Initial Pseudomonas aeruginosa infection in patients with cystic fibrosis: characteristics of eradicated and persistent isolates

Authors


Corresponding author: G. A. Tramper-Stranders, Cystic Fibrosis Centre, KH.01.419.0, University Medical Centre Utrecht, PO Box 85090, 3508 AB Utrecht, the Netherlands
E-mail:g.tramper@umcutrecht.nl

Abstract

Clin Microbiol Infect 2012; 18: 567–574

Abstract

Despite intensive eradication therapy, some CF patients with early Pseudomonas aeruginosa infection rapidly develop a chronic infection. To elucidate factors associated with this persistence, bacterial characteristics of early P. aeruginosa isolates were analysed that were either eradicated rapidly or persisted despite multiple antimicrobial treatments. Eighty-six early infection episodes were studied. First P. aeruginosa isolates from patients with eradication (36) or persistent infection (16) were included; isolates from patients with intermittent infection (34) were omitted from the study. Virulence assays, antimicrobial resistance, cytotoxicity and mutation frequencies were analysed in vitro. P. aeruginosa was genotyped by SNP-array. Transcriptomic profiles of two eradicated and two persistent strains were compared. Nineteen per cent of patients developed persistent infection; 42% achieved eradication. Secretion of virulence factors and mutation frequencies were highly variable among both eradicated and persistent isolates and were not different between the groups. Cytotoxicity was present in 57% of eradicated vs. 100% of persistent isolates (p <0.01). None of the isolates were resistant to antibiotics. The isolates were genotypically highly diverse. Multivariate analysis showed that in vitro determined bacterial characteristics could not predict persistence after first P. aeruginosa infection. Preliminary transcriptomic data showed increased expression of some genes related to a metabolic pathway. The early onset of chronic infection was not associated with (in vitro determined) bacterial characteristics only. Although the persistent isolates were more often cytotoxic, for the individual patient it was not possible to predict the risk of persistence based on bacterial characteristics. Unknown factors such as host-pathogen and pathogen-pathogen interactions should be further explored.

Introduction

Patients with cystic fibrosis (CF) are at risk for chronic pulmonary Pseudomonas aeruginosa infection because of abnormal mucociliary clearance, high pulmonary salt concentrations, and altered bacterial epithelial adherence and internalization [1,2]. Consequently, early infection is intensively treated with combined antimicrobial therapy, leading to, albeit temporal, eradication of P. aeruginosa in the majority of patients. However, amongst 10–40% of patients eradication fails, leading to chronic infection shortly after the first positive culture [3,4].

Identification of those patients at risk for progression to chronic P. aeruginosa infection is difficult. Host factors that cause predisposition to chronic P. aeruginosa infection are female gender, number of dF508 alleles, decreased lung function and absence of Staphylococcus aureus in sputum cultures [5].

Generally, initial P. aeruginosa isolates express multifactorial virulence [2]. Synthesis of a number of virulence factors is controlled by the quorum sensing system. This system is regulated by two transcriptional activators, lasR or rhlR, which are activated by its autoinducer molecules N-3-oxododecanoyl homoserine lactone (C12-HSL) and N-butyryl-l-homoserine lactone (C4-HSL) [6]. Progression to chronic infection is characterized by genetic diversification through numerous mutations, leading to a gradual loss of virulence, production of biofilm and change in metabolic profiles, leading to various phenotypes [7,8]. Pseudomonas aeruginosa adaptation is initiated early in the infection course and is promoted by the CF airway environment, being inflammatory with anaerobic sputum in the conductive airways [9].

Uncertainty exists over whether rapid adaptation leading to loss-of-function mutations causes predisposition to persistence of infection. Enhanced virulence may also be associated with persistent infection. Type III secretion enzymes induce direct epithelial and neutrophil cytotoxicity [10,11]. In non-CF patients suffering from P. aeruginosa ventilator-associated pneumonia, high type III secretion determined risk for persistent infection despite antimicrobial therapy [12]. Additionally, biofilm production is a strong predictor of chronic infection. In a recently published follow-up study, eradication of first P. aeruginosa isolates was higher for non-mucoid isolates (77%) than for mucoid isolates (50%) [13]. However, mucoid isolates are not frequently detected as initial isolates and might possibly be considered as already adapted to the CF airways, despite being the first isolate to be discovered.

