Discussion
The novelty of our analysis is being able to pin down to the microbial species level differences at the severity, inflammatory and molecular phenotype level of severe asthma using sputum metagenomic sequencing of a large cohort of patients. We focused on 3 pathogenic species, namelyHaemophilus influenzae, Moraxella catarrhalis andTropheryma whipplei because of their potential pathogenecity in respiratory diseases. We found reduced bacterial α-diversity of microbial species in the 2 severe asthma groups where there was higher abundance of Haemophilus influenzae and Moraxella catarrhalis in SAn compared to SAs/ex, MMA and HC, and ofHaemophilus influenzae and Tropheryma whipplei in SAs/ex compared to MMA. In terms of inflammatory status, α-diversity was lowest in the mixed granulocytic group followed by a slight reduction in the neutrophilic group, with neutrophilic inflammation associated with higher abundance of Haemophilus influenzae and Moraxella catarrhalis , while eosinophilic inflammation was associated with high abundance of Tropheryma whipplei . In terms of associated molecular pathways, Haemophilus influenzae was most abundant in TAC1 and TAC2, accompanied by an increase in Tropheryma whippleiin TAC1 and in Moraxella catarrhalis in TAC2. The abundance ofMoraxella catarrhalis in TAC1 and TAC3 was very low. The increased abundance of Haemophilus influenzae in both TAC1, an eosinophilic phenotype, and in TAC2, a neutrophilic phenotype, but not in TAC3, a paucigranulocytic phenotype, is of interest. One possibility is that this increased abundance may be due to the reduced phagocytosis of Haemophilus influenzae by lung macrophages from patients with severe asthma (27). Haemophilus influenzae can induce the release of IL1α, IL1β, IL-6, IL-8, MCP-1 and TNFα from human tracheal epithelial cells (28), through the toll-like receptor, TLR2, activation (29), that could lead to neutrophil activation and inflammation.
In both SAn and SAs/ex, the species most positively correlated with sputum neutrophils was Moraxella catarrhalis while in MMA,Haemophilus influenzae was inversely correlated. Moraxella catarrhalis is a known airway pathogen that causes respiratory infections linked to neutrophilic airway inflammation in severe or poorly-controlled asthma (9, 12). Moraxella catarrhalis was also most abundant in the TAC2 molecular phenotype characterized by neutrophilic and inflammasome activation, supporting the possibility that it may be associated with or be responsible for the induction of neutrophilic inflammation in severe asthma. This link betweenMoraxella catarrhalis and neutrophilic inflammation is further strengthened by the highest abundance of Moraxella catarrhalis in TAC2 where it was positively correlated with sputum neutrophilia. This potential activating role of Moraxella catarrhalis is supported by its activation of TLR 2, 4 and 9 (30), induction of IL-6, IL-8 and prostaglandin E2 release through NF-κB and activation in lung epithelial cells (31, 32).
Although Tropheryma whipplei is known to cause Whipple’s disease with gastrointestinal symptoms, it has also been implicated as a cause of aspiration, ventilator-associated and community-acquired pneumonia (33). Tropheryma whipplei in bronchoalveolar lavage fluid has been isolated in asymptomatic immunosuppressed patients suffering from Human Immunodeficiency virus, which was reduced by anti-retroviral therapy (34). The increased abundance of Tropherymawhipplei in SAs/ex was positively associated with sputum eosinophilia and also with pack years of smoking in the TAC1 and TAC2 phenotypes. A previous study has reported the presence ofTropheryma whipplei by PCR in sputum samples of severe asthma patients who were mainly eosinophilic (12). We found thatTropheryma whipplei was most prominent in severe asthmatics who were either current smokers or ex-smokers. These differences between NSA and SAs/ex severe asthma groups in terms of the sputum metagenome is demonstrated here for the first time. Previous studies using 16S ribosomal RNA microarray comparing the lung microbiome of smokers with non-smokers have reported either no differences (35, 36), or a higher abundance of Veillonella and Megasphaera with a decrease in Haemophilus (37). The pathogenic role of Tropheryma whipplei, if any, in SAs/ex remains unclear.
