Abstract
Cystic fibrosis (CF) patients receive chronic treatment with macrolides for their antivirulence and anti-inflammatory properties. We, however, previously showed that Pseudomonas aeruginosa, considered as naturally resistant to macrolides, becomes susceptible when tested in a eukaryotic medium rather than a conventional broth.
We therefore looked for specific macrolide resistance determinants in 333 CF isolates from four European CF centres in comparison with 48 isolates from patients suffering from hospital-acquired pneumonia (HAP).
Minimum inhibitory concentrations (MICs) of macrolides and ketolides measured in eukaryotic medium (RPMI-1640) were higher towards CF than HAP isolates. Gene sequencing revealed mutations at three positions (2045, 2046 and 2598) in domain V of 23S rRNA of 43% of sequenced CF isolates, but none in HAP isolates. Enzymes degrading extracellular polymeric substances also reduced MICs, highlighting a role of the mucoid, biofilm-forming phenotype in resistance. An association between high MICs and chronic azithromycin administration was evidenced, which was statistically significant for patients infected by the Liverpool Epidemic Strain.
Thus, ribosomal mutations are highly prevalent in CF isolates and may spread in epidemic clones, arguing for prudent use of oral macrolides in these patients. Measuring MICs in RPMI-1640 could be easily implemented in microbiology laboratories to phenotypically detect resistance.
Abstract
Ribosomal mutations are frequent in P. aeruginosa from cystic fibrosis patients, owing to chronic oral macrolide use http://ow.ly/B5vC308qfjl
Introduction
The leading cause of morbidity and mortality in cystic fibrosis (CF) patients is progressive respiratory dysfunction, partially due to chronic bacterial infections. Pseudomonas aeruginosa is among the most frequent pathogens in the respiratory tract of CF adults and causes irreversible lung damage [1]. These patients therefore receive repetitive and prolonged treatments with antipseudomonal antibiotics. Many CF patients also receive long-term treatment with macrolides, based on their known antivirulence [2] and anti-inflammatory [3] properties, which improve their respiratory function [4] particularly when chronically infected with P. aeruginosa. Macrolides are considered intrinsically inactive against P. aeruginosa, with high (≥256 mg·L−1) minimal inhibitory concentrations (MICs) measured in the recommended conventional broth [5]. This intrinsic resistance is due to the efflux of macrolides out of the bacteria via multidrug efflux systems [6], among which MexAB–OprM and MexXY–OprM are constitutively expressed in wild-type isolates [7]. Yet, macrolides show low MICs against P. aeruginosa when tested in the presence of serum, bronchoalveolar lavage fluid or culture media used for eukaryotic cell cultures. Enhanced macrolide activity in these media results from an increased permeability of the bacterial outer membrane [8]. This favours the diffusion of macrolides inside the bacteria, where they repress the expression of OprM, the outer membrane porin coupled to MexAB and MexXY efflux transporters (responsible for macrolide efflux), further increasing the macrolide intrabacterial concentration to a level where clinically meaningful activity can be observed [8].
Whether macrolides also show activity against P. aeruginosa in vivo and thereby contribute to the improvement of lung function in infected CF patients remains to be established. Nevertheless, the extensive use of macrolides in this population clearly raises the question of the potential risk of selecting macrolide-resistant P. aeruginosa, thereby decreasing their potential usefulness in CF patients and increasing the global resistance burden.
While macrolide resistance is well described in Gram-positive bacteria (due to target modification (methylation or mutation in domain V of 23S rRNA) and/or active efflux [9, 10]), a single publication reports resistance associated with mutations in domain V of 23S rRNA of six P. aeruginosa isolates from CF patients treated by macrolides [11]. The extent of this problem is thus largely unknown.
In this study, we determined macrolide activity in a collection of 333 P. aeruginosa isolates from CF patients, using eukaryotic cell culture medium (RPMI-1640) in order to detect increases in MICs that could be ascribed to the acquisition of resistance mechanisms (which would have escaped detection if using conventional broth). We included in our analysis 48 isolates collected from intensive care patients suffering from hospital-acquired pneumonia (HAP) [12], in whom chronic exposure to macrolides could be excluded. We compared conventional macrolides (azithromycin and clarithromycin) to ketolides, the next-generation macrolides that are less affected by ribosomal mutations at domain V of 23S rRNA, as they also bind to ribosome domain II [13, 14]. Among ketolides, we used telithromycin (registered in 2001 but restricted use due to safety issues) and solithromycin (a new fluoroketolide [15] that successfully completed phase III clinical trials for community-acquired pneumonia) [16, 17].
