If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Microbiology, Immunology and Infectious Diseases, University of Calgary, Calgary, AB, CanadaDepartment of Biological Sciences, University of Calgary, Calgary, AB, Canada
Department of Microbiology, Immunology and Infectious Diseases, University of Calgary, Calgary, AB, CanadaDepartment of Medicine, University of Calgary, Calgary, AB, Canada
Department of Microbiology, Immunology and Infectious Diseases, University of Calgary, Calgary, AB, CanadaDepartment of Medicine, University of Calgary, Calgary, AB, Canada
Department of Microbiology, Immunology and Infectious Diseases, University of Calgary, Calgary, AB, CanadaDepartment of Medicine, McMaster University, Hamilton, ON, Canada
Department of Microbiology, Immunology and Infectious Diseases, University of Calgary, Calgary, AB, CanadaDepartment of Medicine, University of Calgary, Calgary, AB, Canada
The CF microbiome was assessed during a 56-day cycle of inhaled aztreonam (AZLI).
•
No major changes in the microbiome were observed with AZLI usage.
•
Clinical response was tested for associations with changes in the microbiome.
•
Organisms known to resist AZLI were found in higher abundance in non-responders.
•
This study suggests that the microbiome may serve to personalize therapies in CF.
Abstract
Background
To improve clinical outcomes, cystic fibrosis (CF) patients with chronic Pseudomonas aeruginosa infections are prescribed inhaled anti-pseudomonal antibiotics. Although, a diverse microbial community exists within CF airways, little is known about how the CF microbiota influences patient outcomes. We hypothesized that organisms within the CF microbiota are affected by inhaled-antibiotics and baseline microbiome may be used to predict therapeutic response.
Methods
Adults with chronic P. aeruginosa infection from four clinics were observed during a single 28-day on/off inhaled-aztreonam cycle. Patients performed serial sputum collection, CF-respiratory infection symptom scores (CRISS), and spirometry. Patients achieving a decrease of ≥2 CRISS by day 28 were categorized as subjective responders (SR). The airway microbiome was defined by Illumina MiSeq analysis of the 16S rRNA gene.
Results
Thirty-seven patients (median 37.4 years and FEV1 44% predicted) were enrolled. No significant cohort-wide changes in the microbiome were observed between on/off AZLI cycles in either alpha- or beta-diversity metrics. However, at an individual level shifts were apparent. Twenty-one patients (57%) were SR and fourteen patients did not subjectively respond. While alpha-diversity metrics did not associate with response, patients who did not subjectively respond had a higher abundance of Staphylococcus and Streptococcus, and lower abundance of Haemophilus.
Conclusions
The CF microbiome is relatively resilient to AZLI perturbations. However, associated changes were observed at the individual patient level. The relative abundance of key “off-target” organisms associated with subjective improvements suggesting that the microbiome may be used as a tool to predict patient response - potentially improving outcomes.
]. Indeed, it is possible that inhaled antibiotics such as aztreonam lysine for inhalation (AZLI)(Cayston ®, Gilead) may affect other organisms in the CF lung microbiome. In a previous study, our group examined AZLI's effects on a cohort of treatment naïve patients in an attempt to determine how the microbiome influenced therapeutic response [
]. Individuals who failed to experience improved lung function one year after initiation had microbiomes with a higher relative abundance of organisms expected to resist aztreonam, including Staphylococcus. However, the study was limited by variability in the number of samples/patient, and timing of sample collection. Herein we present the results of a prospective multi-centre study aimed to address these limitations. We hypothesized that AZLI has additional targets within lung microbiome beyond merely P. aeruginosa and that the presence/abundance of these constituent organisms associates with patient response.
2. Patients and methods
2.1 Ethics statement
Institutional approval for the study was approved by the University of Calgary Conjoint Health Region Ethics Board (REB14-0109) as well as at each contributing site. Prior to enrollment all patients received instruction on the study protocol and its requirements and consented in writing.
2.2 Patient recruitment and sample collection
Patients were enrolled from four Canadian CF clinics. Inclusion criteria were age ≥ 18 years, current AZLI usage, ability to expectorate sputum, and possession of a home deep-freezer to store sputum. Exclusion criteria were infection with Mycobacterium abscessus complex, lung transplantation, and receipt of intravenous antibacterials in the 28 days prior to enrollment. Cohort characteristics including demographics, baseline lung function, co-morbidities, chronic cultured pathogens (including all organisms identified in the year before study enrollment as per the Leed's criteria - adapted from P. aeruginosa) [
], and chronic therapies were documented. AZLI usage prior to enrollment was recorded and patients were classified as being on either AZLI monotherapy or receiving chronic sequential aerosolized antibiotics (CSAA- receiving either TIS/TIP or Colistimethate sodium (COL)).
