Molecular Epidemiology and Clinical Impact of Acinetobacter calcoaceticus-baumannii Complex in a Belgian Burn Wound Center

Multidrug resistant Acinetobacter baumannii and its closely related species A. pittii and A. nosocomialis, all members of the Acinetobacter calcoaceticus-baumannii (Acb) complex, are a major cause of hospital acquired infection. In the burn wound center of the Queen Astrid military hospital in Brussels, 48 patients were colonized or infected with Acb complex over a 52-month period. We report the molecular epidemiology of these organisms, their clinical impact and infection control measures taken. A representative set of 157 Acb complex isolates was analyzed using repetitive sequence-based PCR (rep-PCR) (DiversiLab) and a multiplex PCR targeting OXA-51-like and OXA-23-like genes. We identified 31 rep-PCR genotypes (strains). Representatives of each rep-type were identified to species by rpoB sequence analysis: 13 types to A. baumannii, 10 to A. pittii, and 3 to A. nosocomialis. It was assumed that isolates that belonged to the same rep-type also belonged to the same species. Thus, 83.4% of all isolates were identified to A. baumannii, 9.6% to A. pittii and 4.5% to A. nosocomialis. We observed 12 extensively drug resistant Acb strains (10 A. baumannii and 2 A. nosocomialis), all carbapenem-non-susceptible/colistin-susceptible and imported into the burn wound center through patients injured in North Africa. The two most prevalent rep-types 12 and 13 harbored an OXA-23-like gene. Multilocus sequence typing allocated them to clonal complex 1 corresponding to EU (international) clone I. Both strains caused consecutive outbreaks, interspersed with periods of apparent eradication. Patients infected with carbapenem resistant A. baumannii were successfully treated with colistin/rifampicin. Extensive infection control measures were required to eradicate the organisms. Acinetobacter infection and colonization was not associated with increased attributable mortality.


Introduction
The genus Acinetobacter currently comprises 49 validly named species (April 2016) including 5 novel proteolytic and haemolytic species (http://www.bacterio.net/acinetobacter.html) [1]. Some species are closely related, including the members of the so-called Acinetobacter calcoaceticus-Acinetobacter baumannii (Acb) complex [2,3] and the 5 recently described species [1]. Apart from the 49 species, several Acinetobacter strains or groups of strains have been described with presumptive names, some of which are likely to represent potential novel species including 'between 1 and 3' [4] and 'NB14' [5]. Bacteria of the genus Acinetobacter are ubiquitous and are found in soil and water, insects, and in specimens from human and animal origin [6]. The closely related species A. baumannii, A. pittii, and A. nosocomialis of the Acb complex are most frequently isolated from human specimens and are important nosocomial pathogens [2,7]. Identification of Acinetobacter species according to the current taxonomy and using commercial identification systems in the clinical microbiology laboratory is difficult. Meanwhile, A. baumannii has emerged as a major multidrug-resistant (MDR) microorganism in hospitals worldwide. It is one of the six ESKAPE organisms, which cause the majority of hospital-acquired infections and are able to escape current antibiotics [8]. A. baumannii is rarely recovered from the natural environment and infection reservoirs outside human and veterinary hospitals have not been identified [9]. Molecular typing of isolates obtained from hospitals and cities in Europe permitted the identification of 3 clonal lineages of A. baumannii that were associated with numerous outbreaks and were named EU clones I, II and III [10,11]. By multilocus sequence typing (MLST), these clones were allocated to the corresponding clonal complexes 1-3 [12]. These clones have been found worldwide and, are also referred to as worldwide (WW) clones 1-3, while other widespread clones and MLST types have also been identified (http://pubmlst.org/abaumannii/) [13][14][15]. A. baumannii is responsible for a variety of infections, including wound infections, ventilator-associated pneumonia and bloodstream infections. Some studies provided evidence that A. baumannii infections are associated with increased mortality [16]. Carbapenems have long been the antimicrobial agents of choice, but today carbapenem resistance occurs frequently in A. baumannii, and even resistance to colistin, the last resort antibiotic, has been reported [17]. Their capacity for long-term survival in the hospital combined with antibiotic resistance and a potential to dynamic genomic reorganization under selective pressure [18], make A. baumannii and, to a lesser extent, the closely related species A. pittii and A. nosocomialis [19] important nosocomial pathogens. Patients vulnerable to MDR A. baumannii infection are critically ill patients as found in intensive care units. Special categories are patients with trauma e.g., due to accidents, natural disasters or military actions, or burn patients. Outbreaks of A. baumannii isolations were observed among service members injured in the Iraq/Kuwait (Operation Iraqi Freedom) and in Afghanistan (Operation Enduring Freedom) [20][21][22][23][24][25][26]. Although modern medical care has significantly reduced the mortality among thermally injured, the burn wound is still a site of increased susceptibility to opportunistic colonization and subsequent infection remains a major issue of concern [27]. MDR A. baumannii is increasingly associated with these infections [28,29].
In November 2006, a severely burned woman was transferred from Algiers (Algeria) to the burn wound center (BWC) of the Queen Astrid military hospital in Brussels (Belgium). Bacteriological analysis showed that the patient's wounds and nasopharynx were colonized with extensively drug-resistant (XDR) Acb complex (resistant to carbapenems), which spread to other patients. Despite drastic infection control measures, this first outbreak was followed by recurrent episodes of epidemic spread. Here we report the in-depth epidemiological investigation of Acb complex colonization in the BWC, prompted by this first carbapenem-resistant A. baumannii (CRAB) outbreak. During a 52-month period, 48 patients were colonized or infected with members of the Acb complex. Isolates were presumptively identified and investigated for antibiotic susceptibility using the VITEK 2 microbial identification (ID) and antibiotic susceptibility testing (AST) systems. Representative isolates were further characterized at strain and species level using a combination of DNA-based methods. In addition, the strains were investigated for the occurrence of OXA-51-like and OXA-23-like oxacillinase and other genes. Results were interpreted in light of the time-space origin of isolates for analysis of the epidemiology of Acb complex strains in the BWC. Furthermore, the clinical impact of the organisms and infection control measures were evaluated. This study adds novel information with regard to the occurrence of both baumannii and non-baumannii Acinetobacter species in a burn wound setting.

