Differential phenotyping of Brucella species using a newly developed semi-automated metabolic system

Classical phenotyping and metabolic markers of Brucella spp

Although Brucella is a monophyletic genus, apparent differences between its species do exist e.g. host specificity and pathogenicity. Nowadays, Brucella species and biovars are distinguished by a limited number of microbiological tests measuring quantitative or qualitative differences of dye bacteriostasis, hydrogen sulfide production, urea hydrolysis, carbon dioxide requirement, bacteriophage sensitivity and agglutinin absorption. For at least half a century these microbiological procedures have not changed, although various new Brucella species showing variable phenotypic traits have been detected and new diagnostic methods have been developed.

Neither the classical biochemical tests nor antigenic properties and phage-sensitivity can be considered a reliable guide to the identification of Brucella species. Contradictory results were often reported [14]. However, variations in H₂S production, CO₂ requirement, a change in dye tolerance or atypical surface antigens i.e. inconsistent A and M antigens usually do not affect the oxidative metabolic pattern of a strain [15, 16]. Metabolic activities have proven to be stable parameters allowing unambiguous species identification, particularly in strains which show conflicting identities by conventional determinative methods [14, 17–19]. In addition, differing metabolism may help to describe new species [6, 9, 20]. In our series, two strains isolated from foxes in Austria (strain no. 110 and 111) which displayed an atypical metabolic pattern could be identified.

Oxidative metabolic profiles remain qualitatively stable for long periods of time and usually show no change in characteristic patterns after in vivo and in vitro passages [21]. However, quantitative differences in the oxidative metabolic rate of monosaccharides have been observed after multiple passages in vitro [21]. Variants in the oxidative metabolic pattern found among different CFUs of the same strain have been described in varying frequencies depending on Brucella species and biovars [22]. In our experiments, B. suis bv 1 showed the highest intra-strain variability in its enzymatic activity (data not shown).

Despite the stability of the metabolic markers and their consecutive usefulness in diagnostic assays, studies describing the differences in the metabolism of Brucella spp. have not been conducted for decades as the classical laboratory techniques are labour-intensive and very demanding. Especially Warburg manometry which is carried out in a respirometer measuring oxygen uptake has been widely used to determine oxidative metabolic patterns in order to describe and differentiate species, biovars, and atypical strains of the genus Brucella. Formerly, manometric studies on the metabolic activity of brucellae helped to quantitatively define the species classified within the genus [23]. However, due to the demanding techniques applied only a restricted number of strains and reactions were tested and various substrates e.g. D-asparagine, L-proline, adonitol, fructose and glucose were regarded as not useful for species and biovar differentiation [23, 24]. In the comprehensive setting of this study most of these substrates also proved their usefulness.

Manometric studies have confirmed that a characteristic oxidative pattern for Brucella species exists whereas specific profiles for the biovars have not yet been described except for B. suis bv 1-4 [25]. Using the Micronaut™ system we were able to discriminate B. abortus bv 4, 5, and 7, B. suis bv 1-5, B. ovis, B. neotomae, B. pinnipedialis, B. ceti, B. microti and B. inopinata with a specificity of 100%. However, differentiation among the B. melitensis biovars was impossible as, according to their oxidative metabolic activity, they form a very homogenous group. The results of the cluster analysis based on our biotyping data (Figure 3) are in general concordance with the genotyping data acquired by Multiple Loci VNTR (Variable Number of Tandem Repeats) Analysis (MLVA) [26]. Neither biotyping nor genotyping proved a biovar specific clustering in B. melitensis strains [27]. Although we tested a substantial number of biochemical reactions we may have chosen the wrong set of substrates for the differentiation of B. melitensis strains, but the separation of this species in three biovars could also be somehow artificial.

Biotyping of Brucella spp. using commercially available assays

If biological traits such as enzymatic activities are tested all potential variables must be reduced to a minimum to avoid intra- and inter-assay variations which may occur in addition to minimal biological variations. Commercial test systems offer a large number of quality controls both in the production chain and under experimental conditions.

Commercially available microtiter plates coated with various substrates to characterize the metabolic pattern of bacteria have already been used to describe new species of the genus Brucella e.g. the Biolog™ system for B. ceti[28] and the Micronaut™ system for B. microti and B. inopinata[6, 9]. However, comprehensive metabolic studies including all currently known species and biovars are rare.

