Evaluation of Western blot, ELISA and latex agglutination tests to detect Toxoplasma gondii serum antibodies in farmed red deer

Veterinary Parasitology 244 (2017) 154–159

Contents lists available at ScienceDirect

Veterinary Parasitology journal homepage: www.elsevier.com/locate/vetpar

Research paper

Evaluation of Western blot, ELISA and latex agglutination tests to detect Toxoplasma gondii serum antibodies in farmed red deer

MARK

⁎

Kandarp Khodidas Patela, Laryssa Howea, , Cord Heuera, Geoffery William Asherb, Peter Raymond Wilsona a b

Institute of Veterinary, Animal, and Biomedical Sciences, Massey University, Private Bag 11222, Palmerston North 4442, New Zealand AgResearch, Invermay Agricultural Centre, Private Bag 50034, Mosgiel 9053, New Zealand

A B S T R A C T Abortion due to Toxoplasma gondii has been suspected in New Zealand farmed red deer. However, knowledge around the epidemiology and prevalence of T. gondii in farmed red deer is limited. The aim of this study was to firstly, assess the sensitivity and specificity of two commercially available assays, ELISA and latex agglutination test (LAT), for use in deer and secondly, to estimate the sero-prevalence of T. gondii in red deer. A total of 252 sera from rising 2-year-old and adult hinds from 17 New Zealand red deer herds at early and late pregnancy scanning and from known aborted and/or non-aborted hinds were tested for the presence of T. gondii antibodies. Each assays’ sensitivity and specificity was evaluated by both the Western Blot (WB) as a gold standard method and Bayesian latent class (BLC) analysis in the absence of a gold standard. The sensitivity and specificity for WB were 95.8% (95% credible interval: 89.5-99.2%) and 95.1% (95% credible interval: 90.6-98.1%), respectively. For the LAT at the manufacturer’s recommended ≥1:32 cut-off titre, the sensitivity (88.7%, 95% credible interval: 80.8-94.7%) and specificity (74.3%, 95% credible interval: 67.5-80.5%) were lower and higher than the sensitivity (76.2%, 95% credible interval: 66.7-84.5%) and specificity (89.7%, 95% credible interval: 84.5-93.9%) at a ≥1:64 cut-off, using (BLC) analysis. Sensitivity and specificity of the LAT at cut-off titre of 1:32 were estimated to be 84.4% (95% CI: 74.9-90.9%) and 73.5% (95% CI: 65.8-79.9%) against WB. The LAT had better agreement with WB at cut-off titre of ≥1:64 than ≥1:32 (Kappa = 0.63 vs 0.54). At optimised cut-off S/P of 15.5%, the sensitivity (98.8%, 95% credible interval 96.1-99.8%) and specificity (92.8%, 95% credible interval 88.9-95.7%) of the ELISA were higher and lower, respectively, than the sensitivity (85.1%, 95% credible interval 76.2-91.9%) and specificity (98.5%, 95% credible interval 96.9-99.4%) at manufacturer’s cut-off S/P of 30%, from BLC analysis. The sensitivity and specificity of ELISA at S/P cut-off of 15.5% was 91.1% (95% CI: 83.2-96.1%) and 90.7% (95% CI: 85.2-94.7%), respectively, when assessed against WB. The sero-prevalence from ELISA and LAT, at cut-off of S/P 15.5% and ≥1:64, respectively, was not significantly different to that from WB (McNemar’s Chi-square p = 0.21 for ELISA and p = 0.28 for LAT).

The true sero-prevalence in the study population, from its posterior distribution from BLC analysis, was estimated to be 33.0% (95% credible interval: 27.3-38.9%). The evaluated ELISA with optimised cut-off can be used to detect T. gondii exposure in red deer. 1. Introduction In New Zealand (NZ), red deer (Cervus elaphus) farming is an established industry with almost one million deer on approximately 2000 farms. Sub-optimum reproductive performance of rising two-year-old (R2) hinds has been a persistent problem (Asher and Wilson, 2011;

⁎

Corresponding author. E-mail address: [email protected] (L. Howe).

http://dx.doi.org/10.1016/j.vetpar.2017.08.003 Received 29 March 2017; Received in revised form 6 August 2017; Accepted 7 August 2017 0304-4017/ © 2017 Elsevier B.V. All rights reserved.

