Clonal dissemination of the multi-drug resistant Salmonella enterica serovar Braenderup, but not the serovar Bareilly, of prevalent serogroup C1 Salmonella from Taiwan

Background Nontyphoidal Salmonella is the main cause of human salmonellosis. In order to study the prevalent serogroups and serovars of clinical isolates in Taiwan, 8931 Salmonellae isolates were collected from 19 medical centers and district hospitals throughout the country from 2004 to 2007. The pulsed-field eletrophoresis types (PFGE) and antibiotic resistance profiles of Salmonella enterica serovars Bareilly (S. Bareilly) and Braenderup (S. Braenderup) were compared, and multi-drug resistance (MDR) plasmids were characterized. Results Over 95% of human salmonellosis in Taiwan was caused by five Salmonella serogroups: B, C1, C2-C3, D1, and E1. S. Typhymurium, S. Enteritidis, S. Stanley and S. Newport were the four most prevalent serovars, accounting for about 64% of isolates. While only one or two major serovars from four of the most prevalent serogroups were represented, four predominant serovars were found in serogroup C1 Salmonellae. The prevalence was decreasing for S. Choleraeuis and S. Braenderup, and S. Virchow and increasing for S. Bareilly. S. Braenderup mainly caused gastroenteritis in children; in contrast, S. Bareiley infected children and elderly people. Both serovars differed by XbaI-PFGE patterns. Almost all S. Bareilly isolates were susceptible to antibiotics of interest, while all lacked plasmids and belonged to one clone. Two distinct major clones in S. Braenderup were cluster A, mainly including MDR isolates with large MDR plasmid from North Taiwan, and cluster B, mainly containing susceptible isolates without R plasmid from South Taiwan. In cluster A, there were two types of conjugative R plasmids with sizes ranging from 75 to 130 kb. Type 1 plasmids consisted of replicons F1A/F1B, blaTEM, IS26, and a class 1 integron with the genes dfrA12-orfF-aadA2-qacEΔ1-sulI. Type 2 plasmids belonged to incompatibility group IncI, contained tnpA-blaCMY-2-blc-sugE genetic structures and lacked both IS26 and class 1 integrons. Although type 2 plasmids showed higher conjugation capability, type 1 plasmids were the predominant plasmid. Conclusions Serogroups B, C1, C2-C3, D1, and E1 of Salmonella caused over 95% of human salmonellosis. Two prevalent serovars within serogroup C1, S. Bareilly and cluster B of S. Braenderup, were clonal and drug-susceptible. However, cluster A of S. Braenderup was MDR and probably derived from susceptible isolates by acquiring one of two distinct conjugative R plasmids.

Conclusions: Serogroups B, C1, C2-C3, D1, and E1 of Salmonella caused over 95% of human salmonellosis. Two prevalent serovars within serogroup C1, S. Bareilly and cluster B of S. Braenderup, were clonal and drug-susceptible. However, cluster A of S. Braenderup was MDR and probably derived from susceptible isolates by acquiring one of two distinct conjugative R plasmids.

