Introduction

Non-typhoidal salmonellae (NTS) are commonly acquired from contaminated food products and are responsible for numerous foodborne infections worldwide. Although 2659 serovars have been identified worldwide [1], Salmonella enterica serovar Enteritidis (S. Enteritidis) is responsible for the majority of reported cases of salmonellosis [2]. Poultry and poultry products are the main reservoir and most of the strains are often isolated from broiler meat, laying hens, and breeders [2]. In Greece, S. Enteritidis ranked first during the period 2007–2012 among human isolates, while the notification rate in 2012 was among the lowest in 27 European Union member States (≤5 per 100,000) [2–4]. Most NTS infections result in gastroenteritis, but, sometimes, severe invasive infections require antibiotic treatment, such as fluoroquinolones (ciprofloxacin) and third-generation cephalosporins (the drug of choice for children) [5]. Resistance to these drugs have been widely reported for the past two decades [6, 7].

Several molecular methods, such as pulsed-field gel electrophoresis (PFGE) coupled with phage typing or multiple locus variable number of tandem repeats analysis (MLVA), are often used where further strain discrimination is required [8–10], mainly during outbreaks investigations. PulseNet Europe has set up a standardized protocol to ensure that PFGE fingerprints from various laboratories are reproducible and comparable [11]. Additionally, multilocus sequence typing (MLST) is used for epidemiological purposes in order to study the population structure of Salmonella spp. [12].

The aim of the present work was to study the epidemiology of S. Enteritidis in Greece during a 3-year period. Strains isolated from patients, animals, and food products were characterized by (i) antimicrobial susceptibility testing and (ii) genotyping using PFGE. A subset of strains was studied using MLST.

Methods

A total of 88 isolates isolated in Greece between 2008 and 2010 during the official control programs for Salmonella were serotyped as S. Enteritidis. Sixteen isolates were from food of Greek origin (poultry products), originating from six regions, and 72 isolates were from poultry, representing 57 farms and 17 regions. During the same period, 1039 human isolates were serotyped by the National Reference Centre for Salmonella (NRCS) as S. Enteritidis. Eighty of them were randomly selected for this study [3]. Identification of the Salmonella isolates was carried out at the NRCS by serotyping in accordance to the White–Kauffmann–Le Minor scheme [13, 14]. Susceptibility to a panel of 20 antimicrobials was determined using the agar dilution method in Iso-Sensitest agar (Oxoid, Basingstoke, UK). The final plate concentrations (mg/L) were: ampicillin (A) 8 and 128; chloramphenicol (C) 8; colomycin (Co) 8; gentamicin (G) 4; kanamycin (K) 16; neomycin (Ne) 8; streptomycin (S) 16 and 128; spectinomycin (Sp) 64; sulfonamide (Su) 64; tetracycline (T) 8 and 128; trimethoprim (Tm) 2; ciprofloxacin (Cp) 0.125 and 1; nalidixic acid (Nx) 16; furazolidone (Fu) 8; amikacin (Ak) 4; cephalexin (Cx) 16; cephradine (Cr) 16; cefuroxime (Cf) 16; ceftriaxone (Cn) 1; cefotaxime (Ct) 1. Multidrug resistance was defined as previously proposed [15]. PFGE was performed after the digestion of genomic DNA with XbaI and BlnI (Takara, Tokyo, Japan), according to the PulseNet protocol, as previously described [16], with S. enterica serovar Braenderup (H9812) as a DNA size marker. All profiles following XbaI digestion were submitted to the PulseNet Europe database for official PulseNet profile designation; BlnI and composite XbaI–BlnI profile designations were not official. MLST was performed by sequencing seven core genes, as previously described [12]. Data from MLST were submitted to the Salmonella enterica MLST Database (http://mlst.warwick.ac.uk/mlst/dbs/Senterica) in order to define the sequence types (STs). Statistical comparison between resistance rates among isolates was analyzed by the Chi-square test (Microsoft Excel 2007). Fingerprints were analyzed using BioNumerics v6.1 (Applied Maths, Sint-Martens-Latem, Belgium). Dendrograms were constructed using the Dice similarity coefficient and the unweighted pair group method with arithmetic mean (UPGMA), optimization, and position tolerance set at 1.5%. The composite analysis was based on equal weighting of the XbaI and BlnI data and UPGMA clustering.

