BOARD OF GOVERNORS, AMERICAN ACADEMY OF MICROBIOLOGY

Michele S. Swanson, Ph.D., Chair University of Michigan

Martin J. Blaser, M.D. New York University

Donald A. Bryant, Ph.D. Pennsylvania State University

Nancy Craig, Ph.D. Johns Hopkins University School of Medicine

Terence Dermody, M.D. Vanderbilt University

Gerry Fink, Ph.D. Massachusetts Institute of Technology

Stan Fields, Ph.D. University of Washington

James M. Hughes, M.D. Emory University

Steven Lindow, Ph.D. University of California, Berkeley

Margaret McFall-Ngai, Ph.D. University of Wisconsin-Madison

Mary Ann Moran, Ph.D. University of Georgia

Graham C. Walker, Ph.D. Massachusetts Institute of Technology

COLLOQUIUM STEERING COMMITTEE

George Weinstock, Ph.D., Chair The Jackson Laboratory for Genomic Medicine

Brittany Goldberg, M.D. Children's National Medical Center

Nathan Ledeboer, Ph.D., D(ABMM) Medical College of Wisconsin

Eddy Rubin, M.D., Ph.D. ^* Department of Energy Joint Genome Institute

Heike Sichtig, Ph.D. Food and Drug Administration

Footnotes

* Was not able to attend colloquium but reviewed the report as a participant

PARTICIPANTS

Marc W. Allard, Ph.D. Food and Drug Administration

Andrew K Benson, Ph.D. University of Nebraska-Lincoln

Carey-Ann Burnham, Ph.D., D(ABMM) Washington University School of Medicine

Susan Butler-Wu, Ph.D., D(ABMM) University of Washington

Charles Chiu, M.D., Ph.D. University of California, San Francisco

Chris Detter, Ph.D. Department of Defense Defense Threat Reduction Agency

William M. Dunne, Ph.D., D(ABMM), F(AAM, IDSA, CCM) bioMerieux, Inc.

Jennifer Gardy, Ph.D. University of British Columbia

British Columbia Centre for Disease Control

Amy S. Gargis, Ph.D. U.S. Centers for Disease Control and Prevention

Elodie Ghedin, Ph.D. New York University Center for Genomics & Systems Biology

Kelly Hoon Illumina

William Klimke, Ph.D. National Center for Biotechnology National Institutes of Health

Michael Lorenz, Ph.D. University of Texas Health Science Center

Timothy Minogue, Ph.D. U.S. Army Medical Research Institute of Infectious Diseases Division of Diagnostic Systems

Randall J. Olsen, M.D., Ph.D. Houston Methodist Hospital

Gustavo Palacios, Ph.D. U.S. Army Medical Research Institute of Infectious Diseases Center for Genomic Sciences

Adam Phillippy, Ph.D. National Biodefense Analysis and Countermeasures Center

Gary W. Procop, M.D., M.S. Cleveland Clinic

Julie Segre, Ph.D. National Human Genome Research Institute National Institutes of Health

Shanmuga Sozhamannan, Ph.D. The Tauri Group, LLC, supporting the Critical Reagents Program

ACADEMY STAFF

Marina Moses, M.S., Dr.P.H. Director

Chelsie Geyer, Ph.D. Colloquium Post-Doctoral Fellow

Daniel Peniston Colloquium and Public Outreach Program Assistant

Erin Seglem Board of Governors Program Assistant

ASM LEADERSHIP

Joe Campos, Ph.D., D(ABMM), F(AAM) ASM Secretary Children's National Medical Center

RAPPORTEUR

Chelsie Geyer, Ph.D.

Summary

Next-generation sequencing (NGS) has the potential to dramatically revolutionize the clinical microbiology laboratory by replacing current time-consuming and labor-intensive techniques with a single, all-inclusive diagnostic test. Traditional methods for identifying organisms such as mycobacteria, some bacterial species, and fungi in particular are often slow, specialized, and organism specific. Culturing, Gram staining, and biochemical and molecular tests are traditional assays that consume the manpower of the clinical microbiology laboratory. From this battery of tests, relevant treatment guidance for the clinician is not always produced. This has been described elsewhere as a “diagnostic odyssey” or a guessing game for the diagnosis and identification of infectious diseases. Executing diverse clinical tests can waste precious time for a patient and might be the difference between life and death. In the groundbreaking neuroleptospirosis case described by Wilson et al (2014), 38 different diagnostic tests on various sample types (e.g., cerebrospinal fluid [CSF], brain, urine, stool, sputum, blood, serum, plasma, oropharyngeal/ nasopharyngeal swab) were performed before the diagnosis was ultimately made with NGS. This technology can potentially expedite the turnaround time for a result and allow clinically actionable information to be obtained sooner than a traditional laboratory workup would allow. Data generated from an NGS assay would ideally provide a diagnostic and therapeutic decision in a clinically relevant time frame that would positively enhance patient care and outcome. The American Academy of Microbiology (Academy) conducted a two-day colloquium to review the potential future impact of NGS on clinical microbiology and how barriers for implementation can be overcome through suggested recommendations identified by invited colloquium experts. Main topics discussed at the event are highlighted in an mBio minireview article prepared by the NGS steering committee members.

Basic science and applied research coupled with emerging technologies has enabled NGS to transition into the diagnostic laboratory setting to provide clinically actionable results. Insights acquired from NGS methods can be exploited to improve our health as individuals and the greater public health. NGS is poised to broaden our understanding of how microbes interact in different ecosystems, as well as their functioning during health and disease in humans, animals, and environments, including both the built and natural (outdoor) settings. Food safety measures and product quality may also be advanced by NGS technology by allowing faster detection of contaminating pathogens that may arise during manufacturing or processing and subsequently trigger outbreak scenarios. Taken together, the functionalities of this sequencing tool can help achieve the main goal of the One Health Initiative, which is to obtain optimal health for people, animals, and various environments (Figure 1).

Figure 1.

One Health Initiative: the interdisciplinary global health collaboration.

Moreover, NGS will contribute to the Precision Medicine Initiative, a bold and innovative approach for disease treatment, management, and prevention launched by President Obama in his 2015 State of the Union address. At the present time, NGS is being applied to precision medicine to help diagnose human genetic disorders, prenatal disorders, and cancers. For example, Worthey et al (2011) executed whole-exome sequencing, a type of NGS approach, to diagnose a rare immune defect that saved a 15-month-old patient's life. All standard diagnostic tests had been exhausted with no definitive cause for the patient's Crohn's disease-like illness that was coupled with recurrent episodes of sepsis and infection. Exome sequencing revealed a novel, hemizygous missense mutation in the X-linked inhibitor of apoptosis gene that atypically presented as severe gastrointestinal (GI) disease. This mutation directed clinical management to perform a successful allogeneic hematopoietic progenitor cell transplant that cured the patient's GI illness.

Next-generation sequencing (NGS) is a blanket term that collectively refers to high-throughput DNA sequencing strategies that can produce large amounts of genomic data in a single reaction by diverse methodologies. NGS is also referred to in the literature as “deep,” “high throughput,” or “massively parallel” sequencing.

Precision medicine recognizes that a “one-size-fits-all” diagnostic and treatment approach is not successful for every patient due to individual variability and host characteristics, such as genetic composition, lifestyle practices, and exposure to different environments/agents. Another component that contributes to a person's unique biological makeup consists of the microbial communities that reside in niches throughout our bodies. Collectively known as the human microbiome, it is estimated that nearly 100 trillion microorganisms live within our gut, mouth, skin, and other mucosal surfaces. These microorganisms perform many beneficial functions that help maintain our health, such as by combatting offending pathogens that want to inflict harm and by synthesizing essential vitamins and minerals. Therefore, the scope of precision medicine encompasses the study of microbes by NGS for microbiome analyses and pathogen diagnostics. Precision medicine will integrate patient and family medical histories, clinical signs and symptoms of illness/disease, and genomic information whether human or microbial in order to create treatment pathways that are individualized and tailored for each patient. As an example, a customized treatment plan for a septic patient might be life-saving, as sepsis can be caused by a severe bacterial, viral, or fungal infection. It is thought that newer drugs for the treatment of sepsis have failed clinical trials because of the heterogeneity of sepsis as a clinical entity. Although these drugs showed promise on the bench, it is speculated that inherent variability among individuals and disease pathogenesis is the cause for poor therapeutic responses, thereby calling for a more individualized approach that precision medicine and NGS technologies bolster.

NGS offers great opportunities for advancing precision medicine in the clinical microbiology laboratory. With the hope of applying this technology for microbial organism and antibiotic/virulence marker identification for clinical diagnostics, patient care could be dramatically impacted; however, implementing NGS as a routine test in this setting faces significant infrastructure (i.e., computing resources) and bioinformatics challenges. Refining sequencing data to guide clinical decision-making is a complex bioinformatics task, as it involves formulation and interpretation of molecular data in the biological sphere in connection with treatment information in the clinical sphere. Without question, this is an exciting time for the field of clinical microbiology, given the empowering ability of NGS technologies to help us understand and treat infectious disease. This enthusiasm was expressed at the Academy's colloquium, at which invited participants were tasked to answer seven key questions and subquestions developed by the steering committee. Findings from the discussions are summarized in this report, along with real clinical case studies that used NGS for diagnosis and recommendations that address microbiological NGS challenges. The recommendations put forth by the colloquium participants were identified and categorized under the sections below.

Performance Factors and Infrastructure

In order for NGS-based assays to become commonplace in the clinical microbiology laboratory, there is a need for the development of “turnkey” solutions for all phases of testing (e.g., sample preparation, sequencing, data analysis, and result interpretation). The ultimate goal of diagnostic NGS is to place a direct clinical specimen from any matrix into the NGS workflow and generate an actionable result within a reasonable time frame. Continued efforts for direct clinical sample sequencing should be pursued. (Recommendation 2.1)

It is recommended that a distinction be made between diagnostic clinical specificity/sensitivity and analytical specificity/sensitivity when discussing a clinical microbiological NGS test. The qualifiers of “diagnostic clinical” and “analytical” are not interchangeable, and confusion can arise when reporting a laboratory test result. More efforts are also needed to understand the mutation rates and population structure of commonly encountered clinical pathogens in relationship to their effects on NGS sensitivity and specificity as well the use of NGS for molecular epidemiology. (Recommendation 2.2)

To help minimize the cost and bulkiness of NGS hardware implementation, the utility of benchtop and point-of-care (field-able) sequencing platforms should be emphasized to clinical laboratories seeking to engage in this space. These sequencing systems consume less space and are generally less expensive than larger NGS platforms, and data analysis can be completed on a high-end desktop server or even a laptop. (Recommendation 5.1)

Standard Operating and Validation Procedures

A group of stakeholders (e.g., government agencies, clinical laboratory professionals, academia, industry) should be brought together to develop standardized reference materials and data sets that can be used for assay validation and quality control (QC) procedures. To make the validation of bioinformatics pipelines easier, publically hosted “digital” validation test sets could be purchased and evaluated by clinical laboratories. A set of reads that have a known answer could be downloaded and subjected to the lab's bioinformatics components. Also, fully characterized biological reference organisms will be needed to evaluate both the wet and dry NGS processes. (Recommendation 4.1)

