Respiratory Metagenomics: Ready for Prime Time?

Pneumonia is one of the most common infectious reasons for hospital admission, but current standard-of-care pneumonia diagnostics leave much to be desired. In a population-based survey of community-acquired pneumonia in hospitalized adults in the United States, only 38% had a pathogen identified despite exhaustive clinical testing with culture, multiplex PCR, and urinary antigens (1). This diagnostic gap leads to the overuse of broad-spectrum antibiotic agents, contributing to the ever-increasing global burden of antimicrobial resistance, which is outpacing novel antimicrobial agent development and cited by the World Health Organization as one of the top 10 global threats facing humanity (2). Furthermore, as the number of immunocompromised patients steadily increases (3), so does the risk of infection with unusual pathogens often missed by standard microbiologic testing, resulting in delayed or missed diagnoses in our most vulnerable patients (4). There is an urgent need for new respiratory diagnostics that are less biased and more sensitive and provide rapid results.

A sequencing-based approach can overcome many of the limitations of existing pneumonia diagnostics. Metagenomic sequencing permits unbiased assessment of all nucleic acid in biological samples, enabling the detection of potential pathogens, the wider microbiome, and the human host response in a single assay (5). Presently, Clinical Laboratory Improvement Amendments (CLIA) certified metagenomic tests are clinically available for plasma and cerebrospinal fluid (6, 7), but technical and bioinformatic complexity delays turnaround time, and high sensitivity makes it challenging to distinguish signal from noise (8). Respiratory metagenomics (RMg) presents an even greater challenge because the respiratory tract is a nonsterile environment with a well-described microbiome, further complicating analysis and clinical interpretation.

In this issue of the Journal, Charalampous and colleagues (pp. 164–174) conducted a prospective clinical pilot study (9) implementing a previously established RMg workflow (10) in the ICU of an academic hospital (Figure 1). Although others have explored metagenomics for the diagnosis of lower respiratory tract infections (11–15), this study is notably the first to truly implement RMg in a clinical care setting and assess its impact. Generation and analysis of high-complexity metagenomic data involves multiple steps, including nucleic acid extraction, library preparation, sequencing, removal of human reads, taxonomic alignment, and modeling to distinguish pathogens from the background microbiome (5). Here, the authors report a median turnaround time from sample acquisition to result of just 6.7 hours, which is impressive and much faster than traditional culture-based diagnostics. This pilot study had the capacity to perform RMg on three samples per day, so follow-up studies regarding scale-up and cost effectiveness will be important.

Figure 1.

Clinical implementation of respiratory metagenomics. Nanopore respiratory metagenomics were performed on BAL fluid, tracheal aspirates, and pleural fluid samples from intubated patients in the ICU with suspected respiratory infection. Routine clinical testing with culture was simultaneously performed. Sequencing reports were generated after 30 minutes and 2 hours for microbes and resistance mutations, respectively. Median turnaround time from sample acquisition to the reporting of results to clinicians was 6.7 hours. In comparison, preliminary culture results were reported after an average of 29 hours and were finalized in 40 hours. The impacts on clinical care, antibiotic agent prescribing, and infection control involvement were evaluated. AMR = antimicrobial resistance; BAL = bronchoalveolar lavage; ICU = intensive care unit.

Beyond the rapid turnaround time, this study demonstrated that RMg had a compelling impact on clinical management. When clinical testing results were negative, RMg detected a clinically relevant pathogen 19% of the time, often with organisms not initially suspected by clinicians, such as Legionella or Cytomegalovirus, and resulted in a management change. Conversely, RMg had a high level of agreement (93%) when standard culture results were positive. The high sensitivity gave clinicians confidence to discontinue antibiotic therapy when RMg returned negative results and, in some cases, even start early immunosuppression for suspected autoimmune inflammatory pulmonary conditions. RMg had the added benefit of early detection of resistance genes for many organisms, allowing clinicians to appropriately tailor antimicrobial agents more expediently. In total, the authors estimate that RMg contributed to prescribing decisions in a noteworthy 80% of cases.

Compared with standard culture, sequencing-based approaches pair nicely with infection control and public health efforts. With usual practice, in a suspected outbreak, cultured organisms from the microbiology laboratory are sent for whole-genome sequencing to assess relatedness with other isolates. However, with metagenomics, if sequencing depth is adequate and genome coverage is high enough, the sequencing data can be used for proactive and timely infection control interventions. The authors demonstrate this impact with a case of Legionella pneumophila that originated from a bedside water faucet and a confirmed patient-to-patient Klebsiella variicola transmission.

As with any study, there are some limitations. Because the RMg assay studied was exclusively DNA-based, the workflow is unable to detect RNA viruses, which are the most common cause of lower respiratory tract infection in adults and children (1, 16). This could potentially hamper antimicrobial stewardship efforts because a positive viral test result in the setting of negative bacterial testing results can provide the confidence to stop antimicrobial therapy. Ideally, the RMg workflow could be modified to include DNA and RNA sequencing to better capture the most common causes of respiratory infections and enable the detection of emerging viral pathogens that may not be detectable by standard PCR assays. This could also enable profiling of host gene expression, which could inform whether the detected microbes are matched to an immune response consistent with infection. Another consideration for future studies is the inclusion of a noninfectious control group to more rigorously define specificity, optimize the differentiation of pathogens from commensal organisms, and understand the impact on antibiotic agent use when incidental microbiota are detected.

All said, the authors have performed an impressive clinical pilot study of RMg and have successfully moved the needle toward future implementation of this technology in routine clinical practice, in which current methods often fall short. Although some challenges and limitations persist, this study opens the door for future research in assay optimization, launching of a randomized controlled trial, and cost-effectiveness analyses to propel RMg to prime time.