Results

Based on the antimicrobials tested, high resistance levels to nalidixic acid were observed among human and poultry isolates (25 %), and lower resistance levels (p < 0.05) among food isolates (6.25 %). Resistance to ampicillin was observed in two isolates, and resistance to streptomycin, trimethoprim, kanamycin, and spectinomycin was observed in one isolate each. Fifty-nine out of 80 (73,7 %) human isolates, 54 out of 72 (75 %) poultry isolates, and 15 out of 16 (93.7 %) food isolates were pan-susceptible to the 20 antimicrobials tested. Resistance to ciprofloxacin was not observed.

PFGE analysis identified 39 unique profiles using XbaI macrorestriction endonuclease. Submission to the PulseNet Europe database showed that 109 out of 168 isolates (64.9 %) were represented by the eight PulseNet-identified profiles SENTXB.0001, SENTXB.0002, SENTXB.0004, SENTXB.0005, SENTXB.0007, SENTXB.0023, SENTXB.0010, and SENTXB.0040 (Fig. 1). The remaining 32 profiles were represented by one to seven isolates, but did not correspond to PulseNet profiles. The most prevalent PFGE profile was SHADXB.0001 (55/168, 32.7 %). This profile was identified among human, poultry, and food isolates, and was present in every year from 2008 to 2010. The second highest occurring profile was SENTXB.0005 (29/168, 17.3 %), and was also identified during each year from all sample categories (Table 1). PFGE using BlnI restriction endonuclease revealed 48 distinct BlnI pulsotypes among the 168 strains. The predominant BlnI profile was BlnI-0007 and was present in 60 out of 168 (35.7 %) profiles (two from food, 26 from poultry, and 32 from humans). The second most dominant profile was BlnI-0010, with 21 (12.5 %) strains (four from foods, nine from poultry, and eight from humans). The remaining 46 BlnI profiles corresponded to one to nine isolates. The composite dataset divided strains into 80 distinct XbaI–BlnI profiles. The most common patterns were XbaI–BlnI-0028 (13.1 %), XbaI–BlnI-0026 (10.1 %), and XbaI–BlnI-0039 (7.1 %). All the remaining patterns were represented by one to ten isolates (Fig. 1).

Fig. 1
figure 1

Dendrogram of XbaI, BlnI, and composite XbaI–BlnI pulsotypes and characteristics of the 168 Salmonella enterica serovar Enteritidis (S. Enteritidis) isolates examined in this study

Full size image
Table 1 Origin, year, and proportion of PulseNet Europe profiles in Salmonella enterica serovar Enteritidis (S. Enteritidis) isolates, Greece, 2008–2010 (n = 168)
Full size table

A subset of eight strains were chosen for MLST; five of them had been isolated from humans and three from poultry. These belonged to eight different XbaI pulsotypes, with similarity values ranging from 77 to 95 %. Five of them were identical to five PulseNet Europe profiles and three did not correspond to any of them. All eight isolates belonged to ST11 (allelic profile 5-2-3-7-6-6-11). ST11 includes the majority of S. Enteritidis isolates uploaded to the Salmonella enterica MLST Database (Fig. 2).

Fig. 2
figure 2

PulseNet Europe pulsed-field gel electrophoresis (PFGE) types, XbaI–BlnI PFGE types, and corresponding multilocus sequence typing (MLST) types of S. Enteritidis isolates

Full size image

Discussion

To our knowledge, this is the first study that describes the population structure of S. Enteritidis based on PFGE types submitted to the PulseNet Europe database and drug resistance surveillance of this serovar in Greece. Even though the isolates cover the period 2008–2010, we believe that the general condition has not changed substantially during the last several years and the main conclusions are still an ongoing issue.

In general, S. Enteritidis is a serovar with low resistance rates and, thus, multidrug resistance is rarely observed [17]. Our data showed resistance to nalidixic acid mainly among human and poultry isolates (~25 %). There were variations in resistance to nalidixic acid among human isolates observed in European countries for 2012. High rates were reported from Spain (60.3 %), Romania (37 %), and France (31.7 %), whilst low rates were reported in Germany (1 %) and Hungary (2.3 %) [18]. Poultry population data from 2012 in the European Union reported resistance to nalidixic acid ranging from 0 % in Austria to 39.3 % in Poland and 92.6 % in Portugal [18]. Resistance to nalidixic acid in salmonellae is of great concern, as it has been reported to be an indicator for reduced susceptibility to the drug of choice, ciprofloxacin, resulting in possible treatment failure [19]. Consequently, the 25 % of nalidixic acid-resistant S. Enteritidis in this study must be considered as possibly having reduced susceptibility to ciprofloxacin, and, thus, potentially treatment-resistant, even though our results indicated no resistance/partial resistance to ciprofloxacin. The lack of or partial resistance to ciprofloxacin in this study could be explained as resistance to ciprofloxacin was studied for the breakpoints of 1 and 0.125 mg/L and not 0.064 mg/L, which is the current European Union clinical breakpoint [20]. This observation differed in some other European Union countries, where most S. Enteritidis strains, especially PT4, PT1, and PT6, were susceptible to nalidixic acid and other antimicrobial agents. It must be stressed that, in 2012, Greece had the highest levels of antimicrobial consumption in primary care in Europe [21]; this could explain the 25 % of resistance to nalidixic acid among humans in our study.

