Characterization of recombinant Ybgf protein for the detection of Coxiella antibodies in ruminants

Abstract

Q fever remains a One Health problem, posing a zoonotic threat and causing significant economic losses to the livestock industry. The advancement of detection tools is critical to the effective control of infection. In humans, laboratory investigations depend largely on the immunofluorescence assay, considered the gold standard. In contrast, serologic tools routinely used for veterinary screening have several gaps, resulting in interpretations that are frequently misleading. We investigated the potential application of recombinant Ybgf antigen (r-Ybgf), a periplasmic protein described as one of the most immunodominant antigens in humans, in an indirect ELISA. Following successful expression in the prokaryotic system and the preliminary evaluation of immunoreactivity in western blot, we used r-Ybgf to develop an in-house ELISA using serum samples from sheep, goats, and cattle, which were tested in parallel with an Idexx ELISA kit. The results obtained with the 2 tests were compared, and r-Ybgf performed favorably, with 81.8% sensitivity and 90.1% specificity and substantial agreement, as revealed by receiver operating characteristic analysis. Moreover, we evaluated the serologic response against phase I (PhI) and phase II (PhII) antigens, and r-Ybgf antigen induced by vaccination, using phase-specific ELISAs. The dynamics of antibody response showed a significant increase in reactivity against PhI and PhII, but not against r-Ybgf, antigens. This property may be very useful given the absence of a protocol for the differentiation of infected from vaccinated animals.

Coxiella burnetii is an intracellular gram-negative bacterium responsible for Q fever, an infection reported worldwide in a wide host range, including domestic ruminants. ²¹ Infection in livestock is usually subclinical and can result in significant economic losses in livestock as a result of reproductive disorders, such as abortion, metritis, and placental retention. Domestic ruminants, particularly sheep and goats, are considered the major reservoirs and sources of infection and are frequently implicated in human outbreaks.^2,28

C. burnetii can infect humans, mainly through the respiratory tract (a digestive route has also been described from drinking unpasteurized milk); people in close contact with infected ruminants have a greater chance of becoming infected. Seroprevalence rates have been found to be higher among cattle farmers (56.3%) and veterinarians (88.9%) than blood donors (12.7%).^21,24 The disease progresses with a self-limiting flu-like syndrome in a large percentage of cases, although a small percentage of patients (1–5%) develop the chronic form, characterized by hepatitis, pneumonia, and life-threatening endocarditis. ² The largest human outbreak reported was in the Netherlands in 2007–2010 when >4,000 reported cases and hundreds of chronic Q fever forms were reported, highlighting the impact on human health. ⁸ The very low mean infectious dose (ID₅₀ = 1), wide host range, massive post-abortion dissemination, and transport by atmospheric agents (mainly wind) are the main reasons that C. burnetii is classified as a category B bioterrorism agent by the European Center for Disease Control.^10,21 Accordingly, scientific and human health authorities have highlighted the need to improve surveillance and the active monitoring for this pathogen in order to engage veterinary medicine and implement the One Health concept. ²² The latter was coined as a worldwide strategy concerning the multisectoral and multidisciplinary approach to monitor and control public health threats. ²²

Diagnosis in domestic animals is based on direct detection by PCR assay of milk, aborted material, or vaginal swabs. Although this technique may be expensive, it is recommended in suspected cases of abortion. The performance of molecular tools could be affected by false-negative responses given the intermittent shedding of C. burnetii, especially during subclinical forms. ³⁴ Moreover, false-positive results may be obtained using commonly targeted genes, which do not distinguish between C. burnetii and Coxiella-like bacteria. ²⁹ The positivity of a vaginal swab PCR assay in pregnant goats was not necessarily associated with uterine colonization and then with current infection.^1,13

Indirect methods such as the indirect immunofluorescence assay (indirect IFA) and ELISA are preferred as rapid screening tools.^29,34 The indirect IFA can distinguish between acute and chronic infections, and is considered the gold standard test in humans, but commercial kits are not available for veterinary use.^15,37 ELISA is the most widely used test and is recommended by the OIE (World Organisation for Animal Health) for rapid routine screening and large-scale epidemiologic studies in ruminant populations. ²⁹ Similarly to humans, the immune response in ruminants consists of a strong and early reaction (2–3 wk after infection) against phase II (PhII), the avirulent form with a truncated lipopolysaccharide; phase I (PhI; virulent form with full-length lipopolysaccharide) antibodies increase in the following weeks.^4,27,38