Using the Biolog™ GN MicroPlate system (Biolog, CA, USA) based on 44 differentially oxidized substrates, B. melitensis, B. abortus and B. suis isolates could be grouped into taxons identical with the presently recognized species [29]. However, only a restricted number of strains (n = 35) were tested and biovars were not differentiated. In a larger strain collection (n = 71) which included all biovars of the six classical Brucella species only 50% of the strains were correctly identified confirming the poor specificity of this commercially available, substrate mediated, tretrazolium identification technique [30]. López-Merino and colleagues used the Biotype 100™ carbon substrate assimilation system (bioMérieux, Marcy-L'Etoile, France) which comprises 99 carbohydrates, organic acids and other carbon substrates to discriminate B. melitensis, B. abortus, B. suis and B. canis[31]. Using the most discriminating carbon substrates i.e. D-glucose, D-trehalose, D-ribose, palatinose, L-fucose, L-malate, and DL-lactate more than 80% of the B. melitensis and B. abortus strains could be correctly identified. Similar to the Brucella specific Micronaut™ plate designed in this study B. suis and B. canis could not always be discriminated. The limited number of field isolates tested per species may have produced inconclusive results, particularly when only reference strains were available which are well known for atypical phenotypic traits. Future studies on larger strain collections may reveal more unique metabolic profiles suitable for species and biovar differentiation and also helpful to discriminate between B. suis bv 3 and B. canis. Nevertheless, the overall specificity for the identification of Brucella species using the Micronaut™ system reached 99%.

Experimental conditions potentially interfering with bacterial metabolism and influencing biotyping results

Many experimental parameters may influence the metabolic activity of bacteria. For instance, oxidative rates may decrease if Brucella is prepared from 48 hours rather than 24 hours cultures [25] because Brucella is able to adapt to starvation. This effect does not seem to be important in the Micronaut™ system since turbidity is measured reflecting bacterial growth within a period of 48 hours as an indirect parameter for substrate utilization. Consequently, the bacteria have plenty of time to switch on all necessary metabolic pathways. Hence, the metabolic rate of glutamic acid may differ between B. abortus and B. melitensis[32] but after 48 h the substrate is entirely metabolized by both species. For the same reason B. suis has been described as inactive in the metabolism of glutamic acid but our results revealed extensive utilization of this substrate at least for the biovars 3-5 whereas the metabolization was variable in the biovars 1 and 2.

The growth medium can also have an effect on the utilization of substrates and brucellae may operate with alternate metabolic pathways leading to discrepant stimulatory effects in different assays [30]. Therefore, a minimal medium i.e. buffered sodium chloride peptone (from potatoes) solution was used in Taxa Profile™ and Micronaut™ plates to avoid interference with other potential substrates in the culture medium.

The rates of oxidation of various compounds are also strongly dependent on intact bacterial membranes and pH values [33, 34]. In our experiments, asparagines were easily oxidized by most of the Brucella spp., but aspartic acid was not (exceptions were B. suis bv 4, B. microti, and B. inopinata). Furthermore, glutamic acid was oxidized, but intermediates in the pathway, such as α-ketoglutarate and succinate (except for B. microti and B. inopinata) were usually not. Lowering the pH of a reaction mixture containing intact cells of brucellae markedly increased the oxidation rate of these metabolites e.g. L-aspartate, α-ketoglutarate, succinate, fumarate, L-malate, oxaloacetate, pyruvate and acetate [34]. Differences between Brucella species may occur in the pH range at which the bacteria are able to utilize some of the substrates and therefore labile metabolic profiles can be observed [35]. Nevertheless, such reactions may be helpful for the differentiation of species and biovars if assay conditions are stable. The effect of extracellular adjustment of the pH upon intracellular enzymatic reactions can be explained by organic acids permeating the cell more readily when undissociated than when ionized. Hence, a pH change may overcome the permeability barrier for many substrates especially of the Krebs' cycle. For this reason our results do not easily reflect intracellular substrate utilization. In proteomic studies on intracellular brucellae and bacteria grown under stress conditions comparable to the intracellular niche of Brucella, enzymes of the TCA cycle i.e. the succinyl CoA synthetase and aconitate hydratase were found increased [36, 37]. In contrast, intermediates of the TCA cycle such as citrate, isocitrate, α-ketoglutarate, succinate, malate, fumarate were not generally metabolized in vitro or showed variable metabolization in the different species such as oxaloacetic acid.

