Serum Cell-Free DNA-based Detection of Mycobacterium avium Complex Infection

Abstract

Rationale

Mycobacterium avium complex (MAC) is the most common cause of nontuberculous mycobacterial (NTM) pulmonary disease (PD), which exhibits increasing global incidence. Current microbiologic methods routinely used in clinical practice lack sensitivity and have long latencies, leading to delays in diagnosis and treatment initiation and evaluation. A clustered regularly interspaced short palindromic repeats (CRISPR)–based assay that measures MAC cell-free DNA (cfDNA) concentrations in serum could provide a rapid means to detect MAC infection and monitor response to antimicrobial treatment.

Objectives

To develop and optimize a CRISPR MAC assay for MAC infection detection and to evaluate its diagnostic and prognostic performance in two MAC disease cohorts.

Methods

MAC cfDNA serum concentrations were measured in individuals with diagnoses of MAC disease or who had bronchiectasis or chronic obstructive pulmonary disease diagnoses without histories of NTM PD or NTM-positive sputum cultures. Diagnostic performance was analyzed using pretreatment serum from two cohorts. Serum MAC cfDNA changes during MAC PD treatment were evaluated in a subset of patients with MAC PD who received macrolide-based multidrug regimens.

Measurements and Main Results

The CRISPR MAC assay detected MAC cfDNA in MAC PD with 97.6% (91.6–99.7%) sensitivity and 97.6% (91.5–99.7%) specificity overall. Serum MAC cfDNA concentrations markedly decreased after MAC-directed treatment initiation in patients with MAC PD who demonstrated MAC culture conversion.

Conclusions

This study provides preliminary evidence for the utility of a serum-based CRISPR MAC assay to rapidly detect MAC infection and monitor the response to treatment.

At a Glance Commentary

Scientific Knowledge on the Subject

The diagnosis of Mycobacterium avium complex (MAC) pulmonary disease (PD) currently requires the integration of clinical, microbiologic, and radiographic information but ultimately relies on obtaining one or more positive culture results depending on the analyzed sample type and supporting data. This process is slow and complex, as clinical and radiologic information used for diagnosis is nonspecific, and all MAC species that cause respiratory disease can be isolated from environmental sources.

What This Study Adds to the Field

Our results indicate that the detection of MAC cell-free DNA in serum can identify MAC infection with >93% sensitivity and >97% specificity and could thus serve as a rapid means to screen for MAC infections as the initial step in MAC PD diagnoses. Serum MAC cell-free DNA results could therefore potentially supplement or replace slow MAC cultures currently required for MAC PD diagnosis to reduce the time to diagnosis and perhaps improve diagnostic sensitivity.

Mycobacterium avium complex (MAC) is the most common cause of nontuberculous mycobacterial (NTM) pulmonary disease (PD), accounting for 85–90% of all NTM PD cases. NTM PD incidence has increased globally over the past two decades, with MAC responsible for an estimated 90,000 U.S. NTM PD cases in 2017 (1–5). NTM PD is defined by microbiologic evidence (NTM-positive cultures from at least two separate sputum samples or from a bronchial wash/lavage specimen or a lung biopsy with mycobacteria-associated histologic features and a culture-positive sputum or bronchial wash/lavage specimen) and clinical findings (pulmonary or systemic symptoms, radiographic abnormalities, and the exclusion of alternative diagnoses) (6, 7). Broth cultures normally require 12–16 days to identify MAC species that frequently cause NTM PD, but diagnostic sample quality and quantity variations can delay microbiologic diagnosis and the evaluation of therapy responses (8). Rapid and reproducible diagnostic assays are thus needed to improve the speed and accuracy of NTM PD diagnosis and treatment response evaluations.