Commercial ELISA kits are based on a combination of inactivated antigens (obtained from the culture of reference strains in biosafety level 3 laboratories) that react with both PhI and PhII antibodies. ¹² These tests are not able to distinguish between acute or chronic infection and vaccination, and can sometimes lead to conflicting results and misinterpretation. ²¹ The antigenic variation between C. burnetii strains and the complex response of the immune system also contribute to reducing the analytical performance of ELISA kits, which are also affected by cross-reactivity with other pathogens, such as Chlamydia spp., Rickettsia spp., and Bartonella spp.^19,20 Studies using different statistical approaches, such as latent class analysis and informatics software, have shown how the performances of currently used ELISA kits can differ depending on the species, matrix, and manufacturer.^14,17,23

The improvement of current tests is considered by the international scientific community as a challenge that could be solved by new approaches and technologies. Recombinant DNA technology represents a valid but poorly investigated option, given that immunodominant proteins can be produced in large quantities without the need for biosafety requirements. Moreover, when this technology is successfully applied to specific stage-related proteins, highly sensitive assays are obtained, especially when several antigens are used simultaneously in multiplex serology platforms. ¹⁶

Recombinant DNA technology has been reported to improve the serodiagnosis of Q fever in ruminants in 2 papers.^11,31 Heat shock protein B (HspB) was found to be highly specific (97%) and more effective in identifying early infection than an ELISA kit. ¹¹ Sensitivity (Se) and specificity (Sp) of 85% and 69% for sheep, 94% and 77% for goats, and 71% and 70% for cattle, respectively, were obtained for an outer membrane protein (Com1) test. ³¹

Although Ybgf (CBU0092) is considered to be one of the most immunoreactive proteins in humans, it has not been assessed for the detection of C. burnetii antibodies in ruminants.^3,9,12,38 Ybgf, also known as cell division coordinator CpoB, is located in the outer membrane and is involved in the tol-pal transport system. ¹² During the early acute stage of infection, Ybgf was recognized in the sera of humans and mice, and was suggested as a PhII-specific marker. ³⁸ In human medicine, recent attempts to use Ybgf in a recombinant ELISA revealed good Sp (close to 90%) but poor Se (12.5–21%).^16,25 We evaluated the efficiency of this recombinant antigen for the detection of antibodies against Coxiella burnetii in ruminants.

Materials and methods

Amplification, digestion, ligation

C. burnetii DNA, extracted from the Nine Mile RSA 493 PhI strain (DNeasy blood and tissue kit; Qiagen), was kindly provided by the Department of Rickettsiology, Institute of Virology, Biomedical Research Center, Slovak Academy of Sciences, Bratislava, Slovakia. Specific primers containing BamHI and EcoRI restriction enzyme sites (underlined) were designed according to NCBI (https://www.ncbi.nlm.nih.gov/) and excluded the leader peptides: forward primer (5′-TTGGATCCCCAGTGGAAGATATTTCGGCG CAAC-3′); reverse primer (5′-TTGAATTCAGGCGTTGTTGTTGCTGAATCGAC-3′)

The Ybgf gene (Q83F57) was then amplified without its signal sequence in a reaction mixture consisting of 34.3 μL of RNase-free water, 5 μL of CoralLoad buffer 10× (Qiagen), 1.5 μL of MgCl₂ (50 mM), 1 μL of dNTPs (10 mM), 2.5 μL of each primer (10 μM), 3 μL of gDNA (previously quantified at 50 ng/μL), and 0.25 μL of Taq DNA polymerase (Qiagen), to obtain an 822-bp product. PCR cycling conditions were PCR activation at 94°C for 3 min, denaturation at 94°C for 45 s, 35 cycles of annealing at 60°C for 30 s, and elongation at 72°C for 90 s, followed by a final elongation at 72°C for 10 min. The amplification product was electrophoresed in an agarose gel to confirm its size. Digestion was performed at 37°C for 3 h using appropriate restriction enzymes (BamHI, EcoRI; Thermo Scientific), followed by purification, quantification (NanoDrop 2000/2000c spectrophotometer; Thermo Fisher), and ligation into a pGex 6P-1 expression vector, which had been digested previously with the same enzymes.