Although modelling of the intracellular niche of brucellae is not a topic of this study the Micronaut™ system might be helpful to investigate differences in the metabolic activity between the species under various growth conditions. This will allow a much deeper insight in the metabolic changes of intracellular compared to extracellular brucellae and will help to understand survival strategies of the pathogen under starvation, microaerobic and acidic conditions. In this context, a negative correlation between metabolic activity and the relative degree of virulence was observed among B. abortus strains [38]. Avirulent mutants of B. melitensis, B. abortus and B. suis that failed to replicate or survive in macrophages or animal models often had mutations in the carbohydrate metabolism [39]. In our study, B. microti which is not known to be human pathogenic was the metabolically most active species.

Independent of the method used a broad agreement can be observed for the utilization of carbohydrates by Brucella spp. whereas the results of the amino acid metabolism are more variable [3, 16]. Differences in the oxidation rate of different isomers of the same amino acid have been described for short incubation periods, e.g. B. suis and B. melitensis are known to oxidize D-alanine more rapidly than the L-isomer [40] or B. abortus oxidized L-glutamic acid and L-asparagine rapidly whereas relatively slight activity was obtained with the D-isomers [38]. Differences in the metabolization rate could not be used for differentiation in our multi-substrate test. As many substrates were tested at the same time the incubation period was prolonged to 48 hours to ensure that each substrate was completely utilized. With a few exceptions, there are only minor differences in the general utilization of D- and L-isomers of amino acids within the same species [41]. Therefore both isomers of the same amino acid were only included three times in the Micronaut™ plate, i.e. D-/L-proline, D-/L-alanine, and D-/L-serine. In our experiments, opposing metabolic activity could be observed for the different isomers of proline in B. abortus bv 3, B. suis bv 2, and B. canis, for the isomers of alanine in B. canis and B. neotomae, and for the isomers of serine in B. suis bv 1, 2, and 4, B. ovis, B. microti and B. inopinata.

Further, substrate concentration may influence the metabolic activity of Brucella[34, 38]. Although sample volumes are different in Taxa Profile™ and Micronaut™ plates the final substrate concentration is the same. Hence, apparently contradictory results in these two test systems which could be observed in our study cannot be explained by different concentrations of the same compound.

Because of the small volumes used in the Taxa Profile™ plate turbidity could not be measured due to technical limitations. Therefore the indicator phenol red was added to colorimetrically measure respiration. In contrast, in the 96-well Micronaut™ plate turbidity as a measure of bacterial growth was determined. The measurement of respiration instead of growth is much more sensitive since bacteria may respond metabolically by respiring but not by growing [42]. Hence, this effect may have led to differing results for the utilization of the same substrate on the two platforms. However, respiration could not be used in the genus Brucella since some strains are dependent on CO₂ which catalyzes abiotic reduction of the dye.

As most metabolic pathways are encoded within the Brucella genome brucellae might present as fastidious due to slow growth. Although the genome sequence of B. microti is almost identical to that of B. suis with an overall sequence identity of 99.84% in aligned regions, phenotypically these species differ significantly which might be caused by variable gene regulations and different growth patterns [43].

Both respirometry and tetrazolium reduction assays proved that B. abortus is characteristically stimulated by L-alanine, L-asparagine and L-glutamate [30]. In contrast, the Micronaut™ results were heterogeneous for L-alanine in B. abortus strains. The differences in metabolic activity observed between these methods might be caused by the cut-off selected in our experiments. Deduced from the OD values measured with the Micronaut™ system three levels of substrate utilization could be defined: no/weak metabolic activity (-), moderate metabolic activity (+), and strong metabolic activity (++) [Additional file 7]. The different levels of oxidative metabolic activity on amino acid and carbohydrate substrates determined by Micronaut™ agreed with the oxygen uptake levels for most substrates measured by conventional manometric techniques [25]. However, owing to the dispersion of the individual OD values, quantitative differences are of limited practical relevance. The selection of cut-offs which delineated positive and negative metabolic activity greatly contributed to the clarification of the presentation of substrate utilization. Of course, the limit between two activity patterns is rather artificial.