Setting
The Brussels BWC is a department of the Queen Astrid military hospital and functions as a referral center for Belgium. It was built in 1980 and consists of an intensive care unit (ICU), a medium care unit (MCU), and consultation and ambulatory services. The ICU harbored one admission room, eight single-bed rooms with an average occupancy rate of 70% and three bathrooms. The medium care unit harbored 12 double-bed rooms and two bathrooms. All bathrooms were equipped with a balneotherapy facility and 2 ICU rooms and 1 bathroom were equipped with a laminar airflow installation. Hand washing with a detergent skin cleansing solution (Hibiscrub, Zeneca Pharma, Cergy, France) was required between patient contacts. Masks, caps, and gloves were required at all times in the patient rooms and bathrooms. Sterile gloves were required for wound care procedures. Microbiological screening samples were taken at admission, every second day (ICU), twice a week (MCU) or whenever deemed necessary for clinical reasons. Clinical samples consisted of swabs of the wound areas, the nasopharynx and the peri-anal region, urine, expectoration, blood and catheter samples, if applicable. Annually, 300-400 patients were admitted to the BWC, while more than 10,000 consultations were performed each year. Available data, derived from existing medical or administrative files, from all patients admitted to the ICU and MCU of the BWC between November 2006 and February 2011 was analyzed in this retrospective epidemiological investigation, which was coordinated by the Hospital Infection Control Team and the Laboratory for Molecular and Cellular Technology, the BWC supporting research unit. Since data collection required no contact with the patients and no clinical samples other than those relevant for the treatment were collected from patients (only from the environment), this non-interventional retrospective data analysis is not subjected to the Belgian law concerning experiments on the human person (May 7, 2004) and ethics committee approval is not required. Health care professionals consulted and processed the data for their own use, respecting the provisions on professional secrecy. Patient records/information was anonymized and de-identified prior to analysis. Under these conditions, no informed consent of the patients is required.

Definitions
For patients from whom Acb complex strains were recovered on bacterial culture, medical records were further screened for laboratory and clinical signs of infection. The French Society for Burn Injuries (SFETB) diagnostic criteria for infection in burn patients (http://www.sfetb.org/ index.php?rub=textes-officiels&art=doc_ref_8) were used for diagnosis of Acb complex pneumonias, bloodstream infections, burn wound infections and urinary tract infections. In the absence of clear laboratory and clinical signs of infection, patients from whom a positive bacterial culture was obtained, were considered colonized. Colonization on admission was defined as the isolation of an organism within 24 h of their hospitalization. Cross-colonization was defined as the isolation of genotypically identical isolates from patients who were in the same ward of the BWC at the same time. Isolates that were non-susceptible to at least one agent in three or more antimicrobial categories were considered MDR, while extensively or extremely drug resistant (XDR) was defined as non-susceptibility to at least one agent in all but two or fewer antimicrobial categories [30].

Infection control and interventions
Several interventions were made in response to the sudden increased incidence of Acinetobacter colonization in the BWC. In the first place, patients were systematically screened for Acinetobacter on admission (wounds, nasopharynx, groin and peri-anal region), colonized patients were isolated in private rooms with dedicated personnel and equipment, a dedicated infection control task force was set up, an epidemiological investigation was initiated and infection control seminars (with an emphasis on hand hygiene, using alcohol-based hand rubs) were organized. Rapidly, the infection control task force started to realize that these initial measures were insufficient to contain Acinetobacter and on January 11, 2007, the BWC was closed (the closure was announced in a press release) for deep cleaning and specialized decontamination using vaporized hydrogen peroxide (STERIS Corporation, Brussels, Belgium). Admissions were stopped and resident patients were transferred to a back up ICU (initially reserved for disaster response). The BWC was reopened on February 15, 2007, after surface and air samples were found to be negative for Acb complex. In December 2009, a new burn unit, designed and constructed to allow for extensive infection control measures, was put into operation. Each of the 8 new ICU rooms was equipped with a laminar airflow, dedicated personnel and material entry and exit airlocks, an integrated balneotherapy facility and dedicated medical equipment.