SENTXB.0001 was, according to our results, the most common PFGE profile in Greece, comprising 32.7 % of the isolates. This profile is also dominant in Europe; 62.5 % (n = 2210) in England and Wales during 2001–2012 and 56.4 % (n = 6734) of human isolates from other European countries (Austria, Denmark, Spain, Finland, Scotland, England and Wales, Italy, Netherlands, and Germany) (data not shown). The second most prevalent profile, SENTXB.0005 (17.3 %), is also common in Europe; 3.75 % (n = 133) in the UK and 9.3 % (n = 1111) in Europe. An interesting observation is that the third most common profile, SENTXB.0002, in the UK with 13 % incidence (n = 459) and also in Europe with 15.7 % incidence (n = 1880), is relatively rare in Greece, where the rate is only 3.6 %. It is also to be noted that SENTXB.00014, the fourth highest occurring profile in the UK and in Europe (~2.7 %), was absent in Greece, and profile SENTXB.0007 was rare; 0.6 % (n = 1) in Greece, 2 % (n = 71) in the UK, and 1.7 % (n = 205) in Europe. In general, our data suggest mainly similarities but also some differences among Greek and European populations of S. Enteritidis. This diversity may suggest that strains originated from multiple reservoirs and, also, non-appearance of point-source outbreaks. The absence of some human S. Enteritidis PFGE patterns, in particular from food and food animals, suggests that sources other than food animals, such as fresh foods or vegetables, may play a role in clinical illness. Furthermore, some differences among occurring profiles could also be explained by the PulseNet database structure, which is biased towards isolates from outbreak investigations and, consequently, not accurately representative. However, as a strong correlation between SENTXB.0001 and PT4, PT1, and PT6 exists [10], future work involving phage typing could be undertaken in order to further define S. Enteritidis populations in Greece.

Due to the clonal nature of S. Enteritidis [22], the use of two to three additional restriction enzymes for PFGE or combinations of multiple methods like phage typing and MLVA have been proposed when enhanced strain discrimination is needed [9, 10, 23]. BlnI is mainly recommended as the second enzyme for discriminating epidemiologically unrelated strains, as well as partly discriminating XbaI-indistinguishable profiles [8, 24, 25]. Τhis last observation has also been noticed in our study, with 48 different BlnI pulsotypes being generated. The predominant BlnI pulsotype was SENTBL.0007, which was seen in 35.7 % of all isolates. The composite dataset using XbaI and BlnI revealed that the SENTXB.0001–SENTBL.0007 clone was predominant, with an occurrence rate of 12 % (n = 21). Seventeen out of these 21 strains shared the same R-type (nalidixic acid) and were isolated during each year from humans, animals, and foods. This clone was present in Greece throughout the study period and had spread through the food chain among eight different regions in Greece.

MLST has been suggested to be a good discriminatory tool for typing S. Enteritidis [26]. Our results do not support this suggestion, as all the isolates belonged to ST11, which is the most common sequence type for S. Enteritidis [12, 26]. The conserved DNA sequence of the internal fragments of the seven housekeeping genes tested in this study explains the lack of diversity among the strains studied.

The last few years has seen a revolution in the application of whole genome sequencing (WGS) for routine microbiology, as it is replacing current conventional phenotypic and molecular methods [27, 28]. WGS is a promising method that represents the opportunity for a step-change in diagnostic microbiological practice, as it can be used not only for research, but also as an alternative typing method for public health surveillance and outbreak investigation in routine practice [29]. Newer, higher resolution typing methods based on single nucleotide polymorphisms (SNPs) can be/have been developed to discriminate this closely related S. Enteritidis serovar [30]. One of the drawbacks of the current study was that, due to the lack of resources, we were not able to apply WGS methods to conduct the surveillance, but the PFGE and susceptibility data generated showed enough diversity to ensure that future S. Enteritis surveillance should involve both newer WGS-based typing methods and good epidemiological and clinical data to detect sources of infections. This, in turn, will allow the evaluation of control measures to be carried out to reduce resistance in Salmonella, specifically targeting the prudent use of antimicrobial agents by farmers and veterinarians.