Audigé et al., 1999). Reproductive efficiency (number of calves weaned/number of hinds mated) in R2 and mixed age herds (MA) is estimated to be about 75% (Statistics New Zealand, 2016). A preliminary clinical study reported mid-term abortion rates ranging from 2% to 16% across four large rising 2-year-old (R2) deer herds and 4.5% in one mixed aged hind herd (Wilson et al., 2012). Possible causes of gestational losses in deer may include farm management practices, such as crop poisoning (Nordkvist et al., 1984), poor nutrition (Thorne et al., 1976) and infectious causes including Toxoplasma gondii (Dubey et al., 2014; Dubey et al., 2008b; Wilson et al., 2012), Leptospira spp. (Subharat et al., 2010; Trainer et al., 1961),

Veterinary Parasitology 244 (2017) 154–159

K.K. Patel et al.

Japan). The antibody end-point titres to T. gondii used were as per manufacturer’s instructions wherein sera with titre < 1:32 were considered sero-negative and sera with titre equal to 1:32 were considered weak sero-positive and ≥1:64 sero-positive. Data were analysed independently for both cut-offs. A commercial anti-ruminant immunoglobulin G (IgG) based T. gondii ELISA test (Chekit-Toxotest, IDEXX laboratories, Bern, Switzerland) was performed at the Institute of Veterinary Animal and Biomedical Sciences, Massey University on a duplicate of all 252 serum samples, according to the manufacturer’s instructions. The sample to positive control (S/P) ratio was calculated for all samples and the sera were interpreted as sero-positive or sero-negative based on the manufacturer’s recommended cut-off S/P of 30% and an “optimised” S/P, as generated using the receiver operating characteristic (ROC) dataset generated by WinBUGS in BLC analysis as described below.

cervid herpes virus (CvHV) (das Neves et al., 2009a; das Neves et al., 2009b), and bovine viral diarrhoea virus (BVDv) (Passler et al., 2009; Passler et al., 2007) infections. Recently, a clinical investigation of fetal loss in farmed deer, identified T. gondii DNA in brain tissue of aborted fetuses found at the time of scanning, and uteri of non-pregnant and aborted R2 hinds on NZ deer farms (Wilson et al., 2012). That study reported a 40% T. gondii sero-prevalence using a latex agglutination test (LAT) suggesting a possible T. gondii exposure association with abortion. Reichel et al. (1999), using a LAT, reported a 52.5% sero-prevalence in NZ farmed deer with sero-prevalence increasing with age. Collectively, while knowledge around the epidemiology and prevalence of T. gondii in farmed deer is limited, these observations suggest that T. gondii might play a role in reducing reproductive performance. Toxoplasma gondii, a protozoon parasite, is found worldwide and infects warm blooded animals including domestic and wild animals and humans (Dubey and Beattie, 1988). Ingestion of T. gondii from the environment or in foodstuffs, possible venereal or vertical transmission and subsequent infection in animals can cause reproductive losses including abortion, fetal death, early embryonic loss, mummification, and still birth leading to economic losses (Buxton et al., 2007; Freyre et al., 1997; Wanderley et al., 2015; Consalter et al., 2017). Historically, T. gondii has been isolated from fetal and placental tissues of aborting sheep during abortion storms in New Zealand (Hartley et al., 1954), three aborted kids and goats in USA (Dubey, 1981), and from brain and/or liver of six aborted and stillborn lambs in United Kingdom (Beverley and Watson, 1961). In addition, T. gondii has also been isolated from fetuses of white-tailed deer in six of 15 early, and nine of 27 mid-term pregnancies in the USA (Dubey et al., 2008b). Investigation of suspect causative agents involved in fetal loss or clinical or subclinical disease in hitherto unstudied species requires tests assessed for sensitivity (Se) and specificity (Sp) specifically for each species, to inform test choice and interpretation. In New Zealand, the only commercially available assays for use in the diagnosis of toxoplasmosis are an immunoglobulin G (IgG) enzyme linked immunosorbent assay (ELISA) and a LAT. Thus, the primary aim of this study was to assess the sensitivity and specificity of the commercial ELISA and LAT for use in deer. Evaluations used both an in-house western blot (WB) as the gold standard (Sohn and Nam, 1999) and Bayesian latent class (BLC) statistics assuming absence of a gold standard (Joseph et al., 1995). Additionally, results of this study contributed to knowledge of sero-prevalence in farmed deer in NZ.

