An estimated 10 million people infected with Mycobacterium tuberculosis (MTB) develop active pulmonary tuberculosis disease (PTB) yearly (1), and up to one third of the population has latent tuberculosis infection (LTBI) at any time, with a 5–10% lifelong risk of developing PTB (2). LTBI, subclinical TB and active TB disease comprise the spectrum of MTB infection. The ability to identify individuals with PTB, differentiate LTBI from PTB and other mycobacterial infections, and to verify treatment success, is key both for individual patient outcomes and for preventing MTB transmission to others. Sputum culture positivity for MTB is one of the primary reference standards for TB diagnosis, just as the conversion from sputum positive to negative after 2 months of intensive-phase treatment is considered a reference indicator of therapeutic response. However, the slow growth of MTB cultures contributes to significant diagnostic delays, making evaluating sputum culture conversion both time-consuming and laborious. Easy-to-measure host inflammatory biomarkers that can differentiate PTB from other diagnoses or that can be used to support assessments of TB treatment response, would be transformative (3). Levels of c-reactive protein (CRP) and blood monocyte-to-lymphocyte ratio (MLR) may help to identify PTB in high-risk populations, and both decrease during treatment, but their changes are poorly predictive of sputum culture conversion at 2 months (3–8). Interferon-γ release assays (IGRA) measure interferon-γ response to MTB-specific antigens and are useful for identifying individuals with LTBI, however, they cannot be used to differentiate LTBI from PTB reliably, and are not useful for monitoring response to treatment (9, 10).

Cell-free DNA (cfDNA) is circulating fragmented DNA in blood that is actively secreted by living cells or released from dying cells and destroyed microorganisms and other pathogens (11, 12). Interest in cfDNA as a biomarker of altered physiologic and pathogenic states has prompted research into its diagnostic value in pregnancy and organ transplant monitoring and cancer diagnosis (13–15). Elevated levels of host-derived cfDNA can be also be predictive of a poor prognosis in infection (16–18). Both nuclear cfDNA (Nu-cfDNA) and mitochondrial cfDNA (Mt-cfDNA) are host-derived cfDNA, with the former originated from the nucleus and the later from mitochondria distributed throughout the cytoplasm in a cell. They can be released by activated inflammatory cells and only Mt-cfDNA can be released by platelets during infection (19, 20). Studies have shown that both plasma Nu-cfDNA and Mt-cfDNA levels correlate with disease severity and mortality in patients with sepsis (18, 21, 22). Prior research has also demonstrated that among HIV-positive individuals, plasma Mt-cfDNA levels were higher in those with TB and severe inflammation than in those without disease (23). Wiens et al. reported that macrophages infected with more virulent MTB released more mitochondrial DNA into the cytosol, suggesting mitochondrial DNA could be a biomarker indicating innate immune responses in TB (24). Studies have also demonstrated that MTB can cause mitochondrial damage in macrophages, leading to the release of mitochondrial DNA (25). Yet, it remains unclear if the release of Mt-cfDNA from inflammatory cells into the extracellular space, presumably resulting in high circulating Mt-cfDNA level, is related to burden of MTB nor if observed levels are different for LTBI or PTB and other pulmonary mycobacterial diseases. It is also not yet known if elevated Mt-cfDNA levels at PTB diagnosis reverts to baseline after PTB disease is effectively treated.

We hypothesized that plasma Mt-cfDNA is associated with MTB bacterial burden and inflammatory response in TB. In order to explore these ideas we a) examined the association between plasma Mt-cfDNA levels in patients with and without culture-positive PTB; b) evaluated changes in plasma Mt-cfDNA level during PTB treatment and c) evaluated the potential predictive value of Mt-cfDNA levels on subsequent failure of sputum conversion after two months of treatment. Finally, we measured levels of Mt-cfDNA in supernatants from a cell line stimulated with MTB early secretory antigenic target (ESAT)-6 antigen to develop potential supporting evidence for the causative role of MTB on Mt-cfDNA release from macrophages.

In our study of patients with MTB and NTM infections we found that Mt-cfDNA levels were elevated in patients with PTB as compared to those with LTBI and NTM-LD, despite the greater clinical severity of disease among the NTM-LD patients. Notably, plasma Mt-cfDNA was positively associated with inflammatory markers and smear-positivity (a proxy for bacterial burden) in PTB patients, indicating a potential association of this biomarker with disease severity. For patients with culture positive PTB, high and increasing levels of Mt-cfDNA also appeared to have some potential for predicting persistent culture-positivity after 2 months of treatment, but more importantly, low Mt-cfDNA levels at start of treatment was highly predictive of culture negativity after two months of treatment. Furthermore, while evaluating the plausibility of our proposed linkage between plasma Mt-cfDNA and MTB infection, we observed stimulation of THP-1 cells with MTB antigen ESAT-6 is associated with Mt-cfDNA release into the extracellular space of inflammatory cells, supporting the hypothesis for the origin of the increased plasma Mt-cfDNA levels in PTB.

