Validation of ELISA with recombinant antigens in serological diagnosis of canine Leishmania infantum infection

BACKGROUND Dogs are the main peridomiciliary reservoir of Leishmania infantum thus the correct diagnosis of infection is essential for the control of the transmission and treatment as well. However, the diagnosis is based on serological assays that are not fully effective. OBJECTIVE We aimed to establish an effective serological assay for the diagnosis of L. infantum infected dogs using Leishmania-derived recombinant antigens. METHODS Leishmania derived rK39-, rK28-, rKR95-based enzyme-linked immunosorbent assay (ELISA) was standardized using symptomatic and asymptomatic L. infantum-infected dogs. Then 2,530 samples from inquiry in endemic areas for VL were evaluated and the results compared with recommended assays by the Brazilian Ministry of Health (MH algorithm). Further samples from a cohort of 30 dogs were searched. FINDINGS For rK39-, rK28- and rKR95-ELISA the sensitivity was around 97% and specificity 100%. The positivity of these three ELISA in the inquiry samples was 27-28%, around 10% higher than the assays currently in use. When cohort samples were searched, we observed likely false-negative results (> 65%) with supposedly negative samples that turned positive six months later with the assays in use (MH algorithm). MAIN CONCLUSIONS For the diagnosis of L. infantum-infected dogs, rK39-based ELISA showed better diagnostic performance than other assays in use in Brazil and worldwide.


2|13
Leishmania infected dogs typically present high anti-Leishmania IgG antibody titers, and serological methods have been widely used as an indirect detection method. (5) In Brazil, the Ministry of Health incorporated a diagnostic algorithm that uses the dual-path platform (DPP) immunochromatographic test for Leishmania for primary screening and enzyme-linked immunosorbent assay (ELISA) detecting antibodies against the total L. major-like antigen, for confirming the diagnosis of CanL. (8,9,10,11,12) The necessity to confirm the diagnosis after screening by the DPP test was suggested in the study of Figueiredo et al., (10) showing a decrease in sensitivity (75%) and low specificity (72%) in asymptomatic dogs.
On the total L. major-like-ELISA, since it is prepared from the crude lysate of L. major-like parasites, variability is frequent in performance, with reported sensitivities ranging from 11.59% (13) to 100.0% (14) and specificities ranging from 68.0% (14) to 90.74% (13) In this scenario, although there is a faster diagnosis of CanL using the immunochromatographic test as screening, we see the possibility of improving the ELISA that currently employs L. major-like total lysate. It would be desirable to have an ELISA based on a recombinant antigen as it is a more specific protein, and production can be done on an industrial scale. Thus, we propose in this study to evaluate the use of four recombinant proteins (rK39, rK28, rKR95 and rK18) for the serological diagnosis of dogs in ELISA.
The recombinant protein rK39 is a repeated sequence of 39 amino acids highly preserved in L. infantum and L. donovani, related to kinesin and highly expressed in amastigotes. (15) In humans, rK39-ELISA detects active VL and asymptomatic Leishmania infection (16) with sensitivity in the range of 69.4-97.6% and specificity of 81.0-100.0%, depending on geographical area. (17,18,19) In dogs, rK39-ELISA has yielded 95.0% sensitivity and 100.0% specificity. (20) In a multicentric study carried out in Italy, (21) rK39-ELISA presented 97.1% sensitivity and 99.4% specificity. In the latter study, the serological inquiry's performance was not shown clearly and was compared with the indirect immunofluorescent (IFI) anti-Leishmania antibody assay that we know has poor performance.
The rK28 protein, constructed as a chimeric protein based on the fusion of L. infantum rK9, rK26 and rK39 epitopes, presented high sensitivity (99%) and specificity (96%) in the ELISA in samples of dogs from Italy. (22) In human VL, the rK28-ELISA showed a sensitivity of 96. .6% and specificity of 96.2-100.0%. (23,24) In Latin America, immunoenzymatic tests developed with rK28 for the CanL diagnosis showed a sensitivity of 91.0% in the qualitative and quantitative ELISA and specificity of 100% in qualitative and 98.7% in the quantitative assay. (25) Presently other promising recombinant Leishmania antigens that we include in this project became available, and it would be important to study them in other geographical regions, including Brazil, since regional variation in infection features may present.
