Recurrent Pseudomonas aeruginosa blood stream infections (BSI) have been reported in 6% of hospitalised patients with a primary P. aeruginosa BSI [Reference McCarthy and Paterson1]. A haematological co-morbidity or having received steroids in the preceding 30 days, have been identified as patient risk factors for relapse [Reference McCarthy and Paterson1] and other high risk populations include liver, and haematopoietic stem cell transplantation recipients, and acquired immunodeficiency syndrome (AIDS) patients [Reference Korvick2–Reference Mendelson4]. In these cohorts rates of BSI recurrence of up 33% have been described [Reference Korvick2–Reference Mendelson4] and higher mortality is associated with relapse in comparison with the primary BSI episode [Reference McCarthy and Paterson1].

Much of the literature on relapsing Gram negative BSI has focused on coliform bacteria but the correspondence of strain types of P. aeruginosa isolated from primary and recurrent BSI has not been studied. It is therefore unclear if repeated BSI is due to recurrence of the original infecting strain or to a different emergent strain of P. aeruginosa. This study therefore aimed to further characterise the genotypic relationship between the isolates of P. aeruginosa from primary and relapsing BSI episodes in an individual patient.

P. aeruginosa BSI isolates from patients who had blood cultures collected over a 5-year period (January 2008–January 2013) were retrospectively identified from five public or private diagnostic pathology laboratories that service eight tertiary care hospitals and one secondary care hospital in metropolitan Brisbane, Australia. These hospitals provide specialised and broad medical, surgical and intensive care for an urban population of 2·24 million.

Time periods of storage of BSI isolates varied between the laboratories and all stored isolates were included in the initial analyses, with BSI episodes among individual patients defined as the 14 day period after the collection of the sentinel-positive blood culture [Reference McCarthy, Kidd and Paterson5]. A relapsing BSI episode was defined as a subsequent P. aeruginosa-positive blood culture occurring within the study period and outside of the preceding BSI episode. Patients with a relapsing BSI episode or episodes for which both of the BSI isolates were stored were included in the study. Patient data, including name, age, gender, date of birth and hospital of admission, was collected and correlated across the various pathology providers to ensure that cases of recurrent infection were not overlooked. The source of the initial BSI was determined by the treating clinician at the time of notification of the positive blood culture. Permission to undertake the study was obtained from all laboratories involved.

Blood cultures, isolate identification and susceptibility testing were undertaken as previously described [Reference McCarthy, Kidd and Paterson5]. Briefly, each blood culture was incubated using the BD BACTEC™ blood culture system (BD, North Ryde, Australia) with an incubation period of up to 5 days. Bacterial identification and susceptibility testing was achieved using the VITEK 2 system (bioMérieux Australia Pty Ltd, Queensland, Australia). Annual Clinical and Laboratory Standards Institute (CLSI) breakpoints were used to define susceptibility or resistance to the antimicrobial agents tested [6]. From January 2012 in one pathology service and July 2012 in another, antibiotic breakpoint references were changed to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines [7]. A standardised definition for multidrug resistant (MDR) bacteria was applied [Reference Magiorakos8]. Confirmed P. aeruginosa isolates were stored at −80 °C using the Protect Microorganism Preservation System (Thermo Fisher Scientific Australia Pty Ltd, Thebarton, Australia) until transportation to a research laboratory. Upon receipt each isolate was resuscitated from −80 °C storage on Mueller-Hinton agar at 37 °C for 24 h.

All isolates were genotyped with a single nucleotide polymorphism (SNP)-based assay as outlined previously [Reference McCarthy, Kidd and Paterson5]. Briefly, heat denatured DNA preparations from each isolate were analysed using an iPLEX20SNP assay based on the iPLEX MassARRAY matrix-assisted laser desorption ionisation with time of flight mass spectrometry (MALDI-TOF MS) platform (Agena Bioscience, Brisbane, Australia). The iPLEX20SNP assay, comprising 20 informative multilocus sequence typing (MLST) housekeeping gene SNPs, is capable of distinguishing most MLST sequence types (STs) either individually or as belonging to closely related single- or double-locus variant groups [Reference Syrmis9]. ST assignment of the 20-SNP profiles generated in the current study was performed by utilisation of the MLST database as described previously [Reference Syrmis9]. Twenty-SNP profiles, which did not match with any of the listed 20-SNP profiles (i.e. an MLST type could not be assigned) were assigned as novel genotypes. If an isolate had a 20-SNP profile with a singular non-typable allele and a historical isolate was available with an otherwise identical 20-SNP profile, the former isolate was assigned to the same sequence type.

