Serological tests fail to discriminate dogs with visceral leishmaniasis that transmit Leishmania infantum to the vector Lutzomyia longipalpis

Introduction: The control of reservoirs for Leishmania infantum-induced zoonotic visceral leishmaniasis requires the identification of dogs posing a population risk. Here, we assessed the performance of several assays to identify Lutzomyia longipalpis infectious dogs. Methods: We evaluated 99 dogs that were positive for visceral leishmaniasis based on parasite identification. Serological analyses were performed using an enzyme-linked immunosorbent assay, immunofluorescence antibody tests in 1:40 and 1:80 dilutions, rapid dual path platform tests, immunochromatographic assay with a recombinant rK39 antigen, fast agglutination screening tests, and direct agglutination tests. We also performed PCR to analyze peripheral blood and xenodiagnosis. Results: Forty-six dogs infected at least one L. longipalpis specimen. Although the serological test sensitivities were above 85% for detecting L. longipalpis infectious dogs, none showed a satisfactory performance, as both specificity (0.06 to 13%) and the area under the receiver operating characteristic curve (45 to 53%) were low. The PCR results were also weak, with a sensitivity of 30%, specificity of 72%, and an area under the receiver operating characteristic curve of 51%. The infected L. longipalpis proportion was higher among asymptomatic dogs than symptomatic dogs. Among the symptomatic dogs, those with ulceration-free skin diseases were more infectious, with an odds ratio of 9.3 (confidence interval of 1.10 428.5). The larger the number of insects fed, the greater the detected infectiousness. Conclusions: Our study supports the imperative to develop novel technologies for identifying the infectious dogs that transmit L. infantum for the benefit of public health.


INTRODUCTION
Visceral leishmaniasis (VL), also known as kala-azar, is responsible for approximately 60 thousand deaths yearly worldwide.The Phlebotominae sub-family of sandflies transmit VL 1 , with an estimated 200,000 to 400,000 new victims each year 2 .Lutzomyia longipalpis sandflies transmit the protozoa Leishmania infantum to spread the zoonotic form of VL across Central Asia, the Middle East, the Caucasus and Mediterranean regions, West Africa, and the Americas 3,4 .The main L. infantum reservoirs are wild dogs, marsupials, the domestic dog, and humans [5][6][7] .
Zoonotic VL most commonly occurs in Brazil.Moreover, Brazil likely has the only zoonotic VL transmission control program, which is based on the use of insecticide and culling infected dogs 8,9 .However, a systematic review of the use of these strategies, especially the canine reservoir control program, has revealed little if any impact on the prevalence of zoonotic VL 10 .Notably, these programs have not been successful in limiting the urbanization and territorial expansion of the disease in Brazil after more than 30 years of implementation and despite the efforts of sanitation authorities 11,12 .
The failure of the canine reservoir control program is in part due to its inability to correctly identify infected dogs 10,13 .Serosurveys have been performed in areas with a low zoonotic VL prevalence; thus, many non-infected dogs have been culled.Moreover, many infected dogs remain in communities with a high zoonotic VL prevalence.The acknowledgment of such flaws has led some to question the program on both economic and moral grounds 14,15 .Methods to precisely identify not only the infected animals but also those that represent a greater risk of transmitting L. infantum to the vector L. longipalpis are needed immediately.

Animals
The present study involved 99 domestic dogs (58 males and 41 females) of different breeds and ages who were parasitologically confirmed for canine visceral leishmaniasis (CVL) and diagnosed at the Animal Sanity Laboratory at the Federal University of Piauí (UFPI) in Brazil.After a clinical evaluation, the dogs were administered CVL diagnostic exams.None of the dogs were serologically screened, but all were destined to be culled.

Ethical considerations
The study was approved by Ethics Committee on Animal Experimentation at the UFPI, under review number 053/2008.

Clinical evaluation and sample collection
All dogs were submitted to careful clinical evaluation from a veterinary doctor and diagnosed as symptomatic or asymptomatic.Dogs were considered symptomatic if they had at least one clinical sign suggestive of CVL, while asymptomatic dogs appeared completely healthy at the clinical examination.Skin features such as periocular and generalized alopecia, hair loss, seborrhea, and depigmentation in the muzzle were recorded to note the presence of skin disease without ulceration.Afterward, 10 mL of blood was collected from the jugular vein for serology and 40µL of blood was collected from the ear onto filter paper for deoxyribonucleic acid (DNA) extraction.To obtain samples for parasitological examination, bone marrow and popliteous lymph node aspiration were performed, in addition to scraping healthy skin or lesions in the ear or muzzle.The samples were dyed with Giemsa and observed at 100x magnification to visualize parasite amastigote forms.After confirming the infection, the dogs were humanely sacrificed and an aspiration puncture of the liver and spleen were performed.Culling was performed using an endovenous administration of a combination of ketamine, acepromazine, and diazepan.After 15 minutes of sedation, 20mL of 10% potassium chloride was administered endovenously.
Bone marrow, popliteous lymph node, liver, and spleen samples were cultured in NNN medium enriched with Schneider medium (Sigma-Aldrich, St Louis, MI, USA) and incubated at 26°C in a low-temperature biological oxygen demand (BOD) refrigerated incubator.Every 5 days, the cultures were examined under a microscope at 40x to control the Leishmania promastigote forms, repeating this procedure until the 30th incubation day or until a positive result was obtained.The cultures were considered negative after one month of testing.