MAC bacilli and other pathogens release cell-free DNA (cfDNA) into the circulation as they are lysed and cleared by the immune response of their host (9) to allow minimally invasive diagnosis from a blood sample versus a specimen derived from the infection site (10, 11). Conventional molecular diagnostic approaches, including PCR methods, often lack the sensitivity required to consistently detect low concentrations of pathogen-derived cfDNA in blood, particularly when pathogen burden may be low (e.g., early in infection or before the initiation of an antimicrobial response). New clustered regularly interspaced short palindromic repeats (CRISPR)–based cfDNA assay can, however, improve cfDNA detection efficiency to allow the detection of pathogen-derived cfDNA in blood samples for specific diagnoses and timely treatment monitoring (12, 13).

Guide RNA–mediated binding of a Cas12a/guide RNA complex to its target sequence stimulates its collateral cleavage activity to degrade a quenched fluorescent oligonucleotide probe to permit specific, ultrasensitive, and concentration-dependent target DNA detection (14, 15). We hypothesized that serum MAC cfDNA could accurately detect MAC infection and monitor its response to treatment, as serum MAC cfDNA concentrations should reflect real-time MAC burden and be unaffected by sputum variations that can affect current methods. In this study, we therefore developed a sensitive CRISPR assay for serum MAC cfDNA to permit the rapid and accurate non–sputum-based detection of MAC infection, evaluation of its response to treatment, and detection of disease recurrence. Our results indicate that CRISPR MAC serum results accurately detect MAC infection and rapidly and progressively change after MAC-directed therapy initiation, suggesting that serum MAC cfDNA is an effective biomarker of MAC disease and its response to treatment.

Methods

Study Population

Serum samples analyzed in this retrospective molecular diagnostic study were obtained from cryopreserved samples collected from individuals enrolled in the Oregon Health & Science University (OHSU) Northwest NTM Biobank or received by the mycobacteriology laboratory of the Department of Laboratory Medicine at the NIH Clinical Center (Figure 1). OHSU biobank enrollment included collection clinical data from a 12-month period before enrollment and laboratory specimens and patient-reported outcome measures at enrollment. A subset of these patients with MAC PD had serum collected both before and treatment initiation and were evaluated for longitudinal serum MAC cfDNA changes. Individuals were enrolled regardless of their disease stage, comorbidity (cavitary disease, bronchiectasis, and chronic obstructive PD [COPD]), or antibiotic treatment history. Symptom, comorbidity, and treatment data were collected from physician notes. Radiographic findings were derived from radiology reports. Concomitant medication information was obtained from physician notes and the medication fields of electronic records of visits associated with respiratory culture collection (see Table E1 in the online supplement). The discovery cohort for the CRISPR MAC assay analyzed archived sera from 71 individuals enrolled in the OHSU biobank who met American Thoracic Society/European Respiratory Society/European Society of Clinical Microbiology and Infectious Diseases/Infectious Diseases Society of America criteria for MAC PD, including 14 who had samples and data collected at 3 and 6 months after treatment initiation and 16 who had samples collected within 12 months after treatment initiation and who did and did not exhibit culture conversion (Tables 1 and 2). Sera from 15 individuals enrolled in the OHSU biobank who had diagnoses of bronchiectasis or COPD without histories of NTM PD or NTM-positive respiratory cultures were analyzed as a disease control cohort. The validation cohort analyzed archived sera from 12 individuals who met American Thoracic Society/European Respiratory Society/European Society of Clinical Microbiology and Infectious Diseases/Infectious Diseases Society of America criteria for MAC PD, 3 individuals with diagnoses of disseminated MAC disease who had sera archived at the NIH Clinical Center, and an independent set of control subjects with negative NTM culture results who lacked histories of respiratory disease, had diagnoses of COPD or bronchiectasis, or were successfully treated for tuberculosis. Serum samples were processed within 2 hours of collection and then stored at −80°C until shipped on dry ice to the analysis laboratory for DNA extraction and cfDNA analysis.

Figure 1.

