Introduction
Recent studies have implicated the lung microbiome in the occurrence of chronic lung diseases such as idiopathic pulmonary fibrosis (IPF), chronic obstructive pulmonary disease, cystic fibrosis and chronic lung allograft dysfunction (CLAD) in lung transplant recipients [
1‐
4]. There is emerging evidence that early alterations in lung microbiome and/or dysbiosis modulates inflammatory mediators leading to pathogenesis and/or progression of chronic lung diseases including CLAD [
5‐
7]. Our group recently reported that a shift to a
Proteobacteria dominant allograft microbiome was associated with CLAD in lung transplant recipients [
8]. Likewise, microbial adaptations and changes in bacterial diversity have been implicated in progression of fibrosis in IPF subjects [
4]. However, the mechanisms involved in the microbiome-host interactions leading to chronic lung inflammation are not well understood.
Delineation of the microbiome signatures and taxonomic profiles of bacterial communities in various disease states is an important first step but does not directly provide an insight into bacteriome-allograft-host interaction. Bacterial colonization and/or infection leading to pathology often results in physiological changes in the host, including alterations in metabolic profile [
9,
10]. A better understanding of these metabolic shifts associated with a specific infection can improve our understanding of disease pathophysiology through the identification of by-products of host and microbial metabolism, while also providing vital information about the unique metabolites produced with these ever-changing interactions [
11]. In particular, the delineation of bacterial metabolism and the host/allograft inflammatory response is critical to better define factors modulating the local microbiome. Metabolites or metabolomic profiles identified in a defined cohort can also serve as potential biomarkers for disease characterization and/or novel therapeutic targets [
12]. In the context of a potential clinical application, unique metabolome signatures in urine have been found to distinguish
Streptococcus pneumoniae from
Staphylococcus aureus lung infection [
13]. Similarly, lung metabolome analysis from HIV subjects have shown that pyochelin, a siderophore produced by
Pseudomonas aeruginosa, is elevated in HIV-infected individuals compared to HIV-uninfected individuals [
14]. Likewise, specific metabolome pathways have been identified in lung transplant recipients with CLAD [
15], though further investigations are needed to delineate its relevance to CLAD pathobiology. Identification and correlation of novel metabolome profiles associated with specific pathological microbiome signatures may help guide personalized treatment of host disease states [
16].
Dysregulation of muco-ciliary clearance and subsequent increases in bacterial burden are well described in advanced lung disease [
17,
18]. Moreover, colonization of pathobionts in diseased native lungs of single lung transplant recipients (SLTs) may alter and/or contribute to the microbiome of the allografts [
19,
20]. However, to date, the interaction of the metabolome and microbiome in the allografts and native lungs has not been evaluated in patients after lung transplantation. In this study, we utilized the airway microbiome and metabolome signatures of the native lungs and transplanted lungs (allografts) of SLTs as a model system to answer fundamental questions regarding the inherent lung metabolome and its influence on the lung microbiome. We hypothesized that the airway metabolome would correlate with the abundance of distinct microbiome signatures in the native and allografts lungs of SLTs.
Discussion
Emerging evidence suggests a role for the lung microbiome in the pathogenesis of chronic inflammatory lung diseases including CLAD [
5,
8,
40]. In this study, we utilized the airway microbiome and metabolome signatures of the native lungs and transplanted lungs (allografts) of SLTs as a model system to answer fundamental questions regarding the inherent lung metabolome and its influence on the lung microbiome. We characterized the associations between the airway microbiome and metabolome of the allograft and the native lungs of SLTs. In our cohort, we found that the airway microbiome of the native and allograft lungs were distinct with a significantly higher abundance of genus
Pseudomonas and
Acinetobacter (phyla
Proteobacteria) and elevated levels of VEGF and ac-PGP in the allograft. Furthermore, the native lung metabolome differed from the allograft with a higher abundance of sphingosine and sphingosine-like metabolites and its presence negatively correlated with the abundance of bacteria
Pseudomonas and
Acinetobacter. These hypothesis generating observations lay the foundation for future studies to evaluate the cause-effect relationship between the airway metabolome and microbiome.