Cystic fibrosis-specific highly transmissible P. aeruginosa clones are able to chronically infect P. aeruginosa-negative patients as well as replace other P. aeruginosa isolates in the CF lungs. The success of these clones in promoting persistent infection is apparently related to either high virulence or involvement of an optimized metabolic profile in the absence of high virulence [14,15]. Murine P. aeruginosa pulmonary infection models are contradictory with respect to the capacity of non-virulent isolates to persist after initial infection [16–18].

To elucidate bacterial mechanisms that may promote persistence in the very early phase, we explored in vitro characteristics of initial infecting P. aeruginosa isolates. Phenotype, genotype, cytotoxicity and transcriptomic profiles were evaluated in isolates derived from CF patients with evident eradication and from CF patients with evident persistent infection after initial infection.

Methods

Patients, samples and microbiology

The CF centres of the University Medical Centre Utrecht, the Netherlands, and of the Rigshospitalet Copenhagen, Denmark, participated in this study. At each visit patients provided a sputum sample; cough swabs or nasolaryngeal suction specimens were obtained if patients were unable to produce sputum. Within the Dutch CF centre, patients were fully segregated; within the Danish CF centre segregation was based on microbial cultures. Patient data were extracted from CF databases.

Pseudomonas aeruginosa was identified by standard microbiological methods on solid media, by a positive oxidase reaction and by C-390 (9-chloro-9-(4-diethylaminophenyl)-10-phenylacridan) resistance (Rosco, Taastrup, Denmark). Susceptibility testing for aminoglycosides, β-lactams, polymyxins and fluoroquinolones was performed by disk diffusion using Neo-Sensitabs (Rosco). Results were interpreted according to CLSI breakpoints. P. aeruginosa isolates were stored at −80°C till handling. The initial infecting and consecutive isolates were prospectively collected between 2004 and 2008.

Infection status and included isolates

A P. aeruginosa isolate was considered ‘initial’ when it was the first P. aeruginosa ever detected in a CF patient (‘never P. aeruginosa’) or the first isolate detected after at least 1 year of P. aeruginosa-free cultures without (maintenance) anti-Pseudomonas therapy (‘P. aeruginosa free’). The eradication therapy differed between the two centres, being mainly 3 weeks oral ciprofloxacin plus 3 months inhaled colistin for Danish patients or 3 weeks oral ciprofloxacin plus 2 months high-dose inhaled tobramycin for Dutch patients. Patients were cultured at least 3-monthly and immediately after eradication therapy. Infection episodes were categorized as ‘eradicated’ (with no P. aeruginosa returning in cultures within 1 year after one eradication treatment cycle) or as ‘persistent’ (no eradication achieved after the first and consecutive eradication therapies; >50% of cultures P. aeruginosa positive; all the isolates shared an identical genotype). To study isolates in clearly distinct groups, episodes that could not be attributed to one of these categories (intermittent infection) were not included in the study. Phenotypically different isolates from one patient were both included in the analysis.

Characterization of isolates

The initial and consecutive P. aeruginosa isolates were genotyped (SNP-array of core and accessory genome) to establish persistence of infection. All initial isolates were characterized by phenotyping (swimming (flagella) and twitching (pili) motility, protease-, pyocyanin- and C4/C12-HSL concentration), mutation frequency count and IB3-1 CF bronchial cell-line cytotoxicity. Further information can be found in the Supporting Information.

Microarray sample procession and data analysis

Transcriptomic profiles of two eradicated and two persistent isolates were assessed with Affymetrix P. aeruginosa gene chip. These isolates were derived from patients with an identical CFTR genotype (homozygous dF508), antimicrobial eradication therapy (ciprofloxacin and tobramycin) and age (5.67–6.81 years old). Triplicate experiments were performed for each strain, as further explained in the Supporting Information.

Statistical analysis

Phenotype results were categorized according to the 15th, 50th and 85th percentiles. Differences in proportions were estimated by the chi-square test or Fisher’s exact test. Continuous data were analysed with the Student’s t-test or Mann–Whitney U-test, depending on normality. Values are expressed as means ± standard error of the mean (SEM). Univariate- and multivariate analyses were performed to estimate odds ratios (OR) for bacterial- and host-related factors. Factors that showed a p-value <0.15 were included in the multivariate analysis. Measures of association were determined by Pearson or Spearman correlation coefficients. Calculations were performed with SPSS version 15.0 (Chicago, Illinois, USA). To compare the transcriptomic differences between the eradicated and persistent isolates, a loggit-t test comparing the expression data of the two groups was carried out. Further details on microarray analysis can be found in the Supporting Information.