Expression and purification

Ligation product (10 μL was used to transform competent cells of Escherichia coli BL21 C43 [DE3]), grown in lysogeny broth (LB) agar Petri dishes in the presence of ampicillin overnight at 37°C, and verified by tapping in colony PCR to assess the plasmid carrying the insert of the expected length. Plasmid DNA from the positive clones was extracted (NucleoSpin plasmid mini kit; Macherey-Nagel) according to the manufacturer’s instructions, and then Sanger sequenced using PCR primers to verify the authenticity and in-frame orientation of the insert. Positive colonies were inoculated with ampicillin in 500 mL of LB, incubated at 37°C, and induced with 1 mM isopropyl-β-D-1-thiogalactopyranoside during the mid-exponential phase. Bacterial cells were harvested 4 h later by centrifugation, and lysed by physicochemical methods using lysozyme and sonication. The fusion protein was harvested in the soluble fraction, and purified and concentrated under native conditions by 3 chromatography steps performed in columns (glutathione Sepharose 4B resin; Merck) and employing the affinity between glutathione S-transferase (GST) and glutathione. GST cleavage was performed directly on columns with PreScission protease (2 U/mg; Merck), to ensure that the fusion carrier did not interfere with the predicted reactivity. All protein fractions were quantified using the Bradford protein assay (Thermo Fisher) and examined by electrophoresis gels stained with Coomassie brilliant blue R250.

Mass spectrometric analysis

The band was excised with a scalpel blade and destained in a 1.5-mL tube containing a 50 mM ammonium bicarbonate buffer in the presence of 50% acetonitrile (ACN). The reduction and alkylation steps were performed with 10 mM dithiothreitol in a 100 mM ammonium bicarbonate buffer, and 50 mM iodoacetamide in a 100 mM ammonium bicarbonate buffer, respectively. A digestion step with porcine trypsin (Merck) was performed overnight at 37°C. After addition of the stop solution (70% ACN solution containing 1% trifluoroacetic acid), peptides were extracted and analyzed by automated nanoflow reverse-phase liquid chromatography (nanoAcquity UPLC system, Q-TOF premier electrospray ionization tandem mass spectrometer; Waters). The BEH 130 C18 column (200 mm × 75 μm, particle size: 1.7 μm; Waters) was used to separate the peptides.

Data were acquired by spraying the sample at 3.4 kV capillary voltage in the Q-TOF detector and reading spectra from alternating scans at low and high collision energies (4–40 eV). A species-specific C. burnetii RSA 493 database (UniProt Consortium, 9.10.2008 download, 1,824 entries, https://www.uniprot.org/) was consulted, and the data obtained were interrogated with an acceptable false-positive rate of 4%. Identification was established if the sequence coverage achieved was at least 15%. Analysis of spectrometry data was performed as described previously. ³⁰ The aminoacidic sequence of recombinant protein was interrogated for sequence homology to other bacteria using a BLAST search (https://blast.ncbi.nlm.nih.gov/Blast.cgi) with the aim of excluding any cross-reactivity.

Western blot

The sample containing 100 µg of recombinant Ybgf and an equal volume of 2× loading buffer was separated by SDS-PAGE electrophoresis using a 15% polyacrylamide gel before transfer to a nitrocellulose membrane blocked in a 2.5% casein solution. The membrane was cut into 7 strips and incubated with a panel of 7 bovine serum samples (1:100 dilution), kindly provided by Dr. Jens Bottcher (Bavarian Animal Health Service, Germany). True-positive and -negative samples were determined by both serologic (ELISA on serum, Q-fever ELISA 1 and 2 bulk; Dolfinin) and molecular (real-time PCR on individual milk, targeting the C. burnetii transposase gene, IS1111a, accession M80806) methods.^5,6 Each strip was washed with Tween–Tris-buffered saline and incubated with a horseradish peroxidase–conjugated anti-bovine IgG antibody (mouse monoclonal, 1:1,000 dilution; Sigma). Each strip was visualized by incubation in a solution of 3,3′-diaminobenzidine tetrahydrochloride (MilliporeSigma) dissolved in Tris-buffered saline, pH 7.6, and containing 8 µM hydrogen peroxide.