After a ≥14-day washout of inhaled antibiotics, patients nebulized 75 mg of AZLI TID, as licensed, for 28-days, followed by a 28-day period off drug. Patients collected sputum in 15 mL containers on pre-determined days of AZLI cycle; 0 (pre-study baseline sample), 14 (mid-treatment) & 28 (end-treatment), and 42 & 56 (off AZLI or on CSAA). Samples were frozen immediately after collection in patients' home deep-freezers with gel packs to prevent thawing [
] and recorded forced expiratory volume in one second (FEV1) and 6 s(FEV6) using a PiKo-6® spirometer (nSpire Health, Inc). Piko-6® values were compared to formal spirometry done in clinic at enrolment and completion of the study to ensure validity. FEV1 and FEV6 percent predicted were determined using the NHANES method. Patients dropped off sputum samples at the end of the study at their respective sites, transported with gel-freeze packs in insulated bags to prevent thawing. Samples were stored at −80 °C and subsequently shipped to the University of Calgary on dry ice.
2.3 Definitions
Patients were categorized as having mild (FEV1% ≥70%), moderate (FEV1% ≥40 to <70%) or severe lung disease (FEV1% <40%). The primary outcome was decrease in symptom burden as a measure of quality of life (QOL) using the CFRSD-CRISS on day 28 compared to day zero. Raw CFRSD scores were converted to CFRSD-chronic respiratory infection symptom score (CFRSD-CRISS) and patients were deemed subjective responders (SR) if they experienced a decrease in symptom score of at least two [
]. The secondary outcome measure was improvement in objective measures of lung function on day 28 compared to day zero. Patients were defined as objective responders (OR) if they achieved an absolute improvement in FEV1% of ≥2%. This definition of response was selected based on short-term improvements in naïve populations of 2.7–10.8% in previous studies [
]. Notably, minimally clinically significant values associated with patient-reported or spirometric outcomes do not exist for patients on established chronic CF therapies. Cutoff values for both SR and OR were chosen prior to microbiome analysis to generate two groups of comparable size in order to detect differences in a modest sized patient cohort. Exploratory analyses were performed on the following variables to determine their association with changes in the CF lung microbiome; day of AZLI cycle, on CSAA, and number of prior AZLI cycles.
2.4 DNA extractions & 16S rRNA amplification, processing and analysis
Bacterial DNA was extracted from sputum samples as previously described [
]. In brief, amplification of the V3-V4 hypervariable region of the 16S rRNA gene was carried out using reverse and forward barcoded primers on the Illumina MiSeq platform [
] 2013 to assign taxonomy. OTUs were filtered if they did not have bacterial taxonomy and were present as a single read in the entire dataset (singletons), or only present in one sample. A 2500 read minimum sequencing depth cutoff was set a priori. Analysis of OTU tables was carried out using QIIME [
To describe changes in lung bacterial community richness and evenness, we performed Wilcoxon rank sum and Kruskal Wallis tests on Shannon alpha-Diversity Index (SDI). In addition, to account for day of AZLI cycle and for repetitive sampling of patients, a restricted analyses of all samples (with the exception of those on intravenous antibiotics) using a linear mixed effects model randomizing by patient ID followed by a type II Anova using the Kenward-Roger's method was conducted [
]. To describe the changes in lung bacterial community structure, we conducted community-wide Bray-Curtis (BC) beta-diversity analyses and visualized variation in community structure using principal coordinate analysis (PCoA) plots. Read counts were divided by total number of reads to achieve proportional normalization of samples prior to conducting BC community-wide assessments [
]. Restricted analyses accounting for day of AZLI cycle (time) and setting constraints to permutations by patient were also conducted as a sensitivity analysis. Taxa present in greater than one sample and in >1% in total relative abundance were used to analyze genera that are differentially abundant for primary and secondary outcome variables using DESeq2 (test = “Wald” fitType = parametric and Benjamin Hochberg multiple test correction). Log2 of normalized reads were used to visualize data generated by DESeq2, samples with an abundance value of 0 were changed to a value of 1 since log(0) is undefined and log(1) = 0. Change in spirometry and CFRSD-CRISS from day zero were carried out in R-studio using a Wilcoxon rank sum test. Correlations between taxa relative abundance and change in CRISS from day 0 were determined using the Spearman correlation coefficient (r). Change in Pseudomonas relative abundance on day 0 was conducted in STATA v13.1 (Stata Corp, TX, USA) using a Wilcoxon rank-sum test. Matched analysis assessing changes in Pseudomonas relative abundance on day 0 vs day 28 in subjective response status were conducted using a Wilcoxon signed rank-sum. Chi-squared tests were used to determine differences between dichotomous clinical variables including CF related co-morbidities, cultured pathogens, and concurrent therapies. Wilcoxon rank-sum tests were conducted for analysis of continuous variables including median age, BMI, number of previous AZLI cycles.