In brief, we found that CF isolates were significantly less susceptible to macrolides than HAP isolates, with about half of them harbouring mutations in domain V of their 23S rRNA, while no mutations were observed in HAP isolates. MICs of ketolides (especially solithromycin) were less affected by these mutations. Thus, the data document a so-far unreported development of resistance to macrolides in P. aeruginosa collected from CF patients and argue for prudent use of macrolides in this population.
Materials and methods
P. aeruginosa
The common laboratory strain P. aeruginosa PAO1 was used as a reference and strain ATCC 27853 as quality control for susceptibility tests. PAO1-pMES-23S(A2045G) is a PAO1 derivative containing a plasmid pME6031 encoding the whole rRNA operon with an A2045G mutation in domain V of 23S rRNA [11].
Clinical isolates (n=333) were randomly collected from 155 patients (aged 1–59 years) in four European CF centres (Hôpital des enfants malades Reine Fabiola/Hôpital Erasme, Brussels, Belgium (n=88); Hôpital Jean Minjoz, Besançon, France (n=80); Universitätsklinikum Münster, Münster, Germany (n=66); Queen's University of Belfast, Belfast, UK (n=99)) during routine visits. 48 isolates collected from intensive care patients suffering from HAP at the Hôpital Erasme [12] were used for comparison. For UK isolates, those belonging to the Liverpool Epidemic Strain (LES) [18] were previously identified by multilocus sequence typing [19]. Data on file for UK and German isolates also stipulated whether patients were receiving azithromycin regularly (500 mg three times per week continuously in the UK or for intermittent 6-month periods in Germany).
Antibiotics
Antibiotics were obtained as microbiological standards: clarithromycin and azithromycin (Teva, Petach Tikva, Israel), telithromycin (Sanofi-Aventis, Romainville, France), and solithromycin (Cempra Pharmaceuticals, Chapel Hill, NC, USA).
Susceptibility testing
MICs were determined by broth microdilution in cation-adjusted Mueller–Hinton broth (CA-MHB; BD Bioscience, Franklin Lakes, NJ, USA) [5] and RPMI-1640 (Invitrogen, Paisley, UK) adjusted to pH 7.4 and complemented with 10% fetal calf serum [8]. MICs were also measured in the presence of 1) 20 mg·L−1 Phe-Arg-β-naphthylamide (PAβN; a broad-spectrum efflux pump inhibitor) and 1 mM MgSO4 (to strengthen the outer membrane and thereby limit PAβN toxicity [20]), or 2) 20 U·mL−1 alginate lyase, 0.02% DNase and 20 mM MgCl2 (all from Sigma-Aldrich, St Louis, MO, USA), or 100 mg·L−1 Proteinase K (Thermo Fisher, Waltham, MA, USA).
Assessment of outer membrane permeability
Bacteria were incubated in CA-MHB or RPMI-1640 for 4 h, after which 25 µM 1-N-phenylnaphthylamine was added [8]. Fluorescence was read using a Spectramax (Molecular Devices, Sunnyvale, CA, USA) microplate reader (λexcitation/λemission 355/405 nm). Full permeabilisation was achieved by adding 3% Triton X-100 [21] 45 min before reading.
Gene sequencing of rRNA and proteins
The genes encoding domains II and V of the 23S ribosomal subunit or the entire ribosomal proteins L4 and L22 were amplified by PCR (see supplementary table S1 for primers; Eurogentec, Seraing, Belgium) and sequenced (Beckman-Coulter Genomics, Takeley, UK). For domain V, we focused on the 625-bp region where mutations have been previously described (A2058G, A2059G and C2611T [22] in the 23S subunit of Escherichia coli [23], corresponding to positions 2045, 2046 and 2598 in Pseudomonas).
Plasmid construction and transformation in PAO1
Plasmid PAO1-pMES-23S(A2045G) was extracted from its host strain using a GeneJET Plasmid Miniprep Kit (Thermo Fisher) and the gene encoding 23S rRNA was reverted back to its wild-type sequence using a Q5 Site-Directed Mutagenesis Kit (NEB, Ipswich, MA, USA), resulting in plasmid pMES-23S(0). This plasmid was used as a template to introduce each of the point mutations observed in the 23S RNA gene (table 1) and the resulting vectors were transformed into PAO1.