The first outbreak
On November 21, 2006, a 29-year old woman was transferred from the military hospital in Algiers to the BWC of the Queen Astrid military hospital in Brussels. Eight days earlier, she had been severely burned (33% TBSA burned, most 3 rd degree) after a gas explosion in her kitchen. Routine bacteriological analysis, performed immediately upon arrival at the BWC in Brussels, showed that the patient's burn wounds and nasopharynx were colonized with XDR Acb complex, resistant to carbapenems and susceptible to only colistin. When two additional burn wound patients were found to be colonized with XDR Acinetobacter, a limited epidemiological investigation was initiated. This investigation, which was mainly based on repetitive sequence-based PCR fingerprinting (rep-PCR, DiversiLab) of routine Acb complex isolates, revealed that the 3 patients were colonized with an identical CRAB strain. Previously, Acinetobacter was sporadic in the BWC and no carbapenem resistance had ever been observed. This first CRAB outbreak prompted a clinically indicated and partially retrospective epidemiological investigation of Acb complex colonization and infection in the BWC.

Epidemiological investigation
Medical and microbiological records of patients that were colonized/infected with Acb complex were reviewed. The characteristics of the colonized patients were compared with those of the general BWC population. No clinical samples other than those that were relevant for the treatment were taken from patients. Environmental sampling was performed on January 12 and 22, February 9, April 3, 11 and 20, May 10, 11, 14 and 23 and June 18, 2007. In total, 61 (55%) of the 111 health care workers of the BWC, including 29 nurses, 8 physiotherapists, 7 cleaning staff members and 7 MDs, were screened (nose, throat, forehead and peri-anal swabs) on a voluntary basis for transient carriage of Acinetobacter. Environmental samples consisted of 120 swabs of multiple surfaces within the patient environment (including floors, bed sheets, bed side tables, trolleys, sinks, respirators, air conditioning aspiration grids, hemofiltration machines and computer keyboards), 13 air samples (100L) aspirated using a MAS-100 microbial air monitoring system (MBV AG, Stäfa, Switzerland) and tap water samples from all faucets in the BWC (faucet spouts were disinfected with a flame and water was allowed to run for two to three minutes before sampling in sterile 1L bottles). The water feeding the hydrotherapy facilities was sampled every 2 months (routine).

Routine microbiological analysis
Bacteria were isolated from clinical and environmental samples using standard microbiological procedures as described in the Manual of Clinical Microbiology (editions 8-10, issued by the American Society for Microbiology). In short, blood agar, McConkey and mannitol salt agar plates (all purchased from bioMérieux, Brussels, Belgium) were inoculated and incubated overnight at 37°C. Colonies were identified using the gram-negative (GN) and gram-positive (GP) identification cards for the VITEK 2 microbial ID system (bioMérieux, Brussels, Belgium), according to the manufacturer's instructions.

Antibiotic susceptibility testing
The VITEK 2 AST system (bioMérieux, Brussels, Belgium) was used to determine antibiotic susceptibilities (using AST-N155 or AST-N158 cards) of the Acb complex isolates (as identified by the VITEK 2 microbial ID system) and reference strains, according to the manufacturer's recommendations. The Advanced Expert System (AES) (software versions 02.01 to 05.03) was used to analyze the raw antibiotic susceptibility data. Clinical interpretation of minimum inhibitory concentration (MIC) values of antibiotics was based on the Clinical and Laboratory Standards Institute (CLSI) breakpoints.

Typing and species identification by molecular methods
The organisms were characterized at different levels by three methods, i.e. at strain level by rep-PCR, at MLST level by sequencing of seven house keeping gene fragments, and at species level by rpoB sequencing. Of these methods, rep-PCR has the highest resolution and can, in local situations, be used to distinguish strains. MLST has less discriminatory power, but the sequence types can be compared to Internet based databases to assess the global or regional spread of particular MLST types/clones or clonal complexes. Sequence analysis of zones 1 and 2 of the rpoB gene has been proven to be a reliable method for species identification [2].
MLST analysis was performed as previously described [12]. This approach is based on the sequencing of internal regions of seven housekeeping genes. The Acinetobacter MLST database was at the time of analysis located and maintained at the Institut Pasteur (http://www.pasteur. fr/mlst), but has since been moved to http://bigsdb.readthedocs.org/en/latest/submissions. html.
For rpoB gene analysis, the partial sequence of the rpoB gene, i.e. zone 1+2 (861 bp) [34] was determined as previously described [2]. The obtained gene sequences were analysed using BioNumerics software (version 5.1) (Applied Maths, Sint-Martens-Latem, Belgium) and compared to a library of sequences of the validly described Acinetobacter species obtained from EMBL and from reference strains of the LUMC collection.