Infectiousness evaluation
To study infectiousness, a xenodiagnosis was performed with dogs that were sedated using 0.2% acepromazine.L. longipalpis females were obtained from a sandfly replenished colony that was initially established in 1995.First generation L. longipalpis females were kept at 26°C and 92% humidity, used five days after completely hatching, and deprived of any food source.Sixteen to 120 L. longipalpis females per dog were used, with a mean of 55 and median of 60 sandflies per animal.The sandflies were placed inside darkened plastic boxes with an approximate 12cm diameter and 5cm height.The boxes were open on one side, covered with organza tissue, and placed over the internal ear skin for 45 min.The xenodiagnosis was performed after 4:00pm.
Next, the insects were placed inside a BOD incubator with cotton embedded in a 50% sugar solution.After the fifth day of blood repast, the females were dissected on sterile slides for promastigote count under 40x magnification.

Statistical analysis
To verify the accuracy of the tests that identified the dogs that were infectious to the vector, we calculated sensitivity, specificity, and the area under the receiver operating characteristic (ROC) curve for each test and the respective 95% confidence intervals.A Fisher's exact test was performed to evaluate the possible relationship between the results of xenodiagnoses, as a dichotomic variable (positive and negative), and symptomatology, also as a dichotomic variable (symptomatic and asymptomatic).To analyze of the proportion of infected insects in the asymptomatic and symptomatic groups, a chi-square test was performed.The association between clinical signs and vector infectiousness was also evaluated by calculating the odds ratios (using univariate and multivariate analyses) and the Cornfield approximation 95% confidence interval.All analyses were conducted using the statistical package Stata® (College Station, TX, USA).
Of the 99 dogs investigated, 88 were symptomatic and 11 were completely asymptomatic.Xenodiagnosis identified 46 dogs that infected at least one L. longipalpis specimen.Of the symptomatic dogs, 39 (44%) had a positive xenodiagnosis result, while seven (64%) asymptomatic dogs transmitted to the vector; however, this difference was not statistically significant (p-value = 0.34, Fisher's exact test; Table 2).
The proportion of infected sandflies among all 2,646 examined was 5.8%.The proportion was higher among those that fed on an asymptomatic dog (17.6%) compared to those  that fed on a symptomatic dog (4.5%; p-value = 0.00, chi-square test; Table 2).
Of all the annotated clinical signs, only skin disease without ulceration was significantly associated with the likelihood of infectiousness in dogs (i.e., infecting at least one sandfly), with an odds ratio of 9.3 (1.10 -428.5 confidence interval; Table 3).Additionally, in the multivariate analysis, skin disease without ulceration was significantly associated with infectiousness to the vector (Table 3).
Mortality until insect dissection was 49.8% and ranged from 4 to 100%, with a mean of 26.7 and a median of 25 dissected insects per dog that ranged from 3 to 81.The greater the number of insects used or dissected insects for xenodiagnoses, the greater the chance of at least one being infected; Table 4).