MICs of macrolides and ketolides in cation-adjusted Mueller–Hinton broth (CA-MHB) and RPMI-1640 for Pseudomonas aeruginosa PAO1, transformed PAO1 and cystic fibrosis isolates with ribosomal mutations, classified according to the type of mutation detected
Statistical analyses
Statistical analyses were performed using Prism version 7.01 (GraphPad, San Diego, CA, USA).
Ethics
According to a decision of the Ethical Committee of the Université catholique de Louvain, the investigations did not fall under the scope of the law on human experimentation as all isolates had been collected during routine sampling and patients' data had been anonymised before being transmitted to the investigators.
Results
Susceptibility of CF and HAP isolates to macrolides and ketolides
Figure 1 shows the MIC distributions of clarithromycin (figure 1a and d) and azithromycin (figure 1b and e) as measured in CA-MHB and RPMI-1640 for CF versus HAP isolates. As previously described [8], MICs were one to three dilutions lower in RPMI-1640 than in CA-MHB. RPMI-1640 is therefore useful to observe differences in activity among antibiotics and/or isolates. Azithromycin was slightly more active than clarithromycin. HAP isolates were more susceptible than CF isolates, the difference being most apparent for azithromycin in RPMI-1640. All further experiments were performed on a subset of the CF collection selected to cover the whole range of azithromycin MICs. The two conventional macrolides were first compared with two ketolides. Figure 1c and f show the MIC distributions of solithromycin (compared with azithromycin for the same selected CF isolates in figure 1c), with the MIC50/MIC90 of clarithromycin, azithromycin, telithromycin and solithromycin shown in the accompanying table. Ketolides were more potent than macrolides in both media and for both CF and HAP isolates, with solithromycin MIC50/MIC90 being 1–2 log2 dilutions lower than those of telithromycin against CF isolates (in both media) and HAP isolates (in CA-MHB). Yet, CF isolates remained less susceptible to ketolides than HAP isolates.
Minimum inhibitory concentration (MIC) cumulative distributions of macrolides and ketolides in RPMI-1640 and cation-adjusted Mueller–Hinton broth (CA-MHB) for a–c) cystic fibrosis (CF) and d–f) hospital-acquired pneumonia (HAP) isolates. CLR: clarithromycin; AZM: azithromycin; SOL: solithromycin. a) Clarithromycin and b) azithromycin were studied in the whole CF collection. A subset of the collection was then selected to cover the whole range of azithromycin MICs as measured in RPMI-1640 and used to evaluate ketolide MICs. c) The MIC distribution for this subset is compared for azithromycin and solithromycin. The number of isolates included in the study is indicated in each panel. The table below the graphs shows MIC50 and MIC90 values in both media for the subset of 76 CF isolates and the 48 HAP isolates (note that these values are 1 log2 dilution lower for clarithromycin in RPMI-1640 when considering the whole CF collection, but not different for azithromycin).
Outer membrane permeability
We showed that the increased susceptibility of P. aeruginosa to macrolides in RPMI-1640 is due to an increased permeability of the outer membrane in this medium [8]. We therefore investigated whether the lower susceptibility of CF versus HAP isolates could result from a reduced outer membrane permeability (figure 2). As anticipated, outer membrane permeability was higher in RPMI-1640 than in CA-MHB, but independently of the origin of the samples.
Outer membrane permeability of cystic fibrosis (CF) and hospital-acquired pneumonia (HAP) isolates in cation-adjusted Mueller–Hinton broth (CA-MHB) and RPMI-1640. Bacteria were incubated for 4 h in CA-MHB or RPMI-1640 medium and then with 25 µM 1-N-phenylnaphthylamine. Data are expressed as a percentage of the maximum value recorded in the presence of 3% Triton X-100 and are the mean±sd of the values recorded for the different isolates studied. Statistical analysis: two-way ANOVA with Tukey's multiple comparison test (data with different letters are significantly different from one another, p<0.001). The number of isolates included in the study is indicated on the x-axis.
