Molecular Analysis of Acinetobacter baumannii Strains Isolated in Lebanon Using Four Different Typing Methods

This study analyzed 42 Acinetobacter baumannii strains collected between 2009–2012 from different hospitals in Beyrouth and North Lebanon to better understand the epidemiology and carbapenem resistance mechanisms in our collection and to compare the robustness of pulsed field gel electrophoresis (PFGE), multilocus sequence typing (MLST), repetitive sequence-based PCR (rep-PCR) and bla OXA-51 sequence-based typing (SBT). Among 31 carbapenem resistant strains, we have detected three carbapenem resistance genes: 28 carried the bla OXA-23 gene, 1 the bla OXA-24 gene and 2 strains the bla OXA-58 gene. This is the first detection of bla OXA-23 and bla OXA-24 in Lebanon. PFGE identified 11 types and was the most discriminating technique followed by rep-PCR (9 types), bla OXA-51 SBT (8 types) and MLST (7 types). The PFGE type A'/ST2 was the dominant genotype in our collection present in Beyrouth and North Lebanon. The clustering agreement between all techniques was measured by adjust Wallace coefficient. An overall agreement has been demonstrated. High values of adjust Wallace coefficient were found with followed combinations: PFGE to predict MLST types  = 100%, PFGE to predict bla OXA-51 SBT = 100%, bla OXA-51 SBT to predict MLST = 100%, MLST to predict bla OXA-51 SBT = 84.7%, rep-PCR to predict MLST = 81.5%, PFGE to predict rep-PCR = 69% and rep-PCR to predict bla OXA-51 SBT = 67.2%. PFGE and MLST are gold standard methods for outbreaks investigation and population structure studies respectively. Otherwise, these two techniques are technically, time and cost demanding. We recommend the use of bla OXA-51 SBT as first typing method to screen isolates and assign them to their corresponding clonal lineages. Repetitive sequence-based PCR is a rapid tool to access outbreaks but careful interpretation of results must be always performed.

screen isolates and assign them to their corresponding clonal lineages. Repetitive sequence-based PCR is a rapid tool to access outbreaks but careful interpretation of results must be always performed.

Introduction
Acinetobacter baumannii is an opportunistic gram negative pathogen involved in a wide number of nosocomial infections like ventilator-associated pneumonia, bloodstream, urinary tract, wound and meningitis infections frequently associated with a high rate of mortality and morbidity [1]. Outbreaks have been intensively documented worldwide and are usually caused by multidrug resistant strains and more and more carbapenem resistant strains [2,3]. These outbreaks strains mainly belonged to three international clones I, II and III (previously called as European clones [4,5]), but also to different clonal lineages. Karah et al. [2] analyzed the MLST-based global population structure of A. baumannii on 496 isolates. They showed the presence of 26 clones and among them, 18 were international clones and 8 European or Asian restricted clones. The International clone II was the major clone reported in 34 countries in Europe, Asia, Africa Australia, USA, and South America.
To track and monitor these outbreaks, denote strain relatedness and assign an outbreak strain to its corresponding clonal lineage, many typing methods with different intrinsic degrees of resolution are proposed [3] such as pulsed field gel electrophoresis (PFGE) [6], repetitive-sequence-based PCR (rep-PCR) [7], amplified fragment length polymorphism (AFLP) [8], multilocus sequence typing (MLST) [9,10], 3-locus sequence typing (3-LST) [11], bla OXA-51 sequence-based typing (SBT) [12] or Multiple-Locus Variable number of tandem repeat Analysis (MLVA) [13]. Selection of an appropriate genotyping technique is not always easy and depends to the studied objectives. Many authors emphasized the great need to validate the application of each method as well as to harmonize different typing methods by reference networks [3,14]. Among these methods, PFGE is still considered the current gold standard for A. baumannii outbreak investigation at local scale [3]. MLST has a discriminatory power lesser than PFGE and is regarded as the gold standard for large epidemiological and population structure studies. For A. baumannii, two MLST schemes have been proposed: Bartual's MLST [9] and Pasteur's MLST [10]. DiversiLab is a semi-automated form of rep-PCR with a comparable discriminatory power to PFGE [15]. The identification of eight international clones is one of the remarkable advantages of this system [7]. bla OXA-51 SBT has been proposed as a single-locus based typing [12] with a similar discriminatory power to rep-PCR [16] and Bartual and Pasteur's MLST [12,17].
In Lebanon, there have been limited reports studying only local outbreaks in Beyrouth between 2004 and 2007 where bla OXA-58 was the only carbapenem resistance gene identified [18][19][20]. Recently, we have detected four bla NDM-1 producing A. baumannii isolated in Tripoli, Northern Lebanon from Syrian civilians wounded during Syrian war [21].
The present study has a double aim: firstly to compare the performance and effectiveness of four epidemiological typing methods (PFGE, rep-PCR, MLST and bla OXA-51-like SBT), and secondary to get primary results on circulating clones and carbapenem resistance mechanisms in Lebanon by analysis of 42 non duplicate strains conserved on Azm center for research on biotechnology and its application (Lebanese university) and collected between 2006-2012 from different hospitals in Beyrouth and North Lebanon.