It is recommended that different wet and dry bench NGS protocols be created for the detection of etiological agents such as bacteria, viruses, fungi, yeasts, and parasites. Although NGS has the potential to detect all pathogens in a clinical sample, specific protocols would help to advance the transition of NGS into the clinical microbiology laboratory. There needs to be guidance on how to validate and perform QC procedures for these protocols as they pertain to the different pathogens (i.e., what are the unique/pathogen-specific QC metrics that must be considered to ensure the quality of NGS results). (Recommendation 4.2)

Interdisciplinary Teams and Education

To ensure successful utilization of an NGS result, a multidisciplinary team within the clinical or public health laboratory setting should be formed to include the expertise of clinical microbiologists, medical technologists, clinicians, infectious diseases physicians, pathologists, basic research scientists, software developers, and bioinformaticians. This collaborative effort will maximize the strength and interpretation of NGS data. (Recommendation 4.3)

Adoption of NGS into the clinical microbiology laboratory will require clinical microbiologists, medical technologists, and clinicians to receive training in molecular biology and bioinformatics. Future clinical microbiology and public health laboratory professionals will be required to be competent in the field of bioinformatics in order to effectively communicate with bioinformaticians. Beyond general programming skills and bioinformatics knowledge, there needs to be training on understanding and interpreting NGS results. It is recommended that bioinformatics be incorporated into the coursework of medical school students and clinical microbiology/pathology fellowship programs so the students gain familiarity with this diagnostic approach, which is likely to be used during their clinical practice. Exposure to informatics could even begin at the high school and undergraduate levels, since the basic principles are applicable to many fields. (Recommendation 4.4)

Professionals who will use NGS technology should work closely with software developers to create a proficient, streamlined, and more manageable analysis pipeline to provide a quicker return of complete diagnostic information. This collaboration will help in the development of more efficient and user-friendly software programs, interpretable analysis reports, and improved algorithms for genomic data analysis. (Recommendation 6.1)

Interpretation, Sharing, and Management of Data

To greatly assist in outbreak scenarios in both the hospital and the community setting, guidelines or models for responsible data sharing among institutions should be developed and endorsed by a consortium of relevant stakeholders. These models should encourage continual sharing of microbial genomic data and maximize public availability while balancing the need for patient privacy. This balance of sharing data and maintaining privacy is necessary for predictive outbreak detection to work; hence, the public health benefit of using NGS in the clinic can be gained. (Recommendation 3.1)

When NGS data are to be stored remotely such as on a server, in the cloud, or within a database, the information must protect patient privacy and be Health Insurance Portability and Accountability Act (HIPAA) compliant. (Recommendation 5.3)

Sequence data, including raw sequence reads, .FASTQ files, and the complete genomic sequence of the identified pathogen, are large and would consume considerable storage space for a hospital. The assembled sequence should be uploaded to an appropriate database, and only the clinically relevant result should be maintained in the patient's electronic medical file. (Recommendation 5.4)

Clinical Reporting and Billing

Some NGS-based assays, e.g., metagenomic assays, are capable of returning a complex set of results that require careful interpretation by the clinical microbiologist to determine what is clinically actionable and what should be included in the result report to ensure optimal patient care. For example, NGS results should assist the physician in determining what antimicrobial can be used, rather than what the organism is resistant to or what drug is not suitable for treatment. The presence of the gene conferring antimicrobial resistance (AMR) is not evidence of its expression and hence AMR phenotype. There needs to be a way to phenotypically verify the genotypic result generated by NGS. Rapid phenotypic testing methods are currently under development. (Recommendation 5.2)

An information visualization style approach to conducting controlled user studies might help to determine which form of NGS clinical reporting is most effective for clinicians to make a diagnosis or initiate/change treatment for the patient. (Recommendation 6.2)

New regulatory guidelines and insurance reimbursement codes for the use of NGS testing in the clinical microbiology laboratory need to be developed. Insurance billing codes should be revised to enhance the transparency of molecular services that are performed. Additionally, payers such as CMS (Centers for Medicare & Medicaid Services) should review publications showcasing that NGS assays guided or improved diagnostic and therapeutic decisions that could not be made using current laboratory methods. (Recommendation 6.3)

Outcome analyses and clinical trials highlighting the success and cost savings of NGS for the diagnosis of infectious diseases are highly recommended and could serve as justification for reimbursement companies. Incentive and, more specifically, funding, which is not widely available, must be given to initiate such studies. Therefore, an advisory board composed of relevant stakeholders should be created to address this issue. (Recommendation 6.4)

Reference Databases

It is recommended that genomic sequences of emerging microbial pathogens be uploaded to a unified, public database as quickly as possible to allow for community engagement of the data analysis and use of those data to inform other clinical professionals of the pathogens they are encountering in their laboratories. If genomic sequences for high-priority pathogens are routinely deposited, NGS has the potential to serve as the new early warning system for outbreaks that may occur locally, nationally, or internationally. This tactic could help monitor the stability of the outbreak isolate's genome over time and determine if acquisition or removal of genomic information affects diagnostic and therapeutic decision-making. (Recommendation 3.2)

A crucial recommendation is the expansion of curated and regulatory-grade microbial sequence databases in the public domain. Genomic sequence submissions should include high-quality sequence data that are accurately annotated with metadata. These databases should not be a static collection of information but should allow for local, national, and international data exchanges that are in line with agreed standards. Additional databases are not needed, but existing databases should establish standardized quality metrics or curation strategies to promote confidence in clinical decision-making. (Recommendation 7.1)

It is recommended that sequencing efforts be focused on obtaining more pertinent whole genomes for pathogenic fungal, yeast, and parasitic species. (Recommendation 7.2)

Resistance genes should be annotated as a subset within an appropriate existing database. With new genetic mechanisms of resistance frequently arising, these databases would be ongoing projects requiring active curation and reannotation efforts. (Recommendation 7.3)

Approach to the Task

In April 2015, the Academy convened a group of 25 experts at ASM Headquarters to discuss the clinical microbial applications of NGS. A steering committee was appointed prior to the colloquium and was tasked with developing discussion questions that were to be posed at the event and with compiling a list of invitees that had expertise covering the scope of this topic. Invited colloquium attendees included representatives from government agencies, commercial industry, clinical laboratories, and academia, with backgrounds in clinical microbiology, diagnostic development/regulation, food microbiology, public health surveillance, DNA sequencing technologies, bioinformatics, computer science, and database formulation and management. Given that NGS is a field that is constantly evolving, the ASM and the Academy's Board of Governors sought to cast light on the current status of NGS technologies in clinical microbiology practice and to provide recommendations for implementation. Colloquium participants were assigned seven main discussion questions and subquestions that covered the broad categories explained in the statement of task. Participants were divided into working groups that consisted of a variety of expertises in order to answer the discussion questions. Groups reconvened for plenary sessions to review all answers. This report summarizes the discussions held during the colloquium's plenary sessions.

Structure of the Report

The report is divided into seven sections based on the questions that were asked during the colloquium. These discussion questions were developed by the steering committee members prior to the event. Section 1 specifically focuses on the use of NGS for bacterial detection. Although NGS can detect other microbial pathogens, such as viruses, fungi, yeasts, and parasites, the colloquium underscored bacterial identification. Section 2 provides a comparison of NGS technologies to other types of diagnostic assays, including differences in turnaround time, specificity, and sensitivity. Section 3 elaborates on the impact that NGS could have on the clinical and public health microbiology fields, including the use of NGS for metagenomics and human microbiome studies and the identification of unculturable/ difficult-to-culture organisms. Section 4 describes the necessary factors for incorporating NGS into the clinical microbiology laboratory workflow, including standard operating procedures, process validation, and reference materials for validation. This section also discusses the synergy that needs to be created among various groups of professionals to fully realize the benefits of NGS technologies. Section 5 examines the issues encompassing NGS data analysis, interpretation, management, storage, and archiving. Section 6 explains the downfalls of NGS, the challenges of implementation, and suggestions for overcoming these difficulties. Section 7 discusses the need for quality and curated microbial genomic reference sequences and metadata in public repositories. Numerous case studies that demonstrate the utility of NGS as infectious disease diagnostic tests are also highlighted throughout this report.

1a. Status of Virus, Fungus, Yeast, and Parasite NGS Applications

Discussions at the colloquium focused primarily on the use of NGS for the detection and identification of bacterial species, since bacteria are the most commonly encountered microorganisms in the clinical microbiology laboratory (69). A brief synopsis of the status of NGS with other kinds of microorganisms, including viruses, yeasts, parasites, and fungi, is described below.

Much of the pioneering work in establishing NGS as a method for pathogen detection was conducted by virologists. The initial studies using metagenomics by NGS on human clinical specimens unveiled a novel polyomavirus associated with Merkel cell carcinoma and a novel arenavirus that caused fatal febrile illness.). [See Case Study 3, which demonstrates how RNA sequencing and ultimately protein homology were used to discover a new arenavirus that had been transmitted through solid organ transplantation (70, 71).] Briese et al (72) used unbiased pyrosequencing of RNA extracts to evaluate an unexplained outbreak of extremely fatal hemorrhagic fever in southern Africa. Within 72 hours of receiving the clinical samples from outbreak victims, phylogenetic analyses showed the presence of a highly novel viral genetic lineage that was only distantly related to the Old World arenaviruses and was known as Lujo virus (LUJV). The work by Briese et al marks the first application of NGS for pathogen discovery associated with an outbreak of hemorrhagic fever caused by a genetically distinct virus. Other examples of NGS for identification of novel viruses in outbreak settings include the discovery of Bas-Congo virus (BASV), a novel rhabdovirus associated with a cluster of hemorrhagic fever cases in central Africa (73), and novel adenoviruses in an acute respiratory outbreak in a baboon colony with evidence of coincident human infection (74).

Case Study 3

Application of metagenomic RNA sequencing discovers a novel arenavirus responsible for three cases of fatal febrile illness (70).

As evidenced by the abovementioned studies and many others (75, 76, 77, 78, 79), NGS technologies have made significant contributions to the clinical virology field, including diagnostics, discovery, pathogenesis, epidemiology, and genome sequencing, through metagenomic-based sequencing approaches (known as viromes) (13, 79). With viral metagenomics, either DNA or RNA may serve as the input nucleic acid for analysis. Although library preparation is more time-consuming for RNA than for DNA, in some cases, RNA sequencing is more beneficial than DNA sequencing. First, some viruses have an RNA genome which would not be detected if only DNA was extracted and used as the starting material, as exhibited by Case Study 3. Second, performing total RNA extraction/sequencing can capture both DNA and RNA expression, and hence, mRNA sequence can be translated into protein. Amino acid sequence tends to be more conserved than nucleotide sequence and therefore may yield more defined taxonomic information (70, 80, 81).