Identifying individuals with infectious PTB and monitoring treatment response remain critical aspects of TB control worldwide. Of the currently available blood biomarkers for this purpose, CRP performs better than symptom-based methods as a screening tool for PTB in people living with HIV (4), but neither its baseline level nor change over time is predictive of MTB sputum conversion, which despite its limitations, is still considered the reference standard for treatment success (5). Elevated monocyte and platelet counts are correlated with inflammation in PTB and trend down after treatment, but they cannot reliably predict treatment outcomes (37–39). MLR may have utility in identifying TB disease in HIV-infected children and for differentiating active TB from LTBI in the general population (6–8), but persistently high MLRs have not been found to associate 2-month culture-positivity (7, 8). The present study, while limited in size and scope, indicates that Mt-cfDNA levels are elevated in individuals with PTB compared to those with LTBI which reinforces the potential of Mt-cfDNA as a novel biomarker for differentiating PTB from LTBI, possibly in conjunction with existing assays such as the IGRAs. In PTB patients, Mt-cfDNA level was positively correlated with CRP, monocyte count, and platelet count, and was higher in smear-positive patients than in smear-negative patients, suggesting that Mt-cfDNA is correlated with both inflammatory processes and bacterial burden. Our findings are supported by prior research showing elevated Mt-cfDNA to be associated with TB-related immune reconstitution inflammatory syndrome and interleukin-18, a proinflammatory cytokine, in HIV-infected patients (23). From a clinical standpoint, when excluding MTB smear-positive patients who can more easily be diagnosed with PTB using existing diagnostics, Mt-cfDNA level was also statistically different between smear-negative PTB patients and the contacts with LTBI, suggesting a potential role for Mt-cfDNA for differentiating low-bacterial-burden PTB from IGRA positive LTBI patients. However, the presence of outliers with low Mt-cfDNA levels in the smear-negative PTB subgroup suggests that Mt-cfDNA alone might not have sufficient sensitivity in the clinical setting to differentiate all smear-negative PTB from LTBI. However, it may nonetheless play complementary role in the medical decision making model when evaluating these patients. Based on the totality of our results, we believe that human origin Mt-cfDNA has the potential to be used as an indicator of the clinical stage and severity of MTB infection, reflecting the clinical spectrum of MTB infection from LTBI to smear-negative and smear-positive PTB, and its use as a standalone or complementary biomarker in the evaluation of patients with suspected PTB and LTBI warrants further investigation.

Importantly, our study demonstrates that Mt-cfDNA may also have some utility in treatment monitoring for patients with PTB. A high baseline Mt-cfDNA level was found to be moderately predictive of 2-month culture-positivity in PTB patients and it was observed that Mt-cfDNA levels decline after 2 months of treatment. Our observed rate of persistent culture-positivity of 16.4% at 2 months is in line with previous reports (35, 40). Notably, using changes in Mt-cfDNA to stratify patients, we noted that the rate of persistent 2-month culture-positivity was highest in the PTB subgroup with initially high Mt-cfDNA and subsequently increasing levels at 2 months while the rates were lower in other subgroups, with a statistically significant trend. We hypothesize that Mt-cfDNA could be an integrated surrogate marker for bacterial load and inflammation in PTB, which are important factors associated with decreased rates of on-treatment culture-conversion (3, 41, 42). Similarly, decreased Mt-cfDNA levels reflect decreased bacterial load and reduced inflammation following treatment, which could have clinical relevance in indicating successful treatment. To this end, while the exact cause of increased circulating Mt-cfDNA levels is likely complex and multidimensional, our proof-of-concept study suggests the potential value of Mt-cfDNA monitoring the treatment response of PTB.

