Proteomic profile of cystic fibrosis sputum cells in adults chronically infected with Pseudomonas aeruginosa

Discussion

In order to further elucidate the mechanisms of CF-associated tissue damage in adults with chronic respiratory P. aeruginosa infection, this study has for the first time specifically investigated the global proteome and functional activity of the combined population of cells present in CF sputum. As such, it uniquely studies the protein abundance of cells in vivo, thereby taking into account both their natural frequency in the sputum cell population and the effects of intercellular interactions. Thus, it delivers a high degree of clinical relevance to our findings. Moreover, to accurately characterise a consistent CF phenotype for adults chronically infected with P. aeruginosa and avoid bias due to patient individuality, we considered only those proteins detected in all samples within the CF and healthy control cohorts. Thus, our study is more stringent than previously published proteomic studies that either included proteins detected at high frequency in at least one sample [18] or used pooled samples [14–16].

In agreement with previous investigations of CF sputum or BALF [14, 15, 17, 18], we found the CF proteome to be largely distinct from that of the control cohort. IPA predicted that canonical pathways which affect or involve actin cytoskeleton rearrangement directed by the Rho family small GTPases, including RhoA, Rac1 and Cdc42, were over-represented in the CF samples. Thus, our in vivo data confirms previous in vitro observations linking CFTR deficiency with increased RhoA expression [26, 27]. This may be significant for CF lung health as impaired efferocytosis (phagocytosis of apoptotic cells) [28] and reduced expression of inducible nitric oxide synthase [27] resulting from CFTR deficiency have been reported as mediated by RhoA upregulation.

Similarly to previous proteomics studies of cell-free CF sputum/BALF, GO and IPA Bio Function analyses identified upregulation of the immune and inflammatory response [14–18]. In particular, we found increased immune cell movement and activation, and generation of reactive oxygen species. These observations fit with existing knowledge about the CF phenotype, which is typified by chronic inflammation and neutrophilia. Interestingly, an IPA predicted network identified lipopolysaccharide as a likely trigger of inflammation (figure 4). Cell death, growth and proliferation were also predicted as dysregulated in CF, in such a way that both death and proliferation of immune cells were decreased, while both were increased in nonimmune cells. These observations further support the growing understanding of the CF phenotype. Tissue damage and cell death is rampant, while excessive proliferation of respiratory epithelial cells leading to squamous cell metaplasia and dysplasia is common in CF [29, 30]. Delayed neutrophil apoptosis and clearance is thought to contribute to chronic inflammation [2], and our study detected CF-associated differential abundance of matrix metalloproteinase-8, annexin I and nicotinamide phosphoribosyltransferase, which have been implicated in delayed apoptosis and clearing of inflammatory and immune cells in vitro [31] and in vivo [32, 33].

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4

Ingenuity Pathway Analysis predicted network for proteins detected in cystic fibrosis (CF) involved in Inflammatory Response and Cell Movement Bio Functions. The intensity of red or green shading indicates the degree of increased or decreased abundance in CF relative to healthy controls, respectively. A red outline with no internal shading indicates that the protein was detected exclusively in the CF cohort. The network identifies lipopolysaccharide as a likely inflammatory trigger.

Compositional breakdown analysis of sample proteomes found S100-A9, the histone H2B family, lactotransferrin, histone H4, cathepsin G and myeloperoxidase to be consistently present in high abundance in CF and to each be significantly more abundant in the CF cohort than the control cohort. Importantly, all six proteins are known constituents of neutrophil extracellular trap (NET) structures, the release of which by neutrophils exerts antimicrobial activity through the degradation of virulence factors, bacterial killing and antifungal activity [34–36].

As our study focused on sputum cellular proteins we did not look specifically for NET structures, which are extracellular, and therefore do not know if they were present. However, NETs have been found previously in the CF airways fluid and the abundance of free DNA that is characteristic of NETs correlated with obstructive lung function [37]. As well as killing microbes, NETs have been found to directly induce human alveolar epithelial and endothelial cell death, with the predominant cytotoxic effect being attributed to histones and myeloperoxidase [38].

Individually, these six high-abundance proteins may also have significance for CF lung pathology and several have previously been reported as upregulated in CF sputum and/or BALF [14, 15, 18]. Importantly, differential abundance of myeloperoxidase in CF sputum fluid has been correlated with FEV₁ % pred [17], and myeloperoxidase and calgranulin B have been identified in sputum fluid as CF-associated autoantigens [16]. Cathepsin G has been implicated in degradation of surfactant protein A, so reducing the innate pulmonary antimicrobial defence [39], while S100-A9 has been shown to induce production of the gel-forming glycoprotein MUC5AC in human bronchial epithelial cells cultured in vitro, suggesting a potential contribution to pathological mucin hyperproduction in vivo [40]. In a study investigating trauma-induced lung injury, Abrams et al. [41] reported that histones induced NET formation and myeloperoxidase release in vitro, while histone infusion in mouse models resulted in lung structural damage and neutrophil congestion. Additionally, extracellular histones H4 and H3.3 have been implicated in asthma-related inflammation and remodelling, and in chronic obstructive pulmonary disease, respectively [42, 43]. Consequently, our consistent detection of these proteins in such high abundance in the sputum cellular fraction identifies a potential risk to lung health, either through their active secretion and NET formation or following passive release into the sputum fluid via cell death.