3. Results
3.1 Cohort characteristics
Thirty-seven patients were enrolled. Patient demographics, chronic therapies, and cultured pathogens are presented in Table 1 and Supplementary Table S1. Median FEV1% predicted at enrollment was 44% (IQR 32.0–69.5%). During the study period 25/37 (68%) patients were on AZLI monotherapy, 10/37 (27%) cycled with inhaled tobramycin and 2/37 (5%) used inhaled colistin during the off-AZLI cycle. There was no significant difference in age (p = 0.34) or lung function (p = 0.18) in those receiving CSAA vs AZLI monotherapy at baseline. During the 28 days of AZLI therapy, a progressive increase in FEV1% and decrease in symptom burden were observed (Supplementary Fig. S1). Twenty-one patients (57%) met the definition of SR, fourteen (38%) did not and two patients were excluded, as they did not record data on day 28 (Table 1). Nineteen (51%) patients met the definition of OR, fourteen did not (38%) and four were excluded from analysis owing to lack of day 28 spirometry data (Table 1). Concordance between SR and OR status was observed in 12 responders and 6 in non-responders. No significant differences were observed in changes in lung function on day 0 and 28 for patients receiving AZLI monotherapy vs CSAA. No significant differences were noted between the number of patients on IV antibiotics in SR vs patients who did not subjectively respond as measured by Pearson chi-square (p = 0.083). Furthermore, no significant differences were noted based on subjective or objective measures of response and patient demographics, cultured therapies or concurrent therapies with the exception of all SR being pancreatic insufficient (Table1 and Supplementary Table S1). However, our modest sample size limits our ability to detect differences in comparisons.
Table 1Baseline cohort demographics and microbiologic characteristics.
A total of 11,344,146 sequences (median 54,445/sample (IQR 42,853-87,047)) were generated from 183 sputum samples (minimum 2766 and maximum 150,382 reads/sample). Taxonomic summaries were generated for each patient for taxa present in more than one sample and >20% in total relative abundance (this cutoff was selected to simplify visualization) (Fig. 1). Patient sub-plots are organized by SR, demographics and day of cycled AZLI as noted in Fig. 1. Many patients' (i.e. 10, 16, 19, 21, 22, 33) sputum microbiota was dominated by Pseudomonas, while others had much greater heterogeneity in their lung bacterial community structure.