The use of NGS for virus detection has progressed tremendously since initial studies were published in 2008. In fact, clinical viral diagnostics are now more advanced than bacterial NGS diagnostics. Given the high specificity of viral sequence data, presently it is more practical for a viral diagnostic test to be CLIA (Clinical Laboratory Improvement Amendments of 1988) approved than a bacterial NGS assay. Applying NGS to bacterial analysis to make a specific diagnosis can be complicated when performing a metagenomic NGS application, simply because large amounts of commensal bacterial microorganisms can colonize the patient, making determination of the causative organism more difficult. Furthermore, it has been demonstrated that contamination of laboratory reagents that are used for NGS, such as DNA extraction kits or molecular biology-grade water (a common contaminant is Bradyrhizobium sp.), can significantly impact the results obtained using metagenomics (82). In contrast, while certain viruses can occasionally constitute part of the normal microbial flora (e.g., torque teno virus) (83) or be potential laboratory contaminants (e. g. Circoviridae/parvovirus-like hybrid virus [NIH-CQV/PHV] and xenotropic murine leukemia virus-related virus [XMRV]) (84, 85), the high sequence specifici-ties of viruses make them easier to identify, and potentially diagnose, from metagenomic data than bacteria. NGS can detect 95% of viral pathogens (13, 86). This high detection rate cannot be said for bacteria unless targeted methods such as 16S rRNA gene sequencing are used. The identification of reads corresponding to a virus is likely indicative of ongoing disease. For example, if influenza virus is detected in a bronchoalveolar lavage (BAL) fluid specimen of a patient with pneumonia, then the infection was likely caused by this virus. On the other hand, if Staphylococcus aureus is detected in a BAL specimen, it is not guaranteed that S. aureus is the cause of pneumonia. S. aureus could present as a colonizer or commensal oral flora, contamination of the laboratory environment or reagents, or a simple misidentification.

Despite the advancements made with NGS viral diagnostics, there are challenges that viral genera and other nonbacterial organisms face when subjected to the technology. Briefly, there are studies being conducted using NGS to detect viral quasispecies with human immunodeficiency virus (HIV), Ebola virus, and influenza virus (56, 79, 87, 88, 89). These small minor variant viral populations can be clinically important but are often challenging to detect, since viral nucleic acid integrity can be compromised during extraction. Furthermore, parasites, yeasts, and fungi are eukaryotic microorganisms that pose bioinformatics challenges. These organisms possess larger genomes than viruses and bacteria, and their reads can be difficult to distinguish from human reads. [See Case Study 4 to understand how WGS was applied to detect a pathogenic fungus that infected three organ transplant patients (90).] In addition, low-quality reference genomes can be contaminated by other viral, bacterial, fungal, or yeast sequences, and thus it can be taxing to determine if a top alignment hit is truly the correct pathogenic hit or just a database contaminant. More discovery efforts are necessary for parasite, yeast, and fungus NGS in addition to reference databases for these nonbacterial organisms (69, 91, 92, 93, 94). There are various medically important parasites, yeasts, and fungi that do not have high-quality reference genomes, and therefore NGS applications involving these organisms will have a tough time advancing without more concrete and curated reference databases.

Case Study 4

NGS reveals a Coccidioides immitis cluster in three organ transplant patients (90).

As the Precision Medicine Initiative moves forward, NGS has the potential to be incorporated into routine clinical microbiology workups to direct diagnostic and therapeutic decision-making that is specific for each patient. Since this colloquium focused mainly on the clinical bacterial applications of NGS, participants identified five main areas that they believe could benefit from the capabilities of NGS. These include (i) clinical identification from primary samples or a pure culture, (ii) infection control actions, (iii) antimicrobial stewardship, (iv) outbreak investigation in community and hospital settings to guide measures for containment, and (v) pathogen discovery (1, 70, 95, 96, 97).

1b. NGS Platforms: Past and Present Commercial Entities

Competition among sequencing vendors has resulted in sustained improvements to NGS platforms from streamlined setup to sequencing chemistry adjustments and user-friendly data analysis (38). Several commercial NGS platforms are currently available in the sequencing marketplace, and newer systems are on the horizon (3, 7, 22, 25, 33, 38) (Figure 2). The basic functionalities of these platforms include WGS, whole-exome sequencing (WES), metagenomic sequencing, and targeted gene sequencing (Table 1) (1, 3, 31, 98). The optimal sequencing platform for a particular laboratory is highly dependent on the purpose for which it will be used. In the clinical microbiology laboratory, NGS platforms are primarily used for WGS and metagenomic sequencing, of which the latter task does not require knowledge of the possible causative agent. Targeted sequencing is also being used in the clinical laboratory if an etiological agent is suspected at the time of clinical sample collection. Regardless of the application, the capital outlay to purchase and maintain NGS equipment is considerable, and smaller hospitals and institutions may not have adequate funding, particularly without a well-established billing infrastructure in place (38).

Figure 2.

High-throughput sequencing platform examples.

Table 1.

Characteristics of NGS platforms.

Although these platforms differ in their sequencing chemistries and engineering principles, they all perform massively parallel sequencing which yields terabytes of data (13). This common technological feature involves the sequencing of either spatially separated, clonally amplified DNA or spatially separated, single DNA molecules on a flow cell (1, 3, 5, 8, 15, 22, 99). Hundreds of megabases (millions) to gigabases (billions) of DNA sequence data can be produced from a single NGS run, which is in stark contrast to the hundreds of bases produced from an individual sequencing reaction of a targeted region by the Sanger method. Also, the amount of input DNA is dependent on the NGS platform and the intended application, along with the library preparation protocol used to extract the DNA to be sequenced. All of these factors might affect expected results and data output (1).

For instance, the typical DNA requirement ranges for the Illumina MiSeq, Ion Torrent PGM, and PacBio RSII are 50 to 1,000 ng, 100 to 1,000 ng, and ∼1 μg, respectively (100). Newer, single-molecule, real-time sequencing (e.g., PacBio) tends to require a larger quantity of purified DNA to yield optimal results (25). Furthermore, when using the NexteraXT library preparation kit for bacterial WGS on the Illumina MiSeq, 1 ng of input DNA is needed, yet when the TruSeq library preparation kit is applied for DNA extraction, 3 to 5 μg of input DNA is required for analysis.

There was a paradigm shift in sequencing design when the Roche-454 system (previously 454 Life Sciences) entered the market in 2005. This technology involves the construction of adapter-ligated DNA libraries, emulsion PCR, and pyrosequencing (Figure 3) (5, 15). Although the Roche-454 instrument had a higher throughput capacity and a lower sequencing cost per base than the Sanger methodology, interest in this next-generation technology among the scientific community was muted. Critics of the technology identified read length, fidelity, infrastructure requirements, and cost of operation as the main issues; however, these same concerns were also voiced when Sanger sequencing was introduced. Currently, the Roche-454 is considered a legacy platform and has paved the way for more advanced and higher-throughput sequencing platforms as described below (19). The Roche-454 systems, including the benchtop GS Junior and the high-end GS FLX+ sequencers, have been surpassed by other less-expensive, higher-throughput platforms with better scalability, and production will be phased out by 2016 (101).

Figure 3.

Roche 454 GS FLX sequencing.

The Illumina platforms employ bridge or cluster PCR to amplify adapter-ligated DNA libraries on the surface of a proprietary flow cell (5, 8). Illumina technology uses reversible dye terminator sequencing by synthesis chemistry that involves repetitive cycles of single base incorporation, imaging, and dye chemistry termination (Figure 4) (5). Sequencing by synthesis is the most widely adopted NGS technology. Illumina sequencers offer the option of paired-end sequencing that enables reads to be generated from both ends of a single clonal fragment. The forward and reverse reads are aligned as read pairs, which allows for more accurate read alignment and superior indel detection than that provided by a single end of the pair. With paired-end reads, the distance between the forward and reverse reads is known and allows assembly algorithms to reconstruct the sequence for repetitive sections of DNA, such as homopolymeric or AT/GC-rich regions of the genome. In recent years, Illumina has dominated the sequencing market for both microorganisms and larger organisms because of the platform's high sequence throughput, low error rate, and low sequencing cost per base. Illumina has a line of machines serving a multitude of purposes and sequencing power on all scales (15). The Illumina sequencing platforms include the HiSeq X™, HiSeq, NextSeq, and MiSeq series that range in cost from nearly $10 million for the latest HiSeq X™ Ten to $99,000 for the smaller, benchtop MiSeq system (9, 100).

Figure 4.

Illumina Genome Analyzer sequencing, the first of the Illumina platforms to be commercially launched.

The SOLiD (Sequencing by Oligonucleotide Ligation Detection) platforms by Life Technologies, including the 5500 W series genetic analysis systems, use a sequencing approach known as oligonucleotide ligation detection. The sequencing by ligation method utilizes dinucleotide encoding, which measures every base twice. Because of a two-base sequencing method, the SOLiD sequencing systems can obtain base call accuracies as high as 99.99%. Although SOLiD sequencing is well suited for high-accuracy applications such as genome resequencing and polymorphism detection, these systems are fading platforms (Figure 5) (5). Of all available NGS platforms, the SOLiD sequencing method produces the shortest read lengths (35 to 50 bp) and has the longest run times (1 to 2 weeks).

Figure 5.

Applied Biosystems (Life Technologies) SOLiD sequencing by ligation.

Other popular components of the NGS portfolio are the benchtop Ion Torrent platforms by Life Technologies, including the Personal Genome Machine (PGM) and the Ion Proton. Although similar to other platforms in their sequencing by synthesis methodology and amplification by emulsion PCR, Ion Torrent platforms differ from other technologies in the detection step. The release of H⁺ ions during base incorporation is measured instead of using fluorescence or chemiluminescence detection strategies, making the Ion Torrent sequencers very sensitive pH meters. Both the PGM and the Ion Proton have flexible reagent chips that generate different scales of sequencing output (10 Mb, 100 Mb, 1 Gb, 10 to 100 Gb) depending on the user's desired sequencing coverage (62). These platforms are popular because of the low sequencer costs ($80,000 to $145,000) and speed of sequencing. The entire workflow, from DNA extraction to annotation, can be completed in less than 24 hours, making the Ion Torrent sequencers most suitable for targeted sequencing or smaller genome sequencing projects (100).

Pacific Biosciences (PacBio) is an example of a sequencing instrument that delivers “single-molecule real-time sequencing” (38). Nucleotides are labeled with dye and continuously added to the growing DNA strand by use of a highly processive DNA polymerase. Attached to this enzyme is a waveguide detector that permits the constant monitoring of the incorporated dye-labeled nucleotides. The sequencing reaction takes place on zero-mode waveguide nanostructure arrays (102). Because of its long reads, PacBio is suited to a variety of applications such as de novo genome assembly and, hence, the characterization of new organisms along with targeted sequencing to detect base modifications (100). Consensus assembly and long reads are providing new reference sequences in regions of genomes which the short-read sequencers fail to uncover, and longer-read PCR sequencing shows promise for phasing alleles and in viral quasispecies. PacBio may yield long reads of ≥10 kb, but the 87% accuracy of individual reads and the substantial infrastructure requirements are downsides of this platform for clinical applications. The PacBio RSII is a high-end sequencing instrument that is priced at nearly $750,000, a cost that is beyond the reach of the average clinical laboratory. The input DNA requirement of approximately 1 μg is also significantly higher than that of other platforms, which need nanograms or even sub-nanogram quantities (100, 103). Additionally, this quantity may not be obtainable when analyzing clinical samples (25). However, it is well suited for the creation of high-quality reference genomes and produces the most accurate and complete genomes of the current NGS platforms (15).