16S rRNA gene analysis
The 16S rRNA gene sequence was determined and analyzed as previously described [39]. Briefly, the complete 16S rRNA gene was amplified, followed by sequencing reactions using the Big Dye Terminator Sequencing kit (Applied Biosystems, Foster City, Calif.) and analysis of the obtained fragments on the ABI 310 capillary electrophoresis apparatus (Applied Biosystems). Total gene assembling of the obtained fragments, alignment, and clustering were done with GeneBase (Applied Maths, Sint-Martens-Latem, Belgium). The obtained sequences were compared to all known sequences in the GenBank by Blast (National Center for Biotechnology Information, Bethesda, Md.; http://www.ncbi.nlm.nih.gov/blast/index.html).

Treatment of CRAB infections
CRAB infected patients received systemic colistin treatment, which consisted of 3 to 9 million units of colistimethate sodium per day for 5 days (urinary tract infections), 7 days (wound infections) or 10 days (respiratory tract infections). Additionally, infected or heavily colonized burn wounds were treated on a daily basis with a compounded colistin lotion prepared by the local hospital pharmacy, consisting of 0.5 g colistimethate sodium, 0.1 ml lactic acid, 5.7 g paraffin, 3.8 g Tween 80, 6.2 g cetylalcohol, 3.3 g Labrafil and 80 g of sterile water per 100 g.
Because it was suggested that colistin's activity on A. baumannii is increased in the presence of rifampin [40], a daily dose of 600 mg rifampicin was usually added.

Epidemiological investigation
No Acinetobacter spp. were isolated from healthcare workers, water or air samples, but 12 A. baumanni isolates were cultured from the patients' environment (respirators, bed side tables, trolleys, mattresses, air conditioning aspiration grids and the lever and the keyboard of a haemodynamic monitor). A total of 157 Acb complex isolates, consisting of 149 clinical and 8 environmental isolates, which were kept in the bacterial culture collection of the clinical laboratory, were selected for epidemiological analysis using DNA-based approaches. The isolates included at least one isolate from each colonized site of each colonized patient and the hospital environment. The clinical isolates were from wound swabs (predominant site, n = 95), throat swabs (n = 20), nasal swabs (n = 13), sputum (n = 10), urine (n = 3), catheter swabs (n = 3), vaginal swabs (n = 2), feces (n = 2) and blood (n = 1). Of the 157 Acb complex isolates, 131 (83.4%) were identified to A. baumannii, 15 (9.6%) to A. pittii, 7 (4.5%) to A. nosocomialis, 1 (0.6%) to A. calcoaceticus and 3 (1.9%) to potentially novel species. The 26 non-baumannii Acinetobacter isolates were exclusively isolated from wounds (n = 18) and nose/throat (n = 8). None-baumannii Acinetobacter strains were never isolated from expectoration, urine, blood or catheters. Rep-PCR analysis of the 157 selected Acinetobacter isolates revealed, at the 95% cutoff level, 31 genotypes among the Acinetobacter isolates (Fig 1 and Table 2). Rep-PCR genotypes 4, 5, 19 and 24 were only observed in single reference strains. Rep-PCR profiles 16a and 16b were closely related (92% similarity) (Fig 1).
One (type) isolate of each rep-type was identified to species by rpoB sequence analysis. It was assumed by inductive generalization that isolates of the same rep-type belonged to the same species. Thus, the analysis of the partial rpoB gene sequence of type strains allowed for the allocation of 14 of the 31 rep-PCR genotypes to A. baumannii (i.e. by inductive generalization 131/157 isolates, 83.4%). Likewise, 10 rep-types (15 isolates, 9.6%) were identified to A.      pittii, 3 rep-types (7 isolates, 4.5%) to A. nosocomialis and 4 rep-types, representing 1 isolate each (0.6%), to A. calcoaceticus, A. courvalinii (previously named genomic species 14BJ), NB14 and to 'close to A. pittii' (Fig 2). Twenty patients (41.6%) were colonized with 'non-baumannii' Acinetobacter strains. No patients were colonized with both baumannii and non-baumannii isolates. Fourteen rep-PCR genotypes harbored an OXA-51-like gene ( Table 2). Thirteen of these 14 OXA-51-like positive genotypes were identified as A. baumannii by rpoB sequencing; the remaining genotype (Rep30) was identified as A. pittii (Table 2 and Fig 2). The rpoB-based identification of some 'non-baumannii' Acinetobacter species was confirmed by 16S rRNA gene sequence analysis (data not shown).
No less than 12 XDR Acb complex strains were observed. Ten of them were identified to A. baumannii and 2 to A. nosocomialis. Rep-PCR profiles indicated that the 2 A. nosocomialis strains were closely related. All XDR A. baumannii strains were non-susceptible to carbapenems and were imported in the BWC through 5 patients, who were injured and subsequently  (Table 1). Four non-MDR strains were imported in the BWC when patients were transferred from Belgian hospitals (Table 1). One burn wound patient (Patient 1 in Table 1), who was transferred from Algeria where she had been hospitalized for 8 days, was shown to carry 4 genomically diverse XDR A. baumannii strains at admission. Two of them, strains Rep12 and 13, were the most prevalent (21 and 82 isolates, from 9 and 14 patients; respectively) in the BWC. Both strains caused consecutive outbreaks, interspersed with short periods of apparent eradication (Fig 3). Patient 1 was co-infected with MDR P. aeruginosa, ESBL producing Klebsiella pneumoniae, Providencia stuartii and MRSA. Strain Rep13 appeared to have been introduced once through the transfer of Patient 1and remained present  (Table 1 and Fig 3), while strain Rep12 was present over the entire study period, apparently as a result of repeated importations into the BWC (through patients 1, 23, 37 and 46) ( Table 1 and Fig 3). Both strains spread to the MCU after a few months, but did not spread to other departments such as the consultation and ambulatory services where discharged patients come for follow-up consultation. All hospital environmental isolates belonged to the Rep13 genotype. No non-baumannii Acinetobacter outbreaks were observed. Four pairs of patients (P2/3, P20/21, P25/29 and P31/ 34) were colonized by the same A. pittii rep-PCR type (each pair with a different rep-type) ( Table 1). Patients P2 and P3 were hospitalized at the same time in the same ward, and underwent the same specific treatment (skin expansion reconstruction), suggesting that cross-contamination may have occurred. Pediatric patients P20 and P21 stayed in the same ward, but during different periods of time. They did share two specific clinical interventions (adapted pediatric analgesia and application of a biosynthetic wound dressing). Pairs P25/29 and P31/34 resided in the same ward, but at different times, and shared no specific interventions or equipment.
MLST was used to investigate the prevalence of particular A. baumannii MLST types (e. g. EU clones I, II and III) in the BWC. Among the confirmed A. baumannii strains, 9 sequence types (STs) and 3 clonal complexes (CCs), including CC1 (EU clone I), CC2 (EU clone II), and CC10 were observed. MLST revealed that the major epidemic strains in the BWC, Rep12 and 13, were closely related to the EU (or WW) clone I. One burn wound patient (Patient 40), who was transferred from a Moroccan hospital, was shown to carry an XDR representative of the epidemic A. baumannii T strain, which is related to EU clone II. This strain type has been involved in outbreaks in the United Kingdom and the United States and was associated with repatriated casualties of the Iraq conflict [31]. In our study, the T strain, as well as the remaining 28 minor or sporadic genotypes, did not cause noteworthy outbreaks in the BWC.
Eight A. baumannii strains were bla OXA-23-like positive and carbapenem resistant ( Table 2). Two A. nosocomialis and one A. baumannii strain showed carbapenem resistance without detection of bla OXA-23-like . Most isolates of the two major epidemic strains in the BWC (Rep12 and 13) were bla OXA-23-like positive ( Table 2). A more detailed analysis revealed the presence of an AmpC β-lactamase gene and an ISAba1 insertion sequence, upstream of bla AmpC , but not of bla OXA-23-like , in both outbreak strains and of a TEM-1-like β-lactamase gene in isolates of the Rep13 genotype (data not shown). No genes coding for additional OXA-carbapenemases (23, 24 and 58-like), β-lactamases or integron sequences were detected (data not shown) in either of the strains.