To our knowledge, this is the first report characterizing the microbial heterogeneity of the native and transplanted lung in the same individual. The results from our cohort suggests that the allograft’s airway microbiome is distinct with a significantly increased relative abundance of genera
Acinetobacter &
Pseudomonas (phyla
Proteobacteria) (Figs.
1a,
2d and e). In comparison, the microbiome in native lungs of SLTs and healthy individuals have a greater proportion of the phyla
Firmicutes and Bacteroidetes. Alterations in the
Firmicutes/Bacteroidetes (F/B) ratio have previously been proposed as a marker for intestinal dysbiosis in various disease states [
34]. In the lung, where a higher proportion of
Proteobacteria is present than the gut [
33], the
Proteobacteria/Firmicutes (P/F) ratio may be a better marker for dysbiosis, given the predominance of
Firmicutes in the healthy lung [
33]. In our cohort, the allografts had an elevated F/B and P/F ratios compared to the native lungs and normal lung controls suggesting greater dysbiosis. Donor lungs at the time of implantation harbor a different microbiome than the recipient, indicative of greater microbial variation between the allograft and native lung at the time of transplantation. Following transplantation, common thinking is that over time the microbiomes of the allograft and native lung would become similar. However, in our cohort, despite the subjects being more than a year post-lung transplantation, the microbiome remained varied between the allograft and native lung. Likewise, we noted an increased bacterial biomass burden and dysbiosis in the allografts compared to the native and healthy lung controls. These results are concordant with other studies linking elevated bacterial biomass to microbial dysbiosis [
41]. Increased bacterial biomass and colonization with pathogenic bacteria such as
Acinetobacter and
Pseudomonas has been associated with allograft dysfunction in lung transplant recipients [
42].
Cytokine VEGF and tripeptide Ac-PGP were found to be significantly elevated in allograft BAL samples. We and others have shown that Ac-PGP, a matrikine tripeptide, mediates inflammation in acute and chronic lung diseases including CLAD and bacterial infections [
43,
44]. Interestingly, elevated VEGF levels has been linked to lung infections with
Pseudomonas [
45] as well as allograft rejection including post lung transplant primary graft dysfunction and bronchiolitis obliterans syndrome, a form of chronic lung allograft dysfunction [
46‐
48]. Likewise, dysbiosis in the lung and gut has been implicated in heightening inflammatory states in several chronic diseases [
31,
32]. The increase in the ratio of the pro-inflammatory bacteria such as
Pseudomonas and low stimulatory bacteria such as
Prevotella and
Streptococcus are known to upregulate the inflammatory gene expression profile [
5]. Although we see an association between elevated
Proteobacteria the inflammatory markers Ac-PGP and VEGF, additional investigations need to be conducted to establish a causal link between a
Proteobacteria-dominant microbiome and pro-inflammation.
To further understand the functional impact of the microbiome on the host, we conducted metabolome analyses. The airway metabolome differentiated the allograft and the native lung of the participants of our study (Fig.
3b and c). Several of these metabolites had significant negative and positive correlations with various bacterial genera (Fig.
4a, b). In the positive ion mode, the top m/z features that differentiated the allograft from the native lung (high VIP scores) were dominated by sphingosine-like molecules that were found to be increased in the native lungs (Table
2 and Figure
S3A &B). These sphingosine metabolites had a negative correlation with bacterial genera (
Pseudomonas, Acinetobacter) from the phyla
Proteobacteria in the native lungs. Sphingolipids are bioactive lipids known to be part of the plasma membrane lipid bilayer in eukaryotic cells [
49] and are cleaved by sphingosine kinases, with several of these molecules having key roles in the regulation of oxidative stress and immune function [
50]. Sphingosine improves the host response to
Pseudomonas infections and augments neutrophil killing of reactive oxygen species resistant
Pseudomonas [
51‐
53]. An increased presence of sphingosine-like molecules in the metabolome of native lungs may account for the lower
Proteobacteria signature found in them as compared to the allografts. Likewise, some metabolite features had positive correlation with the microbial genera (
Pseudomonas and
Acinetobacter) in the allograft lungs. Metabolite m/z 201.1, a alanine metabolite and m/z 186.115, a threonine metabolite are involved in maintenance of bacterial cell wall structure and cellular stiffness, promote bacterial proliferation [
54,
55] and have regulatory role in T cell activation [
56]. Mummichog analysis of the metabolome suggested the presence of metabolites that were increased in the allograft compared to the native lung (Table
3). Pathways in the amino acid metabolism, fatty acid activation and fatty acid beta-oxidation were increased in the allograft compared to the native lung. Methionine and cysteine metabolism pathways are known to regulate oxidative stress through the methionine/glutathinone trans-sulfuration pathway. Disruption of this pathway can lead to increased oxidative injury. Likewise, increased fatty acid activation can result in mobilization of cell membrane derived lipid signaling molecules such as sphingospine-1-kinase and arachidonic acid derived eicosainoides [
57,
58]. More studies are needed to further understand if the variations in metabolome are in part influenced by the difference in airway microbiome or vice-versa.
