A novel aerosol collection method shows the cough aeromicrobiome of people with tuberculosis is phylogenetically distinct from respiratory tract specimens

Background: Tuberculosis (TB), a major cause of disease and antimicrobial resistance, is spread via aerosols. Aerosols have diagnostic potential and airborne-microbes other than Mycobacterium tuberculosis complex (MTBC) may influence transmission. We evaluated whether PneumoniaCheck (PMC), a commercial aerosol collection device, captures MTBC and the aeromicrobiome of people with TB. Methods: PMC was done in sputum culture-positive people (≥30 forced coughs each, n=16) pre-treatment and PMC air reservoir (bag, corresponding to upper airways) and filter (lower airways) washes underwent Xpert MTB/RIF Ultra (Ultra) and 16S rRNA gene sequencing (sequencing also done on sputum). In a subset (n=6), PMC microbiota (bag, filter) was compared to oral washes and bronchoalveolar lavage fluid (BALF). Findings: 54% (7/13) bags and 46% (6/14) filters were Ultra-positive. Sequencing read counts and microbial diversity did not differ across bags, filters, and sputum. However, microbial composition in bags (Sphingobium-, Corynebacterium-, Novosphingobium-enriched) and filters (Mycobacterium-, Sphingobium-, Corynebacterium-enriched) each differed vs. sputum. Furthermore, sequencing only detected Mycobacterium in bags and filters but not sputum. In the subset, bag and filter microbial diversity did not differ vs. oral washes or BALF but microbial composition differed. Bags vs. BALF were Sphingobium-enriched and Mycobacterium-, Streptococcus-, and Anaerosinus-depleted (Anaerosinus also depleted in filters vs. BALF). Compared to BALF, none of the aerosol-enriched taxa were enriched in oral washes or sputum. Interpretation: PMC captures aerosols with Ultra-detectable MTBC and MTBC is more detectable in aerosols than sputum by sequencing. The aeromicrobiome is distinct from sputum, oral washes and BALF and contains differentially-enriched lower respiratory tract microbes.


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, is evaluated in people with cystic brosis (CF) 7 and viral pneumonia 8 .In the rst 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 uid (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 in uence 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 di cult because it requires bronchoscopy 11 , which is invasive, ethically complex for research purposes only, and expensive, often rendering which 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 lter.PMC is designed to separate aerosols from the URT and LRT into the bag and lter, respectively 14 .This separation occurs after a person coughs into the PMC, at which point air from the anatomical URT dead space (~ 150mL) 15 ows rst into the bag.Air after the 150mL is likely from the LRT and then, due to backpressure from the inelastic bag, directed towards the lter 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 lter microbiota.We compared microbiota in the bag and lter to URT and LRT clinical samples.

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 rst 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 de ated 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 10min.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).

Recovery and processing of aerosols
Aerosols were separately recovered from the PMC bag and lter.After removal under sterile conditions, bags were rinsed and incubated for 15min in 10mL stripping buffer (1% Triton X-100 in 10mM Tris-HCl, pH 8.0) whereas the lter was submerged in 10mL stripping buffer, vortexed vigorously for one minute, and incubated for 30min.Bag and lter 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 .
Potentially contaminating background taxa were identi ed using decontam (version 1.14.0) 23.Taxa identi ed as possible contaminants were not removed from downstream analysis but greyed-out in volcano plots only if identi ed as differential.Data and R scripts are publicly available here.

Ultra on aerosols
Ultra sensitivity on bags and lters 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 lter Ultra results nor were there differences in Ultra-generated IS1081-IS6110 C T s.Ultra-generated sample processing control (SPC) cycle threshold (C T ) values, a measure of PCR inhibition 24 , were similar when across bag and lter washes and when compared across people who had Ultra-positive vs. -negative bags or Ultra-positive vs. -negative lters.A distinct aeromicrobiome is detectable over and above background aerosol The number of reads did not differ in sputum, bags, and lters 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 lters 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 lter samples grouping with BKG.Potential bag-and lter-contaminating taxa included Pseudomonas-, Anaerosinus-, Alistipes and-Kocuria (full list in Table S1, Figure S3), however, none of these were differential in other analyses.
The aeromicrobiome is not comparable to oral wash nor BALF α-Diversity was similar in aerosols and across respiratory uids (p = 0.267, Fig. 3a).

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 lters, 2) bag and lter microbiota are compositionally similar, but differ to a similar extent vs. sputum, oral wash and BALF; and 3) the lter captures Mycobacterium more readily sequenced than in sputum and Mycobacterium itself is comparatively overrepresented in the aeromicrobiome vs. sputum.These ndings 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 uids of people with TB.
The PMC bags and lters 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 lters, suggesting both PMC components may be useful, however, further optimization of aerosol collection procedures (e.g., number of coughs su cient) and processing (to optimise release of captured material) require future investigation.
Diversity and composition metrics (α-and β-diversity) were similar between bag, lters, URT and LRT specimens; suggest a potential lack of separation of PMC-collected aerosol from these anatomical sites.
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 25 and highlighting that abundance in respiratory uids does not necessarily translate into abundance in aerosol (previously described only for MTBC 26 ).Mycobacterium was more detectable by sequencing in lter-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 27 .Collectively, these ndings suggests that, amongst the respiratory ora 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 had limitations.Although the largest to date on this topic that included invasive forms of sampling (bronchoscopy), different ndings might result from larger sample sizes.Although the dependencies on parent studies enhanced the feasibility of our work, it also resulted in not all people receiving all procedures.We also only evaluated people with TB; aerosol from people without TB may have different differential taxa.
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 uids.This work lays a foundation for research on the aeromicrobiome in TB.

Figures
Figures

Figure 1 How
Figure 1

Table 2
Summary of Ultra-based MTBC detection in aerosols, showing Ultra results when applied to the bag and lter 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).