Introduction

Brucellosis is a worldwide zoonotic disease, with an estimated 500 000 new cases annually1. The disease is transmitted to humans by direct contact with infected animals or consumption of infected animal products, in particular unpasteurized milk products2,3. Several species of the genus Brucella can be pathogenic for humans, the most frequently encountered being Brucella melitensis4,5. The infection may present with an acute or insidious onset, with continued, intermittent or irregular febrile illness, fatigue, anorexia, weight loss, headache, arthralgia or generalized aching. The condition may progress to chronic disease with severe complications5. Brucellosis requires correct and protracted antimicrobial treatment6,7. Susceptibility testing is strongly recommended, as several studies have described resistance for commonly used antimicrobial drugs, such as rifampicin and trimethoprim/sulfamethoxazole8,9,10,11,12,13,14. However, resistance testing is not always performed, due to challenges concerning personnel biosafety. Brucella spp. is easily aerosolized and poses a high risk of laboratory acquired infections. A specialized BSL3 laboratory is therefore necessary5,15,16. In Norway, treatment of brucellosis follows WHO guidelines5 and includes a combination of doxycycline and gentamicin or alternatively doxycycline and rifampicin for children over eight years and adults. For younger children and pregnant women, doxycycline is replaced by trimethoprim/sulfamethoxazole17. As no data on resistance profiles from Brucella isolates in Norway currently exist, a relevant question is: are current empirical guidelines for treatment adequate?

Bovine brucellosis was eradicated from Norway in 1953 and brucellosis in sheep, goats and pigs has never been detected18. The infection is sporadically detected in humans, with 0–4 cases reported annually the last 20 years19. Most of the patients have been infected abroad, and commonly brucellosis is detected in migrants or travellers returning from endemic risk areas. Brucellosis is considered to be a re-emerging disease20, and travelling and globalization give reason for increased awareness.

In non-endemic countries as Norway, it is important to assess the epidemiological background of rare imported diseases such as brucellosis. The purpose of this study was to trace the origin of Brucella spp. infecting Norwegian patients in the period 1999 to 2016 using WGS based methods and comparison with publicly available B. melitensis sequences. Molecular epidemiological studies provide information about genetic grounds and origin for bacterial isolates, but such trace-back studies can be challenging as the Brucella bacteria are genetically quite conserved21. Multilocus variable tandem-repeat typing analysis (MLVA) has been the method of choice for epidemiological fingerprinting of Brucella spp15,22. This method is limited to specifically targeted regions in the genome. Whole-genome sequencing (WGS) is a more powerful tool for accurate typing of Brucella spp., as the entire genome of the bacteria can be studied and thereby increase the discriminatory power23. In addition, we have assessed antimicrobial resistance mechanisms in all isolates. This is the first study in Norway where human brucellosis is studied in relation to geographical source and antimicrobial resistance pattern.

Results

Sequencing and phylogeny

In all, 23 clinical isolates and two reference strains (B. melitensis biovar 1 strain 16 M (ATCC 23456) and B. abortus B19 (vaccine strain)) were sequenced (Table 1). The average coverage depth and coverage breadth (nucleotide positions covered with at least 5 reads) are summarized in supplementary information. A phylogenetic tree was made to visualize the relationship between the different isolates of B. melitensis (Fig. 1). In addition, 13 previously published genome sequences of B. melitensis available from NCBI24 were included in this tree. The average genome coverage following alignment using Parsnp was 96.8% (individual values varying between 96.4% and 96.9%). RAxML was used to construct the phylogenetic tree from this alignment. Bootstrap values were included to show the strength of the analysis, and for most branches the bootstrap value was greater than 95/100, showing robustness. Two bifurcations in the tree have a bootstrap quality below 50/100. The B. melitensis 16 M reference sequence (isolate #22) grouped together with known 16 M genomes, as expected, and B. abortus B19 vaccine strain (#21) is the obvious outgroup compared to the other isolates.

Table 1 Brucella melitensis isolates.
Full size table
Figure 1
figure 1

Whole-genome SNP-based phylogenetic tree illustrates origin of human brucellosis in Norway linked to Middle-East, Asia and Africa. A phylogenetic tree based on WGS analysis showing the relationship between 23 clinical isolates of Brucella melitensis isolated from patients in Norway compared to selected existing B. melitensis genomes available in the NCBI Genome Assembly and Annotation. Colours indicate bootstraps values. The Norwegian isolates are marked with dots. B. abortus vaccine strain B19 is included as outgroup.