As the primary antigen presenting cells of the lung, resident alveolar macrophages originate in the embryogenesis period and are sparsely replenished by the bone marrow during adult life [
59]. Studies in lung transplant recipients have shown that most of the alveolar macrophages even 3.5 years after lung transplant are donor derived in the allografts [
60]. An alternate explanation to the difference in the lung microbiome in the native and allograft lungs could be due to the variable antigen presenting/regulatory nature of donor-derived and recipient alveolar macrophage phenotypes in the allografts and native lungs.
Our study has limitations, including sample size and design. We did not find any differences in the microbiome or inflammatory signatures in those diagnosed with CLAD at sampling or at follow-up. However, with a cross-sectional study design and inadequate power, we cannot comment on the dynamic changes in the microbiome and metabolome of native lungs and allografts and their association with chronic lung allograft dysfunction. However, this study is hypothesis generating and provides insightful data to further advance our understanding of microbiome-metabolome interactions in the lung. One of the unique challenges in the field of lung microbiome is presence of low biomass in BAL samples compared to samples from the GI tract [
61,
62]. Interpretation of results obtained from a single low biomass sample without appropriate controls can be misleading. To circumvent this issue, we collected control samples from the separate bronchoscopes used during each A and N sample collection. The bronchoscope control microbiome signatures were found to be dissimilar to the A and N airway microbiome (Figure
S2A, B). Furthermore, the bacterial 16S DNA levels in the bronchoscope control samples were below the lower limit of PCR quantification suggesting absence of bronchoscope contamination (data not shown). BAL samples from normal lung control subjects, collected and processed using a similar methodology were also analyzed and compared to the allograft/native lung microbiome. We processed technical replicates of each biological sample to evaluate variability in both the relative abundance and bacterial 16S DNA quantification. Although, antimicrobial prophylaxis and immunosuppression regimen can also influence the microbiome [
63], in our subjects these were given systemically, and hence would impact both the native and the allograft similarly. While we found that lung allografts harbored a more
Proteobacteria dominant microbiome and higher PGP levels, longitudinal studies are needed to investigate the causal association, role of varied prophylactic antibiotics, immunosuppression on microbial composition in the lung and whether this inflammation translates into allograft rejection. Our healthy controls did not have Ac-PGP measurements. Nevertheless, our previous studies have shown absent or extremely low levels of Ac-PGP in healthy lung airway fluid [
64]. It has been reported that BAL samples obtained from separate geographic regions within the same lung can demonstrate highly dissimilar microbial communities [
33,
65]. Although, it is possible that the variations in microbiome in the native lung and allograft may be related to expected variation in sampling from different regions of the lung, these spatial variations are known to be less significant compared to variations across individuals [
65]. Finally, all metabolites that correlated with bacterial genera were not able to be identified due to the current limitations of untargeted mass spectrometry and the adjusted
P values had limitations due to the smaller samples size [
66]. Although, we found correlations between sphingosine like molecules with bacterial genera in the native lungs, due to the very low concentration of these sphingosine metabolites in the allograft, these bacterial-sphingosine correlations could not be accurately predicted in the allograft and targeted LC-MS analysis was not performed to validate the metabolite. Nevertheless, these observations are hypothesis generating and warrant further validation in future mechanistic studies.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.