The recombinant antigen KR95 is a kinesin-bound protein comprising peptide sequences identified from L. donovani infected patients with no corresponding homology in humans. Besides, although the identity between KR95 and Trypanosoma cruzi kinesin is 79%, there appears to be no cross-reactivity. (26) Finally, rK18 was established as useful for both diagnosis and treatment follow up, with a decrease in antibody concentrations over 180 days after treatment. (27) Parasitological examinations that require invasive procedures were not performed on canine survey samples. Instead, the sera were subjected to the Leishmania direct agglutination test (DAT) which has good diagnostic performance, both in human and canine VL. In a study with patients from Ethiopia and Brazil, DAT had 100% sensitivity and specificity. (28) In CanL in the Mediterranean region, DAT reached 100% sensitivity and 98.9% specificity, (29) while in Venezuela the DAT sensitivity was 92.59% (75.69-99.09%), specificity was 100% (79.4-100.0%). (30) Thus, the DAT was our reference test for comparative purposes with the established protocol by the Ministry of Health and also with ELISAs tests with recombinant antigens.
In this study, we validated the ELISA with recombinant Leishmania antigens, suggesting the rK39 antigen for the diagnosis of CanL combined with immunochromatographic DPP assay or alone. Although the rK39-ELISA test has been evaluated in European dogs, (21,31) demonstrating good sensitivity and specificity, there are no studies on the standardization and validation of the ELISA with a large panel of dog samples in Brazil. Therefore, the knowledge of these data is made necessary to establish evidence on the performance of the CanL serological diagnosis in the studied regions. Further, the rK39-ELISA here shown with better diagnostic performance can be extended for diagnosis and management of CanL worldwide. To calculate the sensitivity, specificity and cutoff of the ELISA with recombinant antigens and to determine the cut-off of the DAT, the following samples were obtained from the IAL/SP, where the parasitological examination, DPP and L. major-like-ELISA (characterized samples) were performed:

Samples
-Sera from 74 symptomatic dogs from areas with the canine transmission of L. infantum and the local presence of the vector for Leishmania (Fernandópolis -SP, Votuporanga -SP and Santa Fé do Sul -SP) attended at the Zoonosis Control Centers were used. These dogs had a positive parasitological exam in sternum bone marrow and popliteal lymph node aspirates on glass slides that were fixed and stained with a modified Romanowsky staining rapid test kit (Panótico Rápido, Laborclin, Brazil). In our study, animals with a positive parasitological (gold standard) result were included regardless of the serological result.
-Sera from 66 healthy dogs from areas without canine transmission of L. infantum, with the absence of the vector for Leishmania (Jundiaí -SP). These dogs were negative by the tests recommended in the algorithm of the Brazilian Ministry of Health (DPP and ELISA-L. major-like) and DAT.
To evaluate whether subclinically VL infected dogs could be detected by recombinant antigen-ELISA, sera from 19 asymptomatic dogs were obtained from areas with the canine transmission of L. infantum and the local presence of the Leishmania vector (Fernandópolis -SP, Votuporanga -SP and Santa Fé do Sul -SP). These dogs were positive by the tests recommended in the algorithm of the Brazilian Ministry of Health (DPP and ELISA-L. major-like) and DAT.
To evaluate if an infection could be detected earlier than the recommended assays currently in use, pairedsamples from 30 dogs in the municipality of Bauru (located in the State of São Paulo) were obtained at day 0 (time 1) and again at 180 days (time 2) for evaluation. These dogs used collars impregnated with deltamethrin (Scalibor -Intervet) to reduce the risk of a new infection during the period of evaluation. The collars were changed according to the manufacturer's recommendation or replaced in case of losses, in order to avoid infection through contact with the vector.
To evaluate if the recombinant antigens cross-react with other diseases, we selected 10 samples from dogs without clinical signs for CanL but reagents for Ehrlichia canis and non-reagents for Borrelia burgdorferi, Anaplasma phagocytophilum, Anaplasma platys and Dirofilaria immitis in SNAP 4DX Plus ® test (IDEXX Laboratories Inc., Westbrook, ME, USA). These animals came from São Paulo, a non-endemic municipality for VL.