Development of indirect ELISA

The recombinant protein was diluted at a concentration of 1 ng/μL in 0.1 M carbonate–bicarbonate buffer (pH 9.6) and used to coat (100 μL) the wells of 96-well plates (Nunc Maxisorp; MilliporeSigma) overnight at 4°C. After blocking with 2.5% bovine casein (300 μL), serum samples diluted 1:100 in PBS–1.25% casein were added to the wells (100 μL) and then incubated for 1 h at room temperature. After a wash step, the monoclonal antibodies anti-goat/sheep IgG–peroxidase (Merck) or anti-bovine IgG–peroxidase (Merck) were diluted (1:1,000 dilution in a volume of 100 μL/well) in the same buffer described previously for incubation of the primary antibody (45 min at room temperature). After the final wash, the color reaction was developed with 100 μL of 3,3′,5,5′-tetramethylbenzidine (TMB) and stopped with 100 μL of 0.2 M H₂SO₄. Optical density was measured at 450 nm (OD₄₅₀). A cutoff value of 0.5 was chosen arbitrarily (mean of negative sera plus 4 SDs) to distinguish between positive and negative samples. ⁷

These working conditions were selected by testing different antigen amounts and different dilutions of positive and negative control sera (crisscross serial dilution analysis). A panel of sera belonging to apparently healthy ruminants and from animals associated with reproductive disorders or abortions (231 small ruminants, 356 cattle, 6-mo to 4-y-old) was chosen randomly for assessing the potential of r-Ybgf against a commercial ELISA (Q fever Ab test kit; Idexx). Serum samples were obtained from herds located in northern Italy with unknown infection status from previous investigations (2020–2021) and stored at the Department of Veterinary Sciences, University of Turin, Italy. Before testing with our recombinant in-house ELISA, all sera were tested with the Idexx ELISA kit, according to the manufacturer’s instructions. The Idexx test has proven to be sensitive and specific in several studies (Se = ~90%; Sp = ~98%)^17,23 and is based on tick-derived and inactivated whole-cell antigens that detect the presence of specific IgG against C. burnetii. ¹² We used the Idexx ELISA as the reference assay for the evaluation and comparison of our r-Ybgf results given the lack of a gold standard assay for the detection of C. burnetii in veterinary medicine.

Statistical analysis

The Idexx ELISA was considered as the reference assay in our statistical analysis (MedCalc v.18.11.3, https://www.medcalc.org/). Se and Sp, with accompanying 95% CIs, were calculated by using the results obtained in a contingency table ³³ :

$$\begin{array}{l}\mathrm{Se}=a/\left(a+c\right);a=\mathrm{true-negative};c=\mathrm{false-positive}.\hfill \\ \mathrm{Sp}=d/\left(b+d\right);d=\mathrm{true-negative};b=\mathrm{false-positive}.\hfill \end{array}$$

The level of agreement between the 2 tests was determined by kappa Cohen coefficient (0.81–1.00 = almost perfect agreement; 0.61–0.80 = substantial agreement; 0.41–0.60 = moderate agreement; 0.21–0.40 = fair agreement; 0.01–0.20 = slight agreement; 0.00 = no agreement). Based on the results obtained, a receiver operating characteristic (ROC) curve was used (MedCalc v.18.11.3; GraphPad Prism 8, https://www.graphpad.com/) to compare the relative Se and Sp of the ELISAs at various cutoffs and design an area under the curve (AUC) representing a statistical summary of overall test performance. ³²

In-house phase-specific ELISA and immunoreactivity evaluation of vaccinated animals