Identification
Forty-two strains were confirmed as A. baumannii by molecular techniques. These strains were isolated in different hospitals in Beyrouth (24 strains) and North Lebanon (18 strains) from various clinical specimens between 2009 and 2012 except one strain isolated in 2006 (Table 1). Beyrouth strains were isolated during epidemiological contexts.

PFGE
Using $80% similarity cut-off as a threshold, PFGE classified our strains in 11 types: 4 clusters (A' to D') and 7 unique profiles (Fig. 1). Cluster A' comprised 28 strains, 27 were bla OXA-23 producing strains and one an imipenem susceptible strain. This cluster contained strains from Beyrouth and North Lebanon. Cluster B' contained two bla OXA-58 producing strains that belonged to the same hospital. Cluster C' and D' contained two and three imipenem susceptible strains respectively.

MLST
MLST typing was performed on all strains. Seven ST(s) were identified ( Fig. 1), two of which were novel and assigned as 284 and 285 by MLST Pasteur. The ST2 was the most predominant, present in 34 strains, followed by ST25, present in 3 strains; other ST(s) were present sporadically in our collection. The eBUSRT analysis of our ST(s) with all identified ST(s) in MLST database (28.03.2014) showed that ST 1, 2, 3, 25, 284 belonged to CC 1, 2, 3, 25 and 33 respectively, whereas ST285 was a singleton, and ST108 shared similarity with ST112 (we did not assign a name for this complex because no more of 2 ST(s) have been yet identified). Interestingly, our novel ST284 was a SLV of ST33 (the founder of CC33).

Comparison of the four typing methods
The visual analysis of the collection (Fig. 1) showed an overall agreement between the different techniques, although some discrepancies have been noticed. The adjusted Wallace coefficient analysis (