Sample size calculation -Aiming at a 95% sensitivity for the rK39-ELISA, rK28-, rKR95-and rK18-tests (32) and considering a 95% confidence interval (95% CI) of ± 5%, the sample size needed was at least 73 VL dogs. Aiming at a 98% specificity, at least healthy 47 controls were required, (32) considering a 95% CI of ± 4%. (33) For the ELISA's validation with recombinant antigens, the sample size was calculated according to Banoo et al. (33) Considering a 95% confidence level, an infection prevalence of 15% and an error of 5%, the sample size defined for the study of each region was 196 dogs. tive DPP ® test, was performed according to the manufacturer's recommendations. The Ministry of Health recommends that only samples positive by DPP should be tested by L. major-like-ELISA for confirmation. However, in this study, all samples were submitted to this ELISA.
-DAT (Institute of Tropical Medicine -Antwerp/Belgic) -DAT was performed at the Instituto de Medicina Tropical, Universidade de São Paulo, according to the manufacturer's recommendations. To calculate the agglutination test cut-off, the same groups (Fig. 2), as reported above, were used. Control samples were diluted in series (factor 2) ranging from 1/20 through 1/20,480, and from the positive and negative sample titers, the receiver operating characteristic (ROC) curve was constructed, obtaining the DAT sensitivity, specificity, and cut-off values. In the canine survey of the evaluated regions, the canine sera were also diluted in series (factor 2) ranging from 1/20 through 1/20,480.
-Evaluation of recombinant antigen-ELISA in laboratories within Leishmania endemic areas -Protocols were distributed to laboratories in Leishmania endemic regions and the technicians trained accordingly. ELISA with recombinant antigens were performed at the laboratory of the IMT/USP, the coordination center, and at laboratories in other locations (Laboratório de Leishmanioses da Faculdade de Medicina Veterinária da UFMT, IAL/ SP, IAL/RP) and Laboratório de Infectologia Veterinária do Hospital-Escola de Medicina Veterinária da UFBA), where leishmaniasis caused by L. infantum is endemic. The results were compared to ensure the functionality of the tests, regardless of the location, provided they were carried out under appropriate conditions. Evaluation of samples from a follow-up study with dogs in the endemic area -To evaluate if the studied assays can detect an early infection in animals, samples from 30 dogs from a cohort of 600 dogs from Bauru -SP collected at day 0 (time 1) and 180 days (time 2) of follow-up were evaluated. Samples from this cohort had been previously tested by the DPP + L. major-like-ELISA tests in Bauru -SP. The samples of the 30 dogs selected were submitted to rK39-ELISA, rK28-ELISA, and rKR95-ELISA. These dogs received collars impregnated with deltamethrin (Scalibor -Intervet) that were changed every 6 months or replaced in case of losses, to avoid infection through contact with the vector during the study.
Evaluation of cross-reactivity with E. canis -To evaluate if the recombinant antigens detect antibody related to other diseases, we selected 10 samples from dogs without clinical signs for CanL but reagent for E. canis in SNAP 4DX Plus ® test (IDEXX Laboratories Inc., Westbrook, ME, USA). These samples were non-reagent for B. burgdorferi, A. phagocytophilum, A. platys and D.
Statistical analysis -With the results of the absorbances obtained in each assay with specific recombinant antigens, the percentage of absorbance of the positive standard (ABS% Positive) was calculated by dividing the absorbance of each sample by the absorbance of the positive standard serum and multiplying by 100. (34) The ROC curves were constructed using the ABS% Positive, obtaining the values of sensitivity, specificity and 95% confidence intervals and determining the cut-off. The agreement of the results of the techniques (two by two) was performed using the Kappa index. The stability of the sensitized plates was evaluated by linear regression of each sample. To compare the reactivity indices of samples with the recombinant antigens, the Friedman test (paired samples) was used, followed by Dunn's multiple comparison test (nonparametric data). The comparison of the sensitivity and specificity of the methods used was made by McNemar's chi-square test and the comparison between proportions by the Fisher's exact probability test or the chi-square test. Statistical analysis was performed using the GraphPad Prism 5 (GraphPad Software Inc., San Diego, CA, USA) and SigmaStat 3.5 analysis system software (Systat Software, Richmond, CA, USA).