Fig. 1Microbiome through cycled AZLI. Taxonomic summaries for taxa present in >20% in total relative abundance (n = 183) for a cohort of 37 patients with chronic P. aeruginosa infection as a function of days on a single cycle of AZLI therapy. Day of cycled AZLI therapy are represented on the x-axis; red text = off AZLI, green = on AZLI, blue = baseline samples off AZLI, black = extra samples collected during clinic visits. Individual patient sub-plots are organized by subjective response (SR) and gender. Samples on acute parenteral antibiotics are indicated by asterisks. Samples collected on acute oral antibiotics are represented by black circles and samples on oral azithromycin are represent by (x). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4. The impact of inhaled antibiotics on the CF lung microbiome
No significant differences were observed in either SDI or BC beta-diversity (PERMANOVA, p = 0.99, R2 = 0.011) (Figs. 2A&B) based on day of AZLI cycle. However, when we conducted restricted analyses and stratified permutations by patient ID we see a significant difference in BC beta-diversity (PERMANOVA, p = 0.009, R2 = 0.011) suggesting that overall the cohort microbiome does not change through cycled AZLI but at an individual level there are differences. Post-hoc analyses were conducted between biologically meaningful time points (for example, baseline vs mid- and end-of-treatment and end-of-treatment vs end-of-off cycle). Analyses were performed by setting constrains to permutations by patient ID. Statistically significant differences were noted between day 0 and day 14 (PERMANOVA, p = 0.007, R2 = 0.0089), day 0 and day 28 (PERMANOVA, p = 0.011, R2 = 0.0069). However, no differences were observed between day 0 and day 42 (PERMANOVA, p = 0.58, R2 = 0.0025), day 0 and day 56 (PERMANOVA, p = 0.58, R2 = 0.0025), day 28 and 56 (PERMANOVA, p = 0.057, R2 = 0.0026). Next, we sought to identify whether there are differences in the microbiome of patients on AZLI monotherapy or CSAA. Although no differences were noted for SDI using both the Wilcoxon rank-sum test (p = 0.083) and the mixed linear effects model (ANOVA, p = 0.42) samples clustered based on treatment regimen as demonstrated by PERMANOVA (p = 0.016, R2 = 0.019). These differences were sustained when we accounted for time of sample collection and constraints were set to permutations by patient ID (PERMANOVA, p = 0.031, R2 = 0.019).
Fig. 2The impact of cycled AZLI on the CF microbiome. A) Changes in Shannon alpha-diversity index (SDI) as measured by Kruskal Wallis test reveal no significance depending on when samples were collected during a single 28-day on/off AZLI cycle. B) PCoA on Bray-Curtis dissimilarities shows no significant difference during a single 28 day on/off AZLI cycle as demonstrated by PERMANOVA analysis (p = 0.99). Samples on systemic antibiotics and extra samples collected out of study period were excluded from this analysis.
We also assessed how prior exposure to AZLI associated with changes in the CF microbiome (Fig. 3). Seven patients (19%) (n = 32 samples) were AZLI naïve at enrolment, ten (27%) (n = 48 samples), had 1–6 cycles and twenty (54%) (n = 91 samples) had ≥7 cycles. Alpha-diversity as measured by SDI trended towards a decreased diversity with increased exposure to AZLI cycles as measured using a linear mixed effects model to account for time of sample collection and repetitive samples collected from individual patients (ANOVA, p = 0.20). In contrast there was significant differences in BC beta-diversity based on number of previous cycles of AZLI (PERMANOVA, p = 0.001, R2 = 0.062) (Fig. 3B) and these differences were sustained using restricted analyses (PERMANOVA, p = 0.034, R2 = 0.062).
Fig. 3Repeated AZLI exposures and the CF microbiome. Seven patients (n = 32 samples) had no prior exposure to AZLI at enrollment. Ten patients (n = 48 samples) had received 1–6 prior cycles of AZLI at enrollment. Twenty patients (n = 91 samples) had received ≥7 cycles of AZLI at enrollment. A) Shannon alpha-diversity index trended towards decrease diversity (ANOVA p = 0.20). B) PCoA on Bray-Curtis dissimilarities showing significant community-wide differences as demonstrated by PERMANOVA analyses (p = 0.001). Only samples free of systemic antibiotics were included in this analysis.
Next, we investigated whether the microbiome associated with clinical response (Supplementary Fig. S2 Fig. 4). Microbiome assessments were conducted on baseline samples (Day 0) to determine if the microbiome can be used as a tool to predict subsequent patient response on day 28. This type of sample analysis could be used to personalize CF therapies. When comparing baseline samples in SR versus patients who did not subjectively respond we found no significant difference in SDI. Similarly, no significant differences were observed in BC beta-diversity as communities did not cluster by response status (p = 0.12, R2 = 0.052) (Fig. 4B). However, patients who did not subjectively respond had a higher abundance of Streptococcus and Staphylococcus and a lower abundance of Haemophilus (Figs. 4C–E). When we re-assessed these observations using alternate and increasingly stringent definitions of SR requiring change in CRISS up to −11 we observed similar trends (Supplementary Table S2). Furthermore, similar trends were observed for differences in taxa using Wald's test as implemented in DESeq2 although with more stringent CRISS cut-off values significance is lost. SR trended towards a higher relative abundance of Pseudomonas on day 0 and day 28 of AZLI therapy in comparison to patients who did not subjectively respond and had a significant decrease in Pseudomonas relative abundance on day 28 (Supplementary Fig. S3E). In addition, we found modest correlations between the relative abundance of Streptococcus, Staphylococcus, and Pseudomonas and change in CRISS from day 0 (Supplementary Fig. S3). However, no associations were observed with the relative abundance of Haemophilus and change in CRISS from day 0 (Supplementary Fig. S3).