Full size image

WGS gave a good resolution, separating all isolates but two; isolates #36 and #37 (Fig. 1). There is virtually no distance between them in the phylogenetic tree, indicating a common source. The isolates originated from family members that developed brucellosis following a common travel, which explains the similarity. Isolates #40 and #41 are also closely related in the tree, and one might assume a common source. However, epidemiological data showed two different countries of origin, Israel and Norway. As infection in Norway is not likely, this discrepancy might be due to incorrect collection of epidemiological data.

When compared with selected isolates of B. melitensis available from NCBI24, the isolates from our dataset clustered in two previously described lineages; the East Mediterranean (EM) clade (genotype II) and the African (Af) clade (genotype III)23,25. Isolates from patients infected in Iraq, Israel, Turkey, Afghanistan, Georgia and Norway clustered in the EM clade25. One of these isolates (# 8) was reported to the Norwegian Surveillance System for Communicable Diseases19 as a case of laboratory infection. Unfortunately, the original isolate was not available for comparative analysis. The isolates in the EM clade further divided into two sub-clusters (A and B) (Fig. 1). In sub-cluster A, published genomes from Asia (China, India and Russia)26,27,28,29,30 clustered together with analysed isolates originating from; a patient infected in Afghanistan, one in Georgia, one with unknown country of infection and one reported as infected in Norway. Sub-cluster B included isolates from patients infected in Iraq, Turkey, Israel and Norway together with the published genome of a isolate originating from a patient from Syria (NCBI 2010724553)31.

The sequenced isolates from the two patients reported as infected in Ethiopia and the three patients infected in Somalia clustered in the African (Af) clade (genotype III) (Fig. 1). Additionally, one sequence originated from a patient with unknown country of infection. A published genome from an isolate originating from Somalia representing this clade in a previous publication (BwIM_SOM-36a)23, clustered closely together with the isolates from this dataset. Isolates matching the West Mediterranean clade (genotype I), the European clade (genotype IV) or the American clade (genotype V)23,25, were not detected in our dataset.

Antimicrobial susceptibility

The obtained MIC values for all tested antibiotics are shown in Tables 2 and 3. All isolates were susceptible for doxycycline, streptomycin and gentamicin. The obtained MIC values for imipenem, chloramphenicol and amoxicillin clavulanic acid are described in Tables 2 and 3. By broth microdilution method, 17 out of the 23 isolates (74%) were intermediate resistant for rifampicin (MIC = 2 µg/mL) and 6 isolates (26%) were resistant (MIC = 4 µg/mL). In parallel, isolates were also tested by gradient strip method for rifampicin10,14,32, and 4 isolates (17%) were intermediate resistant (MIC = 2 µg/mL) while the remaining 19 isolates (83%) were susceptible (MIC = 0.25–1 µg/mL). All isolates were susceptible for trimethoprim-sulfamethoxazole by gradient strip testing, (MIC = 0.032–0.064 µg/mL).

Table 2 Results of AST testing of Brucella melitensis.
Full size table
Table 3 Distribution of MIC values among Brucella melitensis isolates tested.
Full size table

SNP analysis in antibiotic resistance determining genes

The gene and protein sequences for seven known antibiotic resistance-associated genes (rpoB, folA, folP, gyrA, gyrB, parC, parE)33 were analysed for SNP variants comparing the sequenced isolates to the reference strain B. melitensis 16 M (#22) (Table 4). Three different SNP variants were detected in rpoB, a gene associated with rifampicin resistance33,34. The mutations detected in the rpoB gene in our data were located at nucleic acid position 1174 [392-Glu (GAG)◊Asp (GAC)], 1185 [629-Ala (GCG)◊Val (GTG)] and 2953 [985-Ala [GCC)◊Val (GTC)]. These alterations are different from mutations previously described as a cause of rifampicin resistance in Brucella34. Additionally, the SNP alterations were not restricted to the four isolates phenotypically intermediate resistant to rifampicin based on gradient strip testing (#43, 45, 46 and 47). The SNP changes therefore does not seem associated with rifampicin resistance. The observed SNP variants were however observed to be specific for certain lineages and sub-clusters based on WGS analysis; the SNP in position 1174 was detected in two related isolates within the Af clade (#24 and #46), SNP in position 1185 was common for all isolates in the EM clade, and the SNP in position 2953 was restricted to the isolates in sub-cluster A in the EM clade (Fig. 1, Table 4).

