The Bronchial Microbiome and Asthma Phenotypes

Emerging data on the astounding breadth, depth, and functional capacity of the human microbiome are transforming our understanding of the determinants of health and disease (1). Studies of the gastrointestinal microbiome are already moving beyond broad associations, as of microbial diversity with the development of immune function in infants (2) and of “microbial dysbiosis” in inflammatory bowel disease (3), to identification of microbial functions as mediators of particular outcomes (4, 5). The trajectory of studies of the bronchial microbiome is slightly behind, delayed in part by the idea that the subglottic airways are sterile, and by the difficulties in sampling them without contamination from the oropharynx. Indeed, one study of respiratory samples obtained with exquisite attention to avoiding upper airway contamination has challenged the idea of a distinct bronchial microbiome in healthy adults, for bacteria found in the lower airways of six subjects appeared to be diluted reflections of bacteria found also in the upper airways (6).

The situation is different in airway disease, and several studies have reported the distortion in airway bacterial community structure and composition in cystic fibrosis, chronic obstructive pulmonary disease, and asthma (7). Although studies so far largely have reported broad associations, there are hints that members of particular bacterial groups may be important in shaping airway function. Hilty and colleagues reported that the bronchial microbiome among individuals with asthma was “disordered,” with greater abundance of members of the Proteobacteria, particularly Haemophilus species, and lesser abundance of members of the Bacteroidetes, particularly Prevotellaceae (8). In a larger study using protected brush samples, we observed greater bacterial burden and diversity in asthmatic subjects, with again greater representation of Proteobacteria (9), and noted specific family members whose relative abundance correlated with greater bronchial hyperresponsiveness. Several genera among these families possess functional properties of potential significance to asthma (e.g., Sphingomonas, Nitrosomonas, and Oxalobacter). Of greatest relevance to this editorial was the association of an increased abundance of Comamonadaceae, which include members capable of metabolizing steroid compounds. In analysis of induced sputum from individuals with asthma not taking an inhaled corticosteroid and from healthy subjects, Marri and colleagues (10) also found the samples from individuals with asthma to have greater bacterial diversity with Proteobacteria in higher proportion.

As reported in this issue of the Journal, Goleva and colleagues (pp. 1193–1201) conducted a study intended to close the gap between description of asthma-associated bronchial microbiota and identification of particular bacteria and their functions, as they relate to a key phenotypic feature of asthma (11). The phenotypic feature examined—corticosteroid resistance—is of unquestionable clinical importance and may reflect a distinct underlying “molecular phenotype” or “endotype” of asthma (12). To identify bacteria linked to this phenotype, the authors extended their description of the communities identified (by metrics such as diversity and by level of taxonomic classification), to identification of bacterial “outgrowths” linked to corticosteroid resistance. This led them to study the effects of specific species in an in vitro model of corticosteroid responsiveness that included isolated lung macrophages. The study thus aimed at moving from broad associations to identifying cellular and molecular mechanisms by which the bronchial microbiome alters function in its host.

The criteria for classification as corticosteroid-resistant (CR) or corticosteroid-sensitive (CS) were sound—improvement in FEV₁ after treatment with prednisone, 40 mg/day for 7 days, though the high frequency of inhaled corticosteroid use among the subjects enrolled admits the possibility of misclassification. Bronchoalveolar lavage (BAL) fluid obtained from these two groups of subjects with asthma and from 12 healthy control subjects was analyzed for bacterial composition by 454-pyrosequencing. On first take, the findings seem to differ from the previously summarized studies, as no difference was found between the subjects with CR asthma, the subjects with CS asthma, and healthy subjects in community metrics or in detected bacterial composition. A careful look, however, at the proportions of sequences belonging to major phyla (Table E3 in Goleva and colleagues’ online supplement) reveals reassuring similarity to the earlier findings. Proteobacteria made up on average 34% of all taxa identified in the subjects with CS asthma, 24% in the subjects with CR asthma, and only 14% in the healthy control subjects, whereas Bacteroidetes such as Prevotellaceae constituted a greater proportion in the healthy subjects than in either group of subjects with asthma. Although few of the differences were statistically significant, the greater abundance of specific Proteobacteria families in the two groups with asthma (Goleva and colleagues’ Table E4) again echo previous findings for Sphingomonadaceae and Comamonadaceae in adults with asthma, and the finding for Moraxella in upper airway samples from wheezy children by Bisgaard and colleagues (13). However, the current study did not confirm other microbiota also previously observed to be more prevalent among individuals with asthma, such as Neisseriaceae (8) or Mycoplasma (14).