2.3. Western Blot (WB) For the WB, live tachyzoites of a New Zealand strain (S48) of T. gondii maintained in Vero cells (Toxovax®, Schering-Plough Animal Health, Wellington, New Zealand) (Hartley 1975) were used to extract crude antigen as previously described by Harkins et al. (1998) with some minor modifications. Briefly, thirty millilitres of 99% pure tachyzoite suspension containing 1 × 104 tachyzoites/ml was centrifuged at 1500 x g for 10 min to pellet tachyzoites. The pelleted tachyzoites were re-suspended in 2 ml PBS (pH 7.4, Invitrogen, Carlsbad, USA) and disrupted by three cycles of freezing/thawing. This was followed by seven cycles of sonication (Sonics Vibra-cellTM, Sonics & Materials Inc., Newtown, Connecticut, USA) on ice. The sonicated tachyzoites were centrifuged at 12,000 x g for 30 min at 4 °C to remove debris and the supernatant containing the water soluble proteins was collected. The protein content was determined using a nano-spectrophotometer (NanoDrop ND-1000, Thermo Scientific, Wilmington, Delaware, USA) and stored in aliquots. Prepared antigen (2.2 μg/μl) was denatured using 2X lamelli buffer (Sigma-Aldrich, St. Louis, Missouri, USA) containing 4% SDS, 20% glycerol, 10% 2-mercaptoethanol, 0.004% bromophenol blue and 0.125 M Tris-HCL by heating at 100 °C for 10 min. The denatured protein was run through a 12.5% Tris-HCL gel (Criterion, Bio-Rad, Hercules, California, USA) for 90 min at 100 V. Each WB was run with a broad range (10–250 kD) protein marker (Precision Plus Protein™, BioRad), a negative and positive control serum. After separation, the proteins were electrophoretically transferred on to polyvinylidene difluoride (PVDF) membrane (Bio-Rad) for 45 min at 70 V. After confirmation of complete protein transfer through Ponsceu S staining (Sigma-Aldrich), membranes were blocked overnight at 4 °C using 5% blotting solution (5 g skim milk powder with 1% PBS (Phosphate buffer solution)-Tween-20 (PBS-T) (Sigma-Aldrich). Blocked membrane was cut into strips and incubated with test sera, diluted 1:50 in 5% blotting solution with 1% PBS-T, at room temperature for 60 min. After primary incubation with sera samples, strips were washed three times for 10 min using 1% PBS-T. Secondary antibody rabbit anti-deer antibody conjugated to horseradish peroxidase (KPL, Gaithersburg, Maryland, USA), diluted 1:9000 in 5% blotting solution with 1% PBS-T, was added to each strip and incubated for 60 min at room temperature. Strips were washed three times as described above for wash after primary incubation and then incubated at room temperature (22 °C) for five minutes in chemiluminescent solution (Amersham™ ECL Select™ Western Blotting Detection Reagent, GE Healthcare UK Limited, Buckinghamshire, UK) and arranged on a transparent plastic film. The film was exposed for 30 s in luminescent image analyser (LAS-1000, Fujifilm, Tokyo, Japan) and the bands were measured against the standard markers. Presence of bands indicating immunodominant antigens (IDA) of T. gondii were recorded. For this study, due to concerns regarding cross-reactivity with other apicomplexans (Harkins et al., 1998), sera that recognised two or more IDA with molecular weight (MW) between 13–95 kD were

2. Materials and methods 2.1. Samples Sera used for this study were selected from a sample bank from an investigation of abortion in farmed red deer from throughout New Zealand as described by Wilson et al. (2012) and Patel (2016). Blood samples were collected from randomly selected pregnant, non-pregnant, and aborting hinds by jugular veni-puncture into 10 ml vacuum blood collection tubes without anticoagulant, and transported chilled to the Institute of Veterinary, Animal and Biomedical Sciences, Massey University where they were centrifuged at 1512g for 15 min and serum withdrawn and stored at −20 °C. A total of 252 sera were randomly selected from the serum bank from 17 farms and analysed using WB, ELISA and LAT. According to a power analysis, 245 sera were required to determine ELISA Sensitivity and Specificity of ≥85% assuming a T. gondii sero-prevalence of 20% with an absolute precision of 10%, with 95% confidence (Cannon and Roe 1982). 2.2. Serological assays All 252 samples were sent to a commercial pathology laboratory (NZ Veterinary Pathology Ltd, Palmerston North) for T. gondii antibody screening using the LAT (Toxoreagent, Eiken Chemical & Co, Tokyo, 155

Veterinary Parasitology 244 (2017) 154–159

K.K. Patel et al.