Table 4 Mutations detected in known antibiotic resistance determining genes relative to B. melitensis 16 M33.
Full size table

One mutation was detected in folA, a structural gene coding for dihydrofolate reductase and described to be involved in resistance mechanisms for trimethoprim33. The detected SNP at position 73 [217-Arg (CGG)◊Leu (CTG)] was present in all isolates belonging to the Af clade in the current dataset. In folP, the gene coding for dihydrofolate synthase and associated with sulfamethoxazole resistance33, one SNP difference was detected compared to the reference strain at position 631 [211-Phe (TTC)◊Leu (CTC)]. This SNP was detected in all isolates in the EM clade. However, as no T/S resistance was detected by phenotypical analysis, the relevance of these findings is uncertain.

Changes in the genes gyrA, gyrB, parC and parE were also detected (Table 4). These genes are all known as fluoroquinolone-resistance determining genes33, coding for DNA gyrase and DNA topoisomerase respectively33. The described mutation in gyrA did not correspond with mutations related to fluoroqinolone resistance described earlier35. The SNP detected in gyrA in position 1759 [599-Leu (CTG)◊Val (GTG)], was detected in all isolates in the EM clade, but was also detected in one isolate in the Af clade (#1). The mutation in the gyrB gene was detected only in the reference strain B. abortus B19 (#21), and not in any B. melitensis isolates. Four SNP differences were detected in parC, of which two was only present in B. abortus (#21), and one SNP difference was detected in parE. The SNP difference in parE in position 27 [79-Asn (AAC)◊Ser (AGC)] was present in all isolates in the Af clade, and the SNP at position 799 [267-Arg (CGC)◊His (CAC)] in parC was present in a sub-cluster in the Af clade. These SNP differences has to not been previously described.

Discussion

This is the first investigation of brucellosis in Norwegian patients. We sequenced all clinical isolates collected between 1999 and 2016. The results provide valuable information on origin and transmission routes, and support the use of whole-genome sequencing for trace-back analyses of B. melitensis. Whole-genome sequencing is becoming a useful tool for molecular epidemiological studies of infectious diseases, providing better resolution compared to other techniques, like MLVA23. WGS provides the possibility to elucidate the geographical origin of an isolate or an outbreak, given no or unclear information on the probable country of infection for the patient. As the use of sequencing will increase, more genomes will be publicly available, adding further information about phylogeny and molecular epidemiology of brucellosis.

Our data demonstrates that human brucellosis in Norway is linked to travelling and migration from the Middle East, Asia or Africa, and our findings show that travel history and genetic epidemiological sequence data harmonize well. The majority of the isolates clustered in the East Mediterranean clade (genotype II). This is in accordance with results from Germany, Denmark and Sweden, and reflects similar travelling and migration patterns in northern Europe15,23,36. The East Mediterranean clade further divided into two sub-clusters (A and B), separating isolates with origin in the Middle-East from isolates originating from the Caucasus region and Central Asia. These findings are in line with data from other studies23,25. The isolates from patients infected in Ethiopia and Somalia clustered in the African clade (genotype III), as expected. Our findings resembles data based on MLVA analysis from Sweden15.

The five isolates reported from patients infected in Norway clustered in the East Mediterranean clade. One of them was a laboratory infection, while the origin of the remaining four isolates remains unclear. This might be due to clinical misreporting to the Norwegian Surveillance System for Communicable Diseases (MSIS)19, but infection from imported food cannot be ruled out. This raises the question if such sources, especially unpasteurized dairy products, might be a risk factor of consideration. Although the number of reported cases in Norway is low, the incidence of brucellosis in Northern Europe is increasing23,36, hence there is a reason to be alert and have an increased awareness about this disease.

In this study we have also analysed phenotypical resistance patterns and compared the results with WGS data, both to evaluate phenotypic test results and to investigate possible genetic markers that could predict resistance. Analysis of such genetic markers might provide a much needed alternative to phenotypic testing, simplifying the strict laboratory biosafety requirements necessary. To date, there are only few studies comparing phenotypical susceptibility testing with genetic sequencing results. However, these studies have demonstrated discrepancies between phenotypical in vitro susceptibility testing and sequencing37,38,39. Specific mutations in the gene coding for the β subunit of the DNA dependent RNA polymerase, rpoB, enable resistance for rifampicin in Brucella spp. as well as in other bacteria33. WGS data from our study did not reveal any mutations in the two previously described rpoB “hot spot” regions related to resistance34. Resistant or intermediate resistant phenotypes for rifampicin were however detected; by the broth microdilution method where all the 23 tested isolates showed resistant or intermediate resistant profiles, and in lesser extent by the gradient strip method where four isolates were reported as intermediate resistant. Our findings question if the recommended broth microdilution method by the Clinical and Laboratory Standards Institute40, and to some extent the gradient strip method, might overestimate in vitro rifampicin resistance in B. melitensis. This topic needs to be further addressed in larger multicentre studies.