Prompted by its apparent frequency among their subjects with asthma, Goleva and colleagues focused on the possible importance of the outgrowth or expansion of a bacterial genus, defined by its comprising more than 5% of all 16S RNA sequences found in subjects with asthma, and twice as great a proportion as observed in healthy control subjects. Such an expansion of one or more represented genera was remarkably common: it occurred in 33 of the 39 subjects with asthma (85%). This was parsed further by analyzing subjects with bacteria expanded just in the CR or CS group (“unique” bacterial expansions).

This must have seemed a promising approach, for the importance of an outgrowth of a single organism in a disease possibly related to asthma was illustrated in a report of the novel role of Corynebacterium tuberculostearicum in the pathogenesis of chronic rhinosinusitis (15). However, that study differed in that this species was disproportionately increased on average across the subjects with disease compared with healthy control subjects. In Goleva and colleagues’ study, the genus mostly commonly “uniquely expanded” in the subjects with asthma, especially CR subjects, was Neisseria, and even that was expanded in only five subjects with asthma, although the majority of all uniquely expanded bacterial genera were Proteobacteria (their Tables 3 and 4).

Disappointingly, this approach to identifying bacteria that might influence corticosteroid responsiveness seems to have come up empty. Little can be made of the finding that the “subjects with CR and CS asthma with bacterial expansions had significant alteration in their airway microbiome composition as compared with normal control subjects,” for the criteria for expansion required a difference from the airway microbiome of the control subjects. Moreover, no clinical features were found to distinguish the six subjects without bacterial expansions from the other subjects with asthma. Finally, the presence or absence of a bacterial expansion seems unrelated to the major phenotypic feature analyzed—corticosteroid responsiveness. Seventy-four percent (29/39) of all the subjects with asthma studied were CR. So were 83% (5/6) of those without and 73% (24/33) of those with bacterial expansions. Going further, so were 71% (15/21) of those without and 77% (14/18) of those with unique expansions.

So in the end, the case for focusing on a member of the bronchial microbiome as associated with, or causing corticosteroid resistance in, asthma rests on the experiments on the effects of coculturing blood monocytes and BAL macrophages with Haemophilus parainfluenzae on the cellular pathways activated by corticosteroid engagement with the glucocorticoid receptor. Compared with the effects of coculturing with Prevotella melaninogenica, H. parainfluenzae indeed reduces corticosteroid responsiveness in this model system. But whether this finding has anything to do with corticosteroid resistance in vivo is still a very open question. Although a Haemophilus species was found to be uniquely expanded in subjects with CR asthma, this expansion was present in only two of the CR subjects. And if this γ-proteobacterium is suspected of mediating corticosteroid resistance, then the γ-proteobacterium uniquely expanded in an individual with CS asthma, a member of the same family (Pasteurellaceae) as Haemophilus, should have been tested as well, to determine its effects on corticosteroid responsiveness in vitro.

Goleva and colleagues’ study presents much to admire, especially in its a priori definition of an important phenotypic feature of asthma—corticosteroid responsiveness—and in its innovative and detailed approach to identifying bacteria influencing this feature. Its leap to selecting Haemophilus parainfluenzae as a prime suspect seems premature, however, and the case made for any of the bacteria genera identified as responsible must be considered unproven. It is to be hoped, however, that the failure of this study to prove its case against one suspected microbial culprit will not discourage these or other investigators from similarly detailed study of the bronchial microbiome not just of bacteria, but of fungi and viruses as well, in other carefully phenotyped subjects with asthma, and examining possible mechanisms of action of candidate microbes in model systems, as Goleva and colleagues have done.