The DiversiLab automated rapid strain typing platform (bioMérieux) was used to assess the genotypic relationships between study isolates showing indistinguishable 20-SNP profiles. Genomic DNA was extracted using the MO BIO Ultraclean Microbial DNA isolation kit (MO BIO Laboratories Inc., Carlsbad, USA) as per the manufacturer's instructions and repetitive element palindromic polymerase chain reaction (rep-PCR) was performed with the DiversiLab Pseudomonas kit using the previously outlined primer mix and PCR conditions [Reference McCarthy, Kidd and Paterson5]. Genetic relatedness was assessed using DiversiLab software version 3·6 (bioMérieux) with isolates showing fingerprints related by ⩾95% being allocated to the same rep-PCR type.

If a patient's BSI isolates were assigned to the same 20-SNP profile that corresponded to a single ST, or the same novel 20-SNP genotype the isolates were considered related. For isolates where the 20-SNP profiles corresponded to ⩾2 STs, genetic relatedness was confirmed provided the isolates were also allocated to the same DiversiLab rep-PCR type.

An unpaired Mann–Whitney test was utilised to compare continuous variables and all calculations were performed with GraphPad Prism 6 for Windows.

Over the study period there was a total of 942 BSI isolates of P. aeruginosa identified that had been stored by the laboratories. Twenty-eight patients, from nine different healthcare settings, experienced one or more relapsing BSI episodes; 25 had one relapse and three had two relapses. Apart from one individual, the primary and secondary episodes occurred within the same healthcare setting. Community-acquired acquisition was not seen. Further comments will relate to the isolates recovered from the primary and the first relapse episodes only for each patient (56 isolates).

All isolates were susceptible to piperacillin-tazobactam. Four isolates (7%) from three patients showed evidence of meropenem resistance; in two of these patients the relapse isolate exhibited resistance to meropenem compared with the primary isolate, which was sensitive. Overall, only two isolates from a single patient exhibited a MDR phenotype that did not include resistance to meropenem.

The iPLEX20SNP and DiversiLab genotyping results of the primary and recurrent BSI isolates are presented in Table 1. Fourteen patients (50%) experienced recurrent infection with an isogenic (or related) strain of P. aeruginosa and nine (32%) patients were reinfected with a genotypically unrelated strain. The genetic relationship between the initial and relapse isolates from the remaining five patients could not be determined due to the limitations of the typing systems.

Table 1. Infection source, genotypic relationships and time to recurrence of P. aeruginosa isolates from 28 subjects showing relapsing blood stream infection

20-SNP, Sequenom iPLEX 20-single nucleotide polymorphism assay genotype; ST, multilocus sequence type; rep-PCR, DiversiLab repetitive element palindromic PCR genotype; NT, non-typeable.

^a Line, line-related BSI; Skin/ST, skin or soft tissue focus of BSI; Urinary, BSI arising from a urinary tract infection; Unknown, a source of infection could not be identified; Cardiovas, BSI arising from endocarditis; Abdominal, BSI arising from diverticulitis.

^b R, related genotypes; U, unrelated genotypes; I, indeterminant genotypic relationship.

^cSubject G experienced two relapsing episodes of P. aeruginosa BSI. The second episode involved a line related infection comprising 20-SNP type CGTCGGCTCCTCCCCGGGCG (Predicted ST: 910, 1689; rep-PCR type: 13) at 70 days after the primary relapse BSI.

^dSubject J experienced two relapsing episodes of P. aeruginosa BSI. The second episode involved a line related infection comprising 20-SNP type CGCAAGCTACCTCCTGGGTA (Predicted ST: NT) at 322 days after the primary relapse BSI.

^eSubject T experienced two relapsing episodes of P. aeruginosa BSI. The second episode involved a unknown source of BSI comprising 20-SNP type CGTCGGCTCTCCTTTGGGTA (Predicted ST: 147) at 56 days post the primary relapse BSI.

