001 anthrax attacks on the United States postal system, and rapidly became established as a major tool in the emerging field of microbial forensics.¹² The ultimate aim of using genomics as a forensic tool is to generate discriminatory (i.e. exclusive or inclusive) signatures that help narrow the range of potential suspects by establishing connections to the attack material and, equally important, eliminating those with no connection to the attack material. Genomic forensics has significant elements in common with the emerging discipline of genomic epidemiology and with genome-level studies of in vitro strain evolution, both of which rely on the ability to generate large numbers of high-quality signatures to enable the tracing of strain lineages.^13–17 Because it must be defensible in court, information generated by genomic forensics will also be required to meet the standards of admissibility of scientific evidence known as the Daubert standard. Briefly summarized, these rules stipulate that scientific techniques be testable, be subject to peer review and publication, have known error rates, established standards, and be generally accepted by the scientific community. The increasing prevalence of WGS in the scientific community to address epidemiological questions and the proliferation of peer-reviewed papers in this field are promising, but to date no general laboratory standards or accreditation programs have been established for forensic genomics laboratories.

The forensic genomic investigation of the attack materials followed the discovery of colony morphology variants in the spore preparations derived from the letters; each variant was completely sequenced and the genetic variants characterized.¹⁸ The unique mutations present in each of five variants formed the basis for discriminatory PCR assays that were utilized to establish the origin of the spore preparations and exclude other potential sources.

Modern Methods and Approaches

More recently, our consortium conducted a retrospective genomic analysis of the Bacillus atrophaeus var. globigii (BG, a non-pathogenic surrogate for anthrax, vide infra) lineage that shared many of the characteristics of a forensic investigation,²⁰ including differentiation based on non-genomic traits (e.g. colony morphology), source tracing, strain ‘matching’ and signature identification. In that study we retraced the ‘military’ lineage of BG using a combination of 454 and conventional sequencing and finishing, and, based on circumstantial open-source publications from the period,²¹ laboratory-verified phenotypes, and the propagation of genomic signatures over decades, were able to establish that the BW workers at Camp Detrick had deliberately selected a hypersporulating strain for their large-scale growths. Thus, our study not only established where the strains in use today had originated, but was able to assign the ‘intent’ behind the use of a given strain.

In the analysis of evidence, data with different levels of confidence can support different stages of the investigation, from lead generation through prosecution. Because the Amerithrax case was closed due to the death of the primary suspect, the data standards of the genomic investigation were not tested in court. Nevertheless, some basic principles can be stated that were derived from the Amerithrax work and from our own subsequent study.

1. Complement to other forensic techniques – Genomic analysis would not stand by itself in a microbial forensics investigation, but would be complemented by other techniques.^22,23

2. Use case for genomic analysis – Like ballistics or fingerprint analysis, genomic analysis must provide information that can be used to match the evidentiary materials to a data base of reference materials.

3. Confirmation of motive – Incertain cases (e.g. the discovery of hypersporulating BG), the discovery of mutations can provide evidence that reinforces the attribution not only of source, but of the motive behind the selection of a particular variant.

4. Requirements for finishing – Fully finished sequences for all materials may not be required for the ‘lead generation’ stage. Draft sequence (or even raw data), provided that both reference and tester data sets exhibit high consensus, quality and coverage, can generate the discriminatory signatures that support assay development for inclusion and exclusion of evidentiary materials (Fig. 1.1). Given the speed at which WGS can now be conducted, it is expected that discriminatory signatures will be available much earlier in the course of an investigation, with subsequent finishing efforts intended to ‘clean up’ the data for use in court.

1.1.4 Combined microbiology and next-generation sequencing (NGS) applications in bioforensics

NGS analysis was critical for the discovery and exploitation of the molecular targets used in the Amerithrax forensics application; however, just as importantwas the upfront microbiology responsible for elucidation of the independent morphotypes.¹ Current trends toward greater reliance on sequencing and bioinformatic data are reaching fruition and providing insights that once were the purview of classic microbiology and biochemistry.²⁴ As discussed later in this chapter, complete linkage of genotype to phenotype is a necessary aspect toward making this a full-blown reality. In the interim, and given the Amerithrax example, classic characterization methods remain relevant and can help frame the biological questions that sequencing-based technologies are uniquely situated to solve.

