Antibiotic perturbation of mixed-strain Pseudomonas aeruginosa infection in patients with cystic fibrosis

Despite improved survival for people with cystic fibrosis (CF) [1], most still die prematurely from chronic pulmonary infections characterised by recurrent exacerbations, progressive lung function decline, increased treatment requirements and reduced quality of life [2‐6]. The pathophysiology of pulmonary exacerbations is nevertheless poorly understood and a standardised definition of an exacerbation remains elusive [7‐9].

Pseudomonas aeruginosa is the most common pulmonary pathogen in CF and antibiotic treatment of exacerbations directed against this organism is pivotal to patient management [10]. Once P. aeruginosa becomes established within the airways of patients with chronic lung disease, it is usually by a single strain that evolves through micro-adaptation into multiple sub-lineages of the original ancestral clone [11, 12]. There are however, reports of co-infection with two or more distinct P. aeruginosa genotypes in both CF [13‐15] and, non-CF bronchiectasis [12]. In Australia, a predominant shared strain, AUST-02, was detected in 18% of all patients with P. aeruginosa infection nationally and was associated with increased centre visits and intravenous antibiotic courses [13]. Given the high-prevalence and clinical significance of the AUST-02 strain within Australian centres, including its predominance in our own clinic, the primary aim of this study was to assess AUST-02 population stability and determine whether other P. aeruginosa strains emerged during and after treating an exacerbation.

Baseline patient characteristics are summarised in Table 1 (individual data, Additional file 2: Table S1 and Additional file 3: Table S2). All patients had advanced lung disease, had been chronically infected with P. aeruginosa for several years, and with one exception were taking maintenance oral azithromycin with inhaled anti-pseudomonal antibiotics prior to the exacerbation episode. Between-group differences at baseline were not statistically significant. A limited numbers of bacterial co-pathogens were isolated from sputum samples patients at baseline and within prior 3 months (4 patients with methicillin-susceptible Staph. aureus), During the period of exacerbation treatment, all patients received combination intravenous anti-pseudomonal antibiotics and continued taking azithromycin. Inhaled antibiotics were ceased. The median (range) period from completion of intravenous antibiotics and follow-up visit was 42 (13 – 119) days.

Shared P. aeruginosa strain dynamics during and after antibiotic treatment of a pulmonary exacerbation in chronically infected CF patients were investigated by combining in-depth culture and genotyping techniques. Interestingly, this Australian study found that a substantial number of CF patients showed mixed-strain infection, which contrasts with the extensive experience in the UK where similar studies of the Liverpool Epidemic Strain (LES) consistently demonstrated that CF patients have chronic infection with a single strain, which has diverged into multiple co-existing sublineages in the airways [11, 26]. This disparity is possibly related to differences in patient segregation practices between study centres.

Patients with mixed-strain infection showed significant changes in strain composition throughout the entire exacerbation episode and follow-up period, with others typically maintaining the same strain. Almost all other non-AUST-02 isolates (>99%) identified comprised previously recognised Australian shared strains, suggesting that people infected with one shared strain are more likely to be co-infected with other shared strains than unique strains. Alternatively, it can be argued that patients with advanced lung disease spending greater time in hospital are more likely to be exposed to one or more shared strains [18].

Single-strain infection with LES was previously shown to be associated with increased treatment requirements [27]. Our current study provides some evidence that infection with multiple shared strains is associated with greater morbidity than single-strain infection in Australia. In the year before recruitment and the months following their exacerbation episodes, those with mixed-strain infections had more treatment requirements than those with predominant single-strain infection, although this difference did not reach statistical significance. Further studies involving larger patient numbers and multiple centres are needed to confirm this observation. The data also potentially suggest that short-term intravenous antibiotic courses might not be a driver of the evolution of single strains dominating infection in CF patients. However, given that seven patients harboured a single-strain infection throughout the study period, longitudinal studies are required to determine if one shared strain displaces other(s) in mixed-strain infections as suggested by our observations and reported occasionally by others [15, 28, 29].

Patients with single-strain infections had a higher start-of-treatment CRP with improved lung function observed during treatment of the exacerbation. In contrast, although numbers were limited, those with mixed-strain infection had less evidence of an acute inflammatory response at the start-of-treatment and no significant change in lung function was observed. This further highlights the complexity of defining CF exacerbations and determining the timing and duration of antibiotic therapy, particularly in those with advanced lung disease, who now live longer than previously [30].