RESULTS
Initial sample characterization -The samples were characterized according to the diagnostic algorithm recommended by the Brazilian Ministry of Health (MH). This involves initial evaluation with DPP followed by L. major-like-ELISA of DPP-reactive samples. When the sera of 74 symptomatic dogs with positive parasitological examination were submitted to the MH algorithm, the sensitivity was 93.2% (95% CI: 85.1-97.1%) and the specificity 100.0% (95% CI: 94.6-100.0%) with 66 samples from healthy dogs (all samples were negative in DPP).
We also characterized samples by their reactivity in DAT. The ROC curve was constructed from the titers of positive samples (dogs with L. infantum infection) and negative samples from healthy dogs, generating a sensitivity of 97.3% (95% CI: 90.6-99.7%) and specificity of 100.0% (95% CI: 94.6-100.0%) at a 1/640 cut-off point (Fig. 3). Of the 74 parasitologically positive symptomatic VL dogs, the MH algorithm was positive in 69 (sensitivity 93.2%), while the DAT was positive in 72 (sensitivity 97.3%); one sample did not react in either test (McNemar's test, p = 0.3711). Although the difference between the MH algorithm and DAT sensitivity was not significant, given the greater number of samples positive by DAT, we determined to use DAT as a reference to compare the ELISA performance using recombinant antigens.
Evaluation of cross-reactivity with E. canis -After verifying the good performance of the tests with the recombinant antigens, we evaluated the possibility of cross reaction with E. canis, a disease transmitted mainly by ticks and very common in the clinical practice of veterinary routine. Of the 10 animals evaluated, 2 (20.0%, 95% CI: 6.0-46.7%) dogs were reactive in the rKR95-ELISA, one (10.0%, 95% CI: 0.0-42.6%) in the DPP and non-reactive (0.0%, 95% CI: 0.0-32.1%) in L. major-like-ELISA, rK39-ELISA, rK28-ELISA and DAT.

Evaluation of each serological diagnostic strategy for the detection of L. infantum-infected dogs in various
localities -The percentages of the positivity of samples from each studied location, using the DAT, MH algorithm (DPP + L. major-like-ELISA), rK39-ELISA, rK28-ELISA and rKR95-ELISA are presented in Figs 5, 6, and Table  III. DAT had significantly higher positivity than the MH algorithm in cities 1, 2, 9 (McNemar's test, p = 0.0455, p = 0.0002, p < 0.0001, respectively), and total samples (McNemar's test, p < 0.001). The positivity was higher in ELISA with recombinant antigens compared with DAT and MH algorithm in total samples (McNemar's test, p < 0.001) and in most locations (McNemar's test, p < 0.05). The exceptions were DAT versus rK39-ELISA, versus rK28-ELISA and versus rKR95-ELISA in city 2; DAT versus rK28-ELISA in city 4 and DAT versus rKR95-ELISA in city 8; MH algorithm versus rKR95-ELISA in cities 6 and 8; MH algorithm versus rK28-ELISA in city 4. Compared with rK39-ELISA, rK28-ELISA detected lower positivity in cities 4 and 9 (McNemar's test, p = 0.0269, p < 0.0001, respectively). Compared with rKR95-ELISA, rK28-ELISA showed lower positivity in cities 1 and 9 (McNemar's test, p = 0.0304, p = 0.0034, respectively). Compared with rK28-ELISA, rKR95-ELISA showed lower positivity in city 8 (McNemar's test, p = 0.0014). There was no significant difference in positivity between rK39-ELISA and rKR95-ELISA (McNemar's test, 0.0896 ≥ p ≤ 0.7548). All evaluations were performed with the recombinant antigen rK18, however, as its performance was worse than the other antigens of the study, its results were not presented.
Evaluation of the prognostic ability of recombinant antigen ELISA -Since ELISA using recombinant anti-  gens showed higher sensitivity than both the DAT and the MH algorithm, we considered the possibility that the ELISA would allow an earlier detection of the infection than the currently recommended assays. To test this, we evaluated samples' reactivity from a cohort of 600 dogs sampled over an extended period. Considering the MH algorithm results, samples from 30 dogs were selected and distributed in two groups with 15 dogs in each. Group 1 consisted of 15 dogs that at time 1 were considered negative at the initial evaluation for Leishmania infection by the MH algorithm but subsequently became positive six months later (time 2). Group 2 consisted of samples from 15 dogs that seroconverted from DPP negative at time 1 to DPP positive at time 2 but were negative by L. major-like-ELISA at both periods.