We found no significant differences in SDI between OR and non-OR nor did samples cluster by OR status (p = 0.93, R2 = 0.011) (Supplementary Figs. S2 A&B). In addition, no organisms were differentially abundant between OR and non-OR. Similar analyses were conducted on samples collected on day 28 to identify whether treatment induced changes associated with both SR or OR status, however, none were identified.
Previous AZLI exposure at baseline and its effect on subjective response was assessed. No correlation was found between previous AZLI exposure at baseline and subjective response as determined using a Spearman correlation (r = 0.073) suggesting that previous exposure to AZLI does not influence subjective response to therapy. Furthermore, there was no significant difference in SR vs patients who did not subjectively respond based on previous AZLI exposure (p = 0.68).
Fig. 4Microbiome analyses in context of subjective response (SR) to AZLI. Responders were defined as patients who experienced a decrease of at least 2 symptoms on day 28 compared to day 0. A-E) Microbiome analyses of baseline samples only A) Differences in Shannon alpha-diversity index (SDI) between SR and non-SR for baseline samples were not significant as demonstrated by Wilcoxon rank sum test. B) PCoA on Bray-Curtis dissimilarities showing community-wide differences between samples collected from SR vs non-SR for baseline samples; PERMANOVA analyses reveal no community-wide differences (p = 0.12). C-E) Log 2 of normalized reads reveal non-SR have a higher abundance of Streptococcus and Staphylococcus as well as a lower abundance of Haemophilus at baseline as measured by a Wald's test using DESeq2. Samples from patients with missing SR data and on systemic antibiotics were excluded from this analysis.
The CF lungs are colonized by a diverse group of organisms, however, how these complex communities influence disease progression and response to various therapies is poorly understood [
]. It is generally established that patients with more advanced lung disease are infected with a higher abundance of key pathogens such as P. aeruginosa, and therefore have a less diverse microbiome [
]. It has also been demonstrated that the use of parenteral antibiotics for acute pulmonary exacerbations induces transient shifts in the CF microbiota [
Long-term cultivation-independent microbial diversity analysis demonstrates that bacterial communities infecting the adult cystic fibrosis lung show stability and resilience.
]. Furthermore, previous studies have determined that specific components within the microbiome may protect P. aeruginosa from destruction through the production of antibiotic modifying enzymes [
Production of extended-spectrum β-lactamases and the potential indirect pathogenic role of Prevotella isolates from the cystic fibrosis respiratory microbiota.
]. However, little is known as to if/how inhaled antibiotics may trigger changes in the CF lung microbiome and whether certain components in the microbiome influence patient response to different therapies [
Recently, our group analyzed the effects of AZLI on a treatment-naïve cohort of patients by performing a retrospective analysis of our Calgary Adult Cystic Fibrosis Clinic sputum biobank [
]. However, this retrospective study was limited as samples were collected during clinic visits irrespective of timing of AZLI administration and included only 24 patients. In that work, no significant changes in the microbiome following AZLI initiation were observed – contrasting what has been observed during parenteral therapies [
]. This was surprising, given the high concentrations achieved by inhaled antibiotics thereby potentially broadening the spectrum of activity of each agent relative to levels achieved with the same drug via systemic delivery.
