Study of the Lung Microbiome. Have We Reached the End of the Beginning?

Studies of the lung microbiome have lagged behind those of other body sites because of the historical belief that the lungs were sterile. Initial studies applying non–culture-based sequencing techniques demonstrated that the lower respiratory tract has a detectable microbial population that may be altered in disease (1–4). Some of the first organized efforts to examine the lung microbiome were performed in HIV-infected individuals (5–7). These studies primarily focused on outpatients who were not experiencing acute pulmonary complications and explicitly excluded individuals with pneumonias to study the microbiome of healthy individuals. Because HIV-infected individuals remain at risk for bacterial pneumonia as a result of persistent innate and acquired immune deficits, even when treated with antiretroviral therapy, there has been recent interest in alterations of the HIV lung microbiome in the setting of acute pneumonia. Low- and middle-income countries shoulder a disproportionate share of the burden of bacterial pneumonia, and pulmonary infection is a common and frequently fatal complication of HIV infection in Uganda, the site of the study published in this issue of the Journal by Shenoy and colleagues (pp. 104–114) (8). As African HIV-infected populations with less access to antiretroviral drugs and higher burden of coinfections such as tuberculosis have an increased risk for mortality after hospitalization for pneumonia, it is important to understand the role of the lung microbiome in this setting as a potential tool to modify pneumonia risk and outcome.

The current study by Shenoy and colleagues expands on prior work in the HIV-infected population with pneumonia by examining a large cohort of individuals from Uganda undergoing bronchoscopy for acute pneumonia (8). In addition to shedding light on the bacterial communities in the lung during pneumonia in HIV in particular, the work is important for its broader implications for the field of the lung microbiome in general. The work takes us beyond the initial efforts in lung microbiome studies that have primarily used 16S rRNA-based taxonomic approaches. The current study used sophisticated bioinformatics to identify distinct microbial community states (MCS) in the enrolled patients and determined the relationship of these communities to functional outcomes. By integrating the microbiome, host gene expression, metabolomics, and clinical outcomes, and by investigating the relationships between bacteria and other organisms, this study demonstrates how to move beyond simple bacterial description and raises the possibility of tailoring therapies based on microbial dysbiosis.

The discovery of different microbial community states or “phenotypes” of the lung microbiome is novel in this population of HIV-infected patients with pneumonia. The study identified an MCS1 microbial community that was dominated by Pseudomonadaceae in conjunction with Sphingomonadaceae and Prevotellaceae, and a second MCS2 community that exhibited a reciprocal gradient of Streptococcaceae (MCS2A) or Prevotellaceae (MCS2B). The community states differed in relationships between bacteria and other organisms with a high proportion of Mycobacterium-positive cultures in MCS1 and Aspergillus-positive cultures in MCS2B individuals.

These microbial community states were also distinguished by key clinical and functional characteristics, and not just differences in their microbial composition. For example, predicted bacterial metagenomic content differed by MCS. Furthermore, host expression of immune genes also distinguished the microbial community states. The authors investigated the relationship of systemic metabolites to the lung microbiome in a subset of the cohort. They found that 60 metabolites differed significantly between the states. Metagenomic prediction, airway immune response, and circulating metabolites were significantly interrelated, suggesting that the microbiome may be an important biomarker and have a potential causal effect in the lung. Finally, the authors found a trend for differing mortality by MCS.

A similar concept of microbial community states has been reported in the HIV-uninfected population. Segal and colleagues identified two distinct bacterial community pneumotypes in healthy HIV-uninfected individuals (2, 9). One of these pneumotypes resembled background microbiota with low levels of inflammation. A second phenotype resembled the supraglottic microbiota with detection of Veillonella and Prevotella and increases in lung inflammatory response and differences in key metabolites. Combined, these studies suggest that microbiome populations drive the phenotype and potentially influence long-term outcomes, although it is also possible that the phenotypes and outcomes determine the microbial populations.

The current study is commendable in its large size and extensive investigations, but there are some limitations that bear mentioning. Attribution of causality is difficult, as sampling was done at the time of pneumonia. It is possible that the microbiome influenced the immune response or the immune response shifted the microbiome. In addition, almost all patients had received antibiotics, thus limiting knowledge of the preantibiotic microbiome. The lung metabolome was not directly examined, and the metabolic contributions of the host versus the microbes are difficult to untangle. The cohort also provides insight into HIV-infected Ugandan patients with pneumonia; however, the results of this study may have limited applicability to patients in the United States and other Western countries because of differences in patient demographics, laboratory testing, and antibiotic availability, as well as the high HIV–tuberculosis coprevalence in Uganda. The investigators have, in fact, previously reported differences in the lung microbiome of HIV-infected patients with pneumonia in San Francisco and Uganda (10). A strength of the study was the integration of the bacterial microbiome with findings of fungal and mycobacterial cultures, but sequence-based analyses of fungi and viruses were not performed. Finally, the authors did not examine the gut microbiome. A growing body of evidence implicates the importance of the gut microbiome in shaping the immune response in the lung and its response to insults (11–13), and infections in the lung may influence the gut microbiome, as has been shown in animal models of tuberculosis (14).

The implications of the study by Shenoy and colleagues extend beyond the investigation of the lung microbiome during pneumonia in HIV-infected individuals (8). With studies such as the current one, we have hopefully begun to move beyond the beginning studies of simple taxonomic investigations to include microbial function, host and microbe metabolism, and host response in determining the role of the lung microbiome in health and disease. The work confirms the need to look for different phenotypes based on the integration of microbial communities, host response, and clinical outcomes. Future work building on these types of investigations has potential to reshape the way we think about the interaction of microbes and host in health and disease and launch the next phase of investigation of the lung microbiome.