Study participants. (A and B) Clinical information and archived serum from patients with Mycobacterium avium complex (MAC) pulmonary disease (PD) and control subjects without nontuberculous mycobacterial (NTM) disease enrolled in the Oregon Health & Science University (OHSU) Northwest NTM Biobank (A) and patients with MAC PD/disseminated MAC seen at the NIH Clinical Center and non-NTM control subjects enrolled in the OHSU Northwest NTM Biobank (B) were respectively used to generate a discovery and a validation cohort to assess CRISPR MAC assay diagnostic performance. All MAC PD cases met American Thoracic Society/European Respiratory Society/European Society of Clinical Microbiology and Infectious Diseases/Infectious Diseases Society of America disease criteria (6) and included a subgroup of individuals who initiated MAC-directed treatment. Non-NTM control patients had diagnoses of bronchiectasis or COPD but had no histories of NTM PD or NTM-positive respiratory cultures. A subset of OHSU biobank patients with MAC PD with follow-up samples after treatment initiation were evaluated to analyze the serum MAC cfDNA response to treatment. cfDNA = cell-free DNA; COPD = chronic obstructive pulmonary disease; CRISPR = clustered regularly interspaced short palindromic repeats; Rx = medication.

Table 1.

Baseline Demographic and Clinical Characteristics of Control Subjects without Nontuberculous Mycobacterial Disease, All Patients with Mycobacterium avium Complex Pulmonary Disease, and a Longitudinal Treatment Subgroup in the Discovery Cohort

Variable	Control (n = 15)	MAC PD (n = 71)	MAC PD Rx (n = 14)	P Value* (MAC PD vs. Control)
Age, yr, median (IQR)	66 (59–72)	69 (62–76)	71 (55–74)	0.35
Sex (female)	9 (60.0%)	54 (76.1%)	11 (78.6%)	0.34
Bronchiectasis	13 (86.7%)	62 (87.3%)	14 (100%)	1.00
COPD	4 (26.7%)	21 (29.6%)	2 (14.3%)	1.00
Cavitary disease	—	9 (12.7%)	2 (14.3%)	—
Prior NTM treatment	—	14 (19.7%)	1 (7.1%)	—
Enrollment
Culture +, smear +	—	22 (31.0%)	7 (50.0%)	—
Culture +, smear −	—	46 (64.8%)	6 (42.9%)	—
Culture +, smear unknown	—	3 (4.2%)	1 (7.1%)	—
RSS, median (IQR)†	—	63.0 (46.7–76.4)	66.7 (46.7–73.3)	—
Vitality score, median (IQR)†	—	44.4 (30.6–66.7)	55.6 (33.3–66.7)	—

Definition of abbreviations: COPD = chronic obstructive pulmonary disease; IQR = interquartile range; MAC = Mycobacterium avium complex; NTM = nontuberculous mycobacterial; PD = pulmonary disease; RSS = respiratory symptom score; Rx = medication.

^*P values were calculated using the Wilcoxon rank sum test for ordinal variables and the chi-square test or Fisher exact test (two sided) for categorical variables.

^† Quality of Life–Bronchiectasis RSS and vitality score were calculated and scaled from 0 to 100, with higher values indicating fewer symptoms.

Table 2.

Baseline Demographic and Clinical Characteristics of the Validation Cohort

Variable	Control (n = 67)	MAC PD (n = 12)/Disseminated MAC (n = 3)	P Value*
Age, yr, median (IQR)	73 (65–77)	63 (51–71)	0.02
Sex (female)	30 (44.8%)	12 (80.0%)	0.02
Bronchiectasis	2 (3.0%)	10 (66.7%)	0.00
COPD	47 (70.1%)	0 (0%)	0.00
Healthy	18 (26.9%)	—	—
Culture +, smear +	—	6 (40.0%)	—
Culture +, smear −	—	6 (40.0%)	—
Culture +, smear unknown	—	3 (20.0%)	—

Definition of abbreviations: COPD = chronic obstructive pulmonary disease; IQR = interquartile range; MAC = Mycobacterium avium complex; PD = pulmonary disease.