We believe that in addition to providing greater insight into the pathophysiology of CF lung disease, the CF microbiome holds great promise as a biomarker for predicting both prognosis and treatment response [
]. Indeed, a biomarker predictive of improved therapeutic response could enable a microbiome-tailored treatment in precision medicine. Herein for the first time, we observed using an established, validated patient reported outcome tool – the CFRSD, that patients who failed to achieve SR had greater relative abundance of Staphylococcus within their microbiome. Indeed, this nicely complements information obtained from our prior work where patients who failed to achieve a significant improvement in FEV1 within one year of treatment initiation also had higher abundance of Staphylococcus [
]. When tested, we found the MIC of S. aureus and various Streptococcus species derived from individuals with CF on AZLI to be consistently ≥512 mg/L (data not shown) – well above the peak level of aztreonam achieved in sputum during AZLI administration - confirming that these organisms are resistant to AZLI [
]. While no minimally clinically significant change in CRISS score exists for stable patients continuing long-term therapies, we confirmed those observations made using our a priori definition with a range of scores – some of which had even greater significance. We similarly observed that a higher abundance of Streptococcus and lower abundance of Haemophilus at baseline also associated with lack of subjective response. Together these findings and their biologic plausibility suggest that certain components within the microbiome may resist the effects of inhaled antibiotics. Furthermore, when we conducted Spearman correlation analyses on all samples (excluding day 0 samples and samples on systemic antibiotics) we observed a modest correlation between an increase in Streptococcus and Staphylococcus and increase in symptom burden as measured by change in CRISS from day 0. This further supports the notion that these organisms may be associated with lack of response. While earlier studies show a decrease in the burden of Pseudomonas after twenty-eight days of AZLI treatment this has not been previously linked to response status [
]. Interestingly, our study demonstrates an association between SR and decrease in the relative abundance of Pseudomonas where SR had a significant decrease in Pseudomonas on day 28 compared to day 0. In addition, we were interested in determining whether previous exposure to AZLI influenced SR. There was no correlation or difference based on change in CRISS on day 28 and previous exposure to AZLI. Interestingly however, no significant differences were observed in SDI or microbiota composition between OR and non-OR. These differences in microbiome findings between SR and OR may be due to the fact that there was concordance between SR and OR status in 12 responders and 6 non-responders.
Whether the microbiome is being driven towards lower diversity by repeated antibiotic exposures, as has been suggested by others [
], or is merely associated with a sicker cohort remains to be conclusively determined. We observed that the microbiome of those patients who were naïve to AZLI trended towards higher diversity in comparison to those previously treated patients. Furthermore, patients on AZLI monotherapy trended towards higher SDI relative to those receiving CSAA. This may be due to increased exposure to different classes of antibiotics thereby potentially affecting the microbiome. Indeed, community-wide BC beta-diversity analyses, showed samples collected from patients on CSAA clustered together. This is biologically plausible as you would expect patients with increased exposure to antibiotics to have a less diverse microbiome. Cycling with alternating inhaled antibiotics is relatively new in the management of chronic P. aeruginosa infections in CF. A study by Flume et al., found that CSAA with AZLI/TIS reduced exacerbation rates by 25.7% (p = 0.25) compared to TIS/placebo although difficulty recruiting meant this study was underpowered to confirm this difference [
]. The potential for this advantage was most pronounced in those with the greatest burden of lung disease.
Several limitations of our work merit discussion. The inclusion of patients with severe lung disease – those for whom AZLI is predominately used given its role as salvage therapy in Canada – means patients generally have a less diverse microbiome, potentially more impervious to perturbations. This may have limited our ability to detect differences incurred from AZLI usage through a single cycle. Indeed, when we stratified permutations by patient ID we observed differences before and after suggesting that there are intra-patient changes in the microbiome as a result of AZLI. Post-hoc analyses revealed these differences were between samples collected off antibiotics on day 0 and half-way through and completion of a 28-day cycle AZLI cycle but not day 0 and half-way or completely through the 28-day off cycle. This may suggest that the microbiome changes during AZLI therapy and reverts back to baseline as has been observed with parenteral anti-bacterial therapies [
]. As management of CF patients generally involves multiple overlapping therapies, this may confound our findings although there were no significant differences in the types of concurrent therapies patients were on in either subjective or objective responders/non-responders with the exception of all SR being pancreatic insufficient. Our ability to detect differences in cohort characteristics would be limited based on our modest cohort size, however, differences in clinical response have not been reported to be influenced by demographics in other studies of AZLI [
Although we observed differences in SDI and BC beta-diversity for a number of the exploratory analyses in some instances these were not sustained when we accounted for day of AZLI cycle and when constrains were set to permutations by patient suggesting individual differences may contribute to the observed microbiome changes and that microbiome data needs to be analyzed with caution. Further studies with larger cohorts are needed to confirm these findings. Indeed, large numbers of patients are generally required to identify pathogen-outcome associations in CF [
]. Despite the simplicity and short duration of our protocol, recruiting thirty-seven patients proved difficult, primarily as AZLI is generally used as a salvage agent in Canadian populations. Although patient reported outcome tools are commonly used as important endpoints in the evaluation of new CF therapies, their use in continued maintenance therapies to distinguish variable treatment responses have not been explored. While there are no clinically established values to assess the reduction of CRISS units during chronic inhaled therapies we did subsequently confirm our observations using a range of values from −1 to −11, where similar trends were evident. Variability in response to specific therapies is an area of active CF research enabling patient specific tailoring of therapies – we believe this should also apply to anti-infectives [
In addition, we only used sputum to assess changes in the lung microbiome and we acknowledge that there may be biases in such analyses as studies have shown heterogeneity in the lung microbiome based on which region of the lung a sample is derived from and the potential for oral-pharyngeal contamination [
]. However, using invasive techniques to sample multiple regions of the CF lungs is difficult, poses risks and is impossible to safely perform serially and therefore lacks the capacity sputum possesses as a potential biomarker. To obtain a more accurate and comprehensive understanding of the effects of new and emerging therapies on the CF lung microbiome future investigations should prospectively collect samples for subsequent multi “omics” analyses aimed at bettering our functional understanding of the lung microbial communities such as metagenomics, meta-transcriptomics, proteomics, and metabolomics.
