(530) Airway Volatile Organic Compounds (VOCs) Can Achieve Culture-Independent Identification of Nontuberculous mycobacteria in People with Cystic Fibrosis

Background: Nontuberculous mycobacteria (NTM) are frequent pathogens in people with cystic fibrosis (pwCF) with a 10% prevalence in the 2020 US Cystic Fibrosis Foundation Patient Registry. NTM has been increasingly recognized and reported as a threat and can lead to higher morbidity in pwCF. Sputum culture combined with clinical presentation is the current gold standard for diagnosis of NTM disease, a process which can take months.

Analysis of volatile organic compounds (VOCs) originating from airway samples, including sputum and breath, is now recognized as an innovative and non-invasive approach to identify biomarkers associated with specific airway pathogens [1]. Here we tested our hypothesis that VOCs present in the headspace of sputum samples can differentiate between CF patients infected with M. abscess (MABs) and M. avium complex (MAC).

Methods: Sputum samples were collected from CF patients enrolled in the long-term multi-site NTM clinical trials: PREDICT and PATIENCE. Samples were pretreated before being extracted by headspace solid-phase microextraction (HS-SPME) with a DVB/CAR/PDMS SPME fiber. The SPME fibers were thermally desorbed into the two-dimensional gas chromatography time-of-flight mass spectrometry (GC×GCtofMS) for analysis. The resulting data was subjected to unsupervised and supervised data analysis tools including principal component analysis (PCA), t-distributed stochastic neighbor embedding (tSNE), and the Boruta feature selection method. The SPME extraction and GC×GCtofMS analysis were carried out on 46 sputum samples, from 29 CF patients infected with MAC and 17 CF patients with MABs. Out of 46 individuals, 25 patients showed positive culture results at the day of sample collection.

Results: To test the utility of the sputum headspace volatile features to differentiate MAC and MABs, we first built a multivariate model (i.e. Model 1) using a subset of samples showing positive culture results at the day of sample collection (nMAC=15, nMABs=10). The headspace of sputum samples contained over 300 volatile features. After the removal of features associated with plastic contaminants and instrument artifacts, eight features driving the difference between the two groups were selected using the Boruta algorithm. The tSNE in Figure 1a visualizes the eight selected biomarkers showing promising differentiation between patients infected with MAC and MABs.

We were further interested to find the subset of volatile features for discriminating between patients infected with MAC and MABs regardless of their culture results at the day of sample collection. In order to achieve this, two more models including samples with positive culture results in a time window of ±30 days of sample collection (i.e. Model 2, nMAC=19, nMABs=14) and ±90 days of sample collection (i.e. Model 3, nMAC=29, nMABs=17), were built following the same data analysis workflow. Both models showed meaningful separation between MAC and MABs samples and we found a list of molecules common between all three discriminative models (see Fig. 1b).

Conclusions: The result in this work paves the way for the development of a culture-independent diagnosis tool for specific differentiation between CF patients infected with MAC and MABs.

Acknowledgements: Funded by CFF (NICK18P0) and NIH (R01HL146228)

References: [1] Bobak C A, Kang L, Workman L, Bateman L, Khan M S, Prins M, May L, Franchina F A, Baard C, Nicol M P, Zar H J, Hill J E. Breath can discriminate tuberculosis from other lower respiratory illness in children. Sci Rep. 2021;11:2704