^*P values were calculated using the Wilcoxon rank sum test for ordinal variables and the Fisher exact test (two sided) for categorical variables.

All individuals provided written informed consent for the future use of their samples and deidentified data before participation. The Institutional Review Board at OHSU approved the biorepository protocol, and all NIH Clinical Center samples were provided by patients who granted consent under NIH-approved protocols. Individuals were eligible for inclusion in the CRISPR MAC diagnostic study if they had serum collected within 6 months of a positive MAC culture result and before initiating MAC-directed treatment. Samples with all required clinical data were analyzed to evaluate CRISPR MAC diagnostic performance and data from a subset of these patients who initiated MAC-directed therapy and had samples collected during treatment were used to evaluate serum MAC cfDNA changes after treatment initiation.

Blood-based CRISPR MAC Assay

The CRISPR MAC assay was developed and optimized as described in the online supplement (see pp. E2–E7 and Figures E1–E4) using genomic DNA or culture lysates obtained from a commercial source and varying assay conditions to determine reaction temperature, incubation times, and reagent ratios that produced the most robust signal for a given amount of input target and the number of PCR cycles required to maximize target detection. For the optimized CRISPR MAC assay, total cfDNA was extracted from 200 μl serum using the Quick-cfDNA Serum and Plasma Kit (Zymo Research), and 10% (5 of 50 μl) of the extracted cfDNA sample was subjected to PCR amplification in a 25-μl PCR. Finally, 2 μl this completed PCR was transferred to a 30-μl CRISPR MAC assay reaction, and assay signal was detected in a fluorescent plate reader (485-nm excitation, 535-nm emission) after a 20-minute incubation at 37°C in the dark. Samples were analyzed in triplicate to determine mean values, and replicates with coefficients of variation ⩾20% were considered invalid and reanalyzed or excluded from study analyses depending on sample availability for reanalysis. The positive signal threshold determined in the method development process was set at the mean + 3 × SD of the signal detected in the negative control samples to exclude 99.85% of the false-positive signals arising from signal variance (16).

Statistical Analyses

Participant characteristics were summarized using medians, interquartile ranges, and proportions. Patient-reported respiratory symptom scores (RSSs) and vitality scores measured using the Quality of Life–Bronchiectasis questionnaire were calculated, with scores scaled from 0 to 100 (higher values indicating fewer symptoms) (17, 18). The Wilcoxon rank sum test was used to compare MAC cfDNA concentrations between patients with MAC disease and control subjects and among patients with MAC disease with different comorbidities. The Wilcoxon signed rank test was used to compare MAC cfDNA concentrations at different treatment intervals. Categorical variables were analyzed using the chi-square test or the Fisher exact test. All tests were two sided, with an α value of 0.05. All analyses were done using SPSS version 26.0 (IBM SPSS 26.0). GraphPad Prism version 10.0.2 was used for data visualization.

Results

CRISPR MAC Diagnostic Performance

The optimized CRISPR MAC assay detected MAC cfDNA with a limit of detection of 0.125 copies/μl (see pp. E6–E8, Figures E1–E4, and Table E2). CRISPR MAC assay diagnostic performance was first analyzed in a discovery cohort of patients with MAC PD and non-NTM control subjects (Figure 1A and Table 1), who had similar mean ages (P = 0.35, chi-square test), sex compositions (P = 0.34, chi-square test), and bronchiectasis and COPD diagnosis frequencies (P = 1.00, Fisher exact test).