The first in a new breed of ultrafast DNA sequencing technologies is Oxford Nanopore, with nanopore sequencing platforms. A single DNA molecule is guided through a protein nanopore, which results in changes in electrical current across a lipid membrane. Nanopore technology reads native DNA and is predicted to be able to sequence directly from clinical samples with a low abundance of DNA (15). Data generated from the MinION are streamed live to a computer via a USB connection and can potentially provide NGS at the bedside or in the field (104). With Nanopore sequencing, DNA sequence is produced in real time, which enables analysis to be conducted on a continuous stream of long reads (105). Since the MinION sequences a single DNA molecule per pore, it eliminates the library amplification phase and its associated biases, which is a common mechanism used by many NGS technologies. However, at this stage of its development, the MinION still requires the initial library preparation prior to DNA sequencing. The MinION has a low capital cost, and its flow cells that contain the nanopores are to be used only once, which eliminates the tribulations of installation and instrument maintenance. It is anticipated that the MinION will yield 150 Mb of sequence per hour with mean read lengths of ∼2 kb. Despite the long reads, the MinION has an error rate of 15 to 22%, which currently limits its ability to compete with other NGS platforms for many applications, although improvements to the base-calling algorithm are yielding benefits (104, 106, 107, 108). So-called “2D” nanopore reads, where both strands of a read are processed by the nanopore, have error rates below 20% (104, 107). In addition to the MinION, the Oxford Nanopore PromethION and GridION, which provide bench-top throughput analyses and scalable throughput using stackable desktop devices, respectively, are under development. Interestingly, Oxford Nanopore has promised a “pay-as-you-go” model for these devices, eliminating upfront infrastructure costs, which could hasten their clinical adoption. However, there is a lot of uncertainty around Nanopore technology. As with any emerging technology, it is difficult to separate promises from the true capabilities of the instrument (15, 22, 105).

There are many sequencing manufacturers that are competing for sequencing supremacy (109). As more companies enter the NGS race, it is probable that instrument purchase costs will decline and democratization of NGS will occur. Currently, the purchase of NGS instrumentation is a significant financial obligation and commitment (9). It is paramount to assess the needs and demands placed on a laboratory to determine which platform is most suitable and fits within the specified budget (25). If the goal is to incorporate NGS into routine use in the clinical microbiology laboratory, the machine must be able to handle the projected daily volume of isolates and to yield cost-effective and actionable results in a clinically relevant time frame (9, 10, 27, 28, 110).

1c. Main Steps of NGS Workflow

Even though the technical specifics of NGS technologies differ, there are several common steps that are shared among the majority of high-throughput sequencing methods, with the exception of single-molecule real-time NGS as described below (1, 5). Typically, the NGS workflow in a clinical laboratory includes sample collection/ nucleic acid extraction, NGS library preparation, sequencing, data analysis, and data storage (7, 8, 25, 111). The clinical sample, e.g., swab, sputum, stool, urine, cerebrospinal fluid (CSF), or tissue, contains the nucleic acid (DNA/RNA) of interest and is placed in a transport device for delivery to the clinical microbiology laboratory. Clinical specimen types for NGS evaluation will differ depending on the patient's clinical syndrome but should be collected from appropriate sources during disease progression. The type of transport device, storage time, and temperature can influence the quantity and integrity of the nucleic acids present. Many transport mechanisms are not microbial DNA free, raising the possibility of false positives and the introduction of background contaminants.

Nucleic acids can be prepared from clinical samples by using a variety of methodologies, some of which are dependent on the NGS system being used. Suitable extraction methods are essential for a successful result and thereby help to lessen the introduction of biases and false negatives. Once the input nucleic acid (e.g., genomic DNA, reverse-transcribed RNA or cDNA, immunoprecipitated DNA) is obtained, the genetic material is fragmented by different methods, such as sonication, nebulization, or enzymatic digestions (5, 8, 112). These fragments are then ligated to platform-specific oligonucleotide adapters to create a library of overlapping sequences (7). The library is then hybridized to beads or a flow cell, which is followed by clonal amplification, such as emulsion PCR or bridge amplification. Not all platforms require the clonal amplification phase or preparation of a DNA library (8). Enrichment procedures can also be completed at this stage to help select for a specific type of DNA if an organism is suspected. Depending upon the NGS platform, the clonally amplified templates are sequenced by various chemistries, such as pyrosequencing, reversible dye terminators, oligonucleotide probe ligation, and phospholinked fluorescent nucleotides. Sequence data are then analyzed to determine the composition of the DNA sequences for pathogen identification. The final components of the NGS workflow are data release and dissemination of a clinically actionable report. If the patient's sequencing data would need to be reexamined, appropriate NGS analysis files should be stored or archived on- or off-site with patient privacy/confidentiality upheld (7, 10, 25, 28).

2a. Sample to Result: NGS Turnaround Times

Rapid diagnostic testing on direct clinical samples has been documented using PCR- and MALDI-TOF MS-specific methods (123, 138). In addition, BioFire technologies can be applied directly to a clinical sample, eliminating the need for isolated and purified microorganisms. BioFire panels are equipped with the coveted “easy button,” which allows for “sample in, answer out” testing. PCR-ESI/ MS is a technology that can also be applied directly to a clinical specimen. Initially, the application of NGS required the isolation of a pure bacterial species from culture in order to deliver key diagnostic information; however, many groups ventured away from this approach and explored the use of NGS for direct detection from clinical samples (29). If feasible, direct clinical specimen sequencing or metagenomic sequencing could theoretically reduce turnaround times from days or weeks to only a few hours, making NGS a procedure that could be completed within an average clinical laboratory workday (Figure 7). Although studies that applied NGS to crude clinical specimens have been published, there are still hurdles, including contaminating normal human microbiota and low-copy-number pathogens that require further evaluation (29, 83, 95, 139, 140, 141). Because NGS technologies sequence both viable and nonviable organisms in a sample, more efforts are needed to establish a normal baseline versus contaminants versus infectious agents. Moreover, the application of NGS to determine a clinical answer does not require knowledge of the infecting pathogen(s). NGS can establish a cause of infection and provide a potential answer in cases where other technologies may not provide an actionable finding (21, 36, 70, 71, 96, 142).

Figure 7.

Principals of current processing of bacterial pathogens in the clinical microbiology laboratory.

In principle, NGS data should not only detect the invading pathogen but also predict phenotypic resistance through the examination of genetic determinants of AMR. Traditional antimicrobial susceptibility testing requires an extra day of laboratory workup, extending the turnaround time to 3 or more days. Genetic information obtained from an NGS approach would ideally prompt rapid antibiotic treatment decision-making for the clinician and the patient. However, NGS is not at the stage where phenotypic susceptibility data can be extracted regularly or reliably from the genotype, although recent works of Pecora et al (143) and Tyson et al (144) are optimistic about this possibility. Genotypic data do not necessarily correlate to a clinical phenotype, and some types of AMR have nothing to do with genotype, such as intrinsically resistant microorganisms. For example, most Gram-negative rod bacteria are intrinsically resistant to vancomycin because the large-polypeptide antibiotic cannot penetrate the outer membranes of these organisms. Furthermore, detection of antibiotic resistance genes in some organisms, such as mecA in methicillin-resistant Staphylococcus aureus (MRSA), is more reliable than resistance gene detection in more challenging organisms, such as Pseudomonas aeruginosa. Hence, standardized growth-based susceptibility testing and perhaps newer, rapid phenotypic testing methods will likely be necessary to confirm an NGS result in the foreseeable future. A transcriptome-proteome combination could also assist in extrapolating the genotypic and phenotypic connections. Overall, NGS can provide more information than is achievable by other methods.

Sequencing run times continue to progressively decrease as technology evolves. The first bacterial genome to be fully sequenced using Sanger sequencing was Haemophilus influenzae, which required more than 1 year to complete at an estimated cost of $1 million (9, 24, 25, 31, 57, 145, 146). Now, depending on the platform used, an entire bacterial genome can be sequenced in less than 1 day at a cost of less than $100 (10, 31, 39). However, compared to existing molecular diagnostic assays that can take minutes to a few hours, current NGS tests are tremendously slow. Although NGS technologies have longer run times, the trade-off is the comprehensive genomic data that are produced. With this barrage of sequencing data comes the challenge of elucidating tangible information that would be desired by the clinician.

2b. Sensitivity and Specificity Differences between NGS and Other Available Technologies

Just as there are differences in the volume of genomic data generated, there are also differences in the quality of NGS data that are generated from the various platforms. The same confidence in base calling cannot be applied to all technologies. Performance metrics such as read length, accuracy, and sequence output (coverage) vary between platforms and dictate the type of applications that can be performed. Laboratories will have different uses for these platforms, and therefore a comparison of the base calling and error rates among the different NGS systems is not feasible or useful. Because there are systematic biases between NGS platforms, it is difficult to assess the difference in specificity and sensitivity of the platforms in their abilities to detect etiological agents when compared to the detection abilities of other DNA technologies. With each successive generation of sequencing platforms, performance continues to improve along with overall sensitivity. Better-targeted enrichment procedures are needed, and elimination of contaminating host DNA is critical for extracting higher-quality DNA. In general, there is a higher error rate associated with NGS than with traditional Sanger sequencing, mainly due to ambiguity in short read sequence alignment, resulting in inadequate coverage (14, 147). Plentiful short reads are generated, and to accurately assemble a sequence, analyzers rely on redundancy of sequence coverage from the short read structures, which can be problematic in areas with sequence repeats. Despite the huge potential of NGS, it is important to understand that the coverage and accuracy are not 100% and might lead to false positives, false negatives, and the misidentification of indels (112). Regardless of the flaws of NGS, it does have a higher analytical sensitivity for detecting low-frequency mutations than Sanger sequencing, whose sensitivity of detection is only 17 to 25% (148, 149). NGS technology is still maturing, and with additional fine-tuning of sequencing chemistries, optics, and processing algorithms, its accuracy will improve.

3a. Use of NGS in Outbreak Scenarios

Despite concerted efforts to improve infection control practices, outbreaks can occur in the hospital setting. With the implementation of NGS, outbreak isolates may be sequenced in real time and guide infection control efforts and antibiotic stewardship (56). Vital to containing outbreaks is fast identification of the infecting pathogen and contaminated source (24). [See Case Study 8 for a description of a MRSA hospital outbreak in a special-care baby unit and how NGS was successfully applied to determine the source (154).] Because NGS holds promise for faster pathogen detection than traditional laboratory workup, this opens the opportunity for quicker intervention strategies, such as patient isolation, contact precautions, or decolonization. With the ability to extract vital pathogen information, NGS has direct clinical value and can contribute to individual and improved patient management through a more accurate diagnosis and tailored therapy. Ideally, these factors would decrease hospital stays, deter the development of AMR, and decrease mortality rates. NGS can also help promote infection control accountability within and between hospitals and thus help curtail the spread and infection rates of high-risk organisms in health care settings (15, 97, 154, 155). In an increasingly connected world, one could ideally determine if the same bacterial strain was responsible for an outbreak in geographically distinct health care facilities. A downfall of NGS is the lack of standardized guidelines or models endorsed by regulatory agencies for responsible data sharing. There is also hesitation from hospitals due to privacy issues that make open data sharing diffi-cult. If the data are siloed and not shared, there will be limited public health benefit to the use of NGS in the clinic. Besides demonstrating transmission pathways of hospital-acquired pathogens, NGS can provide insight into how bacterial populations respond to drug treatment. Since NGS provides an in-depth interrogation of genomic sequences, genetic variation associated with bacterial virulence or antibiotic resistance could be determined (10, 28, 33, 156). Identifying these key biomarkers is advantageous for infection control teams to enhance their preparedness for current, emerging, or predicted infectious threats (6, 24, 97, 157).