Marianelli et al. has suggested that sequencing of the rpoB gene could be used a s a tool for genotyping and species determination of Brucella spp.41. Our results support this, as mutations in the rpoB gene were detected in several of our isolates. However, we could not find any correlation between these mutations and the phenotypical resistance test results. Noteworthy, the observed mutations distributed consistently within certain clades or sub-clusters in the dendrogram. Similar findings have been previously described (DNA position 1185 and 2953)23,41,42, while the SNP in DNA position 1174 from two isolates in the African clade is a novel finding.

In our cohort, all tested isolates were phenotypical susceptible to trimethoprim-sulfamethoxazole. By WGS analysis, one SNP change was detected in several isolates in folP, the gene related to sulfamethoxazole resistance. Additionally, six of the isolates had a mutation in the folA gene, previously described to be associated with trimethoprim resistance33. The SNP in folP was specific for the East Mediterranean clade, while the mutation in folA was detected in all isolates in the African clade. As these isolates were phenotypical susceptible, the association with resistance seems to be negligible, and the relevance of this finding needs to be further addressed.

Based on our results, occurrence of antibiotic resistance in B. melitensis is very low, and hence, current Norwegian empirical guidelines for treatment seem valid. Although the number of isolates in this study is limited, they originate from different continents and reflect a broad geographical area as well as a long time span. The results must be interpret with care, but our data question the reported challenge of rifampicin resistance described in some studies. There is a need for an updated and standardized methodology for resistance testing of Brucella spp., for establishing breakpoints for additional antibiotics and a need for more research into mechanisms determining resistance in Brucella spp. and possible genetic markers for resistance.

Materials and Methods

Study design and isolates

In this study, all available isolates from culture positive brucellosis patients registered in Norway in the period 1999 to 2016 were included. The Norwegian Institute of Public Health (NIPH) has the national reference function for Brucella spp. in Norway, and collects all the available clinical isolates. Patient data was obtained from the Norwegian Surveillance System for Communicable Diseases19. A total of 23 clinical isolates were included (Table 1). In addition, two reference strains, B. melitensis biovar 1 strain 16 M (ATCC 23456) and B. abortus B19 (vaccine strain), were sequenced and included for analysis. All procedures involving infectious agents were performed according to national regulations in a certified biosafety level 3 laboratory.

Culture and identification

All isolates were cultured from freezing stock on Colombia blood agar plates for 48 hours at 37 °C in an enriched atmosphere before further processing. All isolates had previously been confirmed as Brucella spp. by real time PCR detecting the Brucella specific gene bcsp3143 and by MALDI-TOF MS analysis44. Isolates were confirmed as Brucella melitensis by Bruceladder multiplex PCR45,46,47. Inactivation was performed with NucliSENS® lysis buffer (Biomerioux, Durham, NC, USA).

DNA isolation and whole-genome sequencing

Genomic DNA was extracted by QiaAmp DNA minikit (Qiagen, Hilden, Germany). Concentration of genomic DNA was quantified using Qubit® dsDNA BR Assay kit and the Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Illumina sequencing libraries were generated with the High Throughput Library Preparation Kit (KAPA Biosystems, Wilmington, Massachusetts, USA) following the manufacturer’s protocol. Individual libraries were indexed with NEXTflex barcodes (Bio [SIC] Scientific, Austin, Texas, USA). Sequencing was performed using an Illumina MiSeq platform with 2 × 150-bp paired-end reads (Illumina, San Diego, CA, USA)48.

Genome assembly and analysis

Before assembly, the sequence data was trimmed using Trimmomatic version 0.35 to remove the adapter sequences and trim away low quality ends49. The trimmed sequence reads were then mapped to a reference genome using BWA-MEM (version 0.7.15-r1140)50. B. melitensis biovar 1 strain 16 M with chromosome I (NC_003317) and chromosome 2 (NC_003318) (NCBI Genbank®) was used as reference genome for the mapping. The output of this reference based assembly was processed using SAMtools (version 1.3.1, using htslib 1.3.1), BCFtools (version 1.2, using htslib 1.2.1)51 and Picard Tools (version 2.7.0) (http://broadinstitute.github.io/picard/). First, read mate errors were fixed using SAMtools fixmate, and duplicate reads were marked with Picard Tools. A pileup file was made using SAMtools mpileup with BAQ-recalculation enabled, to avoid indel artifacts. Using BCFtools, the variants (SNPs and indels) were extracted from the pileup file and consensus sequences were created for each of the 23 sequenced isolates and for the two reference strains. The average coverage depth and coverage breadth (nucleotide positions covered with at least 5 reads) were calculated using SAMtools.