CRISPR MAC signal was detected in pretreatment sera of all but one patient with MAC PD but none of the NTM control subjects (Figures 2A and 2B), yielding diagnostic sensitivity and specificity estimates of 98.6% (95% confidence interval, 92.4–100%) and 100% (78.2–100%), respectively. Notably, the serum sample of the only patient with MAC PD with a false-negative MAC CRISPR result had substantial hemolysis, which might have interfered with target detection, but lacked sufficient volume for reanalysis after reprocessing to reduce this interferent. CRISPR MAC signal detected in individuals with M. avium and M. intracellulare infections was similar (Figure 2C). Cavitary disease was more common in smear-positive versus smear-negative MAC PD cases (27.3% [6 of 22] vs. 6.5% [3 of 46]), indicating that smear-positive cases tended to have more severe disease, but CRISPR MAC signal did not significantly differ with either cavitary disease or smear positivity status (Figures 2D and 2E).

Figure 2.

Clustered regularly interspaced short palindromic repeats (CRISPR) Mycobacterium avium complex (MAC) diagnostic assay performance. (A) Culture, smear, and MAC cell-free DNA (cfDNA) data for 71 individuals with MAC pulmonary disease (PD) before treatment initiation. (B–E) CRISPR MAC signal detected in (B) pretreatment sera of 15 control subjects and 71 individuals with MAC PD, (C) individuals infected with M. avium or M. intracellulare, (D) smear-positive and smear-negative individuals, and (E) patients with MAC PD with and without cavitary disease. (F) Culture and MAC cfDNA results for the MAC PD/disseminated MAC, disease control (nontuberculous mycobacterial culture negative with chronic obstructive pulmonary disease [COPD] and/or bronchiectasis), and healthy control groups. (G) CRISPR MAC signal detected in pretreatment sera of the MAC PD, disseminated MAC, disease control, and healthy control groups. (H and I) CRISPR MAC signal differences between (H) healthy control subjects and disease control subjects with COPD or bronchiectasis and (I) patients with MAC PD with M. avium and M. intracellulare infections (median with interquartile range). P values were calculated using two-sided Wilcoxon rank sum tests. For cfDNA, (+), (++), and (+++) indicate low to high CRISPR MAC signal tertiles. Graphs depict mean ± SD of three technical replicates for each sample; the red dashed line indicates the threshold for a positive CRISPR MAC signal. a.u. = arbitrary units; unk = unknown.

Similar results were observed in a validation cohort composed of 12 individuals with MAC PD, 3 with disseminated MAC, 49 with COPD or bronchiectasis (disease control subjects), and 18 without known respiratory disease (healthy control subjects) (Table 2 and Figure 1B), in which CRISPR MAC results yielded MAC PD sensitivity and overall specificity estimates of 91.7% (61.5–99.8%) and 97.0% (89.6–99.6%) (Figures 2F and 2G), after excluding the three disseminated MAC cases. Strikingly, the three highest MAC cfDNA signals detected in this analysis were all detected in samples from patients with disseminated MAC, likely reflecting the increased release of MAC cfDNA from multiorgan MAC infections arising from immunosuppression. No differences were detected in the CRISPR MAC signal of the healthy control and disease control groups, and the two false-positive signals in these groups were both lower than true-positive signals in the MAC PD group (Figure 2H). These two healthy control and disease control false-positive results were respectively observed in serum collected from individuals three months after completing treatment for central nervous system tuberculosis and after hospitalization with moderate to severe COPD and congestive heart failure for acute pneumonia, but neither individual subsequently received a diagnosis of or was treated for NTM infection. Finally, MAC cfDNA signal did not significantly differ in validation cohort of patients with MAC PD with diagnoses of M. avium and M. intracellulare, similar to results detected in the discovery cohort (Figure 2I).

