Background

Tuberculosis (TB) is a serious global health concern, with an estimated 10.6 million cases and 1.3 million fatalities in 2022 [1]. Not only does TB remain challenging to diagnose with an urgent need for non-sputum based tests but the characteristics of cough aerosols from people with TB, which are a determinant of transmission success, are poorly understood [2, 3].

Breath-based detection of Mycobacterium tuberculosis complex (MTBC) DNA is a promising non-sputum method for diagnosing TB, with face masks and blow tubes under evaluation [4,5,6]. PneumoniaCheck (PMC), a cough aerosol collection device, has been evaluated in people with cystic fibrosis (CF) [7] and viral pneumonia [8]. In the first study, CF-related bacteria were in the aerosols of 65% of people sputum-positive for these bacteria (and aerosol did not contain lung commensals found in sputum). In the second study, when pathogen readouts from PMC and bronchoalveolar lavage fluid (BALF) were compared, 66% of aerosols were PCR-positive. PMC, is however, unevaluated in TB where, in addition to detecting MTBC in exhaled aerosols, it could be used to characterise the aeromicrobiome, which may influence contacts’ immune responses.

The microbiota is a topic of increasing interest in TB, where sputum is widely studied. However, the sputum microbiota more closely resembles the upper respiratory tract (URT) than the lower respiratory tract (LRT) [9, 10], which is the primary site-of-disease in TB. Sampling the LRT is difficult because it requires bronchoscopy [11], which is invasive, ethically complex for research purposes only, and expensive, often rendering it unfeasible in large cohorts where TB is prevalent [12]. Aerosols, which are more accessible, could be a useful proxy for studying the LRT, as aerosols partly originate from the LRT [13].

PMC comprises a 250mL air reservoir (bag) attached to a mouthpiece with a filter. PMC is designed to separate aerosols from the URT and LRT into the bag and filter, respectively [14]. This separation occurs after a person coughs into the PMC, at which point air from the anatomical URT dead space (~ 150mL) [15] flows first into the bag. Air after the 150mL is likely from the LRT and then, due to backpressure from the inelastic bag, directed towards the filter [7].

It is thus possible that, in addition to PMC-captured aerosols being useful for TB diagnosis, such aerosols may serve as an alternative to BALF for LRT microbiota characterization and more accurately represent the vehicle of TB transmission, including compared to oral washes and sputum. We therefore evaluated MTBC detection by Xpert MTB/RIF Ultra (Ultra) and the bag and filter microbiota. We compared microbiota in the bag and filter to URT and LRT clinical samples.

Methods

Ethics

The study was conducted in accordance with the Declaration of Helsinki. Stellenbosch University Health Research Ethics Committee (SU-HREC) approved the study (N14/10/136, N16/05/070). People provided a written informed consent.

Recruitment and data collection

This study involved people (n = 16; ≥18 years) enrolled at primary healthcare facilities in Cape Town. People were sputum MTBC culture-positive and not on treatment. Clinical and demographic data were collected.

Sample collection

People first gave aerosols collected using PMC and sputum was then induced. For aerosols, people were asked to take deep breaths to stimulate a cough. While sealing lips around the PMC mouthpiece ridge, with teeth rested on the device notches, people were asked to produce ≥ 30 deep coughs (Fig. 1). Bags were deflated by hand squeezing after each cough. People who had adverse effects like dizziness could rest between coughs and, if they produce sputum during coughing, were given a jar in which to expectorate. DNA sampling background controls (BKG) were collected to identify potentially contaminating taxa and included an unused PMC handled in the same manner as those used by people. Sputum was induced with 5% saline for 10 min. Oral washes (representing URT) and BALF (representing LRT) were also collected in a subset (n = 6) of the 16 people (Supplementary Methods). This subset had been enrolled into a separate study examining site-of-disease immunological signatures (NCT03350048).