Background
Non-typhoidal Salmonellae are major zoonotic pathogens that commonly cause salmonellosis outbreaks. Globally, salmonellosis caused by non-typhoidal salmonellae generally results in about 1.3 billion cases of acute gastroenteritis and 3 million deaths annually [1]. In the United States, Salmonellae cause an estimated 1.4 million cases of salmonellosis and over 500 deaths annually [2]. Multidrug resistant (MDR) Salmonella, the global spread of which is mediated by international food trade and travel, is a global public health issue [3,4]. Often, clonal spread of MDR strains has been observed in particular serovars [4][5][6]. In most instances, resistance genes often associated with integrons and/or transposons are clustered within antimicrobial resistance islands that can be horizontally transferred by conjugative or mobilization plasmids [7].
In serogroup C1, S. Bareilly and S. Braenderup are closely related according to molecular analysis [8,9]. Both serovars have been highly susceptible to antimicrobials since 1971 [10,11] and are frequently isolated from feces of people with food-borne salmonellosis all over the world [12][13][14][15][16]. However, prevalence of both serovars differs between hosts and regions. In Denmark, S. Bareilly was isolated from diverse sources, including humans, animals and animal feed, while S. Braenderup was only found in humans [17]. In a study of a broiler-raising plant in the USA, S. Bareilly was often found in broilers and finished feed; however, S. Braenderup was only observed in hatcheries [18]. In addition, S. Braenderup was commonly isolated from cattle and turtles in Sweden [19], pigs [12] and chicken egg shells [20] in USA. These findings imply that animal reservoirs may be important sources of both serovars in human disease.
In this study, prevalent serogroups and serovars were determined for 8,931 Salmonella isolates collected from 2004 and 2007 in Taiwan. Because of the genetic similarity between S. Bareilly and S. Braenderup [8,9], the two serovars were compared with respect to antimicrobial resistance, resistance genes, PFGE and plasmid profiles. Both serovars disseminated clonally and varied in antimicrobial resistance patterns.

PFGE phylogenetic analysis
The clustering analysis of XbaI-digested PFGE patterns demonstrated genetic differences between S. Braenderup and S. Bareilly and within each serovar ( Figure 1). In S.
Braenderup, all isolates were separated into 2 clusters (I and II) at S = 0.68. Most isolates belonged to cluster I, which was further separated into two subgroups (A and B) at S = 0.84 ( Figure 1A). In cluster A, 19 isolates were separated into 9 PFGE patterns, and 78.9% (15/19) of the isolates were from northern Taiwan ( Figure 1A). In cluster B, 25 isolates were grouped into 4 PFGE patterns, and 72% (18/25) of the isolates were from southern Taiwan ( Figure 1A). S. Bareilly isolates were highly genetically homogenous and shared more than 90% pattern similarity ( Figure 1B).

Antimicrobial resistance profiles
Among six traditional antibiotics tested, S. Braenderup and S. Bareilly isolates were almost all susceptible to chloramphenicol (CHL; 6.7% for S. Braenderup vs 0% for S. Bareilly) and kanamycin (KAN; 4.4% vs 0%) and differed significantly in resistance to ampicillin (AMP, 37.7% for S. Braenderup vs 0% for S. Bareilly), nalidixic acid (NAL; 0% vs 15.7%), streptomycin (STR, 37.7% vs 15.7%), and tetracycline (TET; 33.3% vs 0%) ( Figure 1). Additionally,  Dendrograms were constructed by PFGE-XbaI patterns to determine the genotypes for S. Braenderup (A) and S. Bareilly (B) with corresponding information including the number and size of plasmids, PFGE subtypes, antimicrobial resistance patterns and collection location of each isolate Figure 1 Dendrograms were constructed by PFGE-XbaI patterns to determine the genotypes for S. Braenderup (A) and S. Bareilly (B) with corresponding information including the number and size of plasmids, PFGE subtypes, antimicrobial resistance patterns and collection location of each isolate. The dendrograms were generated by the unweighted pair group method with arithmetic mean (UPGMA) using the Dice-predicted similarity value of two patterns. The BioNumerics version 4.5 statistics program was used with settings of 1.0% optimization and 0.7% tolerance. Symbols of black square and white square represent resistant and susceptible respectively. Plasmids were separated into four groups by size. Ex, 1, 1, 1, 3 indicates that this strain harbored 6 plasmids, one is >90 kb, one is from >50 to <90 kb, one is from >6.6 to <50 kb, and three are <6.6 kb. nine resistance patterns were determined, ranging from susceptibility to all antimicrobials to resistance to four antimicrobials. In S. Braenderup, 7 resistance patterns (S, R2, R4 to R8) were found, and significant differences were observed between cluster A (patterns R2, R4-R8) and B (patterns S and R2) for AMP (77.3% vs 0%), STR (63.6% vs 13%) and TET (54.5% vs 13%). In addition, most isolates in cluster A were MDR (73.7%) while most isolates in cluster B were susceptible (84%). In cluster A, pattern R6 (AMP, TET, and STR) was the predominant and was found in four genotypes (A3, A5, A6, and A7). In S. Bareilly, most isolates were either susceptible (S pattern; 52.9%) or resistant to one (pattern R1 and R2; 31.4% and 9.8%, respectively) or two (pattern R3; 5.9%) antimicrobials. NAL resistant isolates were found in S. Bareilly (patterns R2 and R3) but not in S. Braenderup. Since there were susceptible to levofloxacin (LEV) and moxifloxacin (MOX), NAL resistance may result from a mutation in the gyrA gene, which encodes a subunit of the enzyme DNA gyrase.