Serum MAC cfDNA Changes after Treatment Initiation

Serum MAC cfDNA should decrease with reductions in pulmonary MAC bacilli after the initiation effective treatment regimens. We therefore next evaluated if CRISPR MAC assay signal decreased after treatment initiation in a subgroup of patients with MAC PD who had longitudinal sera collected during multidrug regimens using azithromycin, ethambutol, and rifampin; azithromycin and ethambutol; or azithromycin and at least one other drug (MAC PD medication [Rx] cohort; Figure 1A). All these individuals had positive sputum MAC culture results and bronchiectasis at enrollment, but many (42.9%) had negative acid-fast bacilli smear results consistent with paucibacillary disease (Figure 3A), and only two (14.3%) received diagnoses of cavitary disease (Table 1). This MAC PD Rx subgroup had proportionally more bronchiectasis and fewer COPD cases than the overall MAC PD discovery cohort, although cavitary disease was rare in both groups, as was a history of previous NTM treatment. MAC PD Rx cohort patients also had higher median RSSs and vitality scores, consistent with less severe MAC PD pathology, despite having proportionally more culture-positive and smear-positive results suggestive of greater bacillary load.

Figure 3.

Clustered regularly interspaced short palindromic repeats (CRISPR) Mycobacterium avium complex (MAC) assay monitors treatment response. (A and B) Culture, smear, and MAC cell-free DNA (cfDNA) results (A) and changes in serum CRISPR MAC signal (B) for MAC PD medication (Rx) cohort subjects at the indicated treatment intervals. (C) Change in CRISPR MAC signal and respiratory symptom score (RSS) results for MAC PD Rx cohort subjects at 3 months after enrollment; the circled data point indicates data from a patient with both bronchiectasis and COPD but who revealed culture conversion, showing an RSS decrease despite a marked decrease in CRISPR MAC signal. (D) Culture, smear, and MAC cfDNA results for patients with MAC PD who did and did not respond to treatment within a 12-month follow-up interval, as determined by culture conversion. (E and F) Changes of MAC cfDNA concentrations in (E) responders and (F) nonresponders before and after 12-month treatment. P values were calculated using two-sided Wilcoxon signed rank tests. For cfDNA, (+), (++), and (+++) indicate low to high CRISPR MAC signal tertiles. Graphs depict mean ± SD of three technical replicates for each sample; the red dashed line indicates the threshold for a positive CRISPR MAC signal. a.u. = arbitrary units.

Most MAC PD Rx cohort patients with available follow-up culture results had culture conversions after treatment initiation, with only one remaining both culture and smear positive at follow-up (Figure 3A). Most of these individuals (85.7%) also had significant MAC cfDNA decreases at 3-month follow-up that further decreased to approach the minimum positive signal threshold at 6-month follow-up (Figure 3B). MAC cfDNA signal did not markedly change at follow-up in the two patients with MAC PD who had the lowest values at enrollment, and these results cannot be explained using their available clinical information.

Vitality scores did not significantly improve with CRISPR MAC decreases after treatment, possibly because of the short post-treatment monitoring period (see Figure E5 and Table E1). RSSs in most patients improved by 3-month follow-up (see Figure E5), but these changes did not directly correspond with their MAC cfDNA decreases (Figure 3C), potentially because of individual variation in disease recovery. For example, three patients with MAC cfDNA signal decreases lacked RSS improvements, and two patients revealed contradictory MAC cfDNA signal and RSS changes, including one who had a marked MAC cfDNA signal decrease with a counterintuitive RSS and vitality score decrease (circled data point, Figures 3C and E6). This patient was culture negative at 3-month follow-up, but his RSS and vitality score decreases at 3-month follow-up might have been due to deterioration of his bronchiectasis and COPD.

Subsequent analysis of a second cohort of patients with MAC PD (Figure 3D) who had serum collected at variable times within 12 months after treatment initiation detected serum MAC cfDNA decreases at these time points. These groups did not significantly differ by age, sex, NTM treatment history, or frequency of bronchiectasis, COPD, and cavitary disease (Table 3) or by mean MAC cfDNA concentration before treatment initiation. Smear-negative and smear-positive rates and RSSs also did not differ between these groups, and an observed vitality score difference could be an artifact of sample size, particularly as it was not matched by corresponding RSS and smear result differences. Significant MAC cfDNA decreases were observed only in patients with MAC PD with at least two negative culture results after treatment initiation (responders), although accurate assessment of the MAC cfDNA response in patients without two negative culture results (nonresponders) was limited by the smaller size of this group (Figures 3E and 3F).