Efflux pump inhibition
We previously described that intrinsic resistance of P. aeruginosa to macrolides was in part attributable to constitutive expression of efflux pumps [8]. We therefore examined the influence of the efflux pump inhibitor PAβN on MICs. PAβN had limited effect on azithromycin activity against CF isolates (1 log2 dilution decrease in MICs for isolates with baseline values of 64 or 128 mg·L−1; no effect on other isolates; figure 3a and b), indicating that increased efflux does not account for higher resistance in these isolates. Conversely, PAβN decreased MICs of 2–4 log2 dilutions for HAP isolates, independently of the baseline value (figure 3c and d), designating efflux as the main resistance mechanism in these isolates.
Effect of Phe-Arg-β-naphthylamide (PAβN) on azithromycin activity against a, b) cystic fibrosis (CF) and c, d) hospital-acquired pneumonia (HAP) isolates. Isolates were selected to cover the whole range of minimum inhibitory concentrations (MICs) in cation-adjusted Mueller–Hinton broth (CA-MHB). MICs were measured in CA-MHB in the presence (+) or absence (–) of 20 mg·L−1 PAβN and 1 mM MgSO4. a, c) MIC cumulative distributions (the dotted lines indicate 50% and 90% values). b, d) Change in MIC (number of log2 dilutions reduction) observed in the presence of PAβN for individual isolates classified according to their MIC in control conditions. Data are shown as individual values, with medians and interquartile range; the line joins median values. The table below the graphs shows MIC50 and MIC90 values in both conditions.
Sequencing of domain V of 23S rRNA
In Gram-positive bacteria, high-level resistance to macrolides is mediated by mutations in their ribosomal target. We therefore sequenced domain V of 23S rRNA for all isolates presented in figure 1c and e. Mutations were detected in 33 out of the 76 sequenced CF isolates (43%), while no mutations were detected in the 48 HAP isolates. In CF isolates, six different mutations were observed (A2045G, A2045T, A2046G, A2046T, C2598T and C2598G) at three specific positions and in one to four alleles (table 1).
Effect of mutations on macrolide and ketolide activity
The MIC distributions of azithromycin and solithromycin in isolates showing mutations in domain V of 23S rRNA were compared with those of nonmutated isolates (figure 4). As expected, mutated isolates (figure 4a and b) were less susceptible than nonmutated isolates (figure 4c and d). Interestingly, the MIC distribution of nonmutated isolates from CF patients was superimposable to that of HAP isolates for both drugs and in both media (figure 4c and d), suggesting that mutations in part of the CF collection were responsible for the lower susceptibility observed in figure 1.
a–d) Minimum inhibitory concentration (MIC) cumulative distributions of a, c) azithromycin and b, d) solithromycin in cation-adjusted Mueller–Hinton broth (CA-MHB) and RPMI-1640 for cystic fibrosis (CF) and hospital-acquired pneumonia (HAP) isolates presenting a, b) mutations or c, d) no mutations in domain V of 23 S rRNA. e, f) relationship between the number of mutated alleles and the MIC: e) azithromycin; f) solithromycin. ***: p<0.001. The table shows MIC50 and MIC90 values in both media for mutated (“Yes”) or nonmutated (“No”) isolates. The number of isolates included in the study (same as those used in figure 1c and e) is indicated in each panel.
MICs for individual mutated isolates are shown in table 1 and in figure 4e and f, ordered according to the number of mutated alleles. MICs were in general more elevated in both media against isolates with mutations at position 2045 or 2046 than at position 2598, as well as in isolates with three or four mutated alleles. Solithromycin MICs remained low mainly in isolates mutated at position 2598, independently of the number of mutated alleles. One highly resistant isolate harboured two mutations in the four alleles.
To confirm the role of these mutations in resistance, PAO1 was transformed by plasmids containing each of the ribosomal mutations observed in clinical isolates. In these transformants, azithromycin MICs were higher than in PAO1, reaching 1024 mg·L−1 for those mutated in positions 2045 and 2046, and 256 mg·L−1 for those mutated in position 2598 (values in RPMI-1640). In contrast, ketolide MICs remained low regardless of media.