1.1.5 Deep sequencing to look at genomic variations in microbial populations

Genomic variations, both adaptive and neutral, underlie the evolution of all organisms. In microbes, several factors may contribute to rapid appearance of variants in populations: high population densities, rapid replication rates and high error rates of replication enzymes. However, these variants are present at relatively low frequencies. For instance, the observed frequency of any particular mutant in Escherichia coli is usually <10^–5. In cases where the new phenotype can be selected for, even very infrequent spontaneous mutants can be detected. Often, however, it is not possible to select for a particular phenotype, and in suchcases bacteria must be screened for it. Microbial mutation frequencies vary by loci and may be higher than average in some hypervariable regions, in certain mutant backgrounds and in viral populations. To the best of our knowledge, direct population-based genotyping by WGS has not been performed for bacterial strains. However, due to their smaller genome size as compared with bacteria and eukaryotes, studies have been performed to detect variants in viral populations.

We and others have used NGS technologies, such as 454-based pyrosequencing, for genetic fingerprinting of ‘purified’ variants isolated from a population.^3,20,25,26 From these experiments, we have gained some understanding of the power of NGS technologies to detect variations from draft sequences (see later section), frequencies of true positive/false positive variations and the limitations of the WGS approach for variation detection. It is rather easy to generate a unique genetic fingerprint of a ‘purified’ variant; however, this work has not been performed at the population level (population fingerprinting) to detect variants present as a minor component of a mixed bacterial population.

1.1.6 Policy drivers for NGS

In light of the 2001 anthrax attacks, the United States government has worked to develop a robust whole-of-government approach to detecting and responding to biological outbreak events. The Departments of Defense, Homeland Security, and Health and Human Services have been tasked by a series of Presidential Directives to lead development and coordination of the government’s response capabilities. Neither policy documents from the Executive Office of the President (EOP) norany departments specifically call out nucleic acid sequencing as a requisite technology to enable the response capabilities; however, one can ascertain that this technology must play a larger and larger role in fulfilling the President’s directives to bolster the country’s response capabilities.

One of the more important directives to emerge from the EOP is Homeland Security Presidential Directive 21 (HSPD-21).³³ This document calls for the US to plan and enable rapid response to a biological event, including the capacity to rapidly identify and characterize the threat. The document calls for the US to ‘strengthen laboratory diagnostic capabilities and capacity in order to recognize potential threats as early as possible’. In the case of engineered biological weapons, rationally designed detection methods (polymerase chain reaction-, immunological-, or array-based) will likely not be successful in fully identifying or characterizing a threat, pointing to a critical role for unbiased high-throughput sequencing. Furthermore, the mandate to ‘provide early warning and ongoing characterization of disease outbreaks in near real-time’ indicates that direct identification out of complex samples is far preferable to first culturing or propagating a threat prior to employing an identification technology. In support, former US Secretary of the Navy Richard Danzig has commented that future biological weapon attack samples are not likely to be as pristine as the 2001 anthrax samples³⁴ Furthermore, rapid identification and attribution from a first wave of complex samples will be critical to preventing ‘reload’ or a second wave of attacks by perpetrators.

Based on the foundation laid by HSPD-21, the National Strategy for Countering Biothreats was released in 2009 by the National Security Council.³⁵ Importantly, this document expanded upon HSPD-21 by calling for the promotion of global health security through building a ‘global capacity for disease surveillance, detection, diagnosis, and reporting’ for both natural disease outbreaks and biological weapon attacks. Specifically, the document states that ‘rapid detection and containment of, and response to, serious infectious disease outbreaks— whether of natural, accidental, or deliberate origin—advances both the health of populations and the security interests of States’. The following year, the 2010 National Security Strategy stated that timely and accurate insights on current and emerging biothreat risks were a key enabler to increasing national security³⁶ Overall, the linking of public health and biodefense mission spaces has been a critical driver to whole-of-government approaches to rapid identification and characterization capabilities. In support, the Chairman of the Joint Chiefs of Staff called for the integration of biodefense capabilities. including that ‘the Department of Defense and Department of Homeland Security cooperate on … biological initiatives to maximize complementary research, development, test, and evaluation (RDT&E) and acquisition efforts and to minimize duplicative efforts and enhance technical cooperation’.³⁷ While these documents do not explicitly call for a sequencing-based identification approach, the linking of public health and biodefense mission spaces greatly increases the number of targets which thenation must be prepared to identify, characterize, and respond. Unbiased high-throughput sequencing will likely play an increasingly predominant role in supporting the US government’s capacity to identify and characterize a diverse panel of natural or weaponized biological agents. The time-sensitive nature of response requires that this technology will need to be employed to identify biological agents directly from a variety of complex matrices.