Results

Patients and treatments

Forty-six Dutch and 40 Danish CF patients were included. The mean age of P. aeruginosa acquisition was 7.7 years (range 0.8–24.1) and was not different between the two CF centres. Female patients were over-represented (60% female vs. 40% male), but were on average not younger when acquiring initial P. aeruginosa infection. Sixteen (19%) patients progressed to persistent infection shortly after initial P. aeruginosa isolation. Thirty-six (42%) patients achieved long-term eradication. Fig. 1 shows the participant flow diagram. Two patients with persistent infection with the Dutch highly prevalent clone were excluded from further comparative analysis (see below); thus isolates from 50 patients were analysed.

Figure 1.

 Number of participants and flow diagram. NL, the Netherlands; DK, Denmark. *P. aeruginosa free after first treatment cycle for >1 year. P. aeruginosa intermittently isolated in the year after initial infection; eradication after >1 treatment cycle or eradication after first treatment cycle and re-isolation in the year after initial infection. #No eradication despite several treatment cycles and subsequent genotypes identical.

Patients with persistent infection were older than patients with eradication. Previously treated P. aeruginosa infection episodes (mean 3.4 years before included episode) did not more frequently result in persistent infection. The type of eradication therapy and the CF centre were not determinative for eradication success. Patients with homozygous dF508 mutation did not more often progress to persistent infection and were not younger when progressing to persistent infection. Table 1 shows baseline data of patients with eradication and persistent infection.

Table 1.   Participant data and infection status
 Eradication (n = 36)Persistence (n = 14)OR (95% CI)p-value
  1. Data are indicated as means ± SEM and n (%)

Pseudomonas aeruginosa acquisition age5.7 ± 0.610.6 ± 1.51.3 (1.1–1.5)0.006
Previous single isolation of P. aeruginosa4 (11)4 (29)3.3 (0.7–15.6)0.133
Female gender23 (63.9)6 (40)0.3 (0.1–1.2)0.097
Staphylococcus aureus in previous year24 (71)9 (64)0.8 (0.2–2.8)0.669
Haemophilus influenza in previous year20 (57)7 (50)0.8 (0.2–2.6)0.650
CF Centre Utrecht/CF Centre Copenhagen17 (37)/19 (48)8 (17)/6 (15)0.7 (0.2–2.3)0.530
Treatment
 Ciprofloxacin po/inh. colistin17 (47)8 (57) 0.450
 Ciprofloxacin po/inh. tobramycin9 (25)4 (29) 
 Monotherapy ciprofloxacin po, inh. colistin or tobramycin.5 (14)0 (0) 
 Tobramycin/ceftazidim or piperacillin/tazobactam i.v.1 (3)2 (14) 
 No therapy4 (11)0 (0)  
CFTR mutations
 dF508/dF50821 (58)10 (71)1.9 (0.5–7.2)0.342
 dF508/other12 (33)4 (29)  
 Other/other3 (8)0 (0)  
Sputum producers7 (19)4 (29)1.7(0.4–6.9)0.487

Genotyping

Pseudomonas aeruginosa genotypes were highly diverse in both groups. Seven genotype pairs were observed; the accessory genome of the genotype pairs was distinct except for two genotypes, one type found in both the eradicated and the persistent group and one type found only in the eradicated group. The relatedness of bacteria (based on 16 binary SNP codes from the core genome, flagellin and exoenzyme S/U genes; calculated by eBurst algorithm) is shown in Fig. S1. Two patients with persistent infections carried the Dutch CF highly prevalent genotype A418 [19]. These isolates were omitted from further comparative analysis because of probable distinct transmitting and infecting potential compared with non-clonal, often environmental isolates.

Fifty-two per cent of isolates were exoS positive, and 10%exoU positive. The exoS gene was not more often encountered in persistent isolates (OR, 1.12; 95% CI, 0.33–3.84; p 0.860). There were no differences in distribution of alleles of the core and accessory genome and in distribution of one or more genomic islands between the eradicated and persistent isolates.