Because the Idexx ELISA did not distinguish the stage of infection or vaccination, we chose phase-specific antigens to detect antibodies against PhI and PhII antigens independently in vaccinated and control animals, as has been done in both humans and animals. We coated Nunc Maxisorp 96 multiwell plates with PhI and PhII antigens (Q-fever ELISA 1 and ELISA 2 bulk; Dolfinin). We used a final dilution of 1:5,000 for the PhI antigens and a dilution of 1:500 for the PhII antigens, as used by other authors. ⁶ The coated plates were covered overnight at 4°C, then emptied and incubated with 1:100 diluted sera from vaccinated animals for 90 min at 37°C. After 5 washing steps, the protein G peroxidase conjugate (1:5,000 dilution; MilliporeSigma) was incubated for 90 min. The color reaction was developed with TMB and stopped within 20 min with 0.2 M H₂SO₄. Optical density was measured at 450 nm (OD₄₅₀). The cutoff value was set at 20% of the positive control (OD₄₅₀ cutoff is 0.30 for PhI and 0.45 for PhII), as described previously. ⁶ From 10 cattle (3–8-mo-old) vaccinated with Coxevac (Ceva), we collected blood samples at various times: time of vaccination (t0), revaccination 21 d after the first dose (t21), 42 d (t42), and 63 d (t63). The control group consisted of 2 seronegative, unvaccinated cattle bred on the same farm. The same sera were tested simultaneously with our r-Ybgf ELISA, and the results obtained were statistically analyzed using the Wilcoxon test (R, https://www.r-project.org/).

Results

As was confirmed by sequencing, the gene encoding amino acids 31 to 30 of the Ybgf protein (accession Q83F57) was correctly cloned into the pGEX-6P-1 plasmid. Moreover, the insert was successfully expressed in E. coli in the form of a fusion protein of ~60 kDa (31 kDa for r-Ybgf, 29 kDa for GST), as shown on SDS-PAGE stained with Coomassie (Fig. 1). As shown by the huge differences in protein expression patterns between the uninduced and induced bacterial lysates, gene regulation and subsequent protein production were under the control of the lac operon (Fig. 1).

Figure 1.

SDS-PAGE analysis of recombinant Ybgf (r-Ybgf): 1 = lysate of non-induced bacteria; 2 = lysate of induced bacteria; 3 = total extract of transformed r-Ybgf Escherichia coli cells; 4 = total extract after incubation with glutathione Sepharose resin; 5–7 = purified (affinity chromatography) and cleaved Ybgf.

The SDS-PAGE of the purified and cleaved protein revealed a major band of the correct molecular weight (~35 kDa), although weak and shorter bands, comparable to subproducts or E. coli proteins, were also visible in the eluted fractions (Fig. 1). When purified under native conditions, the protein appeared to be quite soluble and concentrated mostly in the supernatant.

Mass spectrometry analysis allowed further identity confirmation of the protein separated by electrophoresis and contained in the band. The physical and chemical properties, such as isoelectric point and molecular mass, were identical, indicating the absence of aminoacidic residues modified elsewhere. The peptides obtained were consistent with the protein sequence available in the database (entry Q83F57_COXBU), reaching a coverage of 61.63% (Suppl. Table 1).

After quantification by a Bradford assay, the first eluate (2.8 mg/mL) was used to coat 96-well ELISA plates; the second eluate (1.5 mg/mL) was used for western blot analysis. Immunoblotting analysis was performed with true-positive and -negative sera and showed good immunoreactivity of the purified r-Ybgf. The antigen was specifically recognized by 3 of the 4 positive sera; it did not react with negative serum samples from healthy cattle. We did not observe any nonspecific bands (Fig. 2).

Figure 2.

Recombinant Ybgf immunoreactivity against Coxiella burnetii–positive and –negative serum samples. Strips 1–3 were incubated with negative samples; strips 4–7 were incubated with samples positive for Coxiella antibodies. The antigen recognized specific antibodies in 3 of 4 positive sera; it did not react with negative samples. We did not observe nonspecific bands of undesired molecular weight.