Specific to biological weapon attacks is the necessity to attribute the attack to state or non-state actors, enabling an international military or diplomatic response. Attribution for the 2001 anthrax attacks took years, and, in fact, the case was not formally closed by the Department of Justice until 2010.²³ The National Strategy for Countering Biothreats and the 2010 National Security Strategy both explicitly call for an expanded capability to attribute biological attacks to perpetrators, specifically by advancing the field of microbial forensics. In parallel, the National Research and Development Strategy for Microbial Forensics was released in 2009.³⁸ This document specifically called out the need to ‘continue to support the development of rapid and cost-effective high-throughput sequencing and closure technologies that can be used to generate high confidence whole genome sequence data and genetic variation data for any known and unknown microorganism’. Specifically, high-throughput sequencing must be developed as an attributional tool to enable a much faster response capability than current state-of-the-art technologies.

1.2.1 Historical perspective of sequencing

1.2.2 Metagenomics

The term ‘metagenomics’, first coined by Handelsman et al. in 1998,⁸⁴ refers to the study of community genomics, as opposed to traditional microbiology in which isolate genomes are sequenced and analyzed as distinct entities. The idea is to study a collection of genes sequenced from an environment. In other words, metagenomics is ‘the application of modern genomics techniques to the study of communities of microbial organisms directly in their natural environments, bypassing the need for isolation and lab cultivation of individual species’.⁸⁵ These efforts have shown that the great majority of biodiversity in the microbial world was missed by traditional cultivation-based methods.⁸⁶

Early efforts, led by the Pace group, utilized PCR of 16S ribosomal RNA sequences to explore microbial diversity outside the biases induced by culturing.^86,87 As all bacterial and archaeal cells contain 16S ribosomal RNA, this highly conserved non-coding gene was ideal for identification of community members.⁸⁶ Early on, before they were technologically possible, the Pace group proposed implementation of true metagenomic studies.⁸⁷ Initial reports of cloning bulk DNA isolated from environmental samples were published in 1991,⁸⁸ suggesting that interactions among environmental microbes are much more complex than previously noted among laboratory cultures, despite a focus on non-protein-coding, highly conserved genes. Comparisons of functional genes from grasses were soon after reported,⁸⁹ as well as studies on the phylogeny of environmental microbial communities based on 16S.⁹⁰ Needless to say, with a decade of existence under its belt, by the mid-1990s metagenomics had become an exploding field of research.

1.2.3 Technological innovations on the horizon for NGS

Anticipating the ‘next big thing’ in any technological field is always a risky business; however, there are several groups bidding to lead the next revolution in genomic sequencing. These endeavors include expanding on currently commercialized technologies (such as pyrosequencing, ligation and hybridization), exploring new methods for single molecule detection, often utilizing nanopores, and a few completely new ideas, such as electron microscopy and nanowires.

1.2.4 The data analysis challenge

A major challenge with utilizing sequencing as either a clinical diagnostic tool or a means of deriving information from pathogens in a rapid response situation in the biodefense context is the ability to process and handle the vast quantities of data generated by each of the sequencing platforms and to turn that information into results that can be acted upon by either a practicing clinician or a policy maker. In the case of bioterrorism, the objective will be to generate investigational leads as rapidly as possible to allow the process of attribution to occur, with the hope that any suspects might be apprehended quickly. In the case of a true biological warfare scenario, attribution would help determine how a nation should respond. For a clinician, the goal would be to diagnose the disease, describe the pathogen, and determine how and with what to treat an infected patient. In all three cases, robust data handling, management and analysis procedures are required to provide the information needed. The process should be automated to run efficiently and effectively; also, the reports generated should be easily understandable. Wherever possible, analytical procedures should highlight how the identified organism differs from previously characterized strains. Ideally, centers performing in such rapid response scenarios should have information flows pipelined, such that as much analysis as possible is performed automatically with outputs readily available to the analyst. At a minimum, a pipeline for rapid response should include the following modules:

All of these steps require high-quality reference sequences and/or well-curated, annotated data bases with functional descriptions of each of the various genes. Conformity in annotation between each reference data base entry, particularly harmonization of annotations across reference sequences in GenBank, would be highly desired (see Section 1.4.5 on annotation).