Case Study 8

WGS used to monitor the transmission of MRSA between the community and hospital (154).

NGS can not only help monitor, detect, and prevent hospital outbreaks, it can also initiate appropriate public health interventions when outbreaks occur in the community and abroad (158). [Case Study 9 showcases the discriminatory power of NGS over pulsed-field gel electrophoresis (PFGE) in tracking a salmonellosis outbreak that afflicted more than 44 states (157).] For example, NGS can provide insights into the source and mode of pathogen transmission associated with a foodborne disease outbreak and can subsequently stimulate a proper public health response in the community, such as a food recall (70, 159, 160, 161). Current bacterial typing methods used to examine isolates involved in outbreak situations include PFGE, multilocus sequence typing (MLST), and multilocus variable-number tandem-repeat analyses (MLVA). Roughly one-third of all outbreaks in the United States do not have a determined source because of the low resolution of PFGE, the gold standard method used for trace backs and source tracking (162, 163). These analyses target only selected regions of the microbe's genome, and thus sequencing a whole genome provides superior resolution and detailed linkages for isolates responsible for outbreaks (159, 164). Essentially, NGS provides confidence in the matches among relatively few isolates across time and space, which allows regulators to intervene early and provide links from past isolates from known foods and environmental samples (61, 158).

Case Study 9

Tracking of a salmonellosis outbreak source using NGS (157).

Outbreak investigators can unambiguously evaluate the relatedness of isolates, and with this information combined with an estimated mutation rate, their evolution from a common ancestor can be calculated (33). The isolate's evolution time and epidemiological data such as the patient's admittance date to the hospital can help predict if a transmission event occurred, which can then direct better targeting of infection control resources (144). NGS has the capacity to identify micro-evolutionary differences among clinical outbreak isolates and essentially “rule in” or “rule out” the links between isolates that would otherwise be indistinguishable by existing approaches (33, 143, 156, 162, 163, 165). Extensive mining of NGS data might reveal novel infectious agents and/or new targets to help bolster outbreak investigations involving highly clonal pathogens. Taken together, the rapid and open release of genomic data has the power to transform our response to outbreaks, especially when data are shared globally and in real time as with the GenomeTrakr project housed at the National Center for Biotechnology Information (NCBI, NLM, NIH) (http://www.ncbi.nlm.nih.gov/bioproject/183844) (105, 158, 166, 167).

3b. Impact of NGS on Metagenomics and Human Microbiome Studies

NGS also impacts human microbiome studies by characterizing the microbes that reside on the outer and inner surfaces of our bodies (168, 169). It is estimated that 90% of the human microbiome cannot be cultured using current technology (31, 56). 16S rRNA metagenomic sequencing, shotgun metagenomic sequencing, and microbial meta-transcriptomics are NGS methods that have the potential to investigate the human microbiome to determine the wealth of microbes that live with us either symbiotically or antagonistically (47). These NGS methods have the potential to evaluate the intrapersonal diversity of culturable and unculturable microorganisms and how their compositions change with various situations or diseases, such as admittance to the hospital, receiving a transplant (e.g., solid organ, bone marrow, fecal), and vaccination (47, 158, 170). The gene content of the microbes that inhabit specific human niches is approximately 100-fold higher than that of the human genome, encoding various physiological and biochemical functions that could contribute to health or disease states (13). If alterations observed in the microbiome are predictive of disease, perhaps interventions to prevent the onset of disease could occur (46). This available information could be distilled and, in turn, stimulate the development of new in vitro diagnostic devices such as a multiplex PCR assay. Although this technology has a multitude of benefits, one concern is the overaggressive interpretation of NGS data, keeping in mind that not every species detected is clinically significant. In general, interpreting microbiome studies is challenging because a standardized methodology has not been established.

Beyond the metagenomic studies of the human body, NGS can be used to research the microbial composition and complex biodiversity of different environments, such as soil, mines, oceans, built environments, and others. A minute fraction of microorganisms isolated from the environment have been able to grow successfully in pure culture. Thus, we lack a true understanding of the diversity that exists on Earth. Metagenomic NGS studies have the potential to discover unknown organisms, etiological agents, and potentially new and original pathogens (171, 172). For metagenomic investigations, DNA is extracted from environmental samples and direct NGS is applied, eliminating the need to culture the organisms beforehand (172). With the hopes of using NGS to discover novel species and subspecies, it is likely that a reevaluation of current taxonomic classification will need to occur (6). An examination of microbial populations and diversity in both clinical and environmental isolates can be performed faster and cheaper with NGS.

4a. Cost of NGS Technology

Start-up costs and the necessary infrastructure (i.e., storage and retrieval services) to operate an NGS platform constitute a substantial in-house investment that can range from $80,000 to $10 million (5, 6, 24, 38, 173). These prices can easily extend beyond the reach of an average diagnostic laboratory. With the advent of benchtop NGS instrumentation, these sequencers will likely be a better fit for sequencing in the clinical laboratory environment and individual research labs, with their reduced platform size and cost (7, 25, 30). Sequencing systems such as the Illumina MiSeq and the Life Technologies Ion Torrent PGM deliver high-throughput sequencing on the benchtop. However, informatics constitutes a major portion of the NGS cost and is highly dependent on the throughput of the instrument. Sequencing platforms with higher-throughput abilities generally require large Unix-based servers with multiple, high-speed core processors and maintenance by trained information technology professionals.

Smaller sequencing platforms with lower throughput can function on high-powered Windows or Unix-based desktop systems (147).

The cost of performing sequencing has decreased dramatically over the past 15 years and has even surpassed Moore's law (10, 174) (Figure 9). In contrast, bioinformatics remain a costly component of the analysis process. If an “easy button” or automated data analysis could be designed, then it is anticipated that bioinformatics costs will decline. There are also operational expenses associated with NGS instrumentation, including service contracts for machine maintenance, consumables for library preparation and sequencing, and the labor cost of the medical technologist performing the test (6). Ancillary equipment or robotics could be used to run various applications or to streamline processes, which can be major unforeseen costs (38). For example, it is highly recommended that the integrity and quantity of the DNA be evaluated following the library preparation, such as with a bioanalyzer and fluorometer, respectively. Devices used for enrichment procedures may also be vital for completion of an NGS assay and need to be factored into the total costs. Therefore, users need to consider the whole process, from sample to sequence interpretation, and the equipment necessary to maximize throughput and prevent sample backlogs (100). Despite these additional expenses, the sequencing cost per base is still significantly lower than that of Sanger sequencing (39). As competition among sequencing vendors intensifies, it is predicted that the price of sequencers, consumables, and analysis will steadily decline. NGS remains a relatively expensive technology compared to other diagnostic assays such as MALDI-TOF MS, which can be performed at a lower cost.

Figure 9.

Dramatic decreases in the cost per megabase (Mb) of sequenced DNA compared to the expectation predicted by Moore's law.

With appropriate scale and efficiency, the U.S. government instituted a goal for the sequencing industry to obtain a $1,000 human genome, which has now been accomplished with select sequencing platforms that are currently available (1, 2, 3, 10, 31). An average bacterial genome consists of approximately 2.8 million nucleotides. This size is less than one thousandth the size of a (haploid) human genome, which is composed of nearly 3 billion nucleotides (7, 9, 33). It is therefore predicted that a whole bacterial genome could be sequenced for $1 in the not too distant future (38, 174). Bacterial genomes are considerably smaller than eukaryotic genomes and house only one haploid chromosome. Despite a smaller genomic content, bacterial species are extremely diverse, with 10% to 40% of their genomes being dispensable due to mobile and accessory elements such as plasmids and transposons (15), which can be gained or lost over time due to selective pressures. In addition to the range of target genome sizes, factors such as genome complexity (e.g., high GC content, homopolymer regions), the anticipated level of completeness, and the availability of a reference strain for comparison could also possibly influence the price of a genome sequencing project as special protocols and/or additional technician time may be required.

NGS must appeal to hospital microbiology laboratories and public health laboratories, with cost being a significant factor impeding its widespread adoption. Depending on the volume of isolates encountered daily in a hospital laboratory, costs can be saved by multiplexing, indexing, or batching isolates with barcodes rather than performing sequencing on individual isolates as they come into the laboratory (6, 38, 43, 175). A unique index “bar-code” sequence of 8 to 12 nucleotides that usually makes up part of a platform-specific adapter is added to each DNA fragment created during library preparation and is linked to the organism that is to be sequenced. These indexes allow for downstream in silico sorting of the sequences, enabling multiple isolates to be sequenced in one run, which maximizes investment in the instrument and is economically sound. However, with barcoding, there are problems with false positives, cross-contamination, and indice bleeding with off-the-shelf methods.

Nonetheless, there are differences in opinion as to how this technology should be marketed by the sequencing companies. It narrows down to a cost-versus-value analysis. From one perspective, NGS technologies offer a cost-effective catch-all for the diagnosis of infectious diseases. Operationally, NGS is cost efficient, because ideally no other tests would need to be performed once the assay has been optimized and validated. However, there are formidable hurdles to the development of an optimized, analytically and clinically relevant NGS assay. From a different perspective, NGS can be perceived as an expensive test but a value-added test that would fill the gaps from the diagnostics already conducted. With this viewpoint in mind, NGS is meant to enhance the standard of care when a clinical need is identified and results from other tests are elusive.

4b. Standard Operating Procedures for NGS Workflow

Genomic sequencing data generated in the clinical microbiology laboratory need to ultimately translate into meaningful information for patient care and intervention. In order to incorporate NGS into the standard of care, there is a need for consensus practice guidelines for different types of infections. Different protocols for both the wet and dry laboratory components of the process that are specific for the detection of etiological agents, including bacteria, viruses, fungi, yeasts, and parasites, should be created. All protocols must be optimized and internally validated prior to clinical implementation. During validation, specifications for essential performance characteristics (see below) need to be established (14, 37, 39, 176).

Before NGS is used routinely by the broader clinical and public health laboratory community, it is anticipated that NGS will first be adopted by larger academic medical centers and reference laboratories that have the significant financial support and computational infrastructure required for the use of NGS. Furthermore, it is expected that the first applications of NGS used regularly in the clinical setting will be limited to a small number of targeted assays and panels that are validated to perform those specific applications. The development of turnkey solutions, e.g., workflows that include automated sample preparation, standardized bioinformatics pipelines, and market authorization from FDA, would help promote the adoption of NGS in the clinical microbiology and public health laboratory space. NGS assays provide the user with large amounts of data that require analysis by clinical laboratory professionals to identify and prioritize the findings that are clinically relevant. The development of fully automated sequence interpretation tools would allow for an easier transition into the diagnostic laboratory and could reduce bioinformatics analysis costs (7, 10, 28, 159, 160).