Table 3.

Baseline Demographic and Clinical Characteristics of the Mycobacterium avium Complex Pulmonary Disease Discovery Cohort of Patients Who Did and Did Not Respond to Treatment

Variable	Responders* (n = 12)	Nonresponders† (n = 4)	P Value‡
Age, yr, median (IQR)	67 (62–71)	68 (64–70)	0.90
Sex (female)	9 (75.0%)	3 (75.0%)	0.73
Bronchiectasis	12 (100.0%)	4 (100.0%)	—
COPD	4 (33.3%)	2 (50.0%)	0.60
Cavitary disease	1 (8.3%)	2 (50.0%)	0.14
Prior NTM treatment	4 (33.3%)	1 (25.0%)	1.00
Enrollment
Culture +, smear +	2 (16.7%)	2 (50.0%)	0.25
Culture +, smear −	10 (83.3%)	2 (50.0%)	0.25
RSS, median (IQR)§	40.7 (35.2–51.9)	70.4 (57.4–81.0)	0.12
Vitality score, median (IQR)§	26.2 (22.2–33.3)	61.1 (55.6–72.2)	0.02

Definition of abbreviations: COPD = chronic obstructive pulmonary disease; IQR = interquartile range; NTM = nontuberculous mycobacterial; RSS = respiratory symptom score.

^*Responders were culture converters (two or more negative culture results) after 12-month treatment.

^† Nonresponders remained culture positive after 12-month treatment.

^‡ P values were calculated using the Wilcoxon rank sum test for ordinal variables and the chi-square test or Fisher exact test (two sided) for categorical variables.

^§ Quality of Life–Bronchiectasis RSS and vitality score were calculated and scaled from 0 to 100, with higher values indicating fewer symptoms.

Taken together, these data indicate that CRISPR MAC results can distinguish patients with MAC PD from healthy and disease control subjects and rapidly detect serum MAC cfDNA changes after MAC-directed therapy initiation, suggesting its potential for detecting MAC infection and treatment evaluation after future validation studies.

Discussion

Herein we describe the performance of a novel CRISPR MAC assay that detects serum concentrations of MAC cfDNA to rapidly identify MAC infection and monitor its response to treatment, two essential clinical needs for MAC disease management. Sputum-based assays for MAC diagnosis and evaluation are suboptimal, particularly for patients receiving treatment, who normally produce little or no sputum. MAC species are slow growing and usually require 12 to 16 days to yield detectable positive growth results in broth culture (8), but real-time PCR assays used for rapid MAC diagnosis have only 71% diagnostic sensitivity (19). No blood-based assay is routinely used to diagnose MAC disease, and an investigational diagnostic assay that detects the IgA response to a MAC-specific glycopeptidolipid has only moderate (84.3%) sensitivity (20) and is unlikely to effectively monitor treatment responses because of the persistence of these IgA concentrations after infection. The CRISPR MAC assay, however, detected MAC PD cases with overall 97.6% (91.6–99.7%) sensitivity and 97.6% (91.5–99.7%) specificity in this study.

MAC disease treatment typically extends 12 months past culture conversion and routinely requires 18 months to complete (21). Evaluating MAC responses to treatment also relies on qualitative or semiquantitative tests (e.g., sputum culture and radiography) and clinical assessments (e.g., reductions in MAC-related symptoms and improvements in quality-of-life measurements), which have limited utility for evaluating treatment success (22). Notably, a substantial fraction of patients with MAC disease still exhibit symptoms and radiographic findings consistent with MAC PD after treatment completion, and ⩾50% of these patients have subsequent NTM-positive respiratory cultures within 12–36 months (23), suggesting that they have residual disease after treatment completion.

