Comparative evaluation of the immunodominant proteins of Brucella abortus for the diagnosis of cattle brucellosis

Background and Aim: The present serodiagnosis of brucellosis in livestock is based on the whole cell or smooth lipopolysaccharide of the Brucella organism in which specificity is hampered by the cross-reactivity, especially with the antibodies against Yersinia enterocolitica O:9 organism. The problem can be addressed by screening for better immunodominant antigens. Hence, the present study was undertaken to screen protein antigens of Brucella abortus for their diagnostic potential in cattle brucellosis. Materials and Methods: Protein antigens of B. abortus (n=10) non-reactive to antibodies against Y. enterocolitica O:9 were selected, expressed in Escherichia coli, assessed the reactivity of expressed recombinant proteins by Western blot, standardized indirect-enzyme-linked immunosorbent assay (ELISA) for detecting Brucella antibodies in cattle serum, and comparative evaluation was done. Results: All the selected protein antigens were expressed and in the Western blot with Brucella antibodies positive cattle serum, six recombinant (Brucella protein 26 [BP26], Cu-Zn Superoxide dismutase [SodC], B. abortus I-1885, Serine protease, Bacterioferritin, and Brucella Lumazine Synthase [BLS]) proteins showed reaction whereas none of the proteins showed reactivity with Brucella negative cattle serum. ELISA has been done using known Brucella positive and negative cattle sera samples (n=113 each) in which the performance of recombinant proteins in diagnosing brucellosis was in the order of BP26 > BLS > SodC followed by rest of the proteins. BP26 based ELISA was found to be better with area under the curve as 0.953, and diagnostic sensitivity, diagnostic specificity, and Youden’s index of 90.27%, 95.58%, and 0.8584, respectively, with the excellent agreement (k=0.85). Conclusion: BP26 could be a potential diagnostic antigen among the immunodominant proteins of B. abortus in ruling out Y. enterocolitica O:9 infection while diagnosing brucellosis in cattle herds.

The disease is usually represented by abortion, reduced fertility, and reduced milk production in ruminants. The common species involved in causing brucellosis in cattle is B. abortus.
Since the discovery of the brucellosis, the conventional serological tests, namely, Rose Bengal plate test (RBPT), standard tube agglutination test (SAT), enzyme-linked immunosorbent assay (ELISA), and complement fixation tests which utilize the whole Brucella cell and/or its smooth lipopolysaccharide (S-LPS) fractions are being used in diagnosis [2]. Cross-reactivity arises due to the resemblance of Brucella O-polysaccharide, a component of LPS with corresponding epitopes of Yersinia enterocolitica O:9, Salmonella urbana group N, Vibrio cholerae, Francisella tularensis, Escherichia coli O157, and Stenotrophomonas maltophilia [3]. In particular, serological interference induced by Y. enterocolitica O:9 Available at www.veterinaryworld.org/Vol.14/March-2021/35.pdf has complicated the eradication of brucellosis in some European countries such as France, Belgium, United Kingdom [4,5] with the prevalence of Y. enterocolitica ranging from 18% to 58% in cattle [6]. In European Union, 15% of the herds in regions free from brucellosis were infected with Y. enterocolitica O:9 [7]. After 1990, the isolation of Y. enterocolitica O:9 from cattle has become a regular phenomenon not only from European Union but also from other parts of the world including New Zealand [8]. As diagnosis plays a key role in disease control/eradication programs, there is a strong need for tests that can avoid cross-reactivity, in particular against Y. enterocolitica O:9. This can be overcome by screening antigens other than surface antigens like S-LPS. Various proteins of Brucella spp. have been investigated for their use as a diagnostic antigen, replacing S-LPS either in the form of the purified native protein(s) or synthesized recombinant protein(s). Different proteins were reported from different research groups for their capacity in diagnosis; however, still, there is a scope for further screening.
The present study was aimed to screen better protein antigens for diagnosing cattle brucellosis which is void of cross-reactivity, especially against antibodies of Y. enterocolitica O:9. ELISA was done with these antigens and different cattle sera samples and compared for better performance. To the best of our knowledge, it is the first study comparing ten recombinant proteins that are non-reactive to Y. enterocolitica O:9.

Study period and location
The work was conducted at ICAR-NIVEDI from January 2016 to March 2018. The cattle sera were collected from Karnataka, India.