Other bacterial species were also noted to be differentially-abundant, being mainly reduced in the severe asthma groups of SAn and SAs/ex when compared to MMA and/ or HC. These include species of Prevotella, Veillonella, Neisseria and Rothia, of which we know little in terms of their potential lung pathogenecity. SAn and SAS/ex were different from MMA with a decreased abundance of Prevotella intermedia and Rothia mucilaginosa consistent with the previous report of decreased abundance of the genus Prevotella in asthma compared to non-asthmatic individuals (7, 38). Most of these species were negatively correlated with sputum neutrophilia. One bacterial species, Haemophilus parainfluenzae, was reduced in abundance in SAn and in TAC1 and TAC2, with negative correlation with sputum neutrophilia in SAn and MMA, and in TAC1 and TAC3. Our data indicate that Haemophilus parainfluenzae is decreased in current smokers compared to ex-smokers; however, since this same species is reduced also in the smoking/ex-smoking group on OCS compared to those not on OCS, it may be related to OCS therapy. Haemophilus parainfluenzae has been reported (39) to increase IL-8 expression with activation of p38 MAPK and inhibit corticosteroid responses in alveolar macrophages from corticosteroid-resistant asthma patients, thus may be involved in induction of corticosteroid resistance.
Overall, the microbiome was stable when repeated at one year in severe asthma with no shift in α or β-diversity. In an unbiased microbiome-driven clustering to identify severe asthma phenotypes using the same microbial data, we found 2 distinct robust phenotypes that exhibited relative overtime stability (40). This stability does not exclude the possibility of changes in the microbiome during acute exacerbations of asthma as occurs in children where an increase in gram-negative microbes have been detected in induced sputum (41). Out of these 2 phenotypes (40), one had worse asthma outcomes, more sputum neutrophilia and greater enrichment of the gammaproteobacteria,Haemophilus influenza e and Moraxella catarrhalis . Further analysis of this phenotype has led to its association with differentially-expressed genes, in particular TNFα and related regulatory genes, the IL-1 family of interleukins, Toll-like receptors and inflammasomes (42).
We highlight some technical issues in our analysis. First, the sequencing technology of using paired-end 100 base-pairs at high sampling depth provides less accurate taxonomy profiling compared to reads of 150 base-pairs or longer at lower depth. However, we have used MetaPhlAn2 because it provides a more sensible taxonomic composition for respiratory microbiome experts compared to a protein-based method such as Kaiju (43). Secondly, the use of a marker-based approach may lead to exclusion of samples for downstream analysis due to their lack of marker genes. This is likely due to sputum having high amounts of host DNA, contributed by neutrophils and neutrophil extracellular traps resulting in low level of microbial reads, with lack of marker genes for estimating taxonomic composition, thus excluding some samples in the downstream analysis. Finally, for differential analysis, DESeq2 was used because it took into account the compositional nature of the metagenomic data, with the addition of pseudocount to overcome lack of taxa shared by all samples.
In conclusion, this sputum metagenomic study of severe asthma patients during a stable state has revealed an important association of reduced bacterial α-diversity at the species level to neutrophilic airway inflammation and neutrophil and inflammasome activation whereHaemophilus influenzae and Moraxella catarrhalis were in greater abundance. On the other hand, in eosinophilic inflammation, bothHaemophilus influenzae and Tropheryma whipplei were most abundant with no Moraxella catarrhalis found. Tropheryma whipplei was mostly linked to the smoking/ex-smoking severe asthma and to eosinophilic inflammation. These bacterial species may play an important role in influencing the severity and inflammatory phenotype of severe asthma, particularly linked to neutrophilic inflammation, and further studies are needed to confirm the importance of these bacteria.