Interventions
Identification of contaminated (bed-side) equipment as potential sources of Acinetobacter transmission led to the implementation of enhanced hospital environmental decontamination procedures. Unfortunately, even specialized decontamination using vaporized hydrogen peroxide, which necessitated a temporary closure of the BWC (from January 11 to February 15, 2007), did not result in noticeable changes in Acb complex colonization (Fig 3). Acinetobacter (cross-)colonization remained an important issue and the major epidemic A. baumanni strains (Rep12 and 13) were still very much present in the BWC after decontamination (Table 1 and Fig 3). The move to a new burn unit did not result in a considerable decrease of the incidence density of Acinetobacter colonization for which colonization prevalence rates were 3.45% (13 of 377 patients) and 2.95% (11 of 373 patients) in the 12 months before and after the move, respectively. However, whereas the admission of two patients colonized with representatives of the epidemic EU clone I (Patients 1 and 23 in Table 1) had led to outbreaks in the 'old' BWC, in the new BWC, the admission of two similar patients, also colonized with representatives of EU clone I (Patients 37 and 46) and one patient (P40) carrying a representative of the epidemic T strain (strain P40-01 and the T strain have nearly identical rep-PCR profiles) did not result in outbreaks (Table 1 and Fig 3). Although we are not able to provide statistical data to support this hypothesis, it seems that the stricter infection control measures, made possible by adapted hospital architecture and the introduction of surveillance methods, resulted in the successful containment of these renown epidemic strains.