6. Conclusion
Our study suggests that the CF airways microbiome was relatively stable under AZLI exposure, although, modest changes were observed in individual patients. Importantly, we observed that patients whose sputum had higher abundance of Staphylococcus and Streptococcus, and lower abundance of Haemophilus were less likely to experience a SR but not an OR to AZLI. Interestingly, SR also showed a significant decrease in the relative abundance of Pseudomonas on day 28 compared to day 0. Our data supports that a patient's microbiome may be used to predict treatment responsiveness hence enabling personalization of therapy.
Funding
This work was supported by Gilead Sciences (IN-CA-205-1343).
Conflicts of interest of statment
AAH, BQ, NA, MLW, DGS, WL, ILL, and LR report no conflicts. YB reports having performed advisory board work for Vertex Canada and provided consultant services to Prometic Life Sciences. MDP has received research support from CF Canada and CIHR. MDP, HRR, and MGS received supporting funding for this work from Gilead Sciences.
Contributions
MDP, YB, BQ, WL served as local site investigators and recruited patients for participation. AAH received samples and data, and was responsible for its coalition and processing. Sample processing and statistical analyses were performed by AAH, NA, MLW, ILL, LR, MGS and MDP. BJW performed susceptibility testing. AAH wrote the initial draft of the manuscript, and all authors contributed to its revision. MGS and MDP designed the study. MDP supervised the study, and serves as guarantor of the work.
Acknowledgments
The authors gratefully acknowledge the participation of patients and staff of the Calgary, Edmonton, Montreal, and Vancouver adult CF clinics. YB holds the IRCM Gosselin-Lamarre Chair in clinical research. The authors thank Mr. Austin Nguyen and Ms. Jasper Greysson-Wong for their assistance in collating clinical outcome information.
Change in clinical symptoms from baseline (day 0). A) Change in CFRSD-CRISS as a measure of patient's quality of life (QOL). Samples where QOL data was not collected were filtered out. A decrease in symptom burden indicates a positive outcome. B) Change in FEV1 as a measure of lung function through a single on/off cycle of AZLI. Samples missing lung function values were filtered out of these analyses. Differences were detected using Wilcoxon rank sum test in R-studio.
Microbiome analyses in context of objective response to AZLI. OR were defined as patients who experienced an improvement of ≥2% in FEV1 on day 28 relative to day 0. A) Differences in Shannon alpha-diversity index between OR and non-OR were not significant as demonstrated by Wilcoxon rank sum test. B) PCoA on Bray-Curtis dissimilarities showing community-wide differences of samples collected from OR vs non-OR. PERMANOVA reveals no community-wide differences (p = 0.93). Samples on systemic antibiotics, and samples where patients did not record lung function values on day 28 were excluded from this analysis.
References
Lipuma J.J.
The changing microbial epidemiology in cystic fibrosis.
Long-term cultivation-independent microbial diversity analysis demonstrates that bacterial communities infecting the adult cystic fibrosis lung show stability and resilience.
Production of extended-spectrum β-lactamases and the potential indirect pathogenic role of Prevotella isolates from the cystic fibrosis respiratory microbiota.
30028-1&id=gr1.jpg){kind=link}
30028-1&id=gr2.jpg){kind=link}
30028-1&id=gr3.jpg){kind=link}
30028-1&id=gr4.jpg){kind=link}