1.3.1 Traditional methods of characterizing known pathogens

NGS and sequencing-based analyses have dramatically transformed the landscape of everything from diagnostics to evolutionary biology.¹¹⁸ The need to characterize pathogens and to define the questions NGS can help solve remains. Currently, classic methods for nucleic acid characterization have been applied to guide more recent sequence-based efforts. However, as sequencing costs continue to decline the trend may reverse sooner than later. Overall, these classic methods for strain/speciesdifferentiation have their route in both genotypic and phenotypic characterization. Current phenotypic characterization tools evolved from classic microbiological techniques for bacterial/viral identification. Methods such as metabolic profiling, Gram staining, morphology and fatty acid typing evolved from classic biochemistry analytics to the current Omnilog, VITEK, electron microscopy and MIDI instruments used today. Similarly, the advent of Sanger sequencing allowed discrimination and identification based on nucleic acid signatures or genotype.¹¹⁹

1.3.2 Genomic standards and viral characterization

Language and vernacular are one of the critical aspects of communication, and, in the genomics context, shared vocabulary results in an understanding of data set quality and completeness. Classification terms of ‘finished’ or ‘draft’¹⁴¹ were adequate in the Sanger era, when genomic data were prevalent but not overwhelming. The need for these classic standards to be redefined arose as data generation became more expeditious than the downstream analysis and the term ‘draft’ spanned a large spectrum of genomic coverage and completeness. Chain et al.¹⁴¹ made the first attempt at tackling this issue and breaking down the ‘draft’ and ‘finished’ statuses for bacterial and eukaryotic genomes into bins-based metrics of coverage and post-sequencing processes applied to the data set. Draft nomenclature was broken into ‘standard draft’, ‘high-quality draft’, ‘improved high-quality draft’ and ‘annotation-directed improvement’, the major segregation being a coverage criterion of >90% between standard and high-quality drafts and remaining terms defined by application of specific bioinformatics processes. The parsing of ‘finished’ into ‘non-contiguous finished’ and ‘finished’ reflects more the necessity to say ‘good enough’ for some eukaryotic genomes for which resolution of highly repetitive and/or intractable gaps may not be resolved by conventional means. These terms are accepted as standards for genome submission; however, one set of organisms is ostensibly absent, viruses.

Defining common genomic standards is a critical issue moving forward in the characterization of viral nucleic acids; however, this issue is mired in complexity not found in bacteria or eukaryotes. While viruses are generally less complex than bacteria in terms of genomic content/architecture,¹⁴² the notion of quasi-species and application to viral populations and RNA viruses^143,144 creates a situation that does not lend itself to standardization. Specifically, the idea that within any given sample a spectrum of different genotypes exists does not allow similar criteria to those defined above to be applied. In fact, the existence of quasispecies within a viral population suggests that a new lexicon must be developed, specific to viral genomes, before the redefinition of the bacterial and eukaryotic standards can be resolved. Should terms such as consensus sequence or criteria such as population representation be instituted to describe viral genomes? That is beyond the scope of this chapter; however, this issue needs to addressed, and in short order.

1.3.3 Changing landscape of bacterial genetics: whole genome sequencing (WGS) and linking phenotype to genotype

Characterization of known pathogens encompasses defining the pathogen’s virulence and other adaptive features and establishing a link between the phenotype and the genotype. Bacterial genetics, the study of heredity and variation in bacteria, entails the understanding of how the genetic information stored (encoded in DNA) is expressed and regulated during the interplay of gene functions; i.e. in bacterial metabolism, growth, reproduction and faithful transmission of genetic material from generation to generation. To state it simply, it is the understanding of how the bacterium’s phenotype is determined by its genotype. The study of bacterial genetics involves three steps: (1) isolation of mutants with defined changes in phenotypes, (2) genetic/physical mapping of mutations and identification of the gene/locus of interest, and (3) verification of the genetic link to the phenotype by complementation of the mutant with a wild-type copy of the gene, typically expressed on a plasmid. All of these steps are time-consuming, taking anywhere from months to years, and various strategies to aid in these steps and minimize the time frame have been developed over the years, especially in model, genetically malleable organisms. Isolation of mutants and mapping of mutations have been the most arduous of these steps, especially in organisms in which genetic manipulations are arduous or are nonexistent.