Clinical impact
Eight (16.7%) of the 48 Acb complex infected (n = 3) and colonized (n = 5) patients died, including the index patient (Patient 1). All were colonized or infected with A. baumannii (Table 1), including 7 with CRAB. With the exception of patients 1 and 28, deaths occurred in debilitated middle-aged to elderly patients with a Baux score (comparative indicator of burn severity expressed as: age + %TBSA burned) ranging from 99 to 118 (Table 1). Deaths could not directly be attributed to A. baumannii infection or colonization. Three CRAB colonized/ infected (at the time of death) patients (Patients 1, 4 and 6 in Table 1) died of multiple organ failure related to sepsis (confirmed by positive blood culture) with P. aeruginosa (unrelated strains) and one from MRSA sepsis (Patient 7). Four patients, who were once colonized with A. baumannii, but were no longer at time of death, died of causes related to a heart condition (Patient 13), Serratia marcescens sepsis (confirmed by positive blood culture) (Patient 28), blood culture negative sepsis (Patient 17) and ventilator-associated pneumonia (VAP) with P. aeruginosa (Patient 23) ( Table 1). One patient (Patient 27, Table 1) survived a culture-confirmed XDR A. baumannii bloodstream infection. It is, however, plausible to assume that prolonged A. baumannii infection and colonization represented a contributing factor in the death of some of these patients.