Phenotyping

Virulence factors.  Two of the eradicated isolates were mucoid. All bacteria were susceptible to investigated antibiotics. Pili, flagella, protease, pyocyanin, C4-HSL and C12-HSL expression were highly variable in both groups and were not significantly different between the groups. Phenotype figures and distributions are shown in Table 2 and Fig. 2(a). The persistent isolates showed more often motility and protease values ≥p50; 32% (eradicated) vs. 14% (persistent) had one or more virulence factors <p15. Twitching and swimming motility were correlated (Spearman r 0.298; p 0.033); only twitching was correlated with protease (Spearman r 0.403; p 0.003). Pyocyanin secretion was not related to any of the other virulence factors. C4-HSL was correlated with protease (Spearman r 0.309; p 0.026) but not with the other virulence factors and cytotoxicity. C12-HSL was not related to any of the virulence factors.

Table 2.   Phenotype figures of eradicated and persistent isolates
 Eradication (37)Persistence (14)OR (95% CI)p-value
  1. Data are indicated as mean ± SE or n (%), unless otherwise stated.

Swimming motility (range 0–64 mm)37.2 ± 2.443.56 ± 3.01.04 (0.98–1.10)0.159
 N < p15 (29)7 (19)0 (0)0.168
 N ≥ p50 (41)20 (54)8 (62)0.843
 N > p85 (50)3 (8)5 (39)0.021
Twitching motility (range 0–41 mm)25.4 ± 2.329.8 ± 3.41.03 (0.98–1.08)0.308
 N < p15 (2)7 (19)1 (7)0.407
 N ≥ p50 (31)17 (46)10 (71)0.104
 N > p85 (38)5 (14)4 (29)0.236
Protease (range 0–29 mm)23.8 ± 1.026.0 ± 0.81.14 (0.93–1.40)0.024
 N < p15 (21)7 (19)1 (7)0.419
 N ≥ p50 (26)16 (43)12 (79)0.024
 N > p85 (28)10 (27)3(21)0.733
Pyocyanin (range 0.10–24 mg/L)8.5 ± 1.17.5 ± 1.50.97 (0.88–1.08)0.622
 N < p15 (0.42)4 (11)3 (21)0.376
 N ≥ p50 (7.50)19 (51)7(50)0.931
 N > p85 (15)6 (16)1(7)0.657
One or more virulence factors (motility, protease, pyocyanin) ≥p5030 (81)14 (100) 0.169
One or more virulence factors (motility, protease) ≥p5026 (70)14 (100) 0.023
C4-HSL (range 0–15.17 μM)7.28 ± 0.638.04 ± 0.761.06 (0.89–1.27)0.496
C12-HSL (range 0–1156 nM)188.74 ± 23.56301.57 ± 340.621.00 (0.99–1.01)0.128
Streptomycin mutation frequency (median and range)0.29*10−8 (0.02–124.0*10−8)0.15*10−8 (0.05–13.05*10−8)0.96 (0.84–1.01)0.563
Rifampicin mutation frequency (median and range)1.39*10−8 (0.46–790.16*10−8)1.16*10−8 (0.57–2.49*10−8)0.54 (0.22–1.32)0.175
Hypermutation (>20-fold)3 (8.6)0 (0) 0.545
Cytotoxicity (% lysis)17.8 ± 4.428.8 ± 8.31.01 (0.99–1.03)0.226
Figure 2.

 (a) Virulence factor assay results with medians. (b) Cytotoxicity (% lysis with medians) on IB3-1 cell layer (dotted line represents 5% cut off) (• eradicated isolates and bsl00066 persistent isolates).

Mutation frequencies.  Hypermutation (>20-fold mutation frequency) was observed in 9% of the eradicated and 0% of the persistent isolates. A greater than five-fold increase in mutation frequency was observed in 16% of isolates. Isolates with a >20-fold mutation rate showed decreased swimming and twitching capacities (p 0.022 and 0.002) and lower protease secretion (p 0.047); pyocyanin secretion was not significantly decreased. A more than five-fold mutation frequency increase was associated with a decrease in cytotoxicity (22.9 ± 4.79% vs. 8.21 ± 1.92%, p 0.006).