Additional studies in ketolide-resistant isolates
Some CF isolates showed reduced susceptibility to ketolides in RPMI-1640 (table 1), suggesting the presence of other resistance mechanisms. We therefore sequenced the genes encoding domain II of 23S rRNA and the ribosomal proteins L4 and L22 [24] in the 19 isolates with solithromycin MICs ≥64 mg·L−1, but no mutations were found. A cooperation with efflux could also be ruled out as MICs were not reduced by PAβN for most isolates (table 2). CF isolates are often mucoid and easily form biofilms with a matrix rich in alginate, DNA or proteins that prevents antibiotic access [25]. Solithromycin MICs were therefore also measured in the presence of enzymes degrading these polymeric substances (table 2). Alginate lyase, DNase or Proteinase K brought MICs back to values close to that measured for PAO1 in 14 isolates, among which those mutated in position 2598 were all reverted by alginate lyase.
Phenotype and solithromycin minimum inhibitory concentration (MICs) in RPMI-1640 for Pseudomonas aeruginosa PAO1 and mutated cystic fibrosis isolates with solithromycin MIC ≥64 mg·L−1, as measured in control conditions or in the presence of the efflux pump inhibitor Phe-Arg-β-naphthylamide (PAβN) or of enzymes degrading extracellular polymeric substances (alginate lyase, DNase or Proteinase K)
Impact of azithromycin treatment on susceptibility to macrolides in CF isolates
In contrast to HAP patients, many CF patients are chronically exposed to macrolides, which might have contributed to resistance development. In the UK and German collections for which data on macrolide treatments were available, we compared azithromycin MICs in isolates from patients treated regularly by azithromycin versus those with no record of chronic azithromycin use (figure 5). In dot-plot analysis (figure 5b), no significant difference but a trend to lower values in the median or geometric mean MIC was noticed for isolates collected from azithromycin-treated versus untreated patients. However, isolates with low MICs (≤32 mg·L−1) were more frequent in untreated patients (36%) than in azithromycin-treated patients (19%). Conversely, isolates with high MICs (>1024 mg·L−1) were predominant in azithromycin-treated patients (38%) versus untreated patients (27%) (figure 5a). In LES clonal isolates (highly prevalent among CF patients in the UK [26]), MICs were significantly higher for those collected from azithromycin-treated patients (figure 5c). Among these clonal isolates, 39% (five out of 13) of those collected from treated patients showed ribosomal mutations versus 12% (two out of 17) only of those originating from patients who were not reported as chronically treated by macrolides (p=0.04, Chi-squared test).
a) Minimum inhibitory concentration (MIC) distribution and b, c) scatter plots of MICs of azithromycin determined in RPMI-1640 for cystic fibrosis isolates from the UK and German collections stratified according to azithromycin treatment: a, b) whole UK and German collections; c) Liverpool Epidemic Strain clone only. “Yes”: patients chronically treated by azithromycin; “No”: patients not chronically treated by azithromycin. b, c) Data are shown as individual values with medians and interquartile ranges. The number of isolates included in the study is indicated on the x-axis; additional analyses are shown in the tables below the graphs (with 95% confidence intervals in brackets). Statistical analysis: a) p-values for Chi-squared (χ2) analysis of specified MIC ranges given in the table below the graph; b, c) Mann–Whitney test.
Discussion
This study is, to the best of our knowledge, the first to describe the presence of macrolide resistance mechanisms in a large collection of P. aeruginosa from CF patients. It was facilitated by the use of RPMI-1640, in which macrolide MICs are considerably lower than in conventional broth [8]. Approximately half of the sequenced CF isolates did harbour high MICs and ribosomal mutations that were not observed in HAP isolates.
The low MICs observed for HAP isolates in RPMI-1640 are clearly due to the reduced expression of constitutive efflux transporters (as originally described for laboratory strains and a few clinical isolates [8]). Indeed, MIC distributions for these isolates are similar in RPMI-1640 versus CA-MHB supplemented by PAβN. However, this is not the case for CF isolates, for which PAβN only reduces the MIC of the most susceptible isolates, suggesting the selective expression of other specific resistance mechanisms.
As CF isolates with nonmutated rRNA have an MIC distribution similar to that of HAP isolates, we concluded that macrolide resistance in CF isolates is due to mutations in domain V of 23S rRNA. The detected mutations are located in three previously described positions [11], but we also found three additional nucleotide changes. Mutations in these positions confer macrolide resistance in other bacterial species [22] and we confirm using specific mutants that they also increase macrolide MICs in P. aeruginosa. Notably, mutations in positions 2045 and 2046 confer higher resistance than those in position 2598, probably because the first two positions are part of the macrolide-binding site, while the third position rather alters the conformation of the binding site [14]. We also show that resistance levels are globally higher in isolates harbouring mutations in several alleles of the corresponding gene.