Fig. 1
figure 1

How to use pneumonia check (PMC): People seal their lips around the mouthpiece ridge and cough into the PMC. As they exhale, air from the anatomical dead space moves into the inelastic air reservoir (bag), while additional air, primarily from the lower respiratory tract (LRT) is directed towards the filter

Full size image

Respiratory fluids processing

Oral washes and sputum were decontaminated using N-acetyl-L-cysteine (NALC), pelleted (3,217xg), resuspended in 2mL PB (pH 6.8; BD, South Africa) and stored at -80 °C [16]. 1mL raw BALF aliquots were stored at -80 °C for microbiota analysis. MTBC in sputum was detected using the Xpert MTB/RIF (Xpert) and Mycobacteria Growth Indicator Tube (MGIT) 960 liquid culture according to the manufacturers’ methods.

Recovery and processing of aerosols

Aerosols were separately recovered from the PMC bag and filter. After removal under sterile conditions, bags were rinsed and incubated for 15 min in 10mL stripping buffer (1% Triton X-100 in 10mM Tris-HCl, pH 8.0) whereas the filter was submerged in 10mL stripping buffer, vortexed vigorously for one minute, and incubated for 30 min. Bag and filter washes were pelleted (3,217xg) and resuspended in 1.5mL PB. Aliquots of 0.7mL each for MTBC testing using Ultra (version 2, according to the manufacturers’ methods) and microbiota analysis were stored at -80 °C.

16S rRNA gene sequencing and analysis

Microbial DNA was extracted using the QIAamp DNA Mini kit (QIAGEN, Hilden, Germany). 16S rRNA gene sequencing (V4 region, 150 bp read length, paired end) was done on Illumina MiSeq platform [16, 17]. Sequences were analysed using QIIME 2-2020.2 [18]. We included samples with a minimum read count of 1000. Reads were clustered into amplicon sequence variants (ASVs) using DADA2 (version 1.1.6) [19] and taxonomy assigned at 99% similarity against GreenGenes [20]. α-Diversity (Shannon index) and β-diversity (Bray-Curtis dissimilarity index) were calculated using vegan [21]. Non-parametric methods (Mann-Whitney, Wilcoxon, Kruskal-Wallis, or Friedman tests for unpaired and paired comparisons) were used, whilst permutational multivariate analysis of variance (PERMANOVA) was used for β-diversity. DESeq2 [22] was used for differential abundance analyses with Benjamini-Hochberg multiple comparison adjustment. Potentially contaminating background taxa were identified using decontam (version 1.14.0) [23]. Taxa identified as possible contaminants were not removed from downstream analysis but greyed-out in volcano plots only if identified as differential.

Results

Study population

We included 16 people of median (IQR) age of 35 (25, 46) years. 31% (5/16) were HIV-positive, 75% (12/16) tobacco smokers, and 94% (15/16) of mixed ancestry (Table 1). Median (IQR) days to culture-positivity (TTP) was 5 (5, 7).

Table 1 Demographic and clinical characteristics. Data are median (IQR) or n (%)
Full size table

Ultra on aerosols

Ultra sensitivity on bags and filters washes was 54% (7/13) and 43% (6/14, p = 0.568), respectively (Table 2). No sputum culture TTP differences occurred when compared based on the bag or filter Ultra results nor were there differences in Ultra-generated IS1081-IS6110 CTs. Ultra-generated sample processing control (SPC) cycle threshold (CT) values, a measure of PCR inhibition [24], were similar when across bag and filter washes and when compared across people who had Ultra-positive vs. -negative bags or Ultra-positive vs. -negative filters.