A suspected source of infection was identified for 46 (82%) of the 56 P. aeruginosa isolates arising from the primary and secondary cases of BSI (Table 1). Indwelling line-related infections were by far the most common source of infection, being implicated in 35 of 46 (76%) positive blood cultures. Other identified sources of isolates included urinary (5), skin/soft tissue (3), abdominal (1) and cardiovascular (2). There were 17 subjects in whom a suspected source of infection was identified for the primary and secondary BSI with both genotyped isolates; 11 of these patients experienced relapsing line-related infections with six showing a related P. aeruginosa strain in both blood cultures. Relapsing BSI arising from the urinary tract and skin/soft tissue sources was also observed in two subjects (C and H, Table 1) but both cases involved unrelated genotypes. Of the four subjects in whom there were apparently differing sources of infection in the first and relapsing BSI episodes, an isogenic strain was identified in three (O, S and AA), while unrelated genotypes were recovered from patient E.

The genotypic relationships of isolates in the context of clinical and laboratory variables for patients are summarised in Table 2. Blood cell count parameters were also available for 50/56 (89%) BSI episodes and showed that 21 (42%) of these originated from subjects who were neutropenic at the time of blood culture collection. The median time to a primary BSI relapse was 49 days (interquartile [IQR] range: 26–101 days). For the 14 patients showing relapsing BSIs with an isogenic P. aeruginosa strain the median time to recurrence was 26 days (IQR: 21–48 days). Conversely, if the primary recurrent BSI involved an unrelated P. aeruginosa strain (nine patients) the median time to relapse of BSI was 91 days (IQR: 70–176 days; P = 0·013). The three subjects (i.e. G, J, T) who experienced a second relapse BSI had P. aeruginosa cultured from their blood at 70, 322, and 56 days after the primary BSI recurrence, respectively.

Table 2. Relationship of clinical and laboratory variables to the genotypes of primary and relapsing P. aeruginosa BSI isolates

The key finding of this study is that recurrent P. aeruginosa BSI can occur from the same or different genotype, with recurrence from a genetically related strain occurring more frequently. Moreover, reinfection from a genetically related strain was associated with a markedly shorter median time to recurrence of 26 days, compared with 91 days for BSI recurrence with an unrelated strain. The most common source for all the BSI episodes studied was line-related infection, and almost half of BSI episodes occurred in neutropenic patients, which suggests that gut translocation may be another potential source of the recurrent BSI.

Although Gram-negative bacteria comprise a substantial component of normal human host microflora, P. aeruginosa is infrequently reported, but may occur at sites other than the bowel, including the perineum, throat, nose and axilla [Reference Deanna and Murray10]. The hospitalised host on the other hand, appears to have increased susceptibility to colonisation by P. aeruginosa often with high rates (>50%) being recorded in intensive care units [Reference Bertrand11]. Colonisation with one or more distinct strain genotypes may occur [Reference Bonten12, Reference Martinez13], and give rise to a secondary site of infection and subsequent bacteraemia [Reference Bonten12, Reference Agodi14]. The finding here that relapsing BSI due to the same strain of P. aeruginosa occurred earlier than episodes due to an unrelated strain may reflect changes in strain populations colonisation of the host due to exposure to the ubiquity and genetic heterogeneity of this species in the natural environment [Reference Kidd15].

In our study, 21 patients were neutropenic at the time of one of the BSI episodes, suggesting that endogenous translocation from the gastrointestinal tract may have been a potential source for some of these infections. Line-related infection causing both primary and relapsing episodes was the most common cause of all BSI cases and the most common source when both episodes were from the same source of infection; the latter were more likely to be due to genetically related isolates. Unfortunately line removal data were not available to determine if line retention contributed to the relapse of BSI. It is noteworthy that intravascular devices were found to be the most common cause of recurrent all-cause Gram-negative bacteraemia by Wendt et al. [Reference Wendt16].

In two relapsing BSI episodes, the isolate had become meropenem resistant in comparison with the previously sensitive isolates in the primary BSI episode. The patient's antibiotic therapy preceding the secondary BSIs was unknown. It would seem likely that antibiotic selection pressure drove the development of antibiotic resistance.

This study has some limitations. First, the number of recurrent P. aeruginosa BSI episodes is relatively small due to a combination of the infrequency of relapse cases and dependency on external laboratories having stored the isolates after processing of blood cultures. Second, sampling bias may have occurred owing to the differing time periods individual laboratories had stored isolates available for study. Finally, only limited clinical data were available on all of the study patients.

In summary, we have shown that relapsing P. aeruginosa BSI due to the same genotype is more frequent, and occurs earlier in time than a subsequent infection from an unrelated genotype. Line related infections and gut related translocations of bacteria are likely important sources of relapsing infection. Development of meropenem resistance in two relapsing BSI episodes was seen.