Standardization of NGS protocols and quality metrics is critical to ensure the validity of NGS test results (37, 39, 121). NGS consists of two major processes, the analytical “wet bench” component and the bioinformatics or “dry bench” component. Both processes require separate considerations for the development of standards (39). Any clinical laboratory test should establish analytical performance specifications (i.e., assay validation) of the assay for performance characteristics, which include accuracy, precision, analytic sensitivity and specificity, reproducibility/repeatability, and reportable range (limit of detection) (37, 43, 176). NGS-specific quality control (QC) metrics (e.g., depth of coverage, quality scores, etc.) should be established during test validation and be used as part of the standard operating procedures (SOPs) to monitor the performance of the assay and ensure accurate results (37, 177). In addition, laboratories developing high-complexity tests, like clinical applications that use NGS, should ensure that the tests are appropriate for the intended clinical use (i.e., establishing clinical validity). Quality assurance guidelines and standards for this technology need to be developed and continually refined with insight from clinical microbiologists (15). Figure 10 emphasizes the key considerations in quality assessment for whole-genome NGS analyses (9).

Figure 10.

Quality assessment considerations for whole-genome NGS analyses.

There are a variety of NGS instruments, applications, sample preparation kits, and approaches to data analysis and result interpretation available; therefore, SOPs will not be the same for each lab using NGS. Nevertheless, clinical laboratories must assess their ability to detect analytes of interest and provide an independent verification of test performance compared to that of other laboratories that use the same or different methods via proficiency testing or external quality assessment exercises. This requires the development of well-characterized proficiency testing materials that are suitable for the diverse range of NGS applications, as well as standard metrics to use as a comparator among laboratories (37). However, there is no current, formal proficiency testing program in place for NGS as applied in the clinical microbiology and public health setting. There is a need for the development of reference materials, such as well-characterized microbial genomic DNA, spiked matrices, and data sets, so that laboratories can run these reference materials for assay validation, performance assessment, and quality control. NIST (National Institute of Standards and Technology) has been granted FDA funding to develop standard microbial reference materials to be used when validating genomic sequencing diagnostics (111, 179). An international, inter-laboratory study involving five national metrology institutes assessed the variability in sequencing of the 16S rRNA gene from two certified genomic DNA references using different sequencing platforms. Participants were instructed to use the methods and protocols available in their laboratories. Although all data sets concurred for the biologically conserved regions, data sets varied in both read length and coverage, which impacted precision when examining the biologically variable positions of the 16S rRNA genes. Biases were also introduced by the individual laboratory's algorithm that aligned the reads to the reference genome (180). Therefore, addressing these inconsistency and reproducibility issues is critical before NGS can make its transition into the clinical laboratory. Allowances/provisions must be made so that SOPs for NGS technologies can evolve over time. Validation and proficiency testing is in its infancy for WGS, but some laboratories have shown reproducibility for resequencing pure culture isolates (158). As new devices and assays are marketed, there needs to be some malleability in proficiency testing procedures or a mechanism in place to change these operations in a relatively alacritous fashion. While it is anticipated that NGS technologies and applications will continue to advance and change over time, the quality management requirements for assay validation, quality control, and proficiency testing or external quality assessment will maintain a high standard (37).

Stakeholders from various sectors should collaborate to advance NGS guidance and standards. For example, the U.S. Centers for Disease Control and Prevention initiated the Next-Generation Sequencing: Standardization of Clinical Testing (Nex-StoCT I and Nex-StoCT II) working groups to develop guidelines for implementing NGS into clinical settings. The first of these guidelines concentrated on four main topics, including test validation, quality control, strategies for proficiency testing among laboratories, and the development and use of reference materials (37). The Nex-StoCT II informatics working group developed recommendations for the development and optimization of a clinical NGS bioinformatics analysis pipeline (181). Although the primary focus of the Nex-StoCT working groups was on human genetic NGS applications, many of the recommendations are appropriate for use in infectious disease testing and diagnosis. Furthermore, College of American Pathologists (CAP) formed an ad hoc committee referred to as the “NGS Work Group,” which developed an 18-item accreditation checklist for NGS standardization in the clinical laboratory (39). A new International Organization for Standardization (ISO) genomics committee has been constituted recently, and global proficiency testing is being conducted by the Global Microbial Identifier (GMI) network (182). Having standards that are published and available for new pipelines and instrumentation is critical to assist laboratories with assay development and to maximize the positive impact NGS can have on clinical genomics and metagenomics.

4c. Formulation of a Team of Specialists for Clinical Translation of NGS Data

In order to shift NGS into the clinical arena, there needs to be open communication and partnerships among clinicians, clinical microbiologists, bioinformaticians, NGS platform technology experts, and software designers (10, 28, 173). Having different areas of expertise, these individuals would have distinct responsibilities, priorities, and ways by which they view and interpret NGS data. There is a pressing need to incorporate this technology into clinical fellowship training programs and pathology training programs. For example, with ASM's CPEP (Committee on Postgraduate Educational Programs) fellowship opportunities, students could be educated in both computational biology and clinical microbiology languages. Washington University School of Medicine has already implemented a Molecular Genetic Pathology fellowship program which functions as a collaborative effort among the Divisions of Anatomic and Molecular Pathology, Laboratory and Genomic Medicine, and Neuropathology. Many medical schools, such as the University of California—San Francisco, provide students with a medical and microbiological informatics lecture series to further prepare them for this era of genomic medicine. A common vernacular or lexicon encompassing both fields needs to be established for the next generation of clinical microbiologists, pathologists, and physicians (1, 31, 121). Since bench-level medical technologists perform the bulk of the work in the clinical laboratory, training opportunities for these individuals must be made available as well. ASCP (American Society for Clinical Pathology) could offer bioinformatics certification exams like those presented for other subspecialties in order to spark interest. Master's level programs in NGS could be designed specifically for medical technologists in order to have adequately trained bench personnel for when NGS technology becomes a routine part of the clinical microbiology laboratory workflow.

Before NGS can reach its full potential in the clinical microbiology laboratory, validation and educational challenges need to be overcome. NGS technologies are not yet at the point where they simplify or accelerate the traditional, multi-step workflow that is currently carried out in the clinical microbiology laboratory. In the initial phases of clinical use, NGS data analysis and reporting will require a holistic combinatorial approach as depicted here.

5a. Software Choices for NGS Analysis

NGS platforms produce massive volumes of data at a high rate, a feature that is both a strength and a limitation of these technologies at the current time. Terabytes of data are typically generated from a single run. New analytic approaches that can mine NGS data sets cannot keep pace with the amount of data produced (5, 110, 183). As displayed in Figure 11, NGS has shifted the sequencing workflow from less upstream sample preparation time to substantial downstream data analysis and management efforts (16). There is a need for advanced, efficient, and user-friendly computational architecture for NGS data analysis (6, 24, 110). Bioinformatics software is often freely available for download or can be purchased from commercial companies or designed by the user (9, 43). See Figure 12 and supplemental tables in studies by Voelkerding et al (5) and Pabinger et al (184) for common examples of NGS software tools used for WGS and other applications. For a more detailed review of diagnostic software tools, see the article by Fricke and Rasko (185). This paper contains a list of seminal publications on bacterial diagnostic applications of WGS along with the analysis method used. The use of NGS analysis programs can be completed in-house within the microbiology laboratory, provided that the necessary hardware infrastructure and training are supplied. Analysis can also occur remotely with cloud-computing services where users rent computing power to evaluate NGS data (16). It is important to acknowledge that software for NGS interpretation is currently a moving target and will continue to change quickly, given that there are very few standard analysis practices at the moment.

Figure 11.

Contribution of different factors to the overall cost of a sequencing project across time.

Figure 12.

Examples of tools used for NGS bioinformatic analysis.

Developed at the University of California—San Francisco (UCSF), sequence-based ultrarapid pathogen identification (SURPI) is a computational pipeline that can detect any pathogen in the GenBank reference database. This software program can be used on stand-alone and cloud-based servers. SURPI can provide a comprehensive identification of bacteria, viruses, fungi, yeasts, and parasites in a clinical sample within a clinically relevant time frame and is capable of diagnosing critically ill patients (21, 96). Illumina has also designed a suite of bioinformatics software that is integrated with each Illumina sequencer. BaseSpace is Illumina's genomic computing environment that carries out many NGS analysis methods, including WGS, de novo assembly, 16S metagenomics, RNA-seq, whole exome/enrichment, and targeted amplicon sequencing. NGS data storage, management, and sharing either through a cloud or on-site are other key features offered by BaseSpace (186). Furthermore, Torrent Suite™ is the data analysis software package that can be integrated with the Ion Torrent platforms. The Ion Reporter software is a component of the Torrent Suite™ that helps to streamline the informatics process by performing primary and secondary analysis and data transfer to a secure Amazon cloud (187).

Ultimately, the choice or design of software and computational analysis pipeline should align with the needs of the individual laboratory and the specific NGS application. Software programs can yield the same quality metrics but will likely differ in performance and potentially results. It is important to maintain records of software versions, dates of use, and changes to the software along with the parameter files used in the analysis phase to generate a complete audit trail (173, 188).

5b. Pipeline for NGS Interpretation

NGS data analysis is not a trivial task and is critical to any sequencing experiment (175). In the clinical laboratory, analysis will equate to a clinically actionable result that has the ability to impact patient care (9, 27, 110). There are three broad categories of NGS analysis tools, including base calling, alignment of the sequence to a reference sequence/ database or de novo assembly, and annotation and variant detection (Figure 13). However, the data analysis process can be directed onto many different paths depending on the needs of the end user. In the microbiological laboratory, steps like full-genome assembly and annotation are not necessarily needed for clinically actionable data (111). Each step of NGS analysis employs a different tool or algorithm, which makes for a lengthy process. There are even differences in algorithms among platforms designed by the same company, such as the Illumina NextSeq versus MiSeq versus HiSeq.

Figure 13.

Outline of informatics pipeline for processing and analyzing data from massively parallel sequencing platforms.

For nontechnical laboratory staff and health care professionals with no bioinformatics training, these tools are complicated, and thus results can remain cryptic without proper education. This issue has to be addressed before NGS technologies can be used routinely in the clinical laboratory.

Once the sequencing phase of the process is finished, raw sequencing reads are subjected to multiple steps beginning with preprocessing to remove adapter and low-quality sequences (38). Other analysis steps include base calling, assembly of the genome, alignment to known and curated genomic sequences, and annotation/comparative phylogenetics. Base calling is initially performed by proprietary software on the platform to account for technology-associated bias (1, 40). Each base call is assigned a level of certainty, similar to a Phred-like score that correlates to its predicted accuracy. Indications of mixed signals or other read errors are filtered or removed from analysis. The quality scores generated from each platform cannot be directly compared, because each platform uses a different algorithm (112). Following base calling, contigs are created by overlapping the numerous short reads (30 to 500 bp depending on the platform) or long reads (>5,000 bp) generated from the millions of sequencing read fragments of the target genome (9). Longer contigs provide a higher depth of sequence coverage that improves the accuracy and sensitivity of detecting pathogens (33). Bioinformatics software will align and compare the contigs against an available database of high-quality sequenced organisms (e.g., Kraken, SURPI, GOTTCHA, OneCodex, CLARK, LMAT, MG-RAST, MetaPhlan, MEGAN). The preferred program for assembling a genome depends on how the data are produced, either long or short read sequencing. For a more exhaustive list of genome assembly programs, see references 15, 189, and 190.