with continuous outcomes based on the latent class analysis methods to determine disease prevalence and parameters for diagnostic test with dichotomous outcomes as described by Joseph and Gyorkos (1996) and Joseph et al. (1995). ELISA S/P values were transformed on a natural logarithmic scale before being analysed in the BCDT software and then back transformed for reporting of results. Each model was run for 50,000 iterations with a burn-in discard period of 5000 iterations. Convergence of the posterior iterates was assessed in WinBUGS by examining history trace plots and by specifying two sets of dispersed initial values and examining the convergence statistic. Additionally, the BLC analyses also provided positive predictive value (PPV) and negative predictive values (NPV) for ELISA and LAT. The influence of the parameter priors on posterior estimates of P, Se, Sp, PPV, NPV of LAT, ELISA and WB was evaluated in a sensitivity analyses using uniform, optimistic and pessimistic priors. To maximise SeELISA, a cut-off S/P for ELISA was selected from the receiver operating characteristic (ROC) dataset generated by WinBUGS. SeELISA, SpELISA, PPVELISA, and NPVELISA at the optimised cut-off are reported and compared with those at the manufacturer’s recommended cut-off.

considered positive for T. gondii. As the S48 strain has the same antigenic structure as other strains of T. gondii (Wastling et al., 1994; Harkins et al., 1998), two positive control red deer sera from naturally infected hinds with a confirmed abortion due to T. gondii (Wilson et al., 2012), and from an S48 vaccinated hind at 30 days post-vaccination which showed consistent strong bands at 13, 23 kD, 28 kD, 30 kD, and 34 kD, with minor bands at 42 kDa, 62 kDa, and 95 kDa acted as internal controls loading and antibody probing. Serum from a red deer with no history of T. gondii infection which did not show reactivity in the WB was used as a negative control. Each sample was examined independently for reactivity by two individuals to confirm a positive or negative response. Discrepant samples were re-examined or re-run if necessary. 2.4. Data analysis 2.4.1. Bayesian latent class (BLC) analyses in the absence of a gold standard test This analysis was used to determine Se, Sp, PPV, and NPV of the three tests and sero-prevalence (P) using a latent class model assuming that WB was not a gold standard test. A posterior distribution was derived combining the observed data with prior knowledge about the parameters (SeELISA, SpELISA, SeWB, SpWB, SeLAT, SpLAT, and P). All three tests were assumed to conditionally depend on the true T. gondii exposure status and with constant test accuracy in the test population. Prior information on Se and Sp for LAT (Dubey et al., 1995; Hokmabad et al., 2011) and WB (Basso et al., 2013) and T. gondii sero-prevalence (P) (Reichel et al., 1999; Wilson et al., 2012) were available from data in past studies on deer and other species and/or from expert opinion from the laboratory applying the test for research purpose (Table 1). For example, T. gondii sero-prevalence recorded earlier in farmed red deer in New Zealand by LAT was 52.5% in 1997 (Reichel et al., 1999) in 40% in 2011 (Wilson et al., 2012). Additionally, as the LAT was suspected as having a lower Se and Sp and hence over-estimated seroprevalence in deer, it was assumed that the most likely sero-prevalence would be 20% and greater than 10%, corresponding to a beta distribution of (α=5.1, β=17.6) (Table 1). Similarly, the most likely SpWB value was assumed to be 95%, from a previous study on pigs (Basso et al., 2013), and 99% from an expert opinion (L. Howe, personal communication 2012), that corresponded to a beta (α=4.18, β=1.2) distribution (Table 1). The priors for the mean of the log-transformed ELISA S/P values for T. gondii sero-positive and sero-negative hinds group was set to three and one, respectively. The standard deviation (SD) for the mean of log-ELISA S/P of both sero-positive and sero-negative was set to one. Independent beta prior distributions were derived for the parameters using Betabuster software (www.epi.ucdavis.edu/ diagnostictests/betabuster.html) to check for their uncertainty using modal or most probable values and 5th and 95th percentiles. The model estimating P, SeELISA, SpELISA, SeWB, SpWB, SeLAT, and SpLAT was fitted to the animal data assuming independence across herds in WinBUGS (Lunn et al., 2000) using the Bayes Continuous Diagnostic Test (BCDT) software package version 3.7 (Division of Clinical Epidemiology, 2015). BCDT software was developed for diagnostic tests

