Imported human brucellosis in Belgium: Bio and molecular typing of bacterial isolates, 1996-2015

Objectives The aim of this study was to characterize by classical biotyping and Multi-Locus variable number tandem repeats (VNTR) Analysis (MLVA) all Brucella spp. derived from human cases in Belgium from 1996 to 2015. Final goals were to determine the species and biovar, to trace-back on genetic grounds the origin of each strain when patient history and risk factors were missing, and to survey for particular trends at the national level. Methods A total of 37 Brucella strains, isolated from 37 patients in Belgium, were analyzed by both classical biotyping and MLVA, and the genetic patterns compared to those of human strains isolated worldwide. Results Classical biotyping revealed that isolates were mainly Brucella melitensis. Most of them belonged to biovar 3, the most abundant biovar in the Mediterranean region. MLVA confirmed that Brucella melitensis is too diverse in VNTRs to be able to make clusters associated to each biovar, but it allowed retrieving precious epidemiological information. The analysis highlighted the imported nature of the strains from all over the world with a dominant part from the Mediterranean countries. Findings of the MLVA11 testing were in line with the travel history of patients coming from Italy, Turkey, Lebanon and Peru. The analysis was particularly useful because it suggested the geographical origin of the infection for 12/16 patients for whom no case history was available. Conclusion Classical biotyping and MLVA analysis are not exclusive but remain complementary tools for Brucella melitensis strain surveillance. MLVA11 is sufficient for Brucella-free countries such as Belgium to trace the geographical origin of infection, but complete MLVA16 is needed to search for links with endemic areas.

Human brucellosis is a mandatory notifiable disease in Belgium. Despite the Brucella-free status and the continuous monitoring, human brucellosis cases are reported yearly [9]. Applied case definition for human brucellosis is that recommended by the European Centre for Disease Prevention and Control [10]. Briefly, a confirmed case will meet both clinical (fever and one other generic symptoms of brucellosis) and laboratory criteria (isolation of the strain from clinical specimens and/or presence of Brucella specific antibodies in serum detected by agglutination tests or ELISA). In case of isolation, the identification of the bacterial species and its biovar is essential for epidemiological follow-up and control of the disease. Classical biotyping is the gold standard to investigate phenotypic characteristics [11]; this method becomes uncertain in case of appearance of mutants within the same species or biovar or the presence of strains with atypical reactions to microbiological tests [12]. Molecular genotyping methods assessing polymorphisms in the omp2A and omp2b genes [13] or by the sequencing of rpoB for example are helpful to confirm the species and the biovars of Brucella [14]. It is known that some human cases can emerge in ObmF countries with the source of contamination retrieved in infected endemic wildlife [15]. In this context, deeper molecular epidemiological tools become particularly interesting to trace-back the infection to identify the geographical source. To fulfill the objective of sub-species discrimination between strains, Variable Number Tandem Repeats (VNTR) has been investigated in Multi-Locus VNTR Analysis (MLVA) by various scientific groups since 2003 [6][16] [17]. This method has proved its efficacy to obtain epidemiological information on Brucella strains, access to public databases and sharing of information between countries. In this study, MLVA was used to characterize Belgian human isolates collected during the national reference activity from 1996 to 2015.

Bacterial isolates
A total of 37 Brucella strains were isolated from 37 patients in Belgium between 1996 and 2015. They represent all human Brucella isolates collected during this period in Belgium. These strains were isolated from blood or cerebrospinal fluid at the first line laboratory or at the National Reference Centre for Brucellosis, Veterinary and Agrochemical Research Centre (CODA-CERVA) in Brussels. Brucella melitensis 16M was used as reference strain to calibrate the VNTR units.

Biotyping
All isolates were identified as Brucella species on the basis of morphology and conventional microbiological procedures according to the OIE manual [18]: requirement of CO 2 and O 2 for growth, urease activity, H 2 S production, sensitivity to thionin (10 and 20 μg/ml), fuchsin (20 μg/ml) and saphranin (100 μg/ml) dyes and agglutination with monospecific antiserum for A and M antigens. Brucella monospecific antisera A and M were obtained from FAO/WHO Collaborating Centre for Brucellosis Reference and Research at the Veterinary Laboratory Agency, Weybridge, UK.