Table 2 Summary of ultra-based MTBC detection in aerosols, showing Ultra results when applied to the bag and filter of PMC. There was no difference in sputum bacillary load (TTP) and PCR-inhibition across all comparisons. Data are % (n/N) or median (IQR) or mean (± SD)
Full size table

A distinct aeromicrobiome is detectable over and above background aerosol

The number of reads did not differ in sputum, bags, and filters in per person comparisons (nor in people who also had oral washes and BALFs sequenced) (Figure S1a-b). Given aerosols have low microbial biomass, we compared the microbiota in aerosol-exposed bags and filters to BKG to check for contamination. α-Diversity was similar between BKG and aerosols (p = 0.430, Figure S2a), but β-diversity differed (PERMANOVA; p = 0.001, Figure S2b), with some bag and filter samples grouping with BKG. Potential bag- and filter-contaminating taxa included Pseudomonas-, Anaerosinus-, Alistipes and-Kocuria (full list in Table S1, Figure S3), however, none of these were differential in other analyses.

Compared to sputum, the aeromicrobiome differs and Mycobacterium is more detectable in the filter

Aerosols vs. sputum

α-Diversity (p = 0.223) was similar yet β-diversity (PERMANOVA; p = 0.001) differed (Fig. 2a-b). Aerosols were compositionally distant from sputum (Figure S4a-b). Bags were Sphingobium-, Corynebacterium- and Novosphingobium-enriched, and Anaerosinus-, Streptococcus- and Neisseria-depleted vs. sputum (Fig. 2c). Filters were Mycobacterium-, Sphingobium- and Corynebacterium-enriched, and Anaerosinus-, Streptococcus- and Neisseria-depleted vs. sputum (Fig. 2d).

Fig. 2
figure 2

Aerosol microbiota is compositionally distinct from sputum: (a) paired α-diversity (Shannon index) using Friedman test for bag, filter, and sputum. (b) β-diversity (Bray-Curtis dissimilarity index), shows distinct clustering of sputum and aerosols. Volcano plots showing differentially abundant taxa in (c) bag vs. sputum, and (d) filter vs. sputum. Taxa that are considered discriminatory appear above threshold (marked by the red dotted line, FDR = 0.2). The size of the dots corresponds to the relative abundance of the taxa

Full size image

Bags vs. filters

α-Diversity and β-diversity (p = 0.223 and PERMANOVA; p = 0.868) were similar between filters and bags (Fig. 2a-b) and differential abundance analyses showed no differential taxa (data not shown).

The aeromicrobiome is not comparable to oral wash nor BALF

α-Diversity was similar in aerosols and across respiratory fluids (p = 0.267, Fig. 3a).

Fig. 3
figure 3

Aerosol microbiota neither similar to oral wash nor bronchoalveolar lavage fluid (BALF): (a) Paired α-diversity (Shannon index) using Friedmann test for bag, filter, sputum oral washes, and BALF (b) β-diversity (Bray-Curtis dissimilarity index) showing distinct clustering patterns for aerosols, sputum, oral washes and BALF. Volcano plots showing differentially abundant taxa in (c) bags vs. oral wash, (d) filters vs. oral wash, (e) bags vs. BALF, and (f) filters vs. BALF. Taxa that are considered discriminatory appear above threshold (marked by the red dotted line, FDR = 0.2). The size of the dots corresponds to the relative abundance of the taxa. Taxa identified as potential contaminants are depicted in grey in the volcano plots

Full size image

Aerosols vs. oral wash

Oral wash β-diversity differed from bags and filters (PERMANOVA p = 0.005 and 0.002, respectively; Fig. 3b). Aerosols were compositionally different to oral washes (Figure S4c) with bags Sphingobium- and Corynebacterium-enriched and Anaerosinus-, Streptococcus- and Prevotella-depleted (Fig. 3c) and filters Mycobacterium-, Pseudomonas- and Sphingobium-enriched and Anaerosinus-, Streptococcus- and Campylobacter-depleted (Fig. 3d).

Aerosols vs. BALF

BALF β-diversity differed from bags and filters (PERMANOVA; p = 0.005 and 0.004 respectively, Fig. 3b). Aerosols were compositionally different to BALF (Figure S4d) with bags Sphingobium-enriched and Anaerosinus-, Streptococcus- and Mycobacterium-depleted and filters Anaerosinus-depleted (Fig. 3e-f).