Chemical and UV mutagenesis have been used to induce various types of mutations, but mapping these mutations without linked selectable markers is difficult, especially if the mutation gives rise to a non-selectable phenotype. To overcome this problem, transposons carrying selectable markers have been used to link mutations or create mutants. Transposon insertions allow facile mapping, cloning and identification of the gene of interest using the antibiotic marker encoded by the transposon. Generally, transposon insertions lead to gene inactivation, gene product loss and inactivation of essential genes, which may be lethal to the bacterium. Transposon insertions often result in phenotypes caused by polar effects on downstream genes, thus creating problems in identifying the true causal variation for a given phenotype. Often interesting mutants are those with altered functions due to protein non-synonymous changes rather than loss of function. Thus, mapping unmarked mutations poses a challenge.

1.4.1 Examples of metagenome sequencing to determine etiologic agents

Identification of disease causal agents has historically relied heavily on the ability to culture the organism in the laboratory and/or use of organism-specific antibodies or sequence-based probes. These methods can be very limiting. For instance, many microorganisms are recalcitrant to laboratory culture. In addition to this, some etiologic agents considered ‘culturable’ have a false negative rate, such as endocarditis caused by staphylococci and streptococci.¹⁵⁰ Serological assays are typically limited to identifying known or closely related organisms, and antigenic drift and shift can result in false negatives. Even highly sensitive PCR-based assays must be continually updated due to signature degradation.¹⁵¹ All of these reasons lead to a need for assays immune to these limitations and/or biases. Prior to the advent of high-throughput sequencing (HTS), high-density oligonucleotide microarrays were used to determine the presence of microorganisms. Syndrome-specific panels showed success in diagnosis of infectious disease.¹⁵² However, any sequence features sufficiently different from the array probe cannot hybridize, and result in false negatives. HTS represents a relatively unbiased approach to detection of causal agents of infectious disease.

The primary example is a study of Honeybee Colony Collapse Disorder (CCD) published in 2007.¹⁵³ In this study, a metagenomic ‘survey’ was conducted, in which numerous healthy colony samples and diseased colony samples were characterized by 454 sequencing, and Israeli acute paralysis virus was identified as being correlated with CCD. Then, in early 2008, a study was published in which HTS was used to identify a novel arenavirus in three transplant patients who had died of an unexplained febrile illness.¹⁵⁴ In this case, all available microbiological, serological, PCR and microarray assays had failed to identify a causal agent of the febrile illness. This report highlighted the power of HTS to detect a novel agent from a limited amount of sequence information. Indeed, of over 100 000 sequence reads, only 14 corresponded to the new virus. The sequences were the basis for design of multiple detection assays, including a RT-PCR assay to detect the virus in tissue samples. Using the same experimental approach, a second novel arenavirus was identified from clinical specimens originating from a highly fatal outbreak in South Africa.¹⁵⁵

Published examples of metagenomic sequencing to detect a causal agent in diseases of animals and humans are already too numerous to be summarized here, despite the recent development of the application. In some cases, as in the case of a novel filovirus that caused an outbreak in Uganda in 2007, metagenomicsequencing was used to follow up when results from traditional diagnostic assays indicated a novel agent. The Ugandan filovirus had produced a positive result in a broadly cross-reactive IgM capture assay followed by mixed results in RT-PCR assays for known filoviruses, so metagenomic sequencing was employed to characterize the novel virus’s genome.¹⁵⁶ In another example, more traditional diagnostic assays and metagenomic sequencing both played a role in detection and identification of human metapneumovirus causing fatal infection of wild Rwandan gorillas.¹⁵⁷ The opposing scenario, in which traditional diagnostic assays completely fail and metagenomic sequencing plays more than just a supporting role, would include a more recent report of astrovirus encephalitis in an immune-compromised teenage boy.¹⁵⁸