Taken together, these findings led us to evaluate the effects of ketolides against CF isolates, as these antibiotics also bind to domain II of 23S rRNA [14]. As expected, CF isolates are more susceptible to ketolides than to macrolides, with solithromycin being more potent than telithromycin, as in other bacterial species [16]. While most of the isolates showing the A2046T or C2598T mutations had low MICs to solithromycin, those harbouring A2045G, A2045T or A2046G mutations generally had higher MICs, in the absence of mutations in domain II of 23S rRNA or ribosomal protein L4 or L22. This resistance could be related to the production of extracellular polymeric substances such as alginate, DNA or proteins, presumably preventing antibiotic penetration.
Importantly, we observed a trend to higher MICs in isolates from azithromycin-treated patients. Specifically in patients infected by the LES clone, mutated isolates with high MICs were more frequently identified from azithromycin-treated patients. This is of concern, as this clone is recognised as multiresistant [27]. Thus, macrolide resistance may, possibly, be more easily selected in the LES background, even though it is acquired by chromosomal mutation and not by gene transfer. Hypermutator variants have been detected, but infrequently in this clone (5–15%) [28, 29]. Cross-infection between patients treated in the same centre is described [18, 30, 31], especially for epidemic clones [32–34], but it is unlikely to markedly contribute to spread resistance here, as mutations are less frequent in LES isolates from nontreated patients.
Macrolides are not given to CF patients for their antibacterial properties against P. aeruginosa. Prescribers may, therefore, consider their use as innocuous with respect to resistance in this organism. Yet, we argue that the highlighted mutations are of concern. First, they may compromise some of the macrolide antivirulence effects that are directly or indirectly dependent on their binding to ribosomes and subsequent protein synthesis inhibition [2, 35]. Second, they may create a reservoir of resistance, possibly compromising the activity of future drugs acting on the same target. In a broader context, the low but persisting serum concentrations of macrolides (especially azithromycin) may aggravate the risk of resistance development, as suggested for Gram-positive pathogens [36]. Providing high, local concentrations in the lung could help mitigate this risk. In this context, liposomes [37] or formulations for inhalation [38] are currently in development for macrolides. Further studies are also needed to define the potential benefit of fluoroketolides such as solithromycin that accumulate in the epithelial lining fluid and show potent anti-inflammatory properties [39, 40].
In the current clinical context, our findings may also be of immediate interest for clinicians in their dialogue with microbiologists. For clinicians, we raise the question of the long-term risk/benefit ratio associated with the widespread strategy of chronic use of oral macrolides in CF patients. For microbiologists, we provide an easy phenotypic method (i.e. testing activity of macrolides in RPMI-1640), applicable on a wide scale, to detect acquired resistance in CF samples. Prospective, comparative clinical trials to further document the impact of macrolide and ketolide exposure of CF patients for selection and spreading of resistance are therefore warranted.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-01847-2016_Supplement
Disclosures
Supplementary Material
H. Traore ERJ-01847-2016_Traore
P.M. Tulkens ERJ-01847-2016_Tulkens
F. Van Bambeke ERJ-01847-2016_Van_Bambeke
F. Vanderbist ERJ-01847-2016_Vanderbist
Acknowledgements
We thank Cempra Pharmaceuticals (Chapel Hill, NC, USA) and R.L. Marvig (Technical University of Denmark, Lyngby, Denmark) for the kind gift of solithromycin and of the PAO1-pMES-23S(A2045G) strain, respectively. J. Turnidge (The University of Adelaide, Adelaide, Australia) gave useful advice for statistical analyses.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Support statement: This work was supported by the Doctiris programme of Innoviris (Région Bruxelloise, Belgium). M-H. Mustafa is a doctoral fellow from the Doctiris programme and F. Van Bambeke is Maître de Recherches of the Belgian Fonds National de la Recherche Scientifique. Cempra Pharmaceuticals (Chapel Hill, NC, USA) provided solithromycin. Funding information for this article has been deposited with the Open Funder Registry.
Conflict of interest: Disclosures can be found alongside this article at erj.ersjournals.com
- Received September 20, 2016.
- Accepted January 15, 2017.
- Copyright ©ERS 2017
References