Discussion
This is the first study in Lebanon providing data about clonality and carbapenem resistance mechanisms of a set of isolates recovered from different hospitals in Beyrouth and North Lebanon. However, this study does not illustrate the overall A. baumannii molecular epidemiology in this country because it does not contain prospectively collected isolates from different hospitals in different Lebanese provinces. Currently, worldwide carbapenem resistant strains are mostly associated with international clone 2 [2], with bla OXA-23 as the main carbapenem resistance mechanism [22][23][24]. Our results stick well to the global situation where the majority of carbapenem resistant strains belonged to ST2, but only two imipenem resistant strains in our collection belonged to ST1 and ST285 (Fig. 1). In several countries, bla OXA-58 [25] or bla OXA-24 [26] have been progressively replaced by bla OXA-23 . In Beyrouth-Lebanon, the studied outbreaks between 2004-2007 [18][19][20] were caused by the three international clones (1 to 3) producing OXA-58 as a main carbapenem resistance mechanism. In the present study, we detected bla OXA-58 in two strains from a hospital in Beyrouth, but bla OXA-23 seemed to be an emerging carbapenemase in Beyrouth and North Lebanon as else. Emergence of bla OXA-23 in Lebanon is linked to the clonal spread of the PFGE type A'/ST2. It is noteworthy that we reported in this study the first detection of bla OXA-23 and bla OXA-24 in Lebanon. Interestingly, the PFGE type A'/ST2 (harboring bla OXA-66 ) dominated heavily in Beyrouth and North Lebanon suggesting an extensively inter-hospital dissemination and thus it could be considered as an epidemic cluster. Beside this epidemic clone, some hospitals had their unique PFGE clone (PFGE type B'/ST2, PFGE type C'/ST2 and PFGE type D'/ST25). In those cases, it is interesting to notice that some clones contained isolates separated by 4 years. These results suggest a successful spread of well-established clones and therefore the urgent need of effective infection control measures to eradicate such bugs. Our A. baumannii strains were analyzed using four epidemiological typing techniques. PFGE was the most discriminating scheme allowing the recognition of 11 types, followed by DiversiLab with 9 types then bla OXA-51 with 8 types and finally MLST with 7 types.
To our knowledge, there is no sufficient use of adjusted Wallace coefficient in A. baumannii typing field, an area which has largely been expanded in the last decades. Many techniques have been proposed with an increasing trend to track this pathogen and assign it to its corresponding international clonal lineage. Therefore, there is an arising need to perform a quantitative comparison between available typing methods using this coefficient or other coefficients to assess their strengths and weaknesses, better understand and validate their fields of application.
PFGE has been considered as a gold standard for outbreak investigations due to its higher discriminatory power which impaired its use for large investigations and population structure studies. As we have shown, PFGE is a very good method to predict bla OXA-51 , MLST and rep-PCR types, but its higher resolution prevents other methods to successfully predict its types. Indeed, PFGE is a time demanding and labor intensive method with many intra and inter laboratory reproducibility problems [27]. ApaI, the classical restriction enzyme used for almost A. baumannii PFGE protocols generates complex DNA patterns with more than 40 fragments [28]. Chang et al., 2013 [28] suggested the use of two other infrequent cutting restriction enzymes (AscI and AsiSI) generating clear patterns with only 10-20 fragments per pattern.
MLST is a highly informative technique which puts isolates in a global context [24] and can directly assign them to their clonal complex. Thus, it is regarded as the method of choice for long term and phylogenetic studies. Although it is a reproducible and portable method, MLST is expensive and time demanding. As expected, MLST was unable to predict PFGE and rep-PCR types because of a discriminatory power lesser than found in PFGE and rep-PCR [12,24]. Besides, MLST could predict very well (84.7%) all bla OXA-51 sequences except the colistin resistant strain that belonged to ST2 and had an OXA-336, a single amino acid variant of OXA-66 enzyme.
DiversiLab is a commercial rep-PCR typing system which benefits from several advantages: rapidity with an ability to investigate large number of isolates, standardization, reproducibility with conserved clustering between laboratories [29], and allowing in house libraries building. For A. baumannii, rep-PCR had revealed a comparable discriminatory power to PFGE [15]. It has been suggested to study A. baumannii population structure with rep-PCR as it can identify eight international clones within 492 isolates from a worldwide collection [7] and generate concordant results with MLST and bla OXA-51 SBT [16,30]. Our rep-PCR results showed an overall agreement with PFGE, whenever some exceptions (Fig. 1) have been noticed which explained the low adjust Wallace coefficient (PFGE predicted 69% of rep-PCR types). Compared to PFGE clustering, rep-PCR grouped differently some strains belonging to the same PFGE type or assembled some strains belonging to different PFGE types in the same rep-PCR types, as reported with some other authors [31,32]. One discrepant result was the strain belonging to ST3 which regrouped with rep-PCR type A, whereas it had a unique PFGE type. When we visually checked the graphs of samples and overlayed functions, we noticed the presence of another peak which was the source of this confusion between strains belonging ST2/rep-PCR type A and strain ST3/rep-PCR type A. Hence, this noticed the importance of careful interpretation of rep-PCR results [31].
The bla OXA-51 gene is an intrinsic chromosomal beta-lactamase gene naturally found in A. baumannii and up to 95 enzyme variants have been identified to date [33]. Sequencing of the entire gene was proposed as a bla OXA-51 SBT [12]. This SBT could correctly identify the eight international clones characterized by rep-PCR [16]. Also, it correlated well with Bartual's MLST [12]. Compared with Pasteur's MLST, SBT could correctly assign the nine clonal complexes in such manner each CC had specific bla OXA-51 alleles [17]. We have found similar results, where each CC or ST had a specific bla OXA-51 variant confirmed by adjusted Wallace coefficient (bla OXA-51 SBT predict 100% all ST). Indeed, the colistin strain (strain 17) carried OXA-336 which differed only by a single amino acid from OXA-66, indicating that bla OXA-51 SBT could correctly identify it as belonging to CC2. Compared with PFGE clustering; bla OXA-51 SBT had a lower discriminatory power. bla OXA-66 characteristically linked to ST2 has been found in 3 PFGE types (A', B' and C') and one unique profile. The bla OXA-64 variant carried by ST25 had characteristically been linked to PFGE type D'. Finally, other bla OXA-51-like variants had unique PFGE profiles.

Conclusion
This report describes the first detection of bla OXA-23 and bla OXA-24 in Lebanon. Although, our collection is unable to give the real picture of molecular epidemiology in Lebanon, it shed lights on circulating clones and on the mechanisms of carbapenem resistance. Other multi center studies are obviously required to better understand the epidemiology of this bug in the country.
Overall, a good concordance with the four typing methods was shown. PFGE and MLST are reference methods in local and long term epidemiological studies respectively, although both methods are time and cost consuming. bla OXA-51 SBT seems to be an excellent choice for initial epidemiological screening of isolates. rep-PCR is a rapid tool to access outbreaks at local scale but careful interpretation of results must be done.

Bacterial strains
A total of 42 non redundant clinical strains of A. baumannii isolated from various clinical samples were collected between 2006 and 2012 from the following hospitals: Rafic Harrii Beyrouth governmental hospital, Tripoli governmental hospital (TGH), Nini hospital, Rahal hospital, Monla hospital and Ain Wazein Hospital (AWH). All the bacterial strains were de-identified and a number was attributed prior to access and analysis. No consent was needed since strains used in this study were those isolated during routine analysis in the different laboratories. The clinical sources of the different strains are noted in Table 1.