2.4.2. Western Blot as a gold standard test Analyses using WB as gold-standard were performed in SAS statistical software version 9.3 (SAS Inc., Cary, NC, USA). The ELISA S/P values were log-transformed on a natural logarithmic scale for normality and analysis purposes and were back transformed for reporting of results. Sensitivity, Sp, PPV, and NPV for both tests were calculated using qualitative results from WB. McNemar’s Chi-square test statistic was performed to statistically compare sero-positive proportions (or sero-prevalences) of WB vs. ELISA, and WB vs. LAT. The Kappa statistic was estimated to check the level of agreement between WB and ELISA and, WB and LAT. Based on the Kappa statistic, the test agreement can be interpreted as < 0.2 = slight agreement; 0.2 − 0.4 = fair; 0.4 − 0.6 = moderate; 0.6 − 0.8 = substantial; > 0.8 = almost perfect (Dohoo et al., 2003). 2.5. Ethics statement All procedures performed on animals were approved by the Massey University Animal Ethics Committee (Protocol 12/34). 3. Results 3.1. Western Blot The estimated sero-prevalence from WB was determined to be 35.7% (90/252; 95% CI: 29.8 − 41.6%). The immunoglobulin G (IgG) antibodies reacted with protein antigens ranging from 13 to 95 kDa on WB as previously reported in sheep, cattle and goats infected with S48, RH, and ME T. gondii strains (Wastling et al., 1994; Conde et al., 2001; Harkins et al., 1998). The most frequent observed band with molecular weight of 30 kDa was present in 91.1% of positive sera, followed by 13 kDa (82.2%), 23 kDa (77.8%), 34 kDa (63.3%), and 28 kDa (62.2%), with 90% (81/90) of positive samples reacting to three or more of these antigens. From the BLC analysis, the SeWB and SpWB were 95.8% (95% credible interval: 89.5% to 99.2%) and 95.1% (95% credible interval: 90.6% to 98.1%), respectively (Fig. 1).

Table 1 Informative beta priors used for estimation of test sensitivity (Se) and specificity (Sp) of Western blot (WB) and latex agglutination test (LAT), and true Toxoplasma gondii seroprevalence in study population in a Bayesian latent class model. Parameter

Mode (%)

95% sure greater than (%)

Corresponding beta prior distribution (α, β)

Sero-prevalence WB sensitivity WB specificity LAT sensitivity LAT specificity

20 95 95 90 80

10 50 50 50 50

5.1, 17.6 4.18, 1.2 4.18, 1.2 5.38, 1.49 7.55, 2.63

3.2. Lat Using the cut-off titre of ≥1:64, 32.5% of sera were positive, whereas a further 14.7% were weak positive with titre of 1:32 (Table 3). Bayesian latent class analysis estimated the SeLAT and SpLAT at a cut-off titre of ≥1: 32 to be 88.7% and 74.3%, respectively (Table 2). At a cut-off titre of ≥1:64, the SeLAT and SpLAT were 76.2% and 89.7%, respectively. When the cut-off was raised from 1:32 to ≥1:64 the PPVLAT increased from 63.6% to 79.2%, and the NPVLAT 156

Veterinary Parasitology 244 (2017) 154–159

K.K. Patel et al.

The agreement between ELISA and WB was substantial at cut-off S/P of ≥15.5% (Kappa = 0.80) and 30% (Kappa = 0.79) (Table 3). When the SeELISA was assessed using WB as gold standard, the selected cut-off S/P of ≥15.5% increased the SeELISA to 91.1% from 78.9% at the cut-off S/P of ≥30%. Conversely, SpELISA was lower at the cut-off S/P ≥15.5% at 90.7% compared with 97.5% at S/P ≥30% cut-off. The selected S/P 15.5% cut-off decreased the PPVELISA from 94.7% to 84.5% but increased the NPVELISA from 89.2% to 94.8% when compared with the S/ P 30% cut-off (Table 3). 3.4. Toxoplasma gondii true sero-prevalence The true sero-prevalence, from its posterior distribution from BLC analysis, in the study population was estimated at 33.0% (95% credible interval: 27.3% to 38.9%). It should be noted that application of weakly informed prior or good prior in BLC analysis had little effect on the posterior median values compared with those obtained using informative priors (data not shown). The change in sensitivity and specificity estimates due to variation in prevalence across herds was very low (data not shown). 4. Discussion

Fig. 1. Western blot patterns of a sub-sample of sera. Negative control sample (a); a negative serum sample (b); positive control sera (e and f) from two naturally infected hinds with confirmed Toxoplasma gondii abortions and two positive serum samples (c, d) showing positive banding patterns to Toxoplasma gondii antigens from 13 to 95 kD.