1.4.2 Some limitations of MGS (metagenome sequencing) as of today

1.4.3 Problems with annotation data base(s) and some potential solutions

A fundamental and widely recognized issue plaguing the expanding genomic information data bases is the creation and propagation of erroneous and misleading annotations. This is not as problematic for well-characterized core metabolic genes. Often annotations are fully automated, relying entirely on comparison with previous annotations: for example, the lpxO gene of Salmonella enterica serovar Typhimurium, whose function was discovered and characterized in 2000 and was subsequently characterized in depth enzymatically to unambiguously identify its function. A cursory analysis of representative annotations of this gene product available in even the relatively well-curated RefSeq database yields the results shown in Table 1.2.

The protein homologues in question are >99.5% identical to the original entry. Most notably, the same gene product is found annotated differently in genomes submitted by the same group in the same study! If anything, it can be argued that the quality of automated annotation is actually decaying as more entries are deposited. Definitive functional evidence for any given gene is likely to come from only a very limited number of sources. Similar confusing annotations can be found with the extremely well-characterized spo0F gene of B. subtilis and related species (see Table 1.3).

The lack of standards in annotation and genetic nomenclature results in considerable loss of valuable time for researchers, particularly during function assignment to given mutations, as researchers must often comb through manydifferent entries for a given gene product. The ‘annotation confusion’ is the combined legacy of decades of idiosyncratic bacterial gene naming and renaming by individual investigators and the proliferation of mis-annotations perpetuated by automated annotation algorithms. Resolution of this issue may come through widespread crowd-sourcing (e.g. Wikigenes) and/or adoption of universal standards for functional annotation combined with effective methods for propagating literature-supported revisions to gene annotations throughout the expanding database of genomic information.

1.4.4 Pathogen discovery process

Classic pathogen discovery strategies typically employ a tiered diagnostic approach for etiologic agent identification in the event of an infection of unknown origin. Samples are screened for the presence or absence of pathogens endemic in the region using relatively inexpensive assays. These assays would include immunoassays, such as antigen-capture or IgM capture ELISAs, as well as realtime PCR-based diagnostics specific to a pathogen group (e.g. a pan-filovirus assay), or may be pathogen-specific. Classical microbiological methods, such as bacterial and viral culture, are also be used for pathogen identification. If these commonly used assays fail to identify the pathogen, a second tier of diagnostic assays will be implemented.

This second tier of assays or nucleic acid characterization tools is more expensive than the initial tier but provides broader detection abilities and/or does not require a priori knowledge of the pathogen. This includes microarray-based screenings, such as the ViroChip^172–175 or the GreeneChip.^152,154 The ViroChip is a microarray that contains probes from all of the known viruses found in GenBank. The GreeneChip is arrayed with oligonucleotide probes designed against known vertebrate viruses, bacteria, parasites and fungi. When these more traditional assays fail to identify the pathogen, NGS may be applied. A sequencing approach is unbiased and generates a significant amount of data. Using this approach, though, can be difficult, since disease causation would still have to be demonstrated. As the cost and time related to NGS decrease, use of microarrays for pathogen detection is becoming less common.

There are multiple examples in the literature of using this tiered approach for novel pathogen identification.^{154,172,173,175–177}

One example involves the identification of a novel adenovirus as the causative agent in a pneumonia outbreak at a US primate center.¹⁷² This outbreak was highly pathogenic, resulting in >80% (19/23) mortality among New World titi monkeys, and at least one researcher and a family member became acutely ill with a respiratory infection. Classical microbiological assays, including bacterial, fungal and mycoplasma culture, and viral respiratory panel assay, did not identify the pathogen. RNA isolated from clinical samples identified adenovirus as the potential pathogen with the ViroChip. Following PCR confirmation, WGS and phylogenetic analysis determined that this new titi monkey adenovirus (TMAdV) is highly diverse from other human and simian adenoviruses.

Another example of this tiered approach is the identification of a novel arenavirus from recent organ transplant recipients.⁴ Three individuals died not long after receipt of organ transplants from a single donor, presenting with febrile illness. Classical microbiological assays, as well as a panmicrobial oligonucleotide microarray analysis,¹⁵² failed to identify the pathogen. NGS and analysis identified an Old World arenavirus related to lymphocytic choriomeningitis virus (LCMV) and identification was confirmed by classical microbial techniques.

