Extended-spectrum β-lactamase (ESBL)-producing bacteria represent a major worldwide threat among drug-resistant bacteria in both hospital and community settings [1]. ESBLs are among the Ambler classes A, confer resistance to β-lactam antibiotics except cephamycins and carbapenems, and are inhibited by clavulanic acid [1]. ESBLs are often located on large plasmids that also harbor resistant genes to other antimicrobial classes with resulting multidrug-resistant isolates [2].

The first ESBLs have evolved by genetic mutation from native β-lactamases TEM and SHV [3][4]. Recently, a novel type of ESBLs, the CTX-M enzymes, emerged worldwide, mostly from Enterobacteriaceae[5, 6]. CTX-M β-lactamases are not closely related to TEM or SHV ESBLs but share high amino-acid identity with chromosomal β-lactamases from Kluyvera spp. [7]. Now, bla_CTX-M-15 is recognized as the most widely distributed CTX-M enzyme [8]. It is derived from CTX-M-3 by a substitution of Asp-240-Gly which increases its catalytic efficiency against ceftazidime [9]. bla_CTX-M-15 are encoded on plasmids belonging to the incompatibility group IncF [10]. In the upstream region of CTX-M genes an insertion sequence element, ISEcp1, is commonly present and is likely responsible for the transposition process of the genes [11].

E. coli is among the most prevalent causes of hospital-acquired and community-acquired bacterial infections and their resistances to antimicrobial agents have become a serious concern for healthcare providers [5]. Phylogenetic analyses have classified E. coli into four main phylogenetic groups (A, B1, B2, and D). Commensal isolates belong mainly to A and B1 groups whereas virulent extra-intestinal pathogenic E. coli (ExPEC) are essentially from the B2 and D groups [12, 13]. ExPEC harbor numerous virulence factors including α-hemolysin, cytotoxic necrotizing factor, adhesins and iron acquisition systems [12]. The spread of bla_CTX-M-15 has been mainly associated with the dissemination of a particular clone of E. coli ST131 belonging to phylogenetic group B2 [14, 15]. Recently, an E. coli clone O25 ST131, producing CTX-M-15, with high virulence potential and belonging to the B2 group, has been reported and represent a major public health problem [14, 15].

Many reports have documented the emergence of ESBL-producing Enterobacteriaceae[16–18]. In Antananarivo, ESBLs were first detected in 2005 from UTI in 9.7% of isolated Enterobacteriaceae[19]. In 2006, outbreaks of CTX-M-15 and SHV-2-producing K. pneumoniae isolates have been described in two pediatric units [20]. More recently, 21.3% of clinical isolates from patients in surgery and intensive care units [21] and 21.2% of intestinal carriage isolates from children hospitalized in a pediatric department of a large teaching hospital [22] were ESBL-producers.

For 49 multidrug-resistant Enterobacteriaceae isolates from Antananarivo, we characterized: i) the genes encoding the ESBLs; ii) the drug resistance genes associated with the ESBL genes; iii) gene cassettes present in the isolates; and iv) the plasmid incompatibility groups of the isolates. We also determined the phylogenetic groups and virulence factors of the E. coli isolates.

We extensively characterized 49 ESBL-producing Enterobacteriaceae collected over a period of 15 months in four hospitals and at the Pasteur Institute Medical Laboratory. Previous studies in Antananarivo have shown resistant bacteria clonal diffusion in hospital settings [20, 30], but among the 49 non-representative ESBL-producing Enterobacteriaceae studied, no clonal isolates have been found.

The bla_CTX-M-15 ESBL gene is considered to be the most prevalent ESBL worldwide [17, 18, 23, 31, 32]. We also found bla_CTX-M-15 to be the most prevalent ESBL in Madagascar, as it was detected in 75.5% of the isolates we studied. A study involving nine Asian countries reported that bla_CTX-M-15 was highly prevalent among ESBL-producing K. pneumoniae isolates (60%, 55/92) [17]. In Tunisia, Dahmen et al. reported that 91% of the ESBL-producing isolates carried bla_CTX-M-15 genes [23]. Our findings are intermediate between those found in Asia and in Tunisia and confirm the predominance of bla_CTX-M-15 among ESBL-producing isolates. In Antananarivo, a previous study conducted in the neonatal units of two hospitals in 2006 documented that a clonal outbreak of K. pneumoniae harbored bla_CTX-M-15 and bla_SHV-2 genes [20]. In 2009, a community-based study of the intestinal carriage of 49 ESBL-producing Enterobacteriaceae demonstrated that the most prevalent ESBL gene was bla_CTX-M-15 (93.9%), followed by bla_CTX-M-3, bla_SHV-12 and bla_SHV-2a[33]. The presences of bla_CTX-M-15, bla_CTX-M-3, bla_SHV-2 and bla_SHV-12 is not surprising as molecular analysis indicated that bla_CTX-M-15 derived from bla_CTX-M-3[6] and bla_SHV-12 from bla_SHV-2[34].