Selection of immunodominant proteins/protein antigens for expression
The immunodominant proteins were identified based on a literature survey concerning their reactivity with Brucella positive serum and non-reactivity with Y. enterocolitica O:9 positive serum. They include Cu-Zn superoxide dismutase (SodC), Serine protease, BAB1-1885, Twin arginine translocation pathway signal sequence domain-containing protein (Twin arginine), Brucella protein26 (BP26), Solute-binding family 5 protein, Leu/Ile/Val-binding family protein, Branched-chain amino-acid ABC transporter substrate-binding protein, Thiamine transporter binding protein, Invasion protein B (InvB), Brucella lumazine synthase (BLS), bacterioferritin (Bfr), malate dehydrogenase (Mdh), VirB12, and Aldehyde dehydrogenase. The selected proteins are outer membrane or periplasmic except for BLS, Bfr, Mdh, VirB12, and Aldehyde dehydrogenase which are cytoplasmic proteins. These identified proteins were further checked through BLASTP for amino acid sequence identity and those showing ≤60% with other cross-reacting pathogens were selected for further expression (Table-1).

Expression of immunodominant proteins
The primers were designed for the selected gene sequences to suit enzyme-free cloning in the pETite N-His Kan vector (Table-2). The respective gene sequences of selected proteins were amplified by PCR with reaction mixture comprising of 10× PCR buffer-5 µL; forward and reverse primers 10 µM each 1 µL; dNTP (10 mM)-1 µL; B. abortus S99 template DNA 2 µL (94 ng); high fidelity Taq polymerase 0.5 µL (M/s. Thermo Scientific, USA); and nuclease-free water were added to make up 50 µL. PCR was performed in the thermal cycler (M/s. Eppendorf, Germany) with the cycling conditions as mentioned in Table-2. The amplified product was run on 1.5% agarose gel containing ethidium bromide (0.5 µg/mL), visualized and documented in UV trans-illuminator.
The PCR amplicons were individually transformed with pETite N-His Kan vector into HI-Control 10G chemically competent cells, respectively, and recombinant colonies obtained were screened by colony PCR. The plasmids were extracted from the positive clones and double pass sequencing of plasmid DNA was carried out commercially (M/s. Eurofins, India) by vector-specific primers. After confirmation by sequencing, pETite N-His Kan vectors containing gene insert were individually transformed into HI-Control BL21 (DE3) and confirmed by colony PCR. The recombinant pETite N-His Kan BL21 clones were induced with IPTG for expression. The expression was optimized for the incubation time (1-5 h after induction) and IPTG concentration (0.5-2.0 mM). The expressed proteins were characterized by SDS-PAGE and purified with Ni-NTA purification system and subsequently dialyzed. The Western blot was carried out with rabbit negative and hyperimmune sera raised against B. abortus S99 antigen, Y. enterocolitica O:9 antigen, and cattle sera (positive and negative for Brucella antibodies confirmed by RBPT and Svanovir Indirect ELISA [Svanova, Sweden] for brucellosis).

Selection of Brucella antibodies positive and negative cattle sera
A total of 573 cattle sera were collected both from brucellosis infected (abortions were common and B. abortus etiology established earlier) and non-infected farms (No abortion and B. abortus etiology absent) and none of the animals were vaccinated for brucellosis. The samples were screened by RBPT [9], Svanovir Indirect ELISA kit for brucellosis (Svanova, Sweden), and Svanovir Competitive ELISA kit for brucellosis (Svanova) following manufacturer instructions, and out of that 114 samples were positive by all the three tests/assays and 322 samples were negative by all the tests for Brucella antibodies. For establishing diagnostic sensitivity (DSe) and diagnostic specificity (DSp), a sample size of 113 positive and 113 negative sera from the above samples by all tests was selected for estimated DSe and DSp as 92%, with an error margin of 5% at 95% confidence interval [10]. Further, the Brucella antibody titers were determined by the STAT/SAT [9]. The titer of selected positive sera from cattle ranged from 1 in 10 to 1 in 640 and hyperimmune serum from rabbit was having more than 1 in 2560 whereas all the negative sera were not having any titer. The negative samples were selected only from non-infected farms where no single positive reactor was present.