1.5.1 Field-able sequencing

Current efforts to place sequencing ‘in the field’ involve the installation and setup of technologies in non-traditional laboratory spaces. Difficulties include the high energy consumption, large computing power required to decipher the data, and the training and equipment required to properly prepare and run samples through the system. While each hurdle is substantial, none are insurmountable.

Current platforms, such as the MiSeq and Torrent PGM, are small enough to fit in a mobile space. In fact, Life Technologies has showcased the portability of its PGM platform by placing it on a bus that travels around the United States (Figure 1.2(a)) and in the trunk of a Mini (Figure 1.2(b)). These mobile laboratories are able to provide enough power to the sequencer during the run for successful completion and then transfer the data generated to cloud-based storage. Of course, an open laboratory, such as in the trunk of a car, is highly susceptible to contamination and so may not be ideal for most applications.

More applicable is the installation of sequencing technologies in non-traditional laboratories, allowing the setup of all ancillary equipment and full training of staff involved. These laboratories can be designed to include electrical power conditioners, allowing for intermittent power access, as well as restricted Internet access and on-site analysis of the data generated by commercial off-the-shelf (COTS) bioinformatics tools.

1.5.2 Sequencing as diagnostic tool

To achieve the next step in disease diagnostics, sequencing-based applications present unique opportunities on the path to broad-spectrum, ‘assay free’ diagnostics. Even in present FDA-approved systems for multiplexed diagnostics, sequence-based recognition is necessary for assay design and function. Molecular identification depends on specifically designed reagents, whereas future systems will likely generate sequence data that will ultimately be compared against reference information in silico to make identification. The common thread between the present and future diagnostic systems is a guiding data base for either assay development or instrument result correlation with a known set of disease signatures.

Such a robust, validated data base does not exist in the public sector, and the FDA-cleared technologies that have been created based on known sequences (i.e. PCR systems, arrays or others) have often relied upon a proprietary data base (or internally verified versions of error-rich public data bases) to drive assay design containing the molecular comparator. In the case of present-day and historical systems, the data base was not large, because a limited number of assay targets were sought and the proprietary data bases could be more easily validated for a very specific set of applications. In the future, a data base driving sequencing-based system diagnosis will likely be much more significant in size, perhaps even approaching GenBank in size. Yet this data base will need to far exceed the presentstate of GenBank in terms of data quality (as do the current, proprietary data bases for diagnostics), as it will serve the needs of agnostic sequencing-for-diagnostics systems and continue to drive developments for assay-based systems.

In today’s molecular comparator systems, probe molecules must be synthesized or generated from a natural system, are fixed after generation and cannot be easily changed. Once incorporated in a multiplex system, addition of new elements is difficult and requires revalidation experiments. In future systems that might rely upon vast (much larger than today’s proprietary diagnostic, sequence-based data bases) high-quality, curated sequence databases, entries would be added (following a pre-defined quality metric) when required. This would enable reconfiguration by changing data base entries, leading to widely available target opportunities for Dx systems of the future. The fact that such a data base would be public also adds to the accessibility of this information for device design, and would therefore enable a potentially greater variety of technical approaches, as more developers could utilize this information.

The data in the present-day GenBank data base serve a very useful function in driving unprecedented scientific discovery, but do not have the consistent quality standards across all entries to serve as a diagnostic data base. The utility of GenBank in the basic research environment shows the power of a vast and centralized genomic repository. Making a high-quality reference data base available for diagnostics would fuel advancement in medical treatment rivaling the advancements in scientific discovery that were fueled by NCBI and GenBank.