Genomes without a close relative in the reference database may not be detected using reference-based methods, necessitating de novo assembly for newly discovered genomes (15, 25, 38). De novo assembly is also useful for the discovery of mobile accessory elements, such as plasmids and phages, which may not be present in the reference genomes. Genomes that are assembled de novo do not use an existing reference scaffold and can be assembled by popular programs such as SPAdes, IDBA-UD, SOAPdenovo, MIRA (which can also be used for reference guided assembly), and Velvet (7, 15, 38, 190, 191, 192). Producing long overlapping reads is essential for de novo assembly since data are not aligned to a reference genome, and choice of assembly software is technology dependent. It is important to understand that a reference database is still necessary to accurately identify the organism even if de novo assembly is used. This may seem counterintuitive at first, but comparing long contigs against a reference database assists in the identification process and discovery of novel variants and genes.

After a set of ordered contigs, if not a whole genome, is obtained, annotation of the draft genome occurs. Most commonly used for bacterial genome annotation are Web-based tools such as the annotation tools RAST and Glimmer (38, 193). Ordered contigs are uploaded to RAST, and open reading frames that most likely represent genes are produced to create a highly annotated genome. The quality of the annotation process is largely dependent on the gene database that is applied. Besides RAST and Glimmer, there are other annotation tools available for users, including DIYA, Prokka, RATT, and BG-7. Additional information on annotation tools is presented in references 194 and 195. With annotated genomes, users are often interested in finding genes of biological importance, such as genes conferring antibiotic resistance and virulence. Furthermore, comparative genome analysis is frequently exploited to align multiple genomes to determine regions of sequence homology or to identify unique regions of the genomes. Examples of software tools that enable annotated genomes to be compared and visualized include BRIG, Mauve, and ACT. Some of these alignment programs such as Mauve can identify SNPs that are useful in phylogenetic analyses and for establishing evolutionary relationships, tasks that are more suitable for public health surveillance and epidemiological studies than for the clinical laboratory (61, 193).

Because unbiased NGS generates sequence data for all types of DNA found in a clinical sample, the data can include organisms that do not prompt immediate action, such as normal microbiota components. With current microbiology practices, and more specifically, culture-based identification, the laboratory will report not only the etiological agent but also such commensal organisms. For example, the laboratory will report “normal respiratory microbiota” for sputum samples and “mixed microbiota” for urine specimens. In parallel to this approach, a result from an NGS-based test that is sent in a final report to the clinician should be similar and include normal microbiota if detected. Nonetheless, there is some skepticism associated with this tactic. Reporting all organisms could inadvertently introduce new organisms that are not necessarily actionable but could be acted upon by certain groups of physicians. This type of reporting could have a detrimental effect on antimicrobial stewardship moving forward, and therefore irrelevant information should be filtered before the report reaches the hands of the physician. For the immediate future, any prediction coming out of NGS and bioinformatics will likely be followed up by real phenotypic tests or some other confirma-tory evidence to verify the prediction.

5c. Determining Clinical Relevancy of NGS by Distillation of Data: Integration into Care

To extrapolate clinical relevancy in a time-efficient manner, an ongoing dialogue between clinicians, clinical microbiologists, and the developers of NGS software is necessary. There should be a gradual and logical transition in the type of data transmitted to the clinician, from language with which they are familiar based on traditional tests and assays toward more complex analyses as training improves and as this dialogue continues. The tremendous amount of NGS output must be condensed into a reportable result, with unnecessary information removed or strongly qualified (6). There will still be uncertainty in the results generated from NGS, even with advancements in the technology and bioinformatics pipelines. The complexity of the data is substantial, and thus it is easy to cherry-pick an answer and make a false conclusion. Finding the most optimal way to communicate an uncertain result is important to relay to the clinician.

NGS methods can both detect and identify infectious agents in a single run. Ideally, information regarding antibiotic resistance mechanisms and virulence determinants should also be derived from the same NGS assay without having to perform additional testing (30). Predicting antibiotic resistance from genetic data is still in the early stages of development, with limited cases showing the direct genotypic-phenotypic link (10, 15, 27, 28, 30, 33, 143, 144, 196, 100). One study performed by Gordon et al (198) highlighted a genomic prediction tool for determining antimicrobial susceptibilities for Staphylococcus aureus. This genotypic tool was comparable to antimicrobial susceptibility testing methods with regards to specificity and sensitivity (198). Antibiotic resistance genes may hold predictive value; however, the genetic mechanisms responsible for a resistant phenotype are not always clear-cut. NGS can determine antibiotic resistance when known resistance genes or mutations are identified but cannot reliably demonstrate resistance when unknown or novel mechanisms of resistance are present in the organism (188). Antimicrobial resistance remains one of the three greatest threats to human health according to the World Health Organization (WHO) (199, 200, 201, 202, 203). Therefore, it is essential to detect resistance quickly and reliably, and in this aspect, NGS shows promise. Assessing virulence gene content from NGS data is also in the preliminary stages of development (30). In the foreseeable future, NGS will be a preferred method for studying virulence and antibiotic resistance mechanisms with the hope of generating knowledge to further diagnostics, therapeutics, and vaccine design (Figure 14) (1, 9, 28, 31, 56, 156, 164, 204).

Figure 14.

Use of microbial genomics for tool development.

5d. Storage of NGS Data and Security Concerns

Another principle challenge of NGS is downstream data management, storage, and sharing. Long-term storage of raw NGS data reads is not reasonable because of the large size of the data files (43). For example, storage of the reads for a single Staphylococcus aureus genome requires approximately 1 GB of space, a size that is comparable to the storage of 10 music albums in MP3 format (33). The 1 GB of storage space for the S. aureus genome is uncompressed, and therefore if the data were compressed to just the variants with respect to the reference (e.g., CRAM), the storage size would be much less. Following base calling, sequence data should be stored in platform-specific output files such as .FASTQ files. Because each NGS platform delivers data in a slightly different format, this can cause difficulties when trying to compare data from various platforms and when uploading data into databases with multiple formats (38). A possible solution is to leverage cloud-based or public repositories, such as the NCBI Sequence Read Archive (SRA), which use standardized formats and offload storage costs. When regulations prohibit off-site storage, raw data and sample metadata (information on the clinical isolate) can be stored locally in a compressed format or replaced with a genome assembly or report of detected genes and variants. In fact, the NCBI, along with the FDA, CDC, and USDA, has already established a template (known as BioSample) for minimal information for pathogen metadata which could at least serve as a starting point for discussions of other organisms or for clinically relevant publicly available metadata (see https://submit.ncbi.nlm.nih.gov/biosample/template/?package=Pathogen.combined.1.0&action=definition). It should be emphasized that when the raw sequencing reads are submitted to the public archive, previous downstream analysis can be independently tested with new scientific algorithms, software, and protocols. Therefore, public repositories could be tools for the evaluation of new scientific understandings. Another information technology infrastructure demand includes long-term data archives, backup storage solutions, and retrieval facilities that would permit traceability to a patient's report (10). An important archival note involves recording the software version, database version, and analysis parameters so that if the data require reanalysis, users will know what was applied previously.

Because data generated from an NGS test is associated with a specific patient, laboratories must ensure that proper measures are in place to uphold patient confidentiality and privacy when these data are internally and externally stored or transferred (129). Consequently, laboratories must maintain the standards required by the Health Insurance Portability and Accountability Act (HIPAA). Because in-house NGS computational infrastructure requires significant financial resources, a popular alternative is cloud computing (7, 38). These services are offered at a reasonable price by technology companies such as Google, Amazon, and Microsoft and are easily scalable depending on the sequencing needs (10). Analysis of data can occur in the cloud, and data can be shared through this platform, although this may be impaired by bandwidth and data transfer capabilities. However, storing patient data in the cloud is not permitted by some institutions, and compliance with HIPAA regulations varies with provider. Rigorous policies regarding the sending and receipt of a patient's clinical genomic data to cloud-based vendors or other health care institutions need to be enforced. Examples of these safety procedures include data encryption, secure data transfer, user authentication, and transmission tracking pathways (39).

6a. Regulatory Issues of an NGS Diagnostic Test

When any new technology, device, or method is implemented into a specific environment, there are challenges that have to be overcome. Incorporation of NGS technology in the clinical laboratory is no exception and has its own set of obstacles (37). In 1976, Congress enacted the Medical Device Amendments, which amended the Federal Food, Drug, and Cosmetic Act (the FD&C Act) to create a comprehensive system for the regulation of devices intended for use in humans. Under section 201(h) of the FD&C Act, “an instrument, apparatus, implement, machine, contrivance, implant, in vitro reagent, or other similar or related article, including any component, part, or accessory…intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease, in man or other animals” is defined as a device. Under the FD&C Act, the U.S. Food and Drug Administration (FDA) ensures that medical devices, including in vitro¹ diagnostic (IVD) devices, provide a reasonable assurance of safety and effectiveness (176). The Center for Medicare and Medicaid Services (CMS) is responsible for regulating laboratories under the 1998 Clinical Laboratory Improvement Amendments (CLIA). CLIA governs the accreditation and certification process for laboratories (181, 207). In the United States, diagnostic tests, similar to other medical devices, undergo evaluation by the FDA to obtain marketing authorization. Determining the appropriate amount of regulatory oversight for NGS technology presents a new challenge. The FDA premarket review process has generally followed a “one test, one disease” paradigm that assesses a test's analytical and clinical performance. NGS technology is essentially an ultimate multiplex test that can detect many infectious agents.

In 2013, the FDA allowed marketing of the Illumina MiSeqDx instrument platform and the Illumina Universal kit, two devices that make up the first FDA-regulated test system that allows laboratories to develop and validate sequencing of any part of a patient's genome. This marketing authorization marked a significant milestone for this technology (96, 208). The FDA also cleared two devices that are used to detect DNA changes in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, which can result in cystic fibrosis (CF), an inherited chronic disease that affects the lungs, pancreas, liver, intestines, and other organs of those who inherit a faulty CFTR gene from both parents (209). These two cleared devices are: Illumina MiSeqDx Cystic Fibrosis 139-Variant Assay and Illumina MiSeqDx Cystic Fibrosis Clinical Sequencing Assay. The Illumina MiSeqDx has not been granted marketing authorization by the FDA for microbial diagnostic use (http://www.access-data.fda.gov/cdrh_docs/reviews/K123989.pdf). In addition, two other sequencing platforms were registered and listed with the FDA as in vitro diagnostic devices in 2014: Life Technologies Ion PGM and Vela Sentosa SQ301 (209, 210, 211). Taken together, establishing an FDA review process for an NGS platform or assay seeking marketing authorization that appropriately balances the interest in doing an adequate review against the interest in minimizing the information and effort involved in preparing a submission able to support a favorable decision has presented challenges, especially because the technology is changing at a rapid pace. By the time one test is cleared or approved, a new test will likely appear on the market that can perform that same test with a quicker turnaround time and at a lower cost. However, this situation is not unique to clinical microbial NGS technology.

