Negative dog sera - Negative sera were from 16 dogs living in the city of La Paz, where no Chagas transmission exists. Dogs were born in the city and never went out in an endemic Chagas region. Negativity was checked by the Chagas STAT-PAK rapid test which is an accurate test for Chagas diagnosis in dogs (Nieto et al. 2009 (link)), and by polymerase chain reaction (PCR) targeting the kDNA of T. cruzi following Fernandes et al. (2001) (link), slightly modified by one of us (Aliaga et al. 2011 (link)). The 16 negative sera were included as negative controls in each of the processed titer-plates.
Positive dog sera - 10 positive dog sera were obtained from dogs originated from the same region of the field sample (see below) and diagnosed positive both by PCR using the same protocol as above, and by the Chagas STAT-PAK rapid test following the manufacturer’s instructions. Then, in each ELISA plate, five-10 of them were included as positive controls to allow the computation of a cut-off value with formula F3 (
| Formula | a | f | Computation | Comment |
|---|---|---|---|---|
| F1 | 2 | 0 | 2 x MEAN of negative controls | - |
| F2 | 3 | 0 | 3 x MEAN of negative controls | - |
| F3 | 1 | 0 | MEAN of negative controls + 0.13 x MEAN of positive controls | Pan et al. (1992) formula |
| F4a | 1 | MEAN + f x SD, with f = 2.197 | Frey et al. (1998) formula. Confidence level (1-α) for t computation: 97.5% | |
| F5a | 1 | MEAN + f x SD, with f = 3.848 | Frey et al. (1998) formula. Confidence level (1-α) for t computation: 99.9 % | |
| F6 | 1 | 3 | (MEAN + 3 x SD) of negative controls | Classen et al. (1987) |
a: for the computation of F4 and F5,j is the number of negative controls used in the plate (16 in the present study) and t is the (1-α)th percentile of the one-tailed Student t-distribution with (j-1) degrees of freedom. Because 16 negative controls were used in the study, and taking into account the confidence level for the computation of the Student t, the f values were 2.197 and 3.848 for F4 and F5 respectively.
IgG-ELISA protocol to detect antibodies against T. cruzi - In the laboratory, tubes containing blood samples of dogs were centrifuged at 3000 rpm for 10 min for plasma separation. The ELISA protocol was from Lauricella et al. (1998) (link) which is routinely used for Chagas diagnosis in dogs (Enriquez et al. 2013 (link)). It was slightly modified as follow: ELISAs were carried out in 96-well micro-titer plates (NUNC Maxisorp, flat bottom) coated with a homogenate of T. cruzi epimastigote culture. The homogenate was prepared as follow: 1 mL of pure culture of epimastigotes (forms cultured at 28ºC in LIT liquid medium) was centrifuged in a 5 mL Eppendorff tube at 4000 rpm at 4ºC for 10 min. The supernatant was discarded, 1 mL of phosphate buffer saline (PBS) at pH7.2 was added and the tube vortexed. This washing operation was realised three times. Then, 1 mL of PBS was added and the tube vortexed. A dilution of 1/1000 of the solution was realised in PBS in carbonate buffer (100 µL of “parasites” in PBS + 9900 µL of carbonate buffer), vortexed, and 100 µL of the solution was then added in each well. The plate was sealed with adhesive plastic sheet and incubated overnight at 4ºC. The following day, the content was discarded by inversion. The plate was washed three times with 120 µL/well of washing buffer (PBS - 0.01% Tween 20). Then each well was loaded with 100 µL of blocking buffer (PBS - 3% skimmed milk REGILAIT, France) and incubated 1 h at 37ºC. Then, the plate was washed three times with 120 µL/well of washing buffer. Dog sera were diluted at 1/100 in dilution buffer (PBS - 1% skimmed milk) in 1.5 mL Eppendorff tubes, vortexed and kept at 4ºC until loaded in the plate. Diluted sera were loaded in duplicate at 50 µL/well and incubated 1 h at 37ºC. The plate was then emptied by inversion and washed three times with 120 µL/well of washing buffer. Anti-dog IgG were diluted at 1/1200 in dilution buffer. Each well was loaded with 50 µL of peroxidase conjugated antibodies anti-IgG and incubated 1 h at 37ºC. Then the plate was emptied by inversion and washed three times with 120 µL/well of washing buffer. Then 50 µL of TMB (3, 3’, 5, 5’ - Tetramethylbenzidine, SIGMA) was added in each well and the plate was incubated for 5 min at room temperature. Then, 50 μL/well of sulfuric acid 1 N were added to stop the reaction and absorbance values were obtained at 450 nm in a microwell plate reader (Multiskan). The mean absorbance of each pair of duplicate sera was calculated. When the difference between both values was more than 30%, the sample was retested (Lauricella et al. 1998 (link)). In total, the 231 dog sera and the controls were processed in seven titer plates.
Cut-off formulas (
Change-point analysis - The whole set of sera was also analysed by change-point analysis which does not need the presence of known positive or negative sera (blind analysis). Change-point analysis is aimed at identifying points in a series where the statistical properties change. In particular, such analysis can be used to detect abrupt steps in the mean level of a series. In the case of ELISA, if absorbance values of a micro-titer plate are ordered in ascending order, negative samples are supposed to be the lower ones in the series while positive ones (if they exist) would be the higher. However, values are not supposed to increase regularly if positive samples exist in the series. Indeed, as positive controls are supposed to be “different” from negative ones, a step, even small, should appear in the series, separating the negative from the positive values. Therefore, change-point algorithms might be used to detect such a change and locate the value where in the series this change occurs. The detected value is therefore a kind of specific cut-off proxy that discriminates between positive and negative samples.
For each of the seven processed titer-plates, absorbance values were first arranged in ascending order and each series was analysed using the R package “changepoint” (Killick & Eckley 2014 ) which detects a change-point if it exists and locates it in the series. In this package, the Pruned Exact Linear Time (PELT) algorithm was selected (Killick et al. 2012) with the CUSUM method as detection option (Page 1954) . The PELT algorithm divides iteratively the series of absorbance values in sub-groups of increasing size. In each, it calculates the minimum of a “cost” function that takes into account the method of detection (Killick et al. 2012) . Minima indicate where the change-points are located within the series. The PELT algorithm can therefore rapidly detect various change-points in a series. The CUSUM method is based on cumulative sums and operates as follow: The absorbance values x are ordered in ascending values (x1,…xn) and sums (S) are computed sequentially as S0 = 0, Si+1 = max (0, Si+ xi - Li), where Li is the likelihood function. When the value of S exceeds a threshold, a change-point has been detected.
Approval for the study was granted by the WHO’s Research Ethics Review Committee (ERC), project #A90281 and by the Comisión de Ética de la Investigación del Comité Nacional de Bioética (CEI-CNB) of Bolivia (letters 3 august 2010 and 21 august 2012).
Free full text: Click here