Characterization of MDR plasmids
The prevalence of plasmid profile determined by plasmid number and size differed between these two serovars.  Figure 1). Plasmids larger than ca.75 kb were only found in resistance isolates of cluster A with the R4 to R8 patterns. Cluster B S. Braenderup isolates and S. Bareilly isolates carried smaller plasmids with the size smaller than 6.6 kb or lacked plasmids. Larger plasmids were further identified as R plasmids by analysis of the antimicrobial resistance profiles of E. coli pir116 transformants, and assigned to type 1 and 2 based on HindIII-restriction patterns (Table 3, Figure 2). Further conjugation, antibiotic resistance and PCR characterization of incompatibility and oriT types, mobile element IS26, class 1 integron, and AMP resistance genes bla TEM and bla CMY-2 were performed for these two plasmid types. Type 1 plasmids were separated into 7 subtypes (1a ~1g) based on differences in plasmid size ranging from 99.1 kb to 137.4 kb and restriction pattern. All plasmids carried bla TEM , replicons F1A and F1B, IS26, and a class 1 integron (Additional files 1 and 2: Figure S1 and S2) with a gene cluster of dfrA12-orfF-aadA2-qacEΔ1-sulI, conferring resistance to trimethoprimsulfamethoxazole (Sxt) and disappearing in plasmid 1 g (Table 3), which apparently coincides with that in the plasmid of S. Typhimurium (Accession number AB365868). The size of R plasmid was associated with antimicrobial resistance and conjugation capability (  (Figure 3). In contrast to a 1.1-kb PCR product in the largest 1a plasmid, 1b, 1d, and 1e plasmids lacked any PCR products; 1e and 1g plasmids presented 3.1 kb PCR products; and 1c plasmid yielded two PCR products with sizes of 3.1 kb and 0.7 kb. These results suggest that the number of IS26 and/or distance between two IS26 elements differed among these type 1 plasmids. In contrast to type 1 plasmids, type 2 plasmids were much smaller in size (77.5 kb and 85 kb) and had higher conjugation efficiencies, ranging from 8.41 × 10 -2 to 1.28 × 10 -1 (Table 3). In addition, type 2 plasmids were the IncI1 plasmid and contained oriT as well as tnpA-bla CMY-2 -blc-sugE (Table 3, Additional files 3: Figure S3).