Standardization of recombinant protein Indirect ELISA (I-ELISA)
I-ELISA was standardized for antigen concentration and serum dilution by checkerboard titration using Brucella positive and negative serum samples. The recombinant proteins which have reacted in Western blot with cattle serum positive for Brucella antibodies were subjected individually for checkerboard titration for antigen concentration and serum dilution [11]. ELISA was done according to standard ELISA protocol. The recombinant proteins as antigen(s) were doubly diluted in a 96 well ELISA plate ranging from 16 µg to 7.8 ng along the row of the ELISA plate (Maxisorb, Nunc, Thermo Fisher Scientific, USA) in coating buffer (PBS-7.2 pH and kept overnight at 4°C. The next day, the ELISA plate was incubated at 37°C, 150× g for 1 h. Further washing and blocking were done with wash buffer (Phosphate-Buffered saline-Tween 20) and blocking buffer (5% Skim Milk powder, 3% Lactalbumin hydrolysate, and 0.1% Tween-20), respectively. For each antigen concentration Brucella positive and negative samples were titred along the column of the plate (1 in 100; 1 in 50; 1 in 25; and 1 in 12.5 in vertical directions). Further, blocked with blocking buffer for 1 h at 37°C. After washing, conjugate (50 uL of 1:6000 dilution of horseradish peroxidase-conjugated anti-bovine IgG) was added and incubated at 37°C for 1 h. The reaction

Evaluation of the suitability of recombinant protein I-ELISA
After optimizing the antigen concentration and serum dilution, ELISA was carried out on 113 each Brucella antibodies positive and negative cattle sera.

Statistical analysis
Following standard definitions, a 2×2 contingency table and receiver operating characteristic (ROC) curve analysis were used to calculate all the diagnostic accuracy measures and confidence intervals based on OD values of the ELISA. The Youden's index was employed for the analysis. The accuracy measures were expressed as DSe and DSp. Fleiss kappa (κ) statistics was used to assess the agreement between two tests, where a value of κ>0.75 is considered excellent, κ=0.40-0.75 is considered fair to good, and κ<0.40 is considered marginal to poor. p<0.05 was considered statistically significant for proportional analyses.
For the ELISA which has shown higher DSe, DSp, and area under curve (AUC) with a particular antigen, percent positivity (PP) {(Sample OD/Positive OD)×100} values were calculated. Further for the ELISA with the same antigen, other measures such as positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio, and negative likelihood ratio were calculated at different assumed prevalence levels (5%, 10%, and 20%).  All the statistical analyses were carried out using MedCalc, version 17.7.2 (MedCalc Software Ltd, Belgium).

Selection of immunodominant proteins for expression
Out of 15 proteins identified, ten proteins were selected further for expression. On BLASTP analysis, five proteins were eliminated due to their marked amino acid sequence identity (>60%) with other pathogens (Table-3).

Expression and characterization of immunodominant proteins
DNA extracted from B. abortus S99 was used to successfully amplify the target genes ( Figure-1), cloned into pETite N-His Kan vector, and transformed into competent HI-Control 10G cells. Plasmid extracted from overnight grown colonies of HI-Control 10G clones containing the respective gene inserts ranged from 102.7 to 191.1 ng/µL which were optimal in their concentration for further subjecting to sequencing. Sequencing by vector-specific primers revealed the nucleotide sequences of the respective insert gene. All the sequences have shown 100% identity with B. abortus bv. 1 str. 9-941 DNA sequence. After confirmation of sequence identity, the plasmids of pETite N-His Kan vector containing individual inserts of twin-arginine, invB, thiB, bp26, sodC, bab-1885, serine protease, virB12, bfr, and bls, of the B. abortus S99 strain were successfully transformed into HI-Control BL21 (DE3) chemically competent cells, evident from the colony PCR with respective band size amplification for screened clones. The yield of the recombinant proteins ranged between 0.029 and 0.260 mg/mL.  Available at www.veterinaryworld.org/Vol.14/March-2021/35.pdf Western blot analysis of purified recombinant proteins with Brucella antibodies positive cattle serum has shown reactivity against BP26, SodC, BAB1-1885, Serine protease, Bfr, and BLS proteins. Twin arginine, InvB, BP26, SodC, BAB1-1885, Serine protease, and BLS proteins reacted with B. abortus S99 antibodies positive rabbit serum. None of the proteins reacted with Y. enterocolitica O:9 antibodies positive rabbit serum, Brucella antibodies negative cattle and rabbit serum (Figures-2 and 3).  Table-5. The Kappa statistic revealed an excellent agreement of 0.850 (0.781-0.918 at 95% CI) with the standard error of 0.0351.