Our results confirmed Ybgf as a candidate for further study, as seen in the western blot analysis. Specifically, r-Ybgf correctly detected 126 of 154 positive ruminants (relative Se = 81.8%; 95% CI: 74.8–87.5) and 390 of 433 negative ruminants (relative Sp = 90.1%; 95% CI: 86.8–92.7; Table 1). We found 43 false-positive and 28 false-negative samples, with no clear differences between the 2 tests (Table 2). The substantial agreement between the 2 assays is underlined by a concordance of 0.88 and Cohen kappa of 0.69 (95% CI: 0.631–0.763; Table 2). The ROC curve analysis yielded an AUC of 0.85 (95% CI: 0.816–0.879), indicating a moderate level of accuracy (Fig. 3) and suggesting a cutoff value of 0.507.

Table 1.

Ruminant serum samples tested for Coxiella burnetii antibodies with Idexx and r-Ybgf ELISAs.

Species	+ Idexx	– Idexx	Total
Sheep/goat
+ r-YBGF	78	14	92
– r-YBGF	16	123	139
Total	94	137	231
Bovine
+ r-YBGF	48	29	77
– r-YBGF	12	267	279
Total	60	296	356
Overall
+ r-YBGF	126	43	169
– r-YBGF	28	390	418
Total	154	433	587

– = negative; + = positive.

Table 2.

A comparison of the diagnostic performances between the r-Ybgf and Idexx ELISAs.

Species	RSe*	RSp*	Concordance	95% CI	Cohen kappa	95% CI	SE
Sheep and goat	83% (73.8–89.9)	89.8% (83.4–94.3)	0.87	0.82–0.91	0.73	0.64–0.82	0.046
Cattle	80% (67.7–89.2)	90.2% (86.2–93.3)	0.88	0.84–0.91	0.63	0.52–0.74	0.054
Overall	81.8% (75.7–87.9)	90.1% (87.3–92.9)	0.88	0.85–0.9	0.697	0.63–0.76	0.034

RSe = relative sensitivity; RSp = relative specificity.

^*Numbers in parentheses are 95% CIs.

Figure 3.

Receiver operating characteristic (ROC) analysis of recombinant Ybgf ELISA obtained by testing 433 negative sera and 154 positive sera. ROC analysis area under the curve was 0.85 (95% CI: 0.816–0.879).

After t21, reactivity against both PhI and PhII increased significantly following booster dose administration (Fig. 4). Three animals were positive for PhI, and 7 for PhII at t63. No animals were positive for r-Ybgf, whose OD increased significantly only at t21 and never exceeded the cutoff OD₄₅₀ value of 0.5 (Fig. 4). No change in reactivity against PhI, PhII, and r-Ybgf antigens was observed in the control group, which consisted of 2 non-vaccinated animals.

Figure 4.

Dynamics of reactivity against phase I (PhI), phase II (PhII) antigens, and recombinant Ybgf (r-Ybgf) in 10 vaccinated animals. The exponential and significant (Wilcoxon signed rank test p < 0.01 [**]) increase in reactivity against PhI and PhII started at t21 (21 d post-vaccination), and several animals tested positive to our in-house phase-specific ELISA (OD₄₅₀ cutoff is 0.30 for PhI and 0.45 for PhII). Seroconversion against Ybgf was limited; reactivity against Ybgf increased significantly only at t21. No change in reactivity against phase I, II, and r-Ybgf antigens was observed in the control group, which consisted of 2 non-vaccinated animals (not shown).

Discussion

We cloned the Ybgf sequence successfully and expressed it in a prokaryotic system. After extraction, purification under native conditions, and cleavage of the GST carrier, an acceptable amount and purity of protein was obtained. Mass spectrometric analysis confirmed the identity of the 35 kDa protein contained in the SDS-PAGE band and excluded the presence of nonspecific proteins.

