Respiratory mycobiome and suggestion of inter-kingdom network during acute pulmonary exacerbation in cystic fibrosis

Abstract

Lung infections play a critical role in cystic fibrosis (CF) pathogenesis. CF respiratory tract is now considered to be a polymicrobial niche and advances in high-throughput sequencing allowed to analyze its microbiota and mycobiota. However, no NGS studies until now have characterized both communities during CF pulmonary exacerbation (CFPE). Thirty-three sputa isolated from patients with and without CFPE were used for metagenomic high-throughput sequencing targeting 16S and ITS2 regions of bacterial and fungal rRNA. We built inter-kingdom network and adapted Phy-Lasso method to highlight correlations in compositional data. The decline in respiratory function was associated with a decrease in bacterial diversity. The inter-kingdom network revealed three main clusters organized around Aspergillus, Candida, and Scedosporium genera. Using Phy-Lasso method, we identified Aspergillus and Malassezia as relevantly associated with CFPE, and Scedosporium plus Pseudomonas with a decline in lung function. We corroborated in vitro the cross-domain interactions between Aspergillus and Streptococcus predicted by the correlation network. For the first time, we included documented mycobiome data into a version of the ecological Climax/Attack model that opens new lines of thoughts about the physiopathology of CF lung disease and future perspectives to improve its therapeutic management.

Figure 1

Diversity and taxonomy summaries of sputa collected from CF patients with a mild (), moderate () and severe () alteration of the lung function measured by FEV1. Dataset from each sputum sample contained an average of 7,909 bacterial reads (ranging from 5,104 to 14,749), and an average of 9,811 fungal reads (ranging from 2,933 to 15,599). Alpha diversity indexes of the bacterial microbiome (a) but not of mycobiome (a′) were positively correlated with FEV1 values. When patients were divided into 3 groups according to FEV1 values^,, bi-dimensional PCoA representations (b,b′) based on Bray-Curtis similarity matrix did not show clustering between groups. However, we observed a high proportion (62% and 66%) of Pseudomonadaceae (mainly composed of Pseudomonas species) among patients exhibiting a moderate and severe disease, while it represents only 32% of the bacterial composition among patients with a mild ventilatory deficit (c). The fungal composition exhibited also some shifts, especially a decrease in OTUs belonging to Malasseziales in samples from patients with severe lung decline (c′). Taxonomy composition were represented at family level of bacterial (c) and fungal (c′) microbiotas; fungal reads that were not identified as family levels are grouped at order or phylum levels.

Figure 2

Co-occurrence network between bacteria and fungi from the ReBoot method. Bacteria and fungi are represented by blue and green circles, respectively. The large circles refer to recognized dominant CF pathogens. Gray lines connecting circles represent strong positive correlation, while red lines represent strong negatively correlation. A negative correlation between a bacteria genus and a fungus genus refers to an increase of bacteria genus abundance and a decrease of fungus genus abundance; a positive correlation to an increase of bacteria genus abundance and an increase of fungus genus abundance.

Figure 3

Inter-kingdom correlation from the ReBoot method. Statistical significance was determined for all pairwise comparisons; only significant correlations (p < 0.05) are displayed. Warm colors (red to yellow squares) indicate positive correlations, and cool colors (dark blue to light blue squares) indicate negative correlations.

Figure 4

Growth curves for in vitro co-cultures of A. fumigatus with S. mitis, or S. oralis compared to culture of A. fumigatus alone. A. fumigatus growth () expressed in CFU/mL is significantly enhanced by S. mitis at day 4 (p < 0.05) and day 5 (p < 0.01). Growth of A. fumigatus plus S. mitis () is significantly higher than growth of A. fumigatus plus S. oralis () at day 5 (p < 0.05). Growth of A. fumigatus was not significantly enhanced by S. oralis excepted at day 2 (p < 0.05).

Figure 5

OTUs relevantly associated with CFPE (a) and FEV1 (b) by bootstrap-enhanced Phy-Lasso. Only genera of bacteria (in green) and of fungi (in blue) selected with a frequency ≥15% are represented. Bars are represented at left when the genus was negatively associated, and at right when the genus was positively associated with CFPE or FEV1. We considered OTU consistently associated to clinical features when its selection frequency was greater than 70% (red lines).

Figure 6

Adaptation of Climax/Attack model (CAM) in CF as inferred from previous studies and our study^,,,,,. According to CAM, two different microbial populations are evolved dynamically in CF lungs, both being potentially composed of anaerobes^,,,,,. Microbiome changes acquired through environmental exposure or salivary aspiration (a) seem to participate to CFPE occurrence leading to an Attack population. The microbial community will return to its original state (Resilience in (b)) or move to a new community considered as stable but composed of a new Climax population with a different microbial community (Adaptation in (b)), according to perturbation forces of the Attack population and its ability to pass through selection filters (c). Selection filters (c) refer to layers that influence evolution of the microbial structure: changes in nutrient, sources-oxygen pressure, pH, level of microorganism growth and virulence, host immune and inflammatory response, antimicrobial treatment pressure. They participate in selecting the population the most adapted to the new airway remodeling, in a circular relationship. According to a microbial community partitioned by carbon source, the Climax population uses amino acids and produces ammonia, while Attack population uses sugar fermentation and produces acid. In this context, Malassezia yeasts which use lipids as carbon sources are unable to ferment sugar, and may have some advantages (cross-feeding) to share ecological niche with anaerobes and Streptococcus. Streptococcus are responsible for sugar fermentation producing amino acid. By the same time, fermentation is decreasing pH that is favorable for expansion of anaerobes which contribute to produce fermentative compounds. Furthermore, Malassezia are able to develop at acid pH. On the other side, Scedosporium are associated with decline of FEV1, which fit well with a role into an advanced Climax population in agreement with their ability to use a wider range of nutritive substrates, to synthesize specialized metabolites including ammonia fermentation, and their high resistance to antifungal drugs. Genera in black refer to our bootstrap-enhanced Phy-Lasso analysis, and genera in grey to previous CAM studies.