These consensus sequences were annotated using Prokka (version 1.11)52. The gene and protein sequences for seven antibiotic resistance-associated genes (folA, folP, gyrA, gyrB, parC, parE, rpoB) were extracted from the GenBank files created by Prokka to explore possible genetic alterations linked to antibiotic resistance and compared to B. melitensis 16 M and other published genomes.

Phylogenetic analysis

A phylogenetic tree was created using the 25 consensus sequences resulting from the reference-based assembly, including the sequenced genomes for the reference strains B. melitensis 16 M and B. abortus vaccine strain B19. In addition, selected existing B. melitensis genomes available in the NCBI Genome Assembly and Annotation report24 were included in the analysis. Assemblies with completion level ‘complete’ or ‘chromosome’ were selected. In total, 13 genomes were selected. All 38 genomes were aligned to each other using Parsnp (version 1.2) from the Harvest suite53. HarvestTools (version 1.2) was used to convert the Parsnp output to a multiple sequence alignment in FASTA format, which was then converted to Phylip format. The resulting multiple sequence alignment was used as the basis for creating a phylogenetic tree. To decide which substitution model to use when building the phylogenetic tree, the function phymltest of R package ‘ape’ was used. This function tests multiple substitution models (using PhyML) and calculates their Akaike Information Criterion (AIC). This showed that the most complex substitution model, GTR + Γ + I, resulted in the lowest AIC. The GTR + Γ + I model has thus been selected for building the phylogenetic tree. A maximum likelihood tree was built using RAxML (version 8.2.9)54,55 with the substitution model GTR + Γ + I. The Newick tree generated by Parsnp was used as the starting tree. To test the robustness of this tree, a bootstrap analysis was performed with replications set to 100.

Antimicrobial susceptibility testing

All isolates were tested for antibiotic resistance to rifampicin (RIF), doxycycline (DOX), streptomycin (STR), gentamicin (GEN), imipenem (IMP), chloramphenicol (CMP) and amoxicillin/clavulanic acid (AMC) by broth microdilution testing according to Clinical and Laboratory Standards Institute (CLSI) guidelines40. The testing was performed in accordance with the working group for AST testing on highly pathogenic bacteria in the EU joint action “Efficient response to highly dangerous and emerging pathogens at EU level” (EMERGE) (http://www.emerge.rki.eu/Emerge/EN/Home/Homepage_node.html). All isolates were suspended in saline water to a 0.5 McF turbidity, and suspended in in Brucella broth adjusted to pH 7.1 ± 0.1 (BD Diagnostic Systems, Hunt Valley, MD, USA). Microdilution plates (Micronaut-AST IMB gramnegative MIC plates from Merlin Diagnostika GmbH, Berlin, Germany) were inoculated and incubated at 35 ± 2 °C with 5% CO2 for 48 h. Quality control assays were performed with Escherichia coli ATCC #25922 and Streptococcus pneumoniae ATCC #49619. Phenotypical resistance to fluoroquinolones was not evaluated in this study.

Rifampicin (RIF) and trimethoprim-sulfamethoxazole (T/S) resistance was additionally tested with gradient strip method. Susceptibility for T/S is only reported from gradient strip method, as testing in broth microdilution is known to generate false resistance due to high concentrations of thymidine in the Brucella broth56. Gradient strip method was performed using Liofilchem® MIC Test Strips (Liofilchem s. r. l., Roseto degli Abruzzi, Italy) as described elsewhere10,14,32. Briefly, a suspension of bacteria adjusted to 0.5 McF units was inoculated on Mueller-Hinton plates supplemented with 5% sheep blood and the gradient strips applied. The plates were incubated at 35 °C ± 2 °C with 5% CO2 for 48 h before reading. MIC values were interpret in accordance with CLSI guidelines for potential agents of bioterrorism40. The following breakpoints for susceptibilities were used: GEN ≤ 4, STR ≤ 16, DOX ≤ 1 and SXT ≤ 2/38. For RIF, CLSI-interpretation of Haemophilus influenzae (fastidious bacteria) was used; S ≤ 1, I = 2, R ≥ 440. Susceptibility breakpoints for IMP, CMP and AMC have not been established for Brucella spp.

Ethical approval

The study was performed in accordance with relevant guidelines and regulations and no experiments were performed on humans and/or human tissue samples. The study was approved by the Regional Committee for Medical Research Ethics in South Eastern Norway (REK Sør-Øst) (reference no 2016/2036).

Accession codes

The genomes included in the present study are available in the European Nucleotide archive (ENA) database (accession no. PRJEB23353)57.