Detection of Mycobacterium tuberculosis cell-free DNA to diagnose TB in pediatric and adult patients

Dear Editor,

Traditionally, diagnosis of pulmonary TB relies on the analysis of sputum. However, this approach is inadequate for patients who cannot readily provide a sputum sample (e.g., young children, some patients living with HIV). Molecular detection of Mycobacterium tuberculosis from alternative sample types (e.g., stool, oral swabs, or nasopharyngeal swabs) has been widely evaluated in children, but yields low sensitivity for those most in need of a reliable diagnostic (i.e., children with smear- and or culture-negative disease or those who cannot produce sputum). This may be because successful M. tuberculosis detection from these sample types may depend on the presence of sputum that has either been swallowed or contacted the nasal or oral mucusa. Diagnostic approaches focusing on cell-free DNA (cfDNA) detection from plasma or urine are appealing because they do not require the presence of sputum. Although a handful of studies in adults with pulmonary TB have documented M. tuberculosis cfDNA detection from urine^1,3 and plasma,^4,5 data for children are completely lacking. Also, the extent to which cfDNA can be detected in patients with a lower bacillary burden is unknown.⁶

We therefore analyzed plasma samples from well-characterized adult and pediatric patients using a quantitative microbial cfDNA detection method developed by Karius Inc. (Redwood City, CA, USA). This direct detection method utilizes unbiased (“shotgun”) metagenomic sequencing for identification and quantitation of microbial sequences in human plasma and has been validated for the detection of cfDNA from a panel of >1,000 bacteria, DNA viruses, fungi, and eukaryotic parasites.⁷ We hypothesized that M. tuberculosis cfDNA could be detected in patients with culture-confirmed TB, and potentially also from those with culture-negative, clinically diagnosed TB, paving the way for a sputum-free diagnostic approach that would be particularly beneficial in children.

We analyzed pediatric patients in Lima, Peru, categorized into one of three groups: 1) culture-confirmed TB (n = 10); 2) culture-negative, clinically diagnosed TB (n = 10); 3) TB symptoms, but TB ruled out by a pediatric pulmonologist on the basis of negative culture, clinical examination, tuberculin skin test, or chest radiograph (n = 10). A single plasma sample was analysed for each child. In children with TB, sample collection occurred prior to treatment. Additionally, for 10 adults with culture-confirmed TB (5 with a high-grade (3+) positive sputum smear, 5 with a negative sputum smear) we analyzed three serial samples taken prior to TB treatment initiation and during the first week of treatment. Blood samples were collected using whole blood tubes with EDTA and centrifuged within 30 minutes at 2,000 x g for 10 minutes. Plasma was isolated prior to transportation to the laboratory via cold chain at 2–8°C. Samples were stored at −80°C.

The study was approved by the Institutional Review Board of the Harvard Medical School (Boston, MA, USA) and the Ethics Committee of the Peru National Institute of Health (Lima, Peru). Guardians and adult participants provided written informed consent. As an additional control group, we analyzed EDTA-plasma samples collected by Karius (Redwood City, CA, USA) from 684 ambulatory, asymptomatic subjects between 2015 and 2020 as controls for test clinical validation.

Aliquots were processed and analyzed by Karius. CfDNA was extracted from 250 μL of ethylenediaminetetraacetic acid (EDTA) plasma, next-generation sequencing (NGS) libraries were prepared, and sequencing was performed using NextSeq^®500 (Illumina, San Diego, California, USA). Human sequence reads were removed, and remaining reads were aligned to a curated pathogen database. For the CLIA (Clinical Laboratory Improvement Amendments) certified laboratory diagnostic test, statistically significant levels of M. tuberculosis cfDNA in plasma relative to real-time negative controls were reported in molecules of cfDNA per microliter (MPM) of plasma as previously described.⁷ A relaxed research-use only (RUO) statistical threshold was also applied, such that M. tuberculosis would be reported if cfDNA derived from this microorganism was observed in ≥3 unique sequencing reads.

Using the standard reporting threshold, among 4 pediatric and 5 adult patients with smear-positive culture-confirmed TB, M. tuberculosis cfDNA was detected in 2/4 (50%) and 3/5 (60%) plasma samples collected at the time of initial TB diagnosis. MPM values were 111 and 53 for the pediatric patients; and 74, 24 and 59 for the adult patients. Using the RUO threshold, M. tuberculosis cfDNA was detected in an additional 1/4 (25%) of these pediatric patients; and 2/5 (40%) adult patients (Table); MPM values for these samples were 11 for the pediatric patient, and 44 and 5 for the adult patients. Other than one pediatric patient with culture-confirmed smear-negative disease (MPM 15 by the RUO method), M. tuberculosis cfDNA was not detected in children or adults with smear-negative, culture-confirmed TB, or in children with culture-negative, clinically diagnosed TB (Table). All pediatric patients in whom TB had been ruled out by clinical and laboratory methods tested negative for cfDNA.

Using the RUO threshold, all five smear-positive, culture-positive adult patients were positive for M. tuberculosis cfDNA at all three time points, and MPM were noted to increase slightly in the on-treatment samples relative to the pre-treatment sample. In the five smear-negative, culture-positive adult patients, one was newly positive for M. tuberculosis cfDNA on the RUO method on Day 3 of treatment (MPM 4), but all other samples were negative. M. tuberculosis cfDNA was not detected at any concentration from the cohort of 684 asymptomatic ambulatory control subjects.

This study demonstrates that M. tuberculosis is detectable using unbiased metagenomic sequencing of cfDNA in plasma samples collected at the time of TB diagnosis. Based on the RUO threshold, detection was 100% sensitive in smear-positive adults and 75% sensitive in smear-positive children. Specificity was 100% in children in whom TB was ruled out and in a large cohort of ambulatory controls. This suggests that M. tuberculosis cfDNA detection is highly specific, such that detection of M. tuberculosis cfDNA on clinical testing should be taken as evidence of M. tuberculosis infection and acted on for treatment decision-making. However, the lower sensitivity of the clinical vs. RUO analysis methods in patients with confirmed TB, and the lower overall sensitivity of the cfDNA approach in smear-negative, culture-positive patients indicate the need for optimization to improve sensitivity. However, it is notable that this analysis was performed without prior capture or enrichment of M. tuberculosis DNA targets from the samples, and without optimization of the Karius microbial cfDNA sequencing method to increase detection of M. tuberculosis DNA. These steps are likely to improve sensitivity and will be addressed in future studies.

Microbial cfDNA sequencing can detect pathogen DNA in plasma from patients with infection at distant sites,^8–10 and these findings demonstrate that M. tuberculosis is no exception. The varied sensitivity by smear status, previously reported in other studies of non-sputum-based molecular diagnostics,^11,12 may explain heterogeneity in M. tuberculosis cfDNA studies to date, and underscores the importance of stratifying sensitivities by smear status when drawing comparisons across studies. Based on this proof-of-concept study, we conclude that unbiased metagenomic sequencing of plasma samples can detect M. tuberculosis cfDNA in pediatric and adult patients with smear-positive, culture-confirmed pulmonary TB. We expect the sensitivity of the method can be improved with optimization.

Acknowledgements

Footnotes

References