A number of proteins detected consistently in CF sputum correlated negatively with lung function measured as FEV₁ % pred. Their biological/clinical relevance and potential for causing injury are discussed more fully in the supplementary material. Longitudinal analysis of patient-matched CF samples showed differential abundances for 10 out of these 17 proteins in the CF cohort following i.v. antibiotic treatment of a pulmonary exacerbation. Interestingly, all the changes in abundance resulted in levels closer to those in the control cohort. Previously, Sloane et al. [17] observed that for four out of 13 adults with CF, their cell-free sputum proteome profiles moved closer to those of healthy controls following antibiotic treatment of an exacerbation. However, the latter observations were made by comparison of two-dimensional gel images rather than implicating and identifying particular proteins. We also found that, although sputum total protein yield decreased in the CF cohort by completion of antibiotic treatment, it had reverted to pre-treatment levels by the clinically stable follow-up time point. This suggests that proteomic changes associated with antibiotic treatment may be temporary.

Overall, the proteins detected as negatively correlating with FEV₁ suggest that the degree of inflammation, neutrophil influx, protease activity and airway remodelling may all contribute to defining lung function. α₁-Antitrypsin and olfactomedin 4 were identified as independent factors negatively correlating with FEV₁, although this does not demonstrate causation of harm. Acute inflammation-associated protein α₁-antitrypsin has previously been identified as a biomarker of pulmonary disease in CF which can be predictive of response to treatment and prognosis [44]. As an inhibitor of neutrophil serine proteases, which contribute significantly to lung destruction in CF, its negative correlation with lung function might seem surprising. However, previous studies observed that plasma deficiencies during inflammation were associated with less severe lung disease in CF [45]. No association between olfactomedin 4 and CF lung health has previously been reported. However, olfactomedin 4 has been shown to downregulate pro-inflammatory responses to bacterial infection and reduce bacterial killing mediated by granule serine proteinases [46, 47], which could have the potential to negatively affect lung health.

There are a number of limitations to our study. As is typical for untargeted exploratory proteomic studies, we used highly sensitive shotgun technologies to obtain quantitative, high proteome coverage analysis of sputum cells as a novel target. The trade-off for such in-depth analysis is a relatively low sample size. Given the sample numbers it was practical to analyse using these methods, we maximised power by restricting the study to adults with CF chronically infected with P. aeruginosa, who experienced pulmonary exacerbation requiring hospitalisation for treatment with i.v. antibiotics. While CF sputum proteome responses to other microbial infections may vary, we considered this cohort most informative as 57% of the adult CF population is chronically infected with P. aeruginosa [48]. As this study investigated only adults the findings cannot be extended to children or individuals not chronically infected with P. aeruginosa. Although proteins reported as potentially damaging were detected as consistently abundant or correlating negatively with lung function in our study, this is not proof of causation of lung injury. Additionally, the limited sample size requires that any potential significance as biomarkers either of lung function or disease severity warrants validation in a larger independent cohort. As an initial stage in this validation process we used commercial ELISAs to confirm findings in our existing cohort for myeloperoxidase and lactotransferrin. Both proteins were considered as candidate biomarkers due to their consistently high abundance in CF sputum, increased abundance in the CF compared with the control cohort and previous associations in the literature with potential for lung damage. Thus, we demonstrated that commercial ELISAs are a practical method for follow-up validation in an independent cohort. However, while cohort size and characteristics limit the scope of our study conclusions, the findings are of exploratory interest and may have future clinical value within a defined commonly encountered CF cohort. Moreover, the investigation of all types of cells in the proportions occurring naturally in sputum is novel and the approach adopted in our study could be informative for other respiratory conditions.

In summary, within the cohort studied, i.e. adults with chronic P. aeruginosa infection, we detected a distinct signature for the CF sputum cellular proteome, and have used pathway and functional analyses to predict its potential clinical impact. In particular, our data highlight the prevalence and predominance of proteins that can be associated with NET structures and have been reported as linked to tissue damage. Additionally, we identified a number of proteins that correlate negatively with lung function, many of which have previously been associated with lung injury. Overall, these findings expand the current understanding of the mechanisms behind P. aeruginosa-associated CF lung disease in adults and identify sputum cell proteins as candidate biomarkers for disease status, prognostic indicators or as stratification parameters for treatment prescription.