Cytotoxicity IB3-1 CF bronchial cell line.  Most isolates displayed moderate or limited cytotoxicity. Absence of cytotoxicity, defined as <5% lysis, was observed among 31% of isolates (43% eradicated and 0% persistent; p 0.002). Cytotoxicity figures can be found in Table 2 and Fig. 2(b). Cytotoxicity was associated with twitching motility (Spearman r 0.378; p 0.006) but not with the other virulence factors, confirming the requirement of pili-mediated bacterial-epithelial contact.

Enter multivariate modelling of phenotypic figures revealed no significant bacterial characteristics associated with persistence. Odds ratios with confidence intervals and p-values are shown in Table 2.

Transcriptomic profiles

Assessment of other factors important for the early establishment of persistent P. aeruginosa infection was performed by transcriptome analysis.

Seventeen genes (Table 3) were differently expressed between the two eradicated and two persistent isolates. Seven genes showed increased expression in the persistent isolates compared with the eradicated isolates. Four of these genes (dctA, PA5167, PA5168 and PA5169) belong to the C4-dicarboxylate transport system.

Table 3.   Differentially expressed genes comparing persistent isolates and eradicated isolates
 p-valueFold changeProduct nameFunctional classes
Upregulated in persistence group
 dctA0.0033.59C4-dicarboxylate transport proteinTransport of small molecules
 PA17110.0302.39ExsEHypothetical, unclassified, unknown, transcriptional regulators; protein secretion/export apparatus
 PA22520.0462.48Probable AGCS sodium/alanine/glycine symporterTransport of small molecules
 PA51670.0102.74Probable c4-dicarboxylate-binding proteinMembrane proteins; transport of small molecules
 PA51680.0072.29Probable dicarboxylate transporterMembrane proteins; transport of small molecules
 PA51690.0122.12Probable C4-dicarboxylate transporterMembrane proteins; transport of small molecules
 tRNA_Phe0.0302.71  
Downregulated in persistence group
 PA06120.011−4.06Repressor, PtrBTranscriptional regulators
 PA06130.024−3.61Hypothetical proteinHypothetical, unclassified, unknown
 PA06330.018−3.44Hypothetical proteinRelated to phage, transposon or plasmid
 PA06350.035−2.48Hypothetical proteinRelated to phage, transposon or plasmid
 PA06360.030−2.48Hypothetical proteinRelated to phage, transposon or plasmid
 PA06370.023−2.42Conserved hypothetical proteinRelated to phage, transposon or plasmid
 PA06380.027−2.24Probable bacteriophage proteinRelated to phage, transposon or plasmid
 PA06390.017−2.20Conserved hypothetical proteinRelated to phage, transposon or plasmid
 PA10960.042−2.09Hypothetical proteinHypothetical, unclassified, unknown
 tRNA_Arg0.022−2.08  

Discussion

This is the first study that describes characteristics of initial P. aeruginosa isolates that were either eradicated or leading to persistent infection in patients with CF. It is important to gain more understanding of the initial infection process to improve treatment of initial P. aeruginosa acquisition in order to subsequently prevent chronic P. aeruginosa pulmonary infection.

While chronic infection isolates are generally characterized by bacterial adaptation and attenuation of virulence, the results of this study show that the very early onset of persistent infection could not be predicted by virulence as measured in vitro.

Patient data

The only patient factor associated with persistence was age; treatment or previous isolation was not associated. Older patients might have more pulmonary damage from previous infections with other microorganisms, facilitating epithelial attachment of P. aeruginosa. Sinus colonization with P. aeruginosa might lead to subsequent lower airway infection [20]. Additionally, co-colonization with other microbial organisms, which might occur more often in older patients, is significant for up- and/or downregulation of important virulence genes [21]. Co-colonization with S. aureus and H. influenzae did not influence persistence rate in this study; however, many microorganisms that are not easily cultured can play an important role in the infection dynamics. Viral infections and even oral microflora lead to increased P. aeruginosa epithelial adherence and renewal of virulence [22]. It can be debated whether the initial isolates actually resided for only a limited period within the CF airways. Because of the time span, albeit small, between routinely performed respiratory cultures and sampling technique errors, it is impossible to estimate the exact colonization time. Therefore, we categorized infection episodes only as eradicated when there were enough follow-up cultures for a period of at least 1 year. Additionally, the differentiation between P. aeruginosa colonization and infection is gradual and uncertain. The majority of patients do not have respiratory symptoms when P. aeruginosa is detected for the first time while airway inflammation is present [13,23].