Discussion
The first XDR Acinetobacter outbreak in our BWC, in 2006, triggered an in depth epidemiological investigation of Acinetobacter colonization and infection. Phenotypic identification tests such as the VITEK 2 system are insufficient for the identification of Acinetobacter isolates to the species or strain level [41]. Definitive species identification by partial rpoB sequence analysis of 157 selected Acb complex isolates (identified by VITEK 2) showed that most BWC Acb complex isolates were A. baumannii. Yet, no less than 20 patients were colonized with 'nonbaumannii' Acinetobacter strains, including 10 strains of A. pittii and 3 of A. nosocomialis, both of which species are closely related to A. baumannii. No spread of 'non-baumannii' Acinetobacter strains was observed.
Analogous to previous observations of P. aeruginosa [42], incidence of Acb complex colonization was proportional to the extent of the burn wound, the age of the patient, and the duration of the stay in the BWC. Nosocomially acquired colonization was demonstrated in 83% of P. aeruginosa colonized patients and in 77.1% of Acinetobacter colonized patients. Although Acinetobacter species have emerged as nosocomial pathogens from the 1970s onward [43], the Acinetobacter colonization rate observed in our BWC from 2006 to 2011 (3.3%) is still considerably lower than that of P. aeruginosa (16%) in the period 1997 to 1998 [42]. In contrast, a 6-year (2003 to 2008) antibiotic susceptibility records review at the US Army Institute of Surgical Research Burn Center revealed A. baumannii as the most prevalent organism (22% of isolates), followed by P. aeruginosa (20% of isolates) [29]. This difference is most likely due to the fact that, in contrast to the US Army burn center, the BWC of the Belgian Army only sporadically repatriated injured military personnel from areas with a high prevalence of MDR organisms. The average time from admission to Acb complex colonization was 14.7 days (range 0-45 days) in our BWC, while in US soldiers wounded in Iraq (2003 to 2005), the median time from injury to identification of Acb complex infection was 6 days (range 3-12 days) [20].
Strain identification to monitor particular strains was done by rep-typing. The discriminatory capacity of rep-PCR is almost as high as other genomic fingerprinting methods used for Acinetobacter typing, including fluorescent amplified fragment-length polymorphism (f-AFLP)-a method by which restriction fragments are selectively fluorescently labeled, amplified and electrophoretically separated-and pulsed field gel electrophoresis (PFGE) genomic fingerprinting [32,[44][45][46]. We opted for the DiversiLab system, a commercially adapted semi-automated rep-PCR system, because of its reproducibility, ease of use and high throughput [47][48][49]. Thirty-one rep-PCR types/clusters including single strains were distinguished among the 157 isolates from the BWC (Fig 1, Table 2), 12 of them containing XDR isolates. The two predominant XDR A. baumannii clones in our BWC (Rep12 and 13), causing repeated outbreaks involving 20 patients, were shown to be related to the EU (or WW) clone I, as confirmed by MLST. MLST was used to allocate rep-PCR genotypes to known epidemic clones of A. baumannii and is an important tool in the study of the population structure and genetic diversity of A. baumannii [31]. An XDR representative of the epidemic T strain, a highly successful outbreak strain of A. baumannii, associated with soldiers returning to the United States or United Kingdom from Iraq [31], was imported into the BWC through a patient transferred from a Moroccan hospital. Neither the T strain nor other rep-types had spread in the BWC (some found in multiple patients had no time-space overlap).
The index patient of the first outbreak carried 4 genomically distinct XDR A. baumannii strains at admission, including the two variants of EU/WW clone I (Rep 12 and 13), indicating that this patient was submitted from a hospital in which these organisms were likely endemic. Recently, Halachev et al. [50] also observed that one patient could be infected with more than one species or strain of Acinetobacter and could carry a multitude of variants of an outbreak strain. Patients carrying Acinetobacter were most often infected with other ESKAPE pathogens such as P. aeruginosa. During the study period, three additional representatives of Rep12 were likely imported in the BWC through the transfer of three patients from North Africa. In total, 10 XDR A. baumannii strains were introduced in the BWC through patients injured and hospitalized in North Africa in the period November 2006 to December 2010. The Brussels BWC was thus shown to suffer from consecutive clonal CRAB outbreaks, likely resulting from consecutive importations of outbreak strains. Of note, it can be difficult in regions with complex endemic situations to decide whether particular strains have persisted in a hospital or are reintroduced from other hospitals.
Carbapenem resistance in A. baumannii is most often mediated by class D oxacillinases (OXAs) [51], combined with low outer membrane permeability [52], and sporadically by metallo-β-lactamases [53]. Carbapenem hydrolysis by OXAs is considered to be too low to confer carbapenem resistance, but association with insertion elements has been shown to increase the expression of these carbapenemases in some cases [54][55][56]. Two types of OXAs were observed within Acinetobacter: intrinsic (chromosomal) OXAs and acquired OXAs. The first intrinsic A. baumannii OXA carbapenemase gene was detected in 2005 and was named bla OXA-51 [57]. Since then, many closely related variants were reported [58,59], which today are referred to as 'bla OXA-51-like '. Carbapenem non-susceptibility was only associated with intrinsic bla OXA-51-like when the insertion sequence ISAba1 was present upstream. The bla OXA-51 gene has been proposed to be a taxonomic marker for A. baumannii [36]. However, this has been challenged by the finding of the gene in A. nosocomialis [60], while we found bla OXA-51-like in an A. pittii clinical isolate. The detection of bla OXA-51-like is thus on its own not reliable for the identification of A. baumannii [61]. Three subgroups of acquired OXAs have been described so far: OXA-23-like, OXA-24/40-like and OXA-58-like [13]. Most isolates of the two major epidemic strains of the BWC, Rep12 and 13 representatives of EU clone I, were bla OXA-23-like positive and carbapenem resistant. MDR OXA-23 producing A. baumannii clones have emerged to cause nosocomial outbreaks worldwide [62][63][64][65][66][67][68][69]. The first CRAB outbreak in our BWC followed the transfer of a patient, with two OXA-23 producing A. baumanni strains, from North Algeria. CRAB strains have become an increasing cause of concern in Algeria, with OXA-23 producing A. baumannii being endemic in the north [70]. Both epidemic EU I clones also carried an appropriately oriented ISAba-ampC configuration, which was recently shown to increase ampC expression, conferring third-generation cephalosporin resistance in A. baumannii [71]. Rep12 and 13 isolates were indeed resistant to Ceftazidime, the most frequently used third-generation cephalosporin in our BWC during the study period. In addition, strain Rep13 also carried a TEM-1-like β-lactamase gene. TEM-type β-lactamases are most often found in Escherichia coli and K. pneumonia. It has been suggested that TEM-1 represents a clinically relevant mechanism of sulbactam (a β-lactamase inhibitor) resistance in A. baumannii [72].
There is a continuing controversy regarding the attributable mortality of A. baumannii infections [73]. From a review of case-control studies it was concluded that inappropriate antibiotic treatment of A. baumannii infections was associated with increased mortality [16]. All Acinetobacter strains we isolated were sensitive to colistin, which allowed for colistin (usually combined with rifampicin) treatments of CRAB infected patients. Since only two patients were treated with colistin monotherapy (without addition of rifampicin), it is impossible to know if the addition of rifampicin implied a clinical benefit. In a recent study [74], a significant increase of microbiologic eradication rate was observed when colistin was combined with rifampicin, but no difference was observed for infection-related death and length of hospitalization. Also taking into account the emergence of rifampicin resistance [75], this might imply that rifampicin should not be added to colistin in routine clinical practice [74]. Except a moderate renal impairment in one patient, no major adverse events related to colistin treatment were observed. From the analysis of 5 studies published between 1999 and 2006 in which patients with pneumonia due to P. aeruginosa or A. baumannii were treated with intravenous colistin, it was reported that de novo nephrotoxicity occured in 8-36% of patients, while neurotoxicity, which was commonly described in the old colistin era, was rarely observed [76].
Sample sizes are too small to perform adequate statistical analysis of the clinical impact of Acinetobacter infection and colonization. Eight (16.7%) out of 48 Acinetobacter colonized or infected patients died. Although no autopsies were performed, there were no indications that Acinetobacter infection and colonization was associated with increased attributable mortality, suggesting that the implemented antibiotic treatment was overall successful. Six of the eight deceased patients had Baux scores ranging from 99 to 118, scores at which predicted mortality is high. A 27-year retrospective study, including 11,109 burn wound patients, showed that the Baux score at which predicted mortality reached 50% was 109.6 in the 2000 to 2008 cohort [77]. This finding is congruent with a previous study that found that although MDR Acb is a common cause of nosocomial infection in the burn patient population, it does not independently affect mortality [78]. In comparison, 8 out of 70 (11.4%) P. aeruginosa colonized patients died and for 3 of them, death was directly attributed to P. aeruginosa infection [36]. 'Non-baumannii' Acinetobacter species, and especially A. nosocomialis and A. pittii, are increasingly recognized as significant causes of nosocomial infection [2,[79][80][81][82][83]. The mortality rate for patients infected with 'non-baumanii' Acinetobacter strains is considered to be lower than for A. baumannii [81,83]. Poor outcome seemed to be associated with MDR strains rather than with A. baumannii per se [81]. With the exception of two closely related XDR A. nosocomialis strains, 'non-baumanni' Acinetobacter strains were relatively susceptible to antibiotics and as a consequence the initial (empiric) antimicrobial therapy was appropriate. None of the 'non-baumannii' Acinetobacter colonized or infected patients died. The mean length of colonization/infection was only 4.9 days for non-baumannii Acinetobacter as compared to 19.1 days for A. baumannii strains (Table 1). In contrast to A. baumannii strains, and more specifically epidemic strains, which were shown to colonize/infect multiple sites of the burned patients, non-baumannii Acinetobacter strains were mostly isolated from wounds. These differences are possibly due to the more pronounced antibiotic resistance of A. baumannii strains.
The natural reservoir of A. baumannii is still unknown and the human body does not seem to be a natural habitat for A. baumannii. In 1999, Berlau and coworkers determined the distribution of 19 genospecies of Acinetobacter on human skin (forehead, forearm and toe webs). Over 40% of 192 healthy volunteers carried Acinetobacter spp. at one or more body sites, with Acinetobacter lwoffii as the most common (61%) colonizer, and the Acb complex was found in only one individual [84]. Skin or nasal carriage of Acb complex was also not detected among healthy US Army soldiers [85,86]. However, certain MDR epidemic strains can survive in the hospital setting. Susceptible, critically ill patients and their environment become a niche or reservoir from which transmission to other patients and environment occurs [7]. Contaminated healthcare workers may be vectors for transmission [87,88], although a screening of the BWC's healthcare personnel during an outbreak did not provide evidence for this. The epidemic Rep13 strain was recovered from the patients' environment on several occasions. Likely, this contamination was secondary to patient carriage. A. baumannii strains have the ability to survive for a long time on dry surfaces under simulated hospital conditions, and this irrespective of their sporadic or epidemic occurrence [89]. The apparent absence of Acinetobacter in healthcare workers, the recovery of the major epidemic A. baumannii strain from the direct environment of colonized patients and the continuous admittance and presence of colonized patients in the BWC suggest that the epidemic strains entered the BWC through colonized patients and that prolonged carriage and hospital environmental contamination were responsible for the spread to other susceptible patients, possibly through contaminated equipment.
Despite systematic screening on admission, draconian hospital hygiene measures based on intensive surveillance methods and a temporary closure of the BWC for specialized decontamination, two imported OXA-23 producing EU I clones persisted in the BWC, causing recurrent outbreaks. Desiccation tolerance and excessive drug resistance may have contributed to their maintenance in the BWC and their propensity to cause outbreaks. It seems that the move to a new BWC with an adapted architecture finally allowed for the implementation of infection control measures necessary for the containment of these epidemic strains. However, there is a lack of evidence linking hospital design and construction with the prevention of nosocomial infection and it seems that other factors, especially the improper hand hygiene of medical staff, might have greater impact [90].

