POINT
The fundamental role of the clinical microbiology laboratory is to detect, to identify, and to characterize pathogens that cause infections in patients, to guide patient management. Over the years, the methods in use have changed dramatically and now include a variety of culture-based approaches, visualization by microscopy, antigen detection, serological methods, and molecular detection. Each of these methods has its strengths and weaknesses and detects a defined spectrum of organisms that are potentially able to be cultivated and/or detected. Improvements to these microbiological assays, such as enhanced culture media and multiplex PCR syndromic panels, have expanded the range of organisms detectable by the laboratory. Along the way, clinical microbiologists have had to adapt their techniques and understand the benefits and limitations of new technologies in order to provide the most accurate information to guide treatment decisions.
The advent of high-throughput sequencing methods, known as next-generation sequencing (NGS), allows the simultaneous characterization of millions of individual DNA fragments (or cDNA fragments from RNA), facilitating genome assembly and metagenomic analysis. The ability to broadly detect all classes of organisms directly from patient samples is the major advantage of metagenomic NGS (mNGS) methods over targeted methods such as multiplex PCR panels. However, significant concerns exist regarding the performance, validity, and clinical significance of the organisms detected using mNGS. These substantial challenges need to be addressed during the clinical development of mNGS assays for pathogen detection. In our experience, these challenges can be overcome if the laboratory is willing to commit the needed resources and has the appropriate expertise, making it feasible for clinical laboratories to perform mNGS for patient testing (1). Furthermore, the clinical benefits of mNGS patient testing in facilitating the diagnosis of neurological infections and informing treatment decisions have already been demonstrated in a prospective, multisite, diagnostic trial (2). In that study of 204 patients with undiagnosed meningitis, encephalitis, and/or myelitis, 13 (22%) of 58 infectious cases were diagnosed by mNGS testing alone, and actionable clinical management and treatment decisions were made on the basis of mNGS results in more than one-half of those cases. The broad applicability of mNGS methods in research laboratories across a variety of infectious syndromes and sample types has also been illustrated in a growing number of case reports and small case series (3). Laboratories have continued to develop and to validate mNGS assays to address a variety of infectious disease syndromes, such as sepsis and pneumonia, demonstrating high analytical sensitivity and detection of more organisms thought to be pathogenic than are detected by conventional methods (4, 5).
Reference and commercial laboratories are now offering mNGS assays in laboratories certified in accordance with the CLIA program, for patient diagnosis using selected sample types. There appears to be substantial demand from providers and patients for mNGS testing, likely based on the hope of uncovering previously undiagnosed infectious disease. In order to support the use of mNGS testing, analytical and clinical validation studies are needed. Validation of an unbiased mNGS assay is a challenging task, since all organism species and subtypes are potentially detectable. This also poses regulatory challenges, as it would be impractical to obtain and to test more than a handful of cases for each infectious organism for purposes of validation. Our solution has been to adopt a “representative organism” approach, using individual species to serve as models for detection of other pathogens within that category (e.g., DNA viruses, RNA viruses, Gram-negative bacteria, Gram-positive bacteria, yeasts, and molds) (1), with the expectation that similar organism types can be expected to behave similarly in the assay. A list of detected species, with the findings confirmed by alternative laboratory testing, can be maintained and expanded over time to establish the spectrum of organisms that can be reliably tested for and reported by the mNGS assay. A model for this approach may be the clinical validation of matrix-assisted laser desorption ionization–time of flight mass spectrometry-based assays for bacterial identification, for which reference databases of reportable organisms may be expanded with additional confirmed strains over time.
A significant potential limitation to the use of mNGS assays for infectious disease diagnostics is the possibility of false-positive results due to contamination. This issue is not restricted to mNGS testing; in fact, microbiology laboratories are very familiar with contamination, especially for culture-based and PCR assays, and have developed specific protocols to deal with contamination events. Sources of “contamination” in mNGS results include detection of normal human body flora or background contaminating organisms in laboratory reagents or the environment, as well as bioinformatic analysis or database errors, such as misannotations, that can lead to organism misidentification. A salient example is the discovery of xenotropic murine leukemia virus-related virus (XMRV) and its initial apparent association with chronic fatigue syndrome (6). Although XMRV was later shown to be a laboratory-derived recombinant virus not present in the human population and the association with chronic fatigue syndrome was disproven, there was emerging demand for clinical testing and even treatment decision-making on the basis of XMRV status. However, we have found that risk mitigation strategies can be effective in avoiding misinterpretation of mNGS results due to contamination; strategies include (i) addition of interpretive notes based on guidance from expert laboratory physicians, (ii) confirmation of organisms newly identified by mNGS using alternative methods, (iii) implementation of a contamination-monitoring strategy, such as a dedicated contaminant reference database, and (iv) institution of a real-time teleconference hosted by experienced laboratory directors and subspecialist physicians (clinical microbial sequencing board) to discuss results in clinical context with treating physicians (2). Ultimately, implementation of these strategies is the responsibility of the clinical laboratory providing mNGS testing, to ensure that mNGS results are used appropriately to guide patient management and care.