The process of validating microbial NGS-based tests faces many obstacles. Establishment of clear validation standards by either national or international regulatory agencies could potentially help with the use of NGS tests in clinical laboratories. The standard reference materials being developed by NIST have the potential to address some of the challenges encountered during the validation process for microbial NGS-based tests (111). Similar to other medical devices, NGS testing is an inherently multi-analytical procedure for which the FDA, as part of its review, requires validation of all components of the system, including nucleic acid extraction, library preparation, sequence generation, sequence analysis, sequence interpretation, and formulation of a reportable result (7). This systems approach would benefit from continued collaboration from regulatory agencies such as the FDA and CMS as well as medical societies such as College of American Pathologists CAP. Another challenge to address is the regulatory and ethical issues that surface with the storage of patients' NGS test results. Such storage must adhere to privacy laws such as compliance with HIPAA requirements. With the capabilities of NGS methods when sequencing a human clinical specimen, it is also possible to uncover an unexpected finding such as a marker for an inheritable disease of which there are no guidelines for appropriate disclosure of such incidental data (212)).

6b. Recommendations for Overcoming Barriers for NGS Implementation

Reimbursement codes need to be standardized or new codes that cover genomics-driven interventions or NGS assays should be created (31). Some suggestions include “sequencing,” “microbiology genome diagnostics,” and “metagenomic sequencing.” It is important to engage insurance companies and inform them about the downstream cost savings of NGS analysis. Case studies and, ideally, clinical trials would be useful and encouraging tools that could be directed towards insurance companies. Success stories showing how an NGS test result decreased a patient's hospital stay or saved a patient's life because the pathogen was identified more quickly and therapy was adjusted appropriately could be highlighted (36). [See Case Study 10, which demonstrates how WGS traced a hospital outbreak of a Klebsiella pneumoniae (K. pneumoniae) carbapenemase(KPC)-producing organism and hence provided actionable insights to decrease further intrahospital transmission (97).] There needs to be an open dialogue between the scientific community and insurance companies as NGS progresses towards the infectious disease diagnostic space. Tactics to convey the value added of NGS to agencies or groups that deal with reimbursement need strengthening (10).

Case Study 10

WGS tracks a hospital outbreak of carbapenem-resistant K. pneumoniae (97).

There are simple, cheap, and routine methodologies that work efficiently in the clinical laboratory at the current time. It is questioned that if these established methods are efficient, how is NGS better than what we have now? NGS signifies a value-added method that has the capabilities of detecting all pathogens from one test, exemplifying its power to transform the clinical microbiology field and potentially eliminate the need for a suite of conventional tests. A major hurdle is getting the medical community and hospital administrators to embrace NGS technology and to realize the clinical value it possesses over current tests performed in the microbiology laboratory.

Because traditional microbiological methods remain ingrained in diagnostic laboratories, hesitation and fear exist on the part of medical technologists, clinicians, and hospital administrators to learn the technology and terminology, which represents another barrier that is preventing acceptance of NGS (28, 33). This is not just a new technology but a new paradigm for how we conduct clinical microbiology, which can make these groups reluctant to engage NGS. Even though NGS is becoming widely available, there are a lack of education and sparse training opportunities, which inhibit the use of this advanced sequencing method (6, 24, 121). Courses in bioinformatics and data analytics could begin at the undergraduate level, with continued training in graduate and professional schools to help mold a new generation of scientists and physicians. Short courses, workshops, or CME (continuing medical education) credit could stimulate interest in NGS and help lessen anxiety for implementation while strengthening users' competence. It would also be beneficial for professional practice organizations to articulate and implement NGS language and evidence-based guidelines to bolster familiarity with this powerful sequencing method.

Numerous benchmarks must be met in order for NGS technologies to be incorporated into the workflow of the clinical microbiology laboratory. Many hurdles still need to be overcome before the medical community will embrace the capabilities of NGS technology and its deployment as an infectious disease diagnostic system (5, 22, 112).

6c. “Diagnostic” Clinical Trials Evaluating Microbial NGS

Several clinical trials have been launched to evaluate the ability of NGS-based tests to detect inherited genetic diseases or cancer alleles. Little progress has been made surrounding clinical trials for the diagnosis of infectious diseases. As shown on the ClinicalTrials.gov website (https://www.clinicaltrials.gov), there are clinical trials currently under way whose purpose is to evaluate the utility of NGS for the detection or characterization of microbes. However, these clinical trials are parallel to research studies in which patients are enrolled and provide consent for NGS analysis, including microbial NGS. Specifically, diagnostic infectious disease NGS clinical trials are minimal to nonexistent. The ClinicalTrials.gov website shows that an overseas company, PathoQuest, is conducting a clinical trial for the evaluation of high-throughput sequencing for screening pathogens in immunocompromised patients with suspected infectious disease (Clinical Trials registration no. NCT02007642). The purpose of the study is to compare the high-throughput sequencing approach of PathoQuest to detect pathogens causing infectious disease to conventional diagnostic methods. Preliminary findings have not been declared on the website or published as of June 2015 (213).

For the appropriate implementation of NGS as a diagnostic tool, clinical trials are necessary to expose strengths and weaknesses of both current NGS applications and traditional clinical laboratory approaches. Although there were early successes of diagnostic NGS trials, financial resources were lacking to continue to perform the clinical trials necessary to generate sustained interest. Such interest relies in part on adoption of the technology as well as a true detailed understanding of the benefits and cost of use (7, 160). These elements are lacking right now for infectious disease NGS clinical trials. Because NGS is a multistep process, it will be necessary to perform clinical trials at all levels of validation (14). Moving forward, outcome studies will ultimately be needed to demonstrate how NGS information genuinely affects and impacts patient care, followed by a clinical trial evaluation (160). Currently, viable funding mechanisms to implement these types of clinical trials are not available.

7a. Database Creation and Maintenance of Whole-Genome Microbial Sequences

Efforts are under way at the FDA for the creation of a microbial reference database (i.e., FDA-ARGOS Public Regulatory-Grade Reference Database; NCBI BioProject 231221 accession no. PRJNA231221) populated with high-quality, accurately annotated genomic reference sequences that will serve as a reliable resource for NGS analysis, both for developers and clinical end users. The coordination, preservation, and updating of FDA-ARGOS are collaborative efforts among different agencies, including the FDA, NCBI, the Department of Defense (DOD) Medical Countermeasure Systems (MCS), the U.S. Army Medical Research Institute for Infectious Diseases (USAMRIID), the Defense Threat Reduction Agency (DTRA), and the Institute for Genomic Sciences at the University of Maryland. This public database is intended to provide the most accurate microbial genomic sequences (i) to inform and guide treatment decisions and (ii) to streamline the regulatory pathway for clearance and approval of NGS platforms for microbe identification by acting as a single-comparator method (111). With existing collections, there are notable gaps in deposited sequences, emphasizing the need for a broader diversity and depth of microbial submissions (27). It is important to have curated and complete reference databases in order to prevent false positives and public health scares, as evidenced by the high-profile “Anthrax in the Subway” article by Afshinnekoo et al (168). For many submissions in the NCBI BLAST databases, uploaded partial sequences can be contaminated, thereby complicating evaluations that are trying to search for novel microbes with low prevalence. Contaminated entries could prevent discovery of such microbes. Also, low sampling depth, common to de novo genome assembly, can result in incomplete and uneven read coverage for an organism of interest and consequently cause unsuspected contamination (80). The regulatory-grade sequences that will make up the FDA-ARGOS database aim to prevent contamination and serve as a major resource for diagnostic sequencing and ongoing surveillance studies.

Another FDA genomic database known as GenomeTrakr is building databases with the NCBI (see example under BioProject accession no. PRJNA183844) for foodborne pathogens for source tracking in real time (105). This database includes isolates from clinical, environmental, and food sources to embrace the One Health Initiative as well as to provide full public access to draft genomes circulating globally. Laboratories at numerous state and federal departments of health have their own GenomeTrakr bioprojects for the isolates that are important for those areas. The FDA NCBI system should work for any human pathogen, and early database building is beginning under NCBI's pathogen identification program, of which GenomeTrakr is one part for foodborne pathogens. Clinical and public health microbiology laboratories are on the front lines and are the first to encounter outbreak isolates and novel pathogens. These environments will play a key role in depositing these new genomes in the expanding knowledge base.

Assembly of reference databases will require expert consensus and feedback from the infectious disease community and the medical community to determine which organismal sequences are necessary to include. Organisms that have a high clinical index of reportability or are observed in 90% of cases in the laboratory should be included within the database along with rule-out organisms. However, there are also organisms that are not commonly seen in the laboratory but are clinically actionable. For example, Leptospira requires clinical intervention but is not regularly detected in the microbiology laboratory. Furthermore, a common skin contaminant, Malassezia, is not pathogenic for healthy individuals. In the context of an immunocompromised patient or neonate, Malassezia can act like a pathogen and cause sepsis. It is not necessary or cost-efficient to include all known environmental organisms, but a variety of these rare and unusual pathogens should be deposited in reference NGS microbial databases. It would be ideal to craft reference databases for each infectious disease etiological agent, including bacteria, viruses, fungi, yeasts, and parasites. The purpose of the databases is to capture the breadth of diversity for each infectious agent category, and thus more than one genome per species should be denoted. The databases should also be able to distinguish between pathogens and their near neighbors. Antibiotic resistance genes, phenotypic susceptibility data, and virulence factors could be represented as subsets within existing databases. For example, the NCBI, as part of a multi-group effort, has established a template to capture antimicrobial susceptibility phenotype information for those organisms submitted to the BioSample database (http://www.ncbi.nlm.nih.gov/biosample/docs/antibiogram/). Additionally, it is imperative to “version” the databases when significant additions or updates are made. Finally, there needs to be a culture repository of these collected reference strains if a reexamination of the isolate is required.

For the greater good of public health, it is important to have NGS data publically available, filtered with metadata, so that patients are not identifiable. From a computational standpoint, the only way that this predictive strategy will work is if all of the genomic data are centralized. If sequence data are isolated in individual hospitals and not shared, this centralized approach will be nearly impossible to undertake (7). Individualized databases would represent only a single city's microbiome instead of that of the whole country or the whole world. Furthermore, the NIH issued a mandate in 2003 to promote data sharing from principal investigators requesting $500,000 or more in grant support. In January 2015, the NIH announced that all grantees are required to publicize large genomic data sets by the time of publication (166)). One would imagine that this requirement would help tremendously in adding genomic sequence information to public repositories, but enforcement has been meek (214, 215).

7b. Representation of an NGS Diagnostic Test Result in a Patient's Medical Record

It was suggested that none of the actual genomic data should be incorporated into the patient's electronic medical record but rather the clinically actionable, interpreted result should be stated in the patient's file. The version of the software and the genomic reference sequence database should also be noted in the medical record. The suggestion that only the actionable result be stored is partly due to the challenges of laboratory management systems and hospital information systems, which largely cannot support the bioinformatics component of NGS. Hospitals can choose to store the bioinformatics data locally or to utilize off-site remote hosting like Amazon cloud support, as long as these systems uphold HIPAA compliance and security requirements. Clinical sequences could be deposited in the database, where personally identifiable information could be held locally and disconnected from the public data. All information in the reference database should be legally nondis-coverable. A sophisticated link from the patient's medical record to the reference database would benefit the broader community (30).