Phenotyping, genotyping and transcriptomics

Our results show a large variability in bacterial characteristics between isolates. There is a trend towards a higher virulence and cytotoxicity, and a lower hypermutation frequency in persistent isolates. But, prediction for the individual patient to progress to persistent infection remains difficult. Other non-investigated virulence factors such as phospholipase C and exotoxins might be involved as well. Interestingly, eradicated isolates tended to be more often hypermutators compared with the persistent isolates. This phenomenon could be illustrated with the mouse model study of Montanari et al. In this model, hypermutable isolates were less efficient in establishing lung infection compared with wild-type isolates; their transmissibility seemed to be reduced [24].

The percentage of cytotoxicity (69%) in our study group corresponds to the data of Jain et al. [25], showing around 50% of early CF P. aeruginosa isolates expressing type III secreting proteins, which is related to cytotoxicity. Increasing the co-incubation time and adding of EGTA, a calcium chelator, did not lead to major changes in cytotoxicity. El Solh et al. [12] investigated the cytotoxicity of P. aeruginosa from patients with ventilator-associated pneumonia and observed a higher neutrophil cytotoxicity for persistent isolates. Although the cytotoxicity was higher, there was a large variability among isolates. The variance was not shown and the odds ratio was close to one. In a subset of 18 isolates from our study, human neutrophil cytotoxicity was assessed and this correlated with IB3-1 cytotoxicity (Pearson r 0.700; p 0.002; data not shown). In an injured lung, CF host features are probably more important in the infection process than bacterial characteristics, unlike in non-CF patients.

Our in vitro experiments do not necessarily reflect the in vivo situation in the CF airways but rather the bacterial ability to display a set of chosen factors in in vitro selected artificial conditions. The persistent isolates could show a different virulence pattern in vivo, or differ in other yet unknown factors. Probably the host–pathogen interaction, leading to up- or downregulation of virulence components, is determinative for persistence of infection. This might be illustrated by the fact that a few patients shared identical bacterial SNP genotypes. These genotypes were eradicated in some patients and persistent in the others, indicating that the genomic background of a certain strain is not always decisive regarding whether it will persist in the CF lung. However, there are some other pathways that might be essential in this process. Transcriptomic analysis showed seven genes with increased expression in the persistent isolates compared with the eradicated isolates. Four (dctA, PA5167-69) of them belong to the C4-dicarboxylate transport (dct) system. The metabolism of C4-dicarboxylates requires a bacterial uptake system. In Escherichia coli, DctA as a symporter is the major carrier in aerobic growth [26]. DctA mutants show very poor growth on C4-dicarboxylates. In Pseudomonas chlororaphis O6, DctA is also required for C4-dicarboxylate utulization and effective root colonization [27]. PA5167-69 are homologous to dctPQM, a transporter with high affinity to C4-dicarboxylates [28]. In P. aeruginosa, there are four homologous operons of dctPQM including PA5167-69 and their function has not been studied. In conclusion, it is likely that the dctA and dctPQM genes are responsible for or partly involved in C4-dicarboxylates uptake in certain conditions. Upregulation of those genes in the persistent isolates indicates that the C4-dicarboxylates transport might be important for early persisting infection of P. aeruginosa. It could also give hints on carbon sources that P. aeruginosa depends on in early CF infection. However, expression of dct genes has to be examined in a bigger sample size to prove their importance.

Conclusion

Only age was related to development of chronic infection after initial P. aeruginosa infection. Although persistent isolates were more often cytotoxic, it is difficult to predict the chance of persistence based only on in vitro bacterial characteristics because of a large variety. Unknown factors such as host–pathogen and pathogen–pathogen interactions should be further explored. Intensive antibiotic eradication therapy for all patients with first P. aeruginosa infection remains advised. Investigations should be carried out regarding whether patients will benefit from adjuvant therapy targeted at virulence to prevent persistent P. aeruginosa infection.

Acknowledgements

The authors thank Ulla Johansen, Tina Wasserman, Alie McArthurnoom, Niels Hoiby, Tanja Pressler, Kristian Fog Nielsen and Barry Benaissa for collection and storage of samples and assistance with or teaching some of the typing methods. This study was partly sponsored by a small grant from the European Society of Pediatric Infectious Diseases.

Transparency Declaration

All authors declare that they have no conflicts of interest.

Ancillary