Discussion
Human salmonellosis was limited to five Salmonella serogroups: B, C1, C2-C3, D1, and E1 (Table 1) S. Typhimurium, S. Stanley, and S. Enteritidis of serogroup D1 being the three most prevalent serovars overall. Although the prevalence of serogroups C1 and C2-C3 were similar (around 11%), 4 prevalent serovars and 2 main serovars were found in serogroup C1 and serogroup C2-C3, respectively. In the present study, a shift in prevalence was observed in these four prevalent serogroup C1 serovars: a rapidly decrease in the prevalence of S. Choleresuis, mainly due to enhancement of sanitation and control of swine in Taiwan, and an increase in prevalence of S. Bareilly and other serovars ( Contrary to earlier reports that S. Bareilly and S. Braenderup are closely related genetically [8,9], resistant to 10 Salmonella bacteriophages [22], and infect immunocompromised patients, differences between S. Braenderup and S. Bareilly were found in the prevalence trend from 2004 to 2007 (Table 1), patients' age group (Table 2), and plasmid profile as well as antimicrobial resistance groups and XbaI-PFGE patterns ( Figure 1A). In addition to genetic differences between these two serovars, differences in animal hosts were also observed in both serovars based on the geographic regions from which they were isolated [13,17,18,23]. In this study, we found that S. Bareilley isolates were highly homogeneous genetically and that S. Braenderup isolates were much diverse in our PFGE and plasmid analysis (Figure 1). This may explain why S.
Unlike MDR S. Choleraesuis isolated from pigs and humans [5,6], S. Braenderup and S. Bareilly isolated from pigs were highly susceptible to antibiotics in 1971 [10]. In addition, in a study of resistance to 11 antibiotics for Salmonella isolated from turtles, S. Bareilly was still susceptible to all antibiotics, and, in contrast, few S. Braenderup isolates were resistant to gentamycin (6/15), sulfisoxazole (6/15) and TET (2/15) [11]. In our study, almost all of the cluster A isolates of S. Braenderup were MDR and associated with large MDR plasmids (Table 3, Figure 1). Although RFLP analysis separated type 1 plasmids into 7 subtypes, based on antimicrobial resistance encoded by these plasmids, 3 subtypes were observed, conferring resistance to AMP and Sxt (1b-1e and 1g), AMP, CHL, Sxt, and TET (1f) and AMP, CHL, KAN, Sxt and TET (1a), respectively (Table 3). Apparently, the dfrA12-orfF-aadA2-qacEΔ1-sulI region of class 1 integrons, which is frequently found in MDR Salmonella [26][27][28], was located on MDR plasmid and conferred resistance to Sxt (Table 3). Insertion sequence IS26 existed in all (Table 3) and differed from plasmids in S. Braenderburg found in Spain [29]. The size change in type 1 plasmids may be due to presence of multiple IS26 elements that may be involved in plasmid rearrangement (Figure 3).
Although conjugation capability of type 2 plasmids was higher than that of type 1 plasmids, we only identified coexistence of type 1 and 2 plasmids in three S. Braenderup isolates, which differed in isolation day and PFGE pattern (Table 3). Isolate 13 with type 1f and 2a plasmids was collected in July of 2004 from Taipei. Isolate 32 with type 1d and 2a plasmids and isolate 36 with 1c and 2b plasmids were collected in March and May of 2005, respectively, from Taichung (Table 3). Only one isolate 44 with a type 1d plasmid was collected before those three isolates, in June of 2004 from Taichung. These results suggest possibly that isolate 32 with A6 genotype and R6 resistance pattern may be derived from isolate 44 with a type 1 plasmid, A4 genotype and R6 resistance pattern by introduction of a type 2 plasmid. Interestingly, type 2 plasmids are IncI1 plasmids, carrying the tnpAbla CMY-2 -blc-sugE structure (Table 3). AmpC β-lactamases are broadly distributed among the Enteribacteriaceae, and plasmid-mediated AmpC β-lactamases include ACC, ACT, CFE, CMY, DHA, FOX, LAT, MIR, and MOX [30]. At least three transposase associated genetic structures for bla CMY include ISEcp1-bla CMY-2 -blc-sugE, ISCR1-bla CMY-9 -yqgF-yqgE and IS26-frdC-frdD-ampR-bla CMY-13 -blc-sugE-IS26 [30].
Recently, bla CMY has been shown to be broadly spread in Salmonella worldwide [29,31,32] and to be present in S. Braenderup [33]. In Taiwan, since we reported the tnpA-bla CMY-2 -blc-sugE structure in S. Choleraesuis in 2004 [34], this transposon-like element has been found in other Salmonella serovars and Enterobacteriaceae [32]. In the present study, we first reported that S. Braenderup harbors tnpAbla CMY-2 -blc-sugE on a type 2 plasmid. Comparing this plasmid with the 138-kb plasmid pSC138 (accession no. NC_006856) of S. Choleraesuis, both are IncI1 plasmids with the tnpA-bla CMY-2 -blc-sugE structure. However, type 2 plasmids were conjugative and much smaller in size due to lack of a 60-kb DNA region with multiple integrons and transposons, which carry MDR genes [35][36][37].