Identification
Isolates were routinely cultured on Blood agar at 37˚C, and stored at -80˚C until further study. Identification to A. calcoaceticus-baumannii complex was initially performed using MALDI-TOF Vitek MS (bioMérieux, Marcy-l'Etoile, France) and confirmation of identification at species level was done by real time PCR of bla OXA-51 gene [34] and partial RNA polymerase b-subunit (rpoB) gene sequencing [35].

Susceptibility testing and investigation of carbapenem resistance mechanisms
Antibiotic Susceptibility testing was determined by the disc diffusion method according to the guidelines of the French Society of Microbiology (www.sfmmicrobiologie.org/). Resistance to carbapenem and colistin were confirmed by determining imipenem, meropenem, doripenem and colistin minimum inhibitory concentration (MICs) by Etest strips (bioMérieux, Marcy-l'É toile, France). Carbapenem resistant isolates were investigated for the presence of carbapenem resistance genes bla OXA-23 , bla OXA-24 , bla OXA-58 , bla OXA-143 , bla ndm-1 by PCR ( Table 3). Presence of the insertion sequence ISAba1 was also screened. The association ISAba1-bla OXA-23 and ISAba1-bla OXA-51 was tested using a combination of primers ISAba1F with reverse primers targeting bla OXA-23 or bla OXA-51 respectively ( Table 3).

Epidemiological typing
Pulsed field gel electrophoresis (PFGE) PFGE using ApaI as a restriction enzyme was done as described previously [36]. DNA fragments were separated in CHEF-DRIII system (Biorad, Marne LA Coquette, France) at 6V/cm and 14˚C for 21 hours with pulse times ranging from 3 s to 20 s. Computer-assisted analysis was performed by using fingerprinting II (Biorad, Marne LA Coquette, France) with the unweighted pair-group method with artithmetic averages (UPGMA) and Dice similarity coefficient for banding pattern comparison. A PFGE type was defined by a cluster of isolates showing $80% similarity MLST MLST was performed according to the Pasteur scheme (http://www.pasteur.fr/ recherche/genopole/PF8/mlst/Abaumannii.html). The internal fragments of seven housekeeping genes (fusA, pyrG, rpoB, rplB, cpn60, gltA and recA) were amplified, then purified and sequenced by an ABI 3130XL DNA sequencer (Applied Biosystems, Foster City, United States). The sequences were compared to the available sequences present in the MLST website. When a new allele or ST was identified, it was submitted and codified by Institut Pasteur MLST Database. The eBURST [37] was used to compare ST to the existed ST(s) and to assign isolates to their clonal complexes. A clonal complex (CC) is defined as a set of similar ST(s) having 6 identical loci among 7, so a CC is formed by the founder ST and its single locus variants (SLV) [10]. rep-PCR Rep-PCR was performed using the automated system DiversiLab, version 3.4 (bioMérieux, Marcy-l'É toile, France) following the manufacturer recommendations. Briefly, bacteria were cultured on sheep blood agar (Oxoid). DNA was extracted using the MoBio Ultra Clean microbial DNA extraction Kit, and adjusted to 50 ng/ml. After extraction, DNA was amplified using DiversiLab Acinetobacter kit, and the amplified DNA was separated and detected by Agilent 2100 Bioanalyser (Agilent Technologies). The resulted fingerprints were analyzed using the DiversiLab software with the modified Kullback-Leibler (KL) as a statistical method and $95% as a threshold to define a cluster of closely related isolates or a rep-PCR type.
bla OXA-51 sequence-based typing (SBT) This typing method consists to sequence the full length (825 bp) of a single locus bla OXA-51 gene. The bla OXA-51 was amplified by external primers OXA-69A/OXA-69B as described [12]. PCR products were purified and sequenced by ABI 3130xl DNA sequencer (Applied Biosystems). The resulted sequences were compared to all variants present in BLAST. When a novel variant was detected, it was submitted to GenBank and assigned by the Lahey database for beta-lactamase classification (http://www.lahey.org/studies/webt.asp).

Concordance between techniques
The online tool (http://www.comparingpartitions.info/) was used to calculate the adjusted Wallace coefficient. This coefficient is an objective and quantitative measure of clustering agreement between the studied techniques which indicates the probability of a pair of isolates assigned at the same type by one technique is also reassigned at the same by the other technique [32,38].

Ethic statement
Not applicable mlst. We thank also Catherine Ramont, Mariam Yehya and Taha Abdo for excellent technical assistance. A part of results of carbapenemase detection was presented at the 9 th International Symposium on the Biology of Acinetobacter, Cologne 2013, poster presentation P5-7.