Discussion
In this study, an attempt was made to evaluate the suitability of immunodominant recombinant proteins that were not reactive to Y. enterocolitica O:9 antibodies for the diagnosis of cattle brucellosis. The immunoproteomics approach could be one of the efficient methods to identify specific immune-proteins among the greater range of proteins expressed by Brucella spp. [12]. Understanding the humoral immune response during Brucella infections will aid in the identification of immunogenic proteins. Appropriate identification and characterization of several immunoreactive proteins concerning seroreactivity can result in a satisfactory diagnostic test [13]. Although individual recombinant proteins have been described by many researchers for the diagnosis of brucellosis, a methodical way of comparing various immunodominant proteins and identification of best among them was undertaken in this study. Using bovine sera, Ko et al. [14] identified immunodominant proteins of B. abortus 1119-3 strain, Kim et al. [15] identified that of B. abortus RB51 and Lee et al. [16] identified that of B. abortus 544. Lee et al. [17] used mice sera for B. abortus 544. From these studies, 61 Brucella proteins which are reactive only with Brucella antibodies and not with any other cross-reacting antibodies, in particular antibodies of Y. enterocolitica O:9 were identified. Further with the criteria of the outer membrane proteins (OMPs) or periplasmic proteins as more quantity of antibodies is expected against OMPs, the list was narrowed down to 11 proteins. However, since Brucella is an intracellular organism, the host immune response is not restricted to surface proteins and cytoplasmic proteins can also induce a higher antibody response [18], other proteins that were documented earlier were also considered with the idea of not missing   out the immunodominant proteins irrespective of OMP or cytosolic. Besides, proteins such as chaperonin GroEL, Ribosomal protein L7/L12, chaperonin GroES [12], and OMP25 [14] were not considered due to their reactivity with cross-reacting antibodies.
Letesson et al. [18] reported that OMP10, OMP16, OMP19, OMP25, OMP36, p15, and p39 were not useful for detecting naturally infected cattle. Thus, a meticulous selection of immunodominant proteins of B. abortus was made and further narrowed down based on the sequence identity with other pathogens.
In the case of allergic proteins, a minimum 35% shared identity, plus a minimum of 80 amino acid overlap length, are the criteria to identify sequences that have potential cross-reactivity between an unknown and a known allergen [19]. SodC, an immunodominant Brucella protein identified by researchers [14][15][16] has shown non-reactivity with Y. enterocolitica O:9 antibodies even though the sequence identity with the organism is 60%. Hence, in the case of proteins identified in this study, the identity threshold was fixed as 60% and above which if there is any sequence identity, those proteins were not considered further.
B. abortus strain 99 was used in this study as OIE recommends this strain [20] for the production of antigens for the diagnostic tests. Moreover, the amino acid sequences of proteins selected in this study are identical among various strains of B. abortus.
The selected proteins (n=10) were all expressed using pETite N-His Kan vector and as much of the earlier literature suggested that expression of BP26 and few other proteins leads to inclusion body formation, in this study directly after expression denaturing method of purification was carried out irrespective of whether the expressed proteins form inclusion bodies or not. There is a possibility for binding background contaminants under native conditions than under denaturing conditions and thus denaturing method of purification helped in yielding pure proteins. Cloeckaert et al. [21] observed a double band for rBP26 against infected sheep sera which could be due to the presence of both preprotein with its signal peptide and cleaved protein without signal peptide. The proteins in this study were expressed without signal sequences.
In Western blot, the additional reactivity of Twin arginine, InvB in rabbit hyperimmune serum compared to the cattle serum in this study might be due to the strong immune response in rabbits by use of adjuvants and frequent administration of killed antigen. The variations in the reactivity can be attributed to the differences like the serum used in the studies such as quantity and duration of exposure to the Brucella organism and stage of the infection. The observation of reactivity of Bfr only in cattle sera and not in rabbit hyperimmune serum needs to be explored further whether this reactivity occurs only in case of sera from infected and not from sera raised against the killed antigen. In addition, non-reactivity of these expressed antigens against Y. enterocolitica O:9 serum of rabbit confirms the right selection of candidate proteins for increasing the specificity of brucellosis diagnosis.