DNA preparation, PCR and MLVA genotyping
A loopful of cultured bacterial cells were dissolved in water, heat treated at 99˚C for 15 min and, after centrifugation, the supernatant was used as DNA template. PCR amplification was performed in a total volume of 25 μL containing 20 ng of DNA, 1x PCR reaction Buffer (Invitrogen), bethain 5M (Sigma), 1U of Taq DNA polymerase rec (Invitrogen), 50 mM of MgCl 2 , 5 mM of each dNTPs and 10 μM of each MLVA locus flanking primers [6]. The following PCR program with the thermocycler iCycler BioRad was used: an initial denaturating step at 96˚C during 5 min followed by 30 cycles of 96˚C for 30 sec, 60˚C for 30 sec, 70˚C for 1 min and a final extension step of 70˚C for 5 min.
For the markers bruce06, bruce11, bruce42, and bruce55 with repeat unit size of 134 bp, 63 bp, 125 bp and 40bp respectively, the PCR fragment size was evaluated by 2% agarose gel electrophoresis. DNA from the reference strain B. melitensis 16 M, for which repeats lengths are known, was used for standardization. A 100-bp ladder (Invitrogen) was used as molecular size marker. Ethidium bromide-stained gels were visualized by UV light and photographed with GeneGenius bio-imaging system (Syngene). For the markers bruce08, bruce12, bruce43, bruce45, bruce18, bruce19, bruce21, bruce04, bruce07, bruce09, bruce16 and bruce30, the PCR products length were defined by capillary electrophoresis with the CEQ 8000 Genetic Analysis System (Beckman Coulter, Indianapolis, IN, USA). The size of each PCR product was then converted to a corresponding tandem repeat number as described by Le Fleche et al. [6].

Analysis of MLVA data
All data were analyzed using BioNumerics version 6.6 software (Applied Maths, Belgium). Clustering analysis was performed using categorical coefficient and the unweighted-pair group method with arithmetic mean algorithm (UPGMA) as indicated previously [17]. Briefly, three distinct character data sets with different weight were defined according to the diversity index of the markers and combined using the composite data set tool provided by Bionumerics. The first one corresponded to panel 1 markers. Each marker of this panel got an individual weight of 2 (total weight for panel 1: 16). The two others form two groups in the panel 2, 2A and 2B. Panel 2A markers got a weight of 1 (total weight for panel 2A: 3) and panel 2B markers got a weight of 0,2 (total weight of panel 2B: 1). The MLVA profile of the isolates was also subjected to a minimum spanning tree (MST) analysis, illustrating the diversity existing within the clusters based on single locus variations (SLV). Units (and not sizes) of each marker were considered for the analysis. The Hunter-Gaston Discrimination Index (HGDI) was calculated for each locus by the use of the online tool V-DICE (http://www.hpa-bioinformatics. org.uk/cgi-bin/DICI/DICI.pl, latest access February 25, 2017).