1.5.3 Evolving the cultural paradigm and legal framework for collaboration and exploitation of emerging technologies

NGS is truly a disruptive technological innovation. Both the extent and scope of fields of practice potentially impacted by NGS are great. By its nature, having the capability to cost-effectively and readily generate genomic information provides foundational information about organisms and communities of organisms. Hence, information generated by NGS has impacts throughout biology and the myriad fields of practice impacted by biology; and, by extension, the political and religious structures of society. A brief overview of the non-technical challenges facing application of NGS is provided below, since a comprehensive review is beyond the scope of this chapter. While significant and appropriate focus is occurring to address aspects of these challenges (i.e. impacts of NGS on medicine¹⁷⁸),other areas of potential impacts are receiving less attention (e.g. intellectual property management¹⁷⁹). Unforeseen impacts may also occur, as was the case with NGS WGS of strains of Vibrio cholerae providing the foundation for a lawsuit by Haitians against the UN humanitarian effort.¹⁸⁰

The non-technical challenges of NGS can be divided into (a) sample acquisition and management, (b) information generation and (c) information management. Illustrative fields of practice that are utilizing NGS include genomic medicine, the broad area of research and development involving biological systems (e.g. wildlife biology, therapeutics, biofuels), criminology and national security. Each of these fields of practice has overlapping and unique challenges to confront in effectively utilizing NGS. Of equal or greater challenge is managing sample acquisition, information generation and information management across fields of practice.

A hypothetical example, employing previously envisioned activities/events, tracking a single sample of a human genome through scenarios of sample acquisition, sequencing and information management is illustrative. Imagine a time in the not too distant future when at birth an infant’s genome is sequenced by public health regulation, as is currently the case for diagnosis of such genetic anomalies as PKU, or voluntarily due to parental interest. The cost of this sequence is greater in the US because many human genes have been patented and royaltiesmust be paid for WGS and follow-on analysis. Of course, this cost may be avoided if the sequencing is performed in a country that does not recognize intellectual property ownership for natural genes and the information is transmitted over the Internet to the physician and the baby’s parents. The information is provided on some digital storage media for future use by the child’s parents as well as the physician, the hospital, and local and national health authorities for some set of specified uses. A professional from the burgeoning field of genomic counseling provides guidance to the parents as to how the child should be raised in order to maximize predicted phenotypic potential while minimizing adverse health effects associated with presence of deleterious alleles. Of course, this counseling does not occur until the parents have taken a basic course in medical genomics so that they do not take any rash action when presented with negative information.

Our hypothetical baby grows and matures, benefiting from the insights provided by predictive genomic analysis. The school authorities have additional data for making class assignments, and prophylactic medical care, diet and treatments for childhood illnesses are tailored to the child’s unique genotype. All goes well with effective management of the child’s genomic information, with the exception of addressing a paternity concern raised by the father, until an adolescent indiscretion brings the teenager to the attention of the authorities. At this point the ‘medical’ genomic information is requested by the law enforcement community. The nascent area of law that manages the interface between medical and law enforcement genomic information management is asked to rule on the request. After some deliberation, a redacted copy of the teenager’s genome sequence is provided to law enforcement and the matter is resolved. This matter is closed until a bodily fluid sample found at the scene of an alleged terrorist event yields a match to the redacted genome held in the national criminal genomic bank. Of course, it is not a complete match, since the entire genome is not available to law enforcement and the conclusions are probabilistic in nature. At this point a third community of practice, the national security community, enters into the picture. Since this is a potential act of terrorism having grave potential impacts for national security, greater latitude is afforded the investigating officials and the entire genome is provided. This transfer of the entire genome to the national security apparatus is not without controversy, since advocacy support groups decry the breach of privacy and racial profiling. Fortunately, the complete genome exonerates our twenty-something case study. But all is not well, since the love of his life he met on Facebook is now asking for a copy of his genome to run a compatibility profile with her own. Further, she is asking him to submit to a metagenomic analysis of his oral cavity and gut before she will share bodily fluids with him, since she has had her high school Better Health through Genomics course.

It is hard to imagine that the legal, ethical and cultural frameworks needed to realize most aspects of the above hypothetical case could be established in a time frame commensurate with availability of the technical capabilities. Aspects of these challenges are explored in the science fiction movie GATTACA (1997); stimulatingand perhaps prejudicing the lay public’s interest and views on application of the technology. Communities of practice are beginning to explore these challenges, such as the World Health Organization’s Ethical, Legal and Social Implications (ELSI) of human genomics component of the Human Genetics Programme focusing on human health. However, rationalization of guidelines and regulations across communities of practice (e.g. health, law enforcement, national security) will add a great additional level of challenge. Therefore, it is anticipated that lack of needed legal and ethical consensus will result in ‘back pressure’ on utilization of NGS technology. NGS is redefining the concept of disruptive technology.