DISCUSSION
For almost 30 years, serological tests for CVL diagnosis have been used extensively for disease control programs in Brazil years 17 .These tests are antibody-based and do not effectively distinguish infected dogs from healthy dogs with an immunological memory of a previous infection, as we have previously shown 16,18 .Therefore, a large number of seropositive, not infective dogs have been culled in Brazil.This policy has high economical and moral costs, now facing increasing community resistance due to its inefficiency 10,19,20 .In addition, dogs correctly found infected may not be actually infective to sandflies and place no risk for the population.
Studies inconsistently report the proportion of dogs with VL that are infectious to the vector L. longipalpis.In Brazil, Deane and Deane 21 observed that 75% of 16 dogs transmitted L. infantum to 24.8% of 238 L. longipalpis specimens, with a mean of 14.9 insects per dog.Sherlock 22 found that 65% of 20 dogs infected 29% of 368 L. longipalpis specimens, with a mean of 18.4 insects per dog.In Europe, Rioux et al. 23 reported having infected all 17 Phlebotomus ariasi that fed on a sick dog.Molina et al. 24 in Spain found that 13/16 dogs were infectious for Phlebotomus perniciosus, with only three dogs infecting over 70% of the sandflies evaluated.Differences in the infectiousness rate could have been due to a number of parasites acquired after an infectious repast (which depends on the infectiousness level of the dog), the type of maintenance diet sugars in the laboratory after repast 25 , temperature and humidity conditions, the sandfly species, and mainly, the time elapsed until dissection 26    result, insects with a high infectiousness load may have died before being examined, while insects that acquired milder infections could have survived until the dissection day.However, this possibility is unlikely due to the short period (5 to 6 days) between feeding on the dogs and insect examination, as there was not enough time for the sandflies to develop high infection loads.
The present study was most concerned with whether every infected dog is actually infectious.From this perspective, infectiousness would depend on a high infectiousness load above a certain threshold in the insects used for xenodiagnosis, as our study suggested when a high number of insects were examined.Our study was not conclusive because only a small number of xenodiagnostic examinations were performed using a high number of insects.Nonetheless, under natural conditions, animals may be exposed to a high number of L. longipalpis bites when they are infectious; therefore, all infected dogs may indeed be infectious to some degree and for some period time, depending on an exposure to a critical amount of sandflies.Notably, xenodiagnostic studies 21,27 have shown that some more infectious animals can be discriminated, while less infectious animals cannot.However, transmission is an exponential function depending on vector daily mortality 28 , which is high, from 15 to 20% 29 ; thus, poorly infectious dogs may not contribute to transmission.These studies suggest that diagnostic tests can distinguish more infectious dogs, which are the actual sources of infection under natural conditions, from less infectious or non-infectious dogs.An elegant study in Ethiopia showed that in fact only a few individuals contribute to transmission 30 .A control program that discriminates infectiousness would identify such animals.
Our study showed that no test could be used to discriminate infectious dogs from those that were previously infected.These serological tests identify antibodies but not parasites, although antibody titers may be related to infectiousness 31 .Our results show that only ICrK39 and FAST could indicate infectiousness.RT-DPP and ELISA, which are used in the Brazilian VL control program 32 , were unable to discriminate the most infectious dogs.Negative DAT, FAST, and ICrK39 results can be used to distinguish some portion of the most infectious dogs, but a significant proportion of these animals will remain in the community.Consequently, although some diagnostic tests can reveal the infectiousness status of the dog population, they are unable to distinguish the infected dogs that actually represent a population risk.
Serological tests cannot be used to indicate infectiousness because they only measure antibodies.PCR could be more useful in this regard.However, PCR yielded some of the worse results.We analyzed only a very small volume of blood, equivalent to what 20 sandflies ingest 33 ; thus, our approach may explain our results.As a result, with a low circulating parasite density 34 , PCR and xenodiagnosis would discriminate different animals, as skin but not blood appears to be the main source of L. infantum transmission to vectors 35 .
In our study, asymptomatic dogs had a greater probability of infecting L. longipalpis than symptomatic dogs, making the correct identification of dogs representing a population risk even more difficult, although the literature is conflicting 31,[36][37][38][39][40] .In contrast, dogs with dermatitis were much more infectious, and this feature was the only clinical sign that had any discriminatory value.Although dermatitis is a low-frequency clinical sign, this information could be useful given the difficulty of identifying infectiousness-related signs.Quantitative buffy coat 41 and molecular tests 40 are potentially useful approaches to identify infectious dogs, as they can directly identify circulating parasites.However, given that VL transmission from dogs likely occurs via the skin 39,40,42 , new approaches to identify infectiousness should focus on skin rather than blood samples.
Vertebrate host infectiousness data vary widely, indicating that this elusive information is unstable, uncertain, complex, and dependent on multivariate factors.In addition, directly measuring infectiousness based on xenodiagnosis is complex, limiting its use on a large scale and for long periods of time.Therefore, to better correctly identify infectious individuals, as we attempted, further investigation is necessary.Such experiments should focus on identifying and controlling the variables that determine the complexity of infectiousness and the development of simple technological alternatives to xenodiagnosis.Until then, the present investigation and any similar future studies will only substantiate the fact that the available methods are inefficient.

de Mendonça IL et al
. -CVL tests do not identify infectious dogs

TABLE 1
Sensitivity, specificity, and the area under the receiver operating characteristic curve for xenodiagnosis of canine visceral leishmaniasis in an endemic region.indirect immunofluorescence reaction in a 1:40 dilution.c indirect immunofluorescence reaction in a 1:80 dilution.d direct agglutination tests.e fast agglutination screening tests.f immunochromatographic test with rK39 antigen.g rapid dual path platform tests.
a 95% confidence interval.b h enzyme-linked immunosorbent assay.i peripheral blood polymerase chain reaction.

TABLE 2
Number and proportion of infected dogs with a positive xenodiagnosis result and the number and proportion of infected insects among all the insects examined in the asymptomatic and symptomatic groups.
. As a

TABLE 4
Effect of increasing the number of Lutzomyia longipalpis females examined or used in xenodiagnosis under the presence of at least one infected female (positive xenodiagnosis).

TABLE 3
Symptoms associated with dog infectiousness to Lutzomyia longipalpis.
de Mendonça IL et al. -CVL tests do not identify infectious dogs