In absence of a gold standard, for the right classification of sera into Brucella antibodies positive and negative, a battery of tests/assays is required. Hence, in this study, RBPT, I-ELISA, and C-ELISA have been employed to select a panel of Brucella antibodies positive and negative sera for evaluating recombinant proteins. A test/assay with a high DSe is useful for "ruling out" a disease if an animal tests negative and a test with a high DSp is useful for "ruling in" a disease if an animal, tests positive. As LPS/whole antigen-based tests were more sensitive and less specific, the test which can contribute to more specificity and nominal sensitivity can provide additional information or can complement the existing test. In this study, the decreasing order of DSe was observed with BP26, BLS, SodC, BAB-1885, Serine protease, and Bfr based ELISA whereas the decreasing order of DSp was observed with BP26, Bfr, BAB-1885, BLS, and SodC based ELISA. DSe of 70.7%-80.5% and DSp of 84.6%-100% [7] DSe of 88.7%, DSp of 93.8% [22], DSe of 96.7%, and DSp of 95.4% [23] in bovines were observed in ELISA based on BP26 antigen; whereas, in this study, DSe of 92.04% and DSp of 96.46% were observed. The differences in DSe and DSp from different studies may be attributed to differences in samples under study, the gold standard against which the test was compared apart from inherent changes in the test procedure adopted. Youden's index combines DSe and DSp and provides a single value for the comparison. Based on Youden's index the performance of the BP26 is superior which is followed by BLS, SodC, and the rest of the proteins.
The AUC, DSe, DSp, and Youden's index based on PP values of BP26 based ELISA were 0.953, 90.27%, 95.58%, and 0.8584 as against based on OD values 0.968, 92.04%, 96.46%, and 0.8850 (Table-4), respectively. Although there is a marginal reduction in all the values based on PP values in comparison to OD values, the calculations based on PP values are more stable. As the DSe is 90.27%, ruling in the individual animal may not be appropriate whereas it can be used as a test for detecting infected herds. PPV and NPV are both dependent on the prevalence of the disease and hence they were not calculated at the preliminary level of selection of the antigen because the number of positive and negative samples was pre-decided in the experiment and not indicative of prevalence level. Moreover, as PPV and NPV are dependent on prevalence, the values change concerning prevalence of the study area and not constant. Hence, AUC, DSe, DSp were considered in identifying the best antigen for ELISA as they do not vary with the prevalence level. However, PPV and NPV at assumed prevalence levels were calculated to assess the utility of the test. The NPV remains high and constant across various prevalence levels of 20%, 10%, and 5% from 97.49% to 99.46% whereas PPV increases from 47.22% to 80.95% from low prevalence (5%) to high prevalence (20%). Furthermore, the LR provides a suitable summary measure, which is independent of prevalence.
BP26 is highly conserved in the genus Brucella, and the recombinant protein from one Brucella species might be used for the serological diagnosis of infections caused by all the Brucella species, provided that the animal host can induce an immune response against this antigen [24]. Letesson et al. [18] documented that antibody responses to proteins were found to be limited to animals that developed an active brucellosis infection and this can be validated further with BP26 ELISA to identify active shedders of Brucella organisms in a herd. In this study, one of the important cross-reacting organism which has created trouble in few of the countries of the world, namely, Y. enterocolitica O:9 was addressed through the selection of Brucella antibodies reacting recombinant proteins but not reacted to experimentally raised hyperimmune rabbit serum to Y. enterocolitica O:9. Further, the identified recombinant protein should also be checked against other cross-reacting antibodies to ensure maximum specificity.

Conclusion
As there are many claims for various recombinant proteins as the best recombinant proteins for diagnosing cattle brucellosis, this study has brought out the ranking and identified BP26 as the best suitable recombinant protein among them. BP26 based ELISA can be used as a screening test for detecting infected herds, to rule out cross-reactivity due to Y. enterocolitica O:9 in infected herds. It can also complement the existing diagnostic tests such as RBPT and sLPS based ELISA.