DISCUSSION
Considering the fundamental role of dogs in the VL epidemiology, diagnostic algorithms are proposed for routine surveillance to detect L. infantum-infected dogs in control programs. Previous studies show unsatisfactory performance (9,35) of the currently used assays in the diagnostic routine in Brazil. This study evaluated ELISA using four different recombinant antigens, rK39, rK28, rKR95 and rK18, hypothesizing that these could provide better diagnostic performance relative to the current strategy. We obtained excellent results with the rK39 antigen (97.3% sensitivity and 100.0% specificity), while rK18-ELISA performed poorly in the analysis. Results generated with recombinant proteins rK28 and KR95 were also satisfactory, with sensitivity and specificity values only slightly lower than those observed with rK39. The results of sensitivity and specificity obtained in our study with the rK39 antigen are close to those found by Scalone et al. (21) (sensitivity of 97.1% and specificity of 99.4%) who evaluated Italian dogs in ELISA test. Different from these results, in a study carried out by Pereira et al., (36) the rK39-ELISA test showed a lower sensitivity of 82%. In that evaluation, the selection criterion of infected animals was only by serological methods (DPP + L. major-like-ELI-SA), not being used another method for the confirmation of the infection, leaving a margin of doubt about the real infection of these dogs that could thus interfere in the sensitivity results. Importantly, when samples from a group of early-infected still subclinical dogs were evaluated, the accuracy (95% CI) of ELISA with recombinant K39, K28, KR95 and K18 was 97.7% (91.6-99.9%), 96.6% (90.0-99.2%), 97.7% (91.6-99.9%) and 84.1% (74.9-90.4%), respectively. These data suggest the detection of infected dogs before the emergence of overt clinical signs.   As the main purpose of the test is its use in serological inquiry in endemic areas for VL, we conducted a search with 2,530 canine samples collected for serological inquiries in four Brazilian states. Although the parasitological examination is the gold standard for CanL diagnosis, in serological surveys of endemic areas, it becomes impracticable to perform this method due to the demand of execution time, invasive collection material and specialized team. Thus, in our study, we used the DAT, which shows good diagnostic performance in CanL, as a reference test for comparing survey samples. We submitted the samples to the MH algorithm and additionally to the DAT. The reactivities were 28.1% for rK39-ELISA, 27.1% for rKR95-ELISA, and 26.6% for rK28-ELISA, compared with 16.6% with the MH algorithm. These data suggested that the tests with these recombinant antigens detect earlier the dogs with subclinical infections. We evaluated a situation substituting total L. major-like ELISA by rK39-ELISA in the MH algorithm, then the reactivity reached 17.4%, thus similar, tending to be higher in this comparison. In both evaluations, both in DPP + L. major-like-ELISA and DPP + rK39-ELISA, the concordances with the DAT, the reference test, were good. Since the total antigen production is a limiting factor in producing diagnostic kits for national demand, the use of recombinant antigens for ELISA kits emerges as a real possibility to replace L. major-like-ELISA.
Analyzing the performance of the different tests in samples obtained from different geographical areas, endemic for zoonotic VL, we observed that the presently used assays had non-satisfactory performance, compared with tests based on recombinant antigens (K39, K28 and KR95), in city 1 (PA), city 2 (PA), city 3 (BA), city 5 (SP), city 7 (SP) and city 9 (MT). The reason may be the intra-specific variability of L. infantum in these locations. The ELISA with recombinant antigens K39, K28 and KR95 showed good and similar performance in the global analysis. Still, analyzing the samples according to geographical regions, we observed variable per-formances with rK28-ELISA and rKR95-ELISA being worse than rK39-ELISA that showed good performance regardless of the geographic area. Thus, rK39-ELISA appears to be a good assay for serological inquiry in any region. We should consider further that the recombinant antigen K39 is based on the sequence of L. infantum (syn. L. chagasi), (15) a species causing zoonotic VL in Brazil. To reinforce our view that recombinant antigens are highly specific, data by Scalone et al. (21) show no cross-reactivity with other diseases using rK39-ELISA.