Results
A total of 37 Brucella strains with an average of roughly two strains annually were isolated in Belgium during the reference activity between 1996 and 2015. Colony morphology, staining, growth characteristics, and slide agglutination with monospecific anti-Brucella sera were used to characterize all isolates (Table 1). These standard bacteriological procedures classified the bacterium at the species and the biovar levels. Globally, biotyping identified 28 strains as B. melitensis biovar 3, three strains as B. melitensis biovar 1, three as B. melitensis biovar 2, one as B. melitensis rough, and two as B. abortus biovar 1. The rough strain could not be typed (strain 16) and two smooth B. melitensis biovar 3 strains showed an atypical susceptibility to dyes and a delayed agglutination with anti-M serum (strain 32 and 37). Susceptibility to phages was tested in the latter strains which confirmed the identification of atypical B. melitensis biovar 3 strains.
All isolates were further characterized by MLVA ( Table 2). The 16 VNTRs were split into three panels (1, 2A and 2B) [17] because of the difference in their diversity index. Three MLVA analyses were run depending on the panels considered (MLVA8 = panel 1, MLVA11 = panel 1 + panel 2A, MLVA16 = panel 1 + panel 2A + panel 2B). A comparison of the Belgian isolates with the strains present in the public repository by MLVA8 and MLVA11 [19] allowed the identification of the associated genotypes (Table 2). Fourteen genotypes were identified with respect to the MLVA8; five were new and received numbering as "184-to 188". These genotypes were all single-locus variants of known genotypes. Nineteen genotypes were identified with MLVA11; eight as new, numbered "361-to 368", seven as single-locus variants and one as a two-locus variant.
MLVA11 data were used to draw a minimum spanning tree (MST) of Brucella melitensis strains comprising the human Belgian isolates and the human strains isolated worldwide [19] (Fig 1). Three main clusters were derived, corresponding to the origin of the isolates: an East Mediterranean, a West Mediterranean and an American cluster. Belgian human strains did not cluster within a homogenous cloud, indicating the diversity and suggesting that these isolates were imported. A more detailed link with the geographical origin of some Belgian isolates could be established (Fig 1 and Table 3). The strains isolated from patients infected in Italy clustered separately from the other strains, and demonstrated genotypes described for Italian strains (genotype 51 for MLVA8 and genotypes 91 and 96 for MLVA11). The strains from patients with a history of contamination in Turkey had genotypes comparable to those     In certain contexts, MLVA16 can be used to cluster the strains regarding to the biovars, [6] [20]. For B. melitensis this seems to be not effective [17] [21]. Despite the use of three levels of analysis (MLVA8/MLVA11/MLVA16), it was not possible to cluster the Belgian B. melitensis strains according to their biovar. In some cases the three described biovars (bv1; bv2 and bv3) were found within the same cluster (S1 Fig). The HDGI for the different MLVA16 loci were calculated and confirmed the high discriminatory power of MLVA (S1 Table).