Conclusions
This study adds to previous observations that hospitals are vulnerable to importation of MDR A. baumannii when patients are transferred from hospitals in countries known to have a high prevalence of resistance mechanisms. It is no surprise that the military BWC in Brussels was confronted with MDR A. baumannii outbreaks. Patients with extensive burn wounds are particularly at risk for infection and military hospitals often admit patients that were repatriated from regions where they were exposed to endemic or emerging MDR pathogens such as OXA-23 producing A. baumannii. This study also underlines the importance of systematic screening upon admission and early application of strict barrier precautions for any transferred patient, whatever the country of origin [91][92][93]. Rapid automated identification and typing methods (e.g. VITEK and DiversiLab systems) are essential tools in the epidemiological investigation and surveillance of (Acinetobacter) outbreaks. The timely identification of highly related or indistinguishable isolates, suggesting transmission from the environment, health care workers or patients, may lead to successful interventions to control major MDR pathogens such as A. baumannii. More elaborate and robust genetic identification and typing methods such as MLST and rpoB gene sequence analysis are of scientific importance. They can be used to assign strains to renown epidemic clones or to perform taxonomic or population structure studies. The setup of bacterial culture collections and fingerprint databases with relevant clinical isolates and reference strains is recommended. The role of non-baumannii Acinetobacter in human infections is increasingly reported thanks to these new molecular biology techniques that allow correct identification of bacteria at the species level. Although numerous patients were shown to be colonized by non-baumannii Acinetobacter strains, these bacteria were not associated with infections or outbreaks in our BWC. Of concern, two closely related A. nosocomialis strains were XDR with resistance to carbapenems. Finally, this study supports the assumption that Acinetobacter infection and colonization is not associated with increased attributable mortality, provided that an appropriate antibiotic (e.g. colistin/rifampicin) treatment is timely initiated.