In the study carried out by Zanette et al., (37) there was cross-reactivity with E. canis, Toxoplasma gondii and Neospora caninum using the Kalazar Detect™ test, which uses the rK39 antigen. However, in our study, the 10 sera of dogs positive for E. canis did not react in the ELISA with the rK39 and rK28 antigens. Although the number of samples was small, these results may discard the possibility of cross-reactivity at least with E. canis of samples positive for rK39-ELISA and rK28-ELISA.
With rKR95-ELISA, although the protein identity of rKR95 was 79% with T. cruzi, no cross-reactivity was seen with human patients' sera. (26) In our study, there was a cross-reaction with two of the ten dogs positive for E. canis.
On the data showing higher positivity of ELISAs with recombinant antigens, around 10% higher than the result obtained with the MH algorithm, we considered indicative of the existence of a considerable number of false-negative results.
To address this possibility, we evaluated animals from a cohort of 600 dogs from an endemic area for L. infantum infection submitted to periodic evaluation. The dogs in this part of the study were previously evaluated through physical examination and serology (MH algorithm). From the negative diagnosis (DPP + L. majorlike-ELISA), the animals received collars impregnated with deltamethrin (Scalibor-Intervet) that were changed according to the manufacturer's recommendation or replaced in case of losses to avoid infection through con-tact with the vector. Thus, the risk of infection by the vector was extremely reduced. Despite being the gold standard in the diagnosis of CanL, the parasitological examination presents variable sensitivity, (38) especially in asymptomatic animals, (39) so we do not evaluate these animals parasitologically.
We selected 15 dogs (Group 1) that were negative using the DPP test at time 1 that became positive using the MH algorithm six months later (time 2). Although the number of samples was reduced, the results were conclusive. Using rK39-ELISA, around 67% of the dogs turned positive among those considered negative by the MH algorithm at time 1. We also evaluated another 15 dogs from the same cohort that were initially negative in DPP in time 1 that turned out to become positive in DPP at time 2 (Group 2). Using rK39-ELISA, around 47% of these samples were positive, bringing concern on the performance of immunochromatographic rapid assays that are in use worldwide, although it was used here in one brand of kit.
These results suggest that ELISAs with Leishmania recombinant antigens detect infections earlier than the MH algorithm in use. All animals in group 1 in the second collection were reactive to the tests recommended by the Ministry of Health and we believe that some were already infected since the first collection, since anti-Leishmania antibodies were detected with the antigens used in our study. We may rule out the possibility of false positive results using ELISA with recombinant antigens considering the results of the present study showing above the analysis with positive samples confirmed by parasitological exams, and the high specificity obtained. These data bring concerns on the efficacy of measures to control L. infantum transmission, implying the maintenance of infected dogs in the disease cycle with a consequent risk for the human and canine population.
To ensure the ELISAs functionality in other laboratories, we evaluated the viability of applying the tests in laboratories of endemic areas for CanL. For this, we counted on the partnership of collaborating institutions that carried out the tests. The results of the concordance analysis were good to perfect and thus, we conclude that the standardized tests at the coordination center IMT can be safely reproduced by other laboratories provided they are submitted to appropriate test conditions. Further, the antigen-coated plates were stable up to 180 days when maintained at -20ºC and 4ºC.
In Brazil, the Leish-Tec ® vaccine is the only one approved for the prevention of Canine Visceral Leishmaniasis, and several debates regarding serological methods for the differentiation of vaccinated and infected dogs have been raised. In our study we did not evaluate groups of vaccinated dogs, but we believe that the antigens validated in the present study do not detect antibodies from animals vaccinated in the ELISA because they have different epitopes. In the study conducted by Campos et al., (40) dogs vaccinated with Leish-Tec ® were not reagent to DPP and L. major-like-ELISA in the 30, 180 and 360 days after the vaccination protocol.
In this study, we validated the ELISA with recombinant Leishmania antigens, suggesting the rK39 antigen for the diagnosis of CanL. This test is efficient, and it seems there is a superiority in sensitivity compared with the tests currently in use in Brazil. Thus, rK39-ELISA used as a single screening and the confirmatory test is a concrete possibility for the diagnosis of dogs infected with L. infantum, both in Brazil and in other locations where the dog is the reservoir. Although the control of CanL in Brazil was our primary interest, considering the increasing importance of VL caused by L. infantum in other parts of the world, the rK39-ELISA here shown with high diagnostic performance can be extended to the diagnosis and management of CanL worldwide.