Discussion
Brucellosis is a worldwide zoonotic disease that affects various domestic and wild animals but some Brucella species are occasionally able to infect humans, considered as accidental hosts. The transmission occurs by aerosol, consumption of raw dairy products or direct contact with contaminated animals. Human brucellosis is characterized by a febrile debilitating disease that can become a chronic infection accompanied with serious complications if not adequately treated. The impact of the infection on health, the mode of transmission and the zoonotic aspect of the disease have driven European countries to impose a compulsory declaration for each confirmed case to the reference centres. In Belgium, like in other OmbF and OBF countries, human brucellosis is limited to few cases every year and its incidence is low. The majority of the human cases diagnosed in Belgium are due to travel in endemic countries and to immigrations because of the absence of animal control measures in many parts of the world. Epidemiological surveillance in Brucella-free countries remains therefore essential. In Belgium, some Brucella species (B.suis, B. ceti and B. pinnipedialis) reside in wildlife and it was recently shown that clustered outbreaks in domestic animals have occurred. Brucella species are genetically close to each other [22] and, despite their host preference, some species are able to cross the host specificity barrier [23] [24]. This zoonotic characteristic inevitably justifies passive surveillance of brucellosis in humans. Classical biotyping is the gold standard for species and biovar attribution of the isolates but the information lacks resolution. In some cases, appearance of mutants that generate genetic diversity is responsible for atypical susceptibility to dyes or delayed agglutination with anti-Brucella sera and thus hinders an accurate identification of the species and biovar [12]. Molecular genotyping based on single gene or multilocus sequence typing to identify locus polymorphisms or other mutations in genes have also been developed to understand the phylogeny of the genus Brucella and try to address the question of species and biovars [25] [26]. These methods are not exhaustive when sub-typing for epidemiological trace-back is necessary. Since 2002, when the first complete genomic sequences of Brucella melitensis and suis [27][28] were released, a number of tools have been developed to differentiate the isolates according to DNA repetitive sequences. A first application of microsatellites fingerprinting to Brucella strains was made in 2003 [16] and was based on nine VNTRs that allowed discrimination of some isolates beyond the biovar level [29]. Afterwards, other sets of VNTR markers with a different speed of evolution were identified that demonstrated higher discriminatory power [6] [30]. The group of Al Dahouk modified the set of VNTR markers of Le Fleche and showed their ability to cluster Brucella melitensis strains in 3 main geographical groups. In our study, the complete VNTR panel for Brucella was analyzed in strains isolated from humans in Belgium from 1996 to 2015 to improve their characterization and evaluate their origin. When compared with biotyping, we could assess the ability of VNTR in the identification of the biovars of Brucella. Previous studies showed that MLVA was able to discriminate biovars of Brucella species such as B. suis and B. abortus [6]. However, it was also shown that the technique has its limitations with exceptions [31] and that it is not possible to discriminate biovars of the B. melitensis species [17]. Our study confirmed that the high genetic diversity regarding MLVA in this species did not allow clustering it according to the biovars. Therefore, classical biotyping and MLVA characterization remain complementary tools for Brucella strain surveillance. VNTRs were interesting to understand the epidemiology of the isolated strains. The study of MLVA11 was effective to establish phylogeographic relations of the Belgian strains with strains present in public databases. For instance, the single strain belonging to the Americas cluster was isolated in a patient that travelled to Peru. It shared the same genotype as 62 strains coming from Peru described by various groups [32] [33]. Six isolates displayed the exact MLVA11 genotypes of dozens of strains coming from Italy described previously [6][12] [17]. While the Italian origin for 5 of these Belgian isolates was known, thanks to the databases, it is suspected that the 6 th patient also contracted the disease in this country. These 6 strains belonged to the West Mediterranean cluster. A last isolate belonging to this cluster was a new genotype based on a single locus variant with Italian strains and was isolated from a patient with a travel history in Spain. The East Mediterranean cluster contained the majority of the strains isolated in Belgium. Nine of them shared the same MLVA11 genotype (genotype 43) with dozens of strains coming from Turkey as previously described [17] [34]. Few strains with the same MLVA11 genotype come from Syria and Irak [17]; this genotype was also observed in strains coming from Lebanon [35]. Seven other strains isolated in Belgium shared the same genotype with a great panel of strains isolated in China [21] and Turkey [17][34] (genotype 42). Information about the genotypes present in countries geographically located between Turkey and China is missing but this genotype 42 seems to be highly represented in the Middle East [17] [35] and has already been isolated in countries like Kazakhstan [17] [36]. We hypothesize that this particularly stable genotype [21] is largely spread from Turkey to China and that MLVA11 is not variable enough to distinguish strains from this region of the world. Our hypothesis is sustained by an MST analysis performed on 446 B. melitensis MLVA15 profiles of strains belonging to the East Mediterranean group [37]. In this study, part of the Turkish genotypes clustered with most of the genotypes observed in China and Kazakhstan, suggesting a common origin between strains from the Asian continent and the East Mediterranean area [37].
When examined with MLVA16 which includes panel 2B composed of five unstable satellites with a high mutation rate, Belgian isolates were found highly diverse; 33 genotypes were distinguished among the 37 isolated strains. They apparently came from all around the Mediterranean region except one strain with a genotype clustering together with strains from Peru. MLVA16 did not add to the information obtained already by MLVA11 that strains isolated in Belgium were coming from a large panel of areas in the world. It seems that increasing the number of markers is particularly useful in case of outbreak investigation in endemic countries. For highly genetic diverse strains such as in the context of imported cases, MLVA11 is in general sufficient to trace the geographical origin of infection. Accessibility of MLVA16 information in public repositories about human strains isolated in Belgium (and worldwide) could help endemic countries to better investigate a possible link and the spread with their cases.
Classical biotyping revealed that the human cases in Belgium are mainly Brucella melitensis, a species never isolated in animals in Belgium. This is in agreement with the fact that B. melitensis is much more common than B.abortus in many countries of Eastern Europe and the Middle East [38]. More precisely, most of the strains belonged to the biovar 3. This is consistent with the fact that the majority of patients travelled to or originated from Mediterranean countries and the Middle East where this biovar is predominant [11]. Since 1996 only two human isolates in Belgium revealed to be B. abortus, both belonging to biovar 1. One of the two patients was coming from Ecuador. To our knowledge, B. abortus is the only species isolated in humans in this country [39]. We do not have any information about the second patient but it is worthwhile to note that Brucella abortus biovar 1 was not enzootic in Belgium during the years 1980-2000. It has only been isolated in the late nineties from cattle imported from Portugal where this biovar was prevalent. Taken together, this information convincingly indicate that the majority of human brucellosis cases in Belgium is associated with travel to Brucella endemic countries as observed in neighbouring countries [40].
Our study is the first characterization of human Brucella isolates in Belgium. It highlights the genetic diversity of the strains consistent with the imported nature of the infections and the use of MLVA as an important tool to improve the understanding of the phylogeography of Brucella.