Conclusions
Over 95% cases of human salmonellosis surveyed in this study were caused by 5

Pulsed-field gel electrophoresis (PFGE)
The PulseNet Standardized Laboratory PFGE Protocol for Molecular Subtyping of Echerichia coli O157:H7, nontyphoidal Salmonella serotypes, and Shigella sonnei [40] was used for analysis of the Salmonella isolates: 10 U of XbaI were used for the restriction digestion. PFGE images were analyzed by using the fingerprint analysis software BioNumerics version 4.5 (Applied Maths). A unique PFGE pattern was defined as one or two DNA bands differing between PFGE patterns of two isolates. A dendrogram was generated by the unweighted pairgroup method with arithmetic mean (UPGMA) algorithm using the Dice-predicted similarity value of two Xbal-digested PFGE patterns.

Plasmid profile analysis
Plasmid profiles of each isolate were determined by the Kado and Liu method [41], and plasmid size was esti-mated by comparison with the plasmids of two S. Choleraesuis strains: OU7085 (50 kb and 6.6 kb) and OU7526 (50 kb and 90 kb).

Restriction fragment length polymorphism (RFLP) and antibiotic susceptibility analysis of the MDR-plasmid
Large plasmids (> 50 kb) of 17 AMP and STR-resistant S. Braenderup isolates were characterized. Plasmid DNA was purified from resistant wild-type isolates by the alkaline lysis method [42] and then transformed into the competent E. coli strain pir116 (STR R ), which was prepared by the CaCl 2 method. Transformants were selectively grown on LB agar plates supplemented with AMP (100 μg/ml) and further tested for resistance to CHL, TET, and KAN, but not for resistance to STR, since the recipient strain was inherently resistant to streptomycin. The antibiotic resistance genes bla TEM , aadA, and bla CMY-2 , class 1 integron as well as the insertion sequence IS26 and its related DNA fragments were amplified using the primers listed in Table  4. The genes bla SHV and bla CTX-M3 and M14 were also detected by the multiplex method [43]. The R-plasmids of each transformant were purified by use of the Geneaid Plasmid Midi Kit (Geneaid, Taiwan) and were digested with Hin-dIII (New England Biolabs, USA) to determine similarity. Plasmid DNA fragments were separated by electrophoresis through a 0.6 % SeaKem GTG agarose gel (Cambrex Bio Science Rockland, Inc., Rockland, ME, USA) at 25 V for 16 h. The PCR product of class 1 integron was purified by DNA Clean/Extraction kit (GeneMark, Taiwan) and sequenced by Mission Biotech co. (Taiwan).

Plasmid conjugation and incompatibility group
Transferability of R plasmids from each RFLP group was determined by performing the conjugation test following a previously described method [44] with NAL-resistant S. Typhimurium LBNP4417 as the recipient strain. Briefly, 0.6 ml of overnight culture of donor strain was mixed with 1 ml of the overnight recipient strain. Then 1 ml fresh LB broth was added, and the mixture was incubated at 37°C with shaking at 100 rpm for 4 h. The bacterial solution was diluted at 10 1 , 10 3 , and 10 5 times with LB broth, and then 100 μl of the diluted solution was plated on MacConkey agar supplemented with AMP (100 μg/ml) and/or NAL (15 μg/ml). Conjugation efficiency was calculated by determining the number of transconjugants relative to the total number of recipients. Four primer sets were used to amplify the oriT regions of the ColE1, F (IncFI), R100 (IncFII), and pSC138 (IncI1-like) plasmids (Table 1). In addition, replicon types of these resistant plasmids were determined as described by Carattoli et al. [45].

Statistical analysis
The difference in the antimicrobial resistance rates between two serovars was analyzed by the independent t test. P values of < 0.05 were considered significant.