In western blot analysis, Ybgf recognized 3 of 4 positive bovine sera (confirmed by both serologic and molecular approaches) but did not react when using bovine negative sera. Ybgf also proved to be a promising marker in our ELISA in testing sheep, goat, and cattle sera, with good agreement with the commercial kit results. However, ROC curve analysis indicated the imperfect performance of our in-house recombinant ELISA, which might be explained by several hypotheses. At least in human testing, Ybgf is a specific marker of PhII, namely acute infection; thus, in the absence of information on the stage of infection in positive animals, the lack of identification in chronic infections is plausible, considering the complex dynamics of the antibody response to C. burnetii. Moreover, our ELISA was directed to a single class of antibodies (IgG) that recognized a single protein. These aspects may explain why some positive sera did not react with r-Ybgf in the western blot analysis and ELISA. On the other hand, the performance of the commercial test used for comparison was not optimal, which, in the absence of a gold standard and reference sera, led to doubts regarding several false-positive samples that could be false-negatives for the commercial kit. Several studies have found that the Se of the commercial ELISA is ~90%, with a Sp of ~99%.^3,9,35,38

Although our in silico analysis excluded cross-reactions or homologies between antigens from other microbes and Ybgf, the existence of shared epitopes with other microbes, and in particular, with phylogenetically close pathogens that are widespread in ruminants, such as Rickettsia spp., Bartonella spp., and Chlamydia spp., must be considered. To fully understand these dynamics and reveal Ybgf cross-reactions against other pathogens, it would be necessary to test, with our in-house ELISA, sera from animals infected with bacteria that may cross-react. In addition, recombinant proteins do not always retain the antigenic properties of native proteins, given the intrinsic inability of E. coli to carry out post-translational modifications, such as glycosylation or methylation, which could help define epitopes. ³⁶

Different performance was found between our ELISA and those developed in recent work using a recombinant Com1 antigen. ³¹ This Com1 antigen study found a Se of 77% and Sp of 72%, without significant differences among species (sheep, goat, and cattle), ³¹ although Com1 is reported to be more sensitive and specific than Ybgf in humans.^16,25,31 This discrepancy might be attributed to different immune responses against these antigens between humans and ruminants, as well as the choice of sera used for the assay. Certainly, our ELISA is not a useful tool for routine testing, but further experiments are needed to compare the results with a larger number of true-positive and true-negative sera. The use of samples from experimentally infected animals could be used to fully define the evolution of the immune response against Ybgf over time. Other candidate antigens might be tested given the ease of producing recombinant antigens; we produced a large amount of Ybgf in a short time, enough to test 120,000 animals.

The OD₄₅₀ values obtained with phase-specific ELISAs and our in-house ELISA were monitored, and a significant increase in seroreactivity against PhI and PhII antigens was observed for each time. At t63, almost all animals were positive for at least one phase, confirming the limited usefulness of whole-cell antigens ELISAs in vaccination campaigns, in which vaccinated animals could be mistaken for infected animals. In contrast, seroreactivity against Ybgf statistically increased only at t21. However, none of the vaccinated animals was positive against Ybgf antigen at any time; the OD₄₅₀ was always lower than the cutoff. The lack of seroconversion against r-Ybgf after vaccination is probably a result of the composition of the vaccine (C. burnetii strain Nine Mile, PhI), which may not elicit an adequate antibody response against our protein. Another explanation could be the inefficiency of r-Ybgf in the identification of both phases of infection. Nevertheless, this aspect requires further investigation to be confirmed in a larger number of animals and over longer periods of time. Given the lack of protocols for differentiation between infected and vaccinated animals (DIVA), this property could be useful as a serologic monitoring tool in vaccination campaigns in which commercial ELISAs are not used.

The circulation of Coxiella in vaccinated farms paradoxically cannot be monitored by serologic tests (vaccinated animals would be confused with infected animals, leading to a positive result).^12,18 During the largest outbreak ever reported in the Netherlands in 2007–2010, serologic tools played only a minor role, given that the Dutch health authorities had implemented a compulsory vaccination campaign in 2008 in response to the increasing number of human cases. ²⁶ Therefore, surveillance was performed with real-time PCR on bulk tank milk. ²⁶ The use of r-Ybgf combined with phase-specific antigen ELISAs could distinguish among the stage of infection and vaccination, representing a valid DIVA protocol. Although our recombinant Ybgf antigen was an attempt to expand the detection tools for this infection, it only showed moderate diagnostic accuracy, which does not support its use as a screening assay.