Human cfDNA originates from cell apoptosis, neutrophil extracellular traps, and the death of inflammatory cells in response to microbes, as well as from active cellular secretion through extracellular vehicles (12). In comparison to Nu-cfDNA, Mt-cfDNA is a thousand-fold higher in copy numbers because every cell, except red blood cells, contains abundant mitochondria with DNA. We found Mt-cfDNA levels better correlate with culture-positive PTB and predict 2-month culture-positivity than Nu-cfDNA. These findings may be mechanistically supported by the relationship we observed between ESAT-6 stimulation and Mt-cfDNA levels in supernatants from activated THP-1 cells. However, human Mt-cfDNA may be elevated by various physical or psychiatric stresses and pathologic insults (17, 43, 44). To investigate the specificity of Mt-cfDNA as a biomarker of PTB, we also included patients with NTM-LD in the study and found that high Mt-cfDNA facilitated the differentiation of culture-positive PTB from NTM-LD. Interestingly, although the NTM-LD group had higher rates of hemoptysis and higher radiographic scores indicating more severe disease among the NTM patients, elevated Mt-cfDNA levels still differentiated PTB from NTM-LD suggesting the possibility of a unique association between PTB disease and Mt-cfDNA level that goes beyond lung pathology. Using MTB strains with different virulence to infect bone marrow-derived macrophages, Wiens et al. reported the strain-dependent infection-related release of mitochondrial DNA into the cytosol (24). Because the levels of several inflammatory markers and bacterial load were similar in the PTB and NTM-LD groups, the differentially higher level of Mt-cfDNA in PTB patients may indicate a specific role for MTB in causing the release of Mt-cfDNA that deserves further investigation.

The present study has some limitations. First, the timing of laboratory testing during the intensive phase of treatment in PTB patients was limited to only two time points and the number of cases eligible for persistence culture-positivity assessment was relatively small. Large-scale studies are warranted to validate the association between dynamic changes in Mt-cfDNA during follow-up and bacterial burden during and after tuberculosis treatment. Second, the enrolled participants were clinically stable HIV-uninfected outpatients at a single institute, so the generalizability of our findings to other populations of different conditions on diverse treatment regimens remains unknown. Further studies on the performance of Mt-cfDNA in differentiating mycobacterial diagnosis and predicting outcomes among PTB and NTM-LD patients with HIV co-infection are needed. In addition, confounders such as diabetes severity and antihyperglycemic medications might affect the treatment outcome in PTB patients but were not available for analysis (45). Our study is also limited by the lack of a healthy control group. In one recently published study, the authors disclosed that patients with multidrug-resistant TB had increased mitochondrial DNA copy numbers in blood leukocytes when compared to healthy individuals, and proposed this finding was a result of oxidative stress in TB patients (46), which may be one of the plausible mechanisms explaining the elevated extracellular plasma Mt-cfDNA levels in our patients with PTB. Further investigations incorporating healthy controls may help confirm the potential role of Mt-cfDNA in differentiating the spectrum ranging from an uninfected population, to LTBI and PTB. Finally, the plasma samples were not double centrifuged and, thus, the level of cfDNA may have been overestimated (31). However, our results should be robust because, using the same protocol, all intergroup and post-treatment comparisons consistently demonstrated that high levels of Mt-cfDNA were associated with PTB-related inflammation and bacterial burden. Nevertheless, the potential impact of differences in cfDNA measurement protocols among different studies should be acknowledged.

In conclusion, we found that Mt-cfDNA level was correlated with inflammatory markers and smear-positivity in PTB. Importantly, Mt-cfDNA, but not Nu-cfDNA, was associated with culture-positive PTB, was associated with sputum culture-positivity at 2 months of therapy, and decreased after treatment. Further studies are warranted to evaluate the potential complementary role of Mt-DNA in PTB identification and treatment response monitoring.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. Requests to access these datasets should be directed to Sheng-Wei Pan, sanweipan@gmail.com.

The studies involving human participants were reviewed and approved by Taipei Veterans General Hospital Institutional Review Board Nos. 2017-12-001C, 2018-10-017A, 2019-07-003C, 2020-07-009AC, 2021-01-010BC. The patients/participants provided their written informed consent to participate in this study.

S-WP, T-YT, J-YF, and W-JS collected data. All authors performed data analysis. S-WP, RS, DC, M-LH, and C-CS wrote the manuscript. J-YF and TR provided critical revisions for the paper. S-WP and J-YF were the guarantors of the paper, taking responsibility for the integrity of the work as a whole, from inception to published article. All authors contributed to the article and approved the submitted version.

S-WP was funded by grants from Taipei Veterans General Hospital [V110C-040] and the Ministry of Science and Technology [107-2314-B-075-057, 108-2314-B-075-001, 109-2314-B-075-094, 110-2314-B-075-077, 111-2314-B-075-059] in Taiwan. TR, DC and AC were funded by NIH NIAID grant # R01AI137681.

The authors thank the Medical Science and Technology Building of the Taipei Veterans General Hospital and the Stein clinical research building of the University of California, San Diego, for providing research facilities. The authors also would like to thank Dr. Peter G Chiles for providing help in the experiment consultation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.