CTX-M genes may disseminate through clonal expansion or horizontal gene transfer [35, 36]. In our study, ISEcp1 was found upstream from bla_CTX-M-15 at variable distances, as was previously described [18]. ISEcp1 was found to be in the vicinity of many bla_CTX-M genes (including bla_CTX-M-15) and was reported to contain sequences resembling a typical promoter region [11]. Then, plasmids carrying bla_CTX-M-15 were assigned to the IncFII, IncFIA or IncHI2 incompatibility group replicons. Association of the bla_CTX-M-15 gene with IncF plasmids carrying the FII replicon in association with the FIA or FIB replicon has been reported previously for isolates in Canada, France, Spain, Tunisia, and the United Kingdom [35, 36]. The first evidence of the association of the FII plasmid with the bla_CTX-M-15 gene was demonstrated by sequencing the entire pC15-1a plasmid from epidemic E. coli isolated in Canada [2]. The IncHI2 plasmid, frequently associated with bla_CTX-M-2 or bla_CTX-M-9, was first identified in Serratia marcescens[10], but rarely reported in association with bla_CTX-M-15.

Like bla_CTX-M-15, bla_SHV-12 is also widely distributed. In our study, 38% of the isolates harbored bla_SHV-12. First described in Switzerland [37] and subsequently found in various continents, including Africa [38], bla_SHV-12 is most often found in Asia [34]. Plasmids carrying bla_SHV-12 were assigned to the IncFII replicon, as previously reported in France [39]. Evolutionary analysis of GenBank sequences indicated that bla_SHV-12 evolved from the branch of bla_SHV-2a[34]. Although it is possible that this transformation occurred in Antananarivo, as bla_SHV-2a was reported in neonatal units in 2009 [20]. It can also be assumed that the local emergence of bla_SHV-12 could be explained by introduction of international clones.

Our antimicrobial susceptibility analysis of the ESBL-producing isolates found highly prevalent resistances to gentamicin (87.7%); tobramycin (93.8%); ciprofloxacin (69.3%) and to trimethoprim-sulfamethoxazole (100%) and confirm the presence of multidrug-resistant isolates in Antananarivo [19, 22]. The finding of multidrug resistance among ESBL-producing isolates is of great clinical relevance due to the severely limited therapeutic options and the high risk of treatment failure in patients infected with these strains.

Genes encoding ESBLs are often associated with determinants of resistance to other antimicrobial agents, including aminoglycosides (aac(6)-Ib), fluoroquinolones (qnr), tetracycline (tetA), and trimethoprim-sulfamethoxazole (sul) and are frequently located on plasmids belonging to the IncF group [10]. In this study, we found the first example in Madagascar of the plasmid-mediated quinolone resistance (PMQR) genes: qnrB (24.5%) and qnrA (2%), and a variant gene aac(6 ^′ )-Ib-cr (67.3%) that encodes an aminoglycoside-modifying enzyme. Qnr gene prevalence was higher in the K. pneumoniae (41.7%) isolates than in the E. coli (25%) isolates, which has been noted by other authors [24, 40]. The aac(6 ^′ )-Ib-cr gene accounted for 94.3% (33/35) of the aac(6 ^′ )-Ib genes detected. This high proportion of aac(6 ^′ )-Ib-cr/aac(6 ^′ )-Ib was also observed in a previous study [40]. The PMQR genes qnr and aac(6 ^′ )-Ib-cr are now recognized to be geographically widespread [24, 25]. These genes have been previously reported to be associated with ESBLs. The horizontal transfer of plasmids harboring genes encoding for ESBLs and PMQR genes could have promoted this co-resistance.

The cassette region could not be amplified by PCR in 23 class 1 integron-containing isolates, which may have been due to the lack of the 3^′CS. The analysis of 25 cassette regions revealed a predominance of aadA and dfrA genes, which confer resistance to aminoglycosides and trimethoprim, respectively. This result correlates with previous studies of African Enterobacteriaceae isolates [27, 41]. The combination of dfrA17-aadA5 (22%) was the one most frequently detected in our study. Similar findings were reported for isolates from Taiwan and Tunisia, as dfrA17-aadA5 was found in 81 of 224 (36%) and in 3 of 4 (75%) E. coli class 1 integrons, respectively [42, 43].

Analysis of the phylogenetic groups and virulence factors of E. coli isolates revealed that most of these isolates belong to group A1. The phylogenetic group A1 consists of commensal enteric E. coli and may therefore be the natural reservoir of pathogenic isolates. Pathogenic E. coli isolates may have derived from commensal isolates by acquiring chromosomal or extra chromosomal virulence operons [44]. Although virulence determinants are considered to be mobile, strain phylogeny and virulence may be linked [45]. The B2 phylogenetic group, which diverges from the commensal isolates, evolved toward extra intestinal virulence by acquiring numerous pathogenic determinants [12].

We also encountered an E. coli isolate belonging to group B2, harboring bla_CTX-M-15 and other resistance genes, and corresponding to the worldwide pandemic clone O25b-ST131. It has been reported that most O25-ST131 isolates are multidrug-resistant, produce CTX-M-15 ESBL enzymes [14] and harbor virulence genes required for pathogenic invasion of hosts. In one study, the genes for adhesins (iha, fimH), siderophores (fyuA, iutA) and the toxin (sat) were found in 95% - 100% of the O25b-ST131 E. coli isolates [14], but typical fimbriae and pilus genes, such as those encoded by the papA allele, were not. In Africa, few data exist on the presence of ST131. In a South African study, 43% of 23 isolates were ST131 [46]; as were 50% of the CTX-M-15-producing E. coli isolates collected in the Central African Republic [13]. The presence of this clone in Antananarivo hospitals is of concern and further studies should be conducted to assess its prevalence.

Our results highlight the dissemination of multidrug resistant Enterobacteriaceae isolates in Antananarivo, in different hospital settings and probably in the community. These findings underline the need for a rational use of antibiotic and for appropriate methods of screening ESBL in routine laboratories in Antananarivo.