Several studies have analyzed host-derived factors in liquid biopsies as indicators of MAC treatment response (24–27). However, these factors (MAC-specific IgA, IL-17, IL-23, and carbohydrate antigen 19-9) provide indirect or nonspecific evidence of MAC disease, may not correspond with MAC burden, or may be influenced by other infections or proinflammatory conditions. Serum MAC cfDNA concentrations, however, reflect the real-time release of MAC genomic DNA from dying MAC bacilli because of the short half-life of cfDNA in the circulation (28). MAC cfDNA concentrations detected in serial blood samples collected during treatment should therefore correlate with changes in MAC burden and function as a reasonable surrogate of MAC clearance and an effective treatment response. Consistent with this hypothesis, serum MAC cfDNA concentrations significantly decreased at three and six months after MAC-directed treatment initiation in patients with MAC PD receiving multidrug treatment regimens in this study and corresponded with RSS increases. This may be noteworthy, as demonstrating a correlation between clinical and microbiological responses to treatment has been problematic.

Caution must be exercised when interpreting these results, however, because of the limited number of patients available for this analysis, but the rate and degree of the serum MAC cfDNA changes observed strongly suggest that it may serve as a useful surrogate measure of an effective treatment response. Larger studies with extended follow-up and MAC culture time-to-positivity results will be required to validate this finding and to determine if MAC cfDNA clearance correlates with sputum clearance of viable MAC bacilli and lasting MAC remission and can thus a serve as an endpoint for effective MAC treatment.

Several factors may limit the interpretation and generalizability of these findings. First, we analyzed cryopreserved serum, which may have influenced cfDNA extraction and stability, as serum is prone to genomic DNA contamination arising from cell lysis during the coagulation and centrifugation procedures (29) and contains nucleases that could degrade serum cfDNA at any stage before its isolation. Both these factors could have reduced CRISPR MAC assay sensitivity or introduced variability. Ethylenediaminetetraacetic acid plasma samples may provide more reliable results, as these samples would not be affected by genomic DNA release during coagulation and would be less affected by nuclease degradation because of their more rapid processing procedure and the inclusion of ethylenediaminetetraacetic acid, which should directly attenuate nuclease activity.

Second, results from patients with MAC PD detected at three and six months after treatment initiation were compared with results from sputum culture, the reference standard, but sputum culture is known to provide false-negative diagnostic results early after treatment initiation, especially among patients who have difficulty producing sputum. Future studies performed in large prospective cohorts will be required to better understand the diagnostic and prognostic performances of serum MAC cfDNA.

Third, none of the sera analyzed in this study were obtained from individuals infected with M. chimaera, which was detected at lower signal intensity in our optimization studies. Our current CRISPR MAC assay could thus exhibit reduced diagnostic sensitivity for M. chimaera versus M. avium and M. intracellulare, although target sequences from all three species exhibited equivalent limits of detection in our optimization studies. Further studies are needed to address this question and potentially to further optimize the CRISPR MAC amplification primers to avoid a potential decreased sensitivity for M. chimaera.

Finally, data available for these patients did not permit the analysis of potential correlations between serum MAC cfDNA signal and pulmonary MAC burden or disease involvement, which could be addressed in future studies that measure sputum MAC bacilli and DNA concentrations and radiological involvement in MAC disease cohorts before and after treatment initiation.

Conclusions

This study provides evidence that a rapid serum-based CRISPR MAC assay suitable for use in clinical settings can sensitively and specifically detect and quantify serum MAC cfDNA to accurately identify MAC infection and monitor its response to treatment. We believe that these results provide strong preliminary evidence for the utility of a serum-based CRISPR MAC assay to rapidly detect MAC infection as well as a potential secondary application for serum-based monitoring MAC treatment response and MAC recurrence.