Although not statistically significant, we observed a transient decrease in the total P. aeruginosa load during the first 6-9 days of intravenous antibiotic treatment, which was reversed by the end of treatment. Similar transient effects have been previously described [31]. However, selection and clonal expansion of a P. aeruginosa population with a specific genotype or particular phenotypic traits may underpin pulmonary exacerbations [11, 15, 32, 33]. Whilst we cannot assign causality of exacerbation here, our in-depth sampling approach enabled the first characterisation of the relative abundance and dynamic nature of mixed strains, which cannot be determined if only a few isolates per sample are genotyped [13‐15]. We found that the relative abundance of AUST-02 declined during anti-pseudomonal treatment [34]. At the same time, the relative abundance of other shared strains increased, suggesting they were under positive selection pressure. This was true especially for AUST-06, which was identified almost exclusively in Queensland previously and was the second most common shared strain identified here [13]. Whilst previous studies have shown within-host microevolution leads to co-existing sublineages of single P. aeruginosa strains emerging over months-to-years [26, 35‐37], this study demonstrates rapid multi-strain turnover in mixed-strain infections within-host during antibiotic treatment of pulmonary exacerbations.

It is difficult to explain the observed temporal dynamics of mixed-strain infection, particularly when considering the antibiotic susceptibility profiles of strains from a single patient treated with meropenem, aztreonam and tobramycin. Even though the AUST-02 relative abundance decreased between start-of-treatment and end-of-treatment, this strain exhibited greater or equal in-vitro resistance to these antibiotics than the AUST-06 and AUST-07 strains that emerged in sputum during treatment. These unexpected results emphasise the poor correlation between in-vitro susceptibility testing and in-vivo response [38‐40] and suggest some strains harbour alternative mechanisms, such as alginate overproduction and adaptive resistance that enable persistence despite aggressive antibiotic treatment of exacerbations [11, 41‐45]. In addition, other virulence determinants may impact the host inflammatory response and co-infection with other CF pathogens [46‐48]. Such factors could potentially favor the selection of AUST-06 during intravenous antibiotic treatment. Furthermore, phenotypic traits could generally be assigned to a particular strain type, but more isolates are required to confirm this observation and relate findings to the clinical course [11, 49].

While more than 2000 P. aeruginosa isolates were genotyped, the results must be interpreted with caution given the small number of patients attending a single centre involved. A further limitation is that sputum was not collected during clinical stability, before the onset of the exacerbation. Therefore, the proportions of strains at clinical stability before and after an exacerbation could not be compared. Sputum may also not be ideal for inferring the overall airway P. aeruginosa load or population composition as it may not represent all lung compartments, and whilst bronchoscopic sampling enables collection of regional P. aeruginosa populations, this method is too invasive for routine use [50]. Fluctuations in the population composition might represent regional changes in P. aeruginosa airway density and local micro-environmental conditions, altered mucus volume or variations in sputum sampling within the lung [37]. The patients had advanced lung disease and were treated with different antibiotic regimens, which may also affect the dynamics and evolution of mixed-strain infections. Furthermore, despite confirming AUST-02 by culture sweep at study entry, using a random culture approach, in two patients who had previously had chronic AUST-02 infection, AUST-02 was not subsequently identified in the genotyping of 48 randomly selected colonies, and only AUST-06 and AUST-01 were detected. It is possible AUST-02 might have constituted a minority of the P. aeruginosa population in these cases at the time of study recruitment, and therefore, were not selected because of limitations in sampling.

Further work will now be conducted to assess the relative abundance of shared strains directly in sputum using high-resolution molecular approaches and with larger study populations across a range of disease severities and other CF centres to validate the findings. This will extend to investigating intra-strain diversity and temporal dynamics, including an AUST-02 strain sub-type (M3 L7) that we described recently [51].

This exploratory study provides novel data characterising the temporal dynamics of a P. aeruginosa mixed-strain population during and after intravenous antibiotic treatment of exacerbations. Various commonly shared strains from throughout Australia, alone or in combination, were identified in individual patients. Together, the data show the rapidly changing strain heterogeneity of pulmonary exacerbations, raising further questions over whether acquiring shared P. aeruginosa strains is a marker or cause of more advanced CF lung disease. Ultimately, the much-needed answers to these questions will assist with refining treatments and existing infection control policies within CF centres.

Within CF airways, multiple co-existing sublineages evolve gradually from a single dominant P. aeruginosa strain overtime. Combining in-depth sputum culture and genotyping techniques, we now also show that patients harbouring the AUST-02 strain may have other unrelated, but commonly shared, P. aeruginosa strains whose relative abundance with one another may change rapidly in a sustained and unpredictable manner during antibiotic treatment.