Brucella strains

Various strains initially tested with the 384-well Taxa Profile™ plates were re-evaluated using the newly developed 96-well plate. In addition, a limited selection of closely related and clinically relevant bacteria was tested, i.e. Acinetobacter lwoffii (DSM 2403), Yersinia enterocolitica O:9 (IP-383 RKI/Paris), Ochrobactrum intermedium (CCUG 24964), O. anthropi (DSM 6882), Enterococcus faecalis (DSM 2570), Escherichia coli (DSM 1103), Pseudomonas aeruginosa (DSM 1117), and Staphylococcus aureus (DSM 2569).

Culture and sample preparation

All strains were grown on Brucella agar for 48 h at 37°C with or without 10% CO₂ depending on the needs of the particular species. Horse serum (10%) was added to the culture medium to facilitate the growth of B. ovis. Colony material was harvested and solubilised in 0.1% buffered sodium chloride peptone (from potatoes) solution and in sterile 0.9% NaCl for use in profile A or C plates and profile E plates, respectively. The turbidity of the bacterial suspension was adjusted to a 2.0 McFarland standard. Each well of the 384- and 96-well plates was inoculated with 25 μl and 100 μl of the respective preparation, respectively. The microtiter plates were incubated at 37°C for 48 h before reading.

Brucella phenotyping

The metabolic activity of Brucella was comprehensively assessed using the Taxa Profile™ system (Merlin Diagnostika, Bornheim-Hersel, Germany) based on 384-well microtiter plates coated with various substrates. The Taxa Profile™ A microtiter plate allows testing of 191 different amines, amides, amino acids, other organic acids and heterocyclic and aromatic substrates [Additional file 1]. The Taxa Profile™ C microtiter plate enables the analysis of 191 different mono-, di-, tri- and polysaccharides and sugar derivates [Additional file 2]. Using the Taxa Profile™ E microtiter plate another 188 substrates to determine enzymatic activity were tested: 95 amino peptidase- and protease-reactions, 76 glycosidase-, phosphatase- and other esterase-reactions, and 17 classic reactions [Additional file 3]. According to manufacturer's instructions supplementary reagents were added to some wells of the Taxa Profile™ E microtiter plates to visualize substrate utilization. In addition, the indicator phenol red was added to all wells of the Taxa Profile™ A and C microtiter plates to optimize detection. The blank value was measured for each biochemical reaction on the same plate and subtracted from measured values. In order to assess inter-assay variability five independent experiments per strain were conducted.

For evaluation of the newly developed Brucella specific 96-well microtiter plate three trials per strain were run independently. Intra-assay variability was assessed with the reference strains testing all substances twice within the same experiment. Since the blank values measured on extra plates proved to be constant a fixed mean value of each substrate was subtracted from the measured data.

Data acquisition and analysis

Turbidity and colour change were measured photometrically using a Multiskan Ascent^® photometer (Labsystems, Helsinki, Finland) at a wave length of 405 nm, 540 nm and 620 nm according to manufacturer's recommendations. Optimal OD cut-off values were empirically adapted from the preliminary test results of the 384-wells Taxa Profile™ microtiter plates.

Stable and discriminatory markers were selected to design a 96-well Micronaut™ plate (Figure 2) to identify bacteria of the genus Brucella and to classify their species and biovar.

Dendrograms were deduced from the biotyping data using SPSS version 12.0.2 (SPSS Inc., Chicago, IL, USA). First of all, three different character data sets were defined following the metabolic activity tested (Taxa Profile™ A ("amino acids"), C ("carbohydrates"), and E ("other enzymatic reactions")). Each character was considered as equal within the particular data set. Both the raw OD data and the binary coded data based on the empirically set cut-off were analyzed using the Pearson coefficient and the categorical coefficient, respectively. Hierarchical cluster analysis was performed by the Ward's linkage algorithm, and a dendrogram was generated. If necessary, analysis was repeated within each cluster for further discrimination. Secondly, a separate data analysis of the 23 Brucella reference strains representing the currently known species and biovars was performed including all biochemical reactions of the Taxa Profile™ system or exclusively the substrates selected for the newly developed plate. Finally, the whole collective of 113 strains tested with the Brucella specific Micronaut™ microtiter plate was analyzed to prove the diagnostic system. An identification table presenting quantitative and qualitative metabolic activity was created [Additional file 7] and the specificity of the test system to differentiate Brucella species and biovars was calculated (Table 1).