Differential abundance across sputum, oral wash, and bronchoalveolar lavage fluid

In BALF vs. sputum, Mycobacterium- was enriched and Campylobacter-, Sebaldella-, and Prevotella-depleted. In BALF vs. oral wash, Mycobacterium- was enriched and Prevotella-, Campylobacter-, and Treponema-depleted (Figure S5a-b).

Discussion

We evaluated detection of MTBC and aeromicrobiome captured by PMC in people with TB. Our data shows (1) MTBC is detectable by Ultra in aerosols in about 57% of TB-positive people with comparable sensitivity in bags and filters, (2) bag and filter microbiota are compositionally similar, but differ to a similar extent vs. sputum, oral wash and BALF; and (3) the filter captures Mycobacterium more readily sequenced than in sputum and Mycobacterium itself is comparatively overrepresented in the aeromicrobiome vs. sputum. These findings show proof-of-concept for a novel sampling method for evaluating the aeromicrobiome, which we show is phylogenetically different from the microbiota in other respiratory fluids of people with TB.

The PMC bags and filters retained aerosols containing MTBC detectable using the WHO-recommended molecular test Ultra. This is consistent with studies that show PMC to capture LRT microbes in other diseases [7, 8]. We did not detect differences in mycobacterial load nor PCR inhibition when comparing bags and filters, suggesting both PMC components may be useful, however, further optimization of aerosol collection procedures (e.g., number of coughs sufficient) and processing (to optimise release of captured material) require future investigation.

Diversity and composition metrics (α- and β-diversity) were similar between bags and filters; suggest a potential lack of separation of PMC-collected aerosols to URT and LRT. As the primary purpose of the bag is to collect URT aerosols (~ 150mL in typical adults) but the volumetric capacity of the bag is 250mL, it is conceivable the bag retains aerosols originating from LRT in addition to URT, resulting in a mixture of LRT and URT microbiota.

The aeromicrobiome differed from sputum, oral wash, BALF microbiota, with aerosol depleted of anaerobic taxa such as Prevotella and Streptococcus, previously described as enriched in TB patients’ sputa and oral washes [16] and highlighting that abundance in respiratory fluids does not necessarily translate into abundance in aerosol (previously described only for MTBC) [25]. Mycobacterium was more detectable by sequencing in filter-captured aerosol than sputum, where sequencing this genus can be challenging even in people with severe pulmonary disease (TB and non-TB mycobacterial disease) [10, 16]. This agrees with a prior study that did sequencing on mask-captured aerosol from people with TB [26]. Collectively, these findings suggests that, amongst the respiratory flora in people with TB, Mycobacterium is especially adept at aerosolization, permitting it to be the dominant taxon in aerosol. Besides the enrichment of Mycobacterium in aerosols vs. sputum, Sphingobium and Corynebacterium were enriched, however, their role in TB airborne survival and transmission requires future evaluation.

Our study has strengths and limitations. Although it is the largest to date on this topic in people with TB and included invasive and expensive forms of sampling (bronchoscopy), different findings might result from larger sample sizes, especially for the diagnostic accuracy analyses. As such as our findings should be regarded as explorative and hypothesis generating. They therefore help justify larger expensive studies which, for reasons of capacity, were not possible in the current work. Potential for a larger study was also constrained due to a dependency on parent studies to refer participants (these studies have now stopped). While this enhanced feasibility and resulted in important comparisons being possible (for example, with site-of-disease fluid), it resulted in not all people receiving all procedures. While in this study no cultures were performed on aerosols because the samples were only enough for Ultra and sequencing, we recommend future studies to explore whether these aerosols can be cultured. We also only evaluated people with TB; aerosol from people without TB may have different differential taxa. Future studies could also compare aerosols in relation to additional samples from the aerodigestive tract like bronchial aspirations.

In summary, PMC captures aerosols which can be used to detect MTBC using WHO-approved molecular tests and sequencing. The taxonomic composition of aerosol differs to that in other respiratory fluids. This work lays a foundation for research on the aeromicrobiome in TB.