With respect to the adoption of mNGS testing in clinical laboratories, the guidelines and standards for clinical NGS testing that were initially developed for oncology and inherited disease applications (7, 8) have been adapted to infectious disease diagnostic performance evaluation (9). Conceptually, mNGS assays for infectious diseases are similar to oncological and noninvasive prenatal testing NGS assays, which are currently FDA approved and in widespread use, in their ability to detect unlimited numbers of sequence variants, and sample-processing steps in generating and analyzing sequence data overlap significantly. Thus, we may be able to leverage the requirements for testing and reporting of results for clinical mNGS assays that have already been established by our colleagues in other fields of medicine.
Quality control (QC) for mNGS assays is critical to ensure accurate performance. Because these are complex multistep tests, several QC metrics are required, including sample controls, external controls, internal controls, library quality metrics, sequencing quality metrics, and contamination controls. Each of these can be defined by the laboratory with acceptable criteria to ensure that sample quality and run quality are adequate and errors have not occurred during the processing steps. Although complex, the establishment and monitoring of robust mNGS QC metrics have been successfully adapted to the clinical microbiology laboratory setting (1, 10).
Analysis of mNGS data relies on the ability of bioinformatic pipelines to accurately classify sequence reads using established databases. Several bioinformatic pipelines have been developed for metagenomic analysis (11). Methods to filter out database errors and inaccurate sequence matches to avoid inadvertent misclassification are often necessary, especially when noncurated databases are used. Such methods include computational host subtraction, masking or removal of known errors, taxonomic classification, and filtering of sequence reads to ensure accuracy using alternative alignment programs. Another challenge is that application and validation of bioinformatic tools for mNGS analysis generally require specialized computational expertise. However, these challenges have been successfully addressed in the field of metagenomics with the development of easy-to-use, automated software with graphical user interfaces, including platforms such as SURPI+, Taxonomer, CosmosID, and OneCodex. In addition, the recent availability of databases that have been extensively curated for accuracy, such as FDA-ARGOS (12), has been useful in ensuring that mNGS results are reliable and correct.
The interpretation of mNGS results in clinical context is a key part of the use of MNGS testing in patient management. An advantage of this unbiased approach is the ability to diagnose rare infections, such as hepatitis E virus-associated meningoencephalitis (13). From a public health perspective, mNGS testing can provide evidence of newly emerging or reemerging pathogens. For example, our clinical mNGS testing of cerebrospinal fluid resulted in identification of the first human case of St. Louis encephalitis virus in California since 1986 (14). Unusual organism detections such as this can be reported to public health agencies, which may be able to assist in follow-up confirmatory testing.
Studies regarding the clinical utility and cost-effectiveness of mNGS assays for pathogen detection are lacking, making this area one in need of further investigation. While mNGS testing is expensive, relative to most traditional microbiological tests, the incremental costs of testing are minimal, compared to the costs of intensive care unit stays or invasive diagnostic biopsy procedures or even relative to the myriad diagnostic tests that are ordered for complex cases. Thus, if even a small percentage of mNGS test results demonstrate a causative pathogen, with effects on subsequent patient care, substantial cost savings for health systems overall are likely. Since mNGS testing for infectious disease diagnosis lacks a specific CPT code, reimbursement for direct testing costs is limited; therefore, the cost impact is best seen in the context of all patient management costs until further cost-benefit data are available to guide reimbursement decisions.
The sequencing data provided by mNGS may be able to provide clinically useful information beyond mere identification of the presence or absence of a potential pathogen. mNGS data have been leveraged to classify infections on the basis of host response, to characterize antimicrobial resistance, and to genotype detected pathogens for infection control and public health purposes. While negative results for typical diagnostic tests can be used to exclude the possibility of infection only for the agents targeted, negative pan-pathogen mNGS results may indicate a lower likelihood of an infectious etiology, hence being potentially useful as “rule-out” results (although indirect tests such as serological assays may also be needed). The absence of detectable pathogen nucleic acids can potentially reassure the treating physician that a severe infection is unlikely to be present. Finally, new technologies such as nanopore sequencing may enable mNGS to be performed with a turnaround time of under 6 h, greatly increasing the window of clinically actionable results afforded by mNGS testing.
mNGS assays for infectious disease diagnosis have been developed in the CLIA environment by specialized laboratories using approaches and guidelines similar to those used for oncological and inherited disease NGS tests. Both published and pending reports indicate that mNGS testing increases the diagnostic yield for infectious diseases in patients, especially those with critical illness and/or immunocompromised status, and may be useful for early identification of emerging pathogens relevant to public health. Clinical mNGS assays that have been validated in CLIA-certified laboratories are now available for a number of infectious syndromes, including meningoencephalitis, sepsis, and pneumonia. Testing of patients who are utilizing high levels of health care resources is likely to be particularly effective, as timely identification of atypical and/or rare infections is possible with mNGS testing. The utility of mNGS may also extend to other applications in the future, including discrimination of causes of infections based on transcriptome profiling (RNA sequencing) of the patient host response, identification of antibiotic resistance genes, and rapid mNGS diagnosis within hours on portable platforms such as the nanopore sequencer. The broad-spectrum, direct detection capability of mNGS testing may also facilitate its use as a rule-out test for excluding infection in patients with suspicion for noninfectious etiologies such as autoimmune disease. With ongoing improvements in technology and eventual regulatory approval, costs and barriers to mNGS implementation will continue to decrease, enabling expanded access to testing by treating physicians.
Steve Miller and Charles Chiu