This is the first evaluation of LAT and ELISA serological tests for Toxoplasma gondii in farmed red deer and used both gold standard and Bayesian latent class analyses techniques. While both assays had reasonable Se and Sp when interpreted according to manufacturer’s recommendations, BLC analysis of the ELISA determined that the optimum SeELISA and SpELISA was achieved when an optimised cut-off S/P of 15.5% was used. At this optimized S/P 15.5% cut-off, the SeELISA and SpELISA were higher than SeLAT and SpLAT at both 1:32 and 1:64 cutoffs, and equivalent to the gold standard SeWB and SpWB as determined from BLC analysis. Hence, the ELISA at the optimised cut-off, performed better than the LAT. This observation supports the need for evaluation and optimisation of tests used for each species. Overseas, T. gondii IgG sero-prevalence has been determined using commercial or in-house ELISA or the widely used MAT assays (Dubey et al., 2008a; Schaefer et al., 2013; Jokelainen et al., 2010). Unfortunately, the MAT assay is not available in New Zealand. Thus, the LAT assay is the preferred assay for New Zealand diagnostic laboratories as it is not species specific and detects both immunoglobulin G (IgG) and immunoglobulin M (IgM). However, the LAT is not without its’ limitations as it requires user interpretation as it involves microscopic observation of agglutination endpoint allowing for possible intra- and inter-laboratory variation. Despite the lower SeLAT at the ≥1:64 titre compared to ≥1:32, the improved concordance between the LAT and WB at the higher cut-off justifies use of ≥1:64 as the recommended cut-off titre for the LAT when testing deer serum for the presence of T. gondii antibodies. The improved concordance may be due to the higher concentration of antibodies leading to clearer agglutination patterns, reducing user ambiguity, and improving the identification of true sero-negative animals. Thus, this data supports that the recommended cut-off titre of ≥1:64 should be used in preference to the lower cut-off in laboratories employing the LAT. However, the lower

decreased from 93.0% to 88.1%, (Table 2). Using WB as gold standard and the cut-off titre of ≥1:32, the SeLAT and SpLAT was 84.4% and 73.5% with PPVLAT and NPVLAT of 63.9% and 89.5%, respectively, and at the cut-off titre of ≥1:64, the SeLAT and SpLAT was 72.2% and 89.5% with PPVLAT and NPVLAT of 79.3% and 85.3%, respectively (Table 3). Using the McNemar’s Chi-square test, the sero-prevalence from LAT was significantly higher (p < 0.001) at the ≥1:32 cut-off when compared to those from WB (Table 3), but not at cut-off titre of ≥1:64 (p = 0.28). Agreement between LAT and WB was substantial at LAT cut-off titre of ≥1:64 (Kappa = 0.63) whereas it was moderate at a cut-off titre of ≥1:32 (Kappa = 0.54) (Table 3). 3.3. ELISA At the manufacturer’s recommended cut-off S/P of ≥30%, using BLC analysis, the SeELISA and SpELISA were 85.1% and 98.5%, respectively, and 29.8% of sera were positive (Table 2 & 3). An optimised cutoff S/P of ≥15.5% was determined using ROC data, yielding a SeELISA and SpELISA of 98.8% and 92.8%, respectively (Table 2, Fig. 2). The area under the ROC curve (AUC), as determined from BLC analysis was 0.98 (95% CI: 0.97 − 0.99). Sero-prevalence using this cut-off, was 38.5% (Table 3). The optimised cut-off decreased the PPVELISA from 96.5% to 87.0% but increased the NPVELISA from 93.2% to 99.4% (Table 2), when compared with the manufacturer’s cut-off S/P of ≥30. The sero-prevalence from ELISA using optimised cut-off S/P of ≥15.5% was not significantly different to that from WB (McNemar’s Chi-square p = 0.21), but was significantly lower (McNemar’s Chisquare p = 0.001) when using the manufacturer’s cut-off S/P of ≥30%.

Table 2 Test sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV) with their 95% credible intervals for LAT and ELISA as obtained from Bayesian latent class analysis with informative priors at optimised and manufacturer’s cut-off. Parameter

LAT

ELISA

Cut-off

Titre 1:32 (manufacturer)

Titre 1:64

S/P:15.5%

S/P:30% (manufacturer)

Sensitivity (%) Specificity (%) Positive predictive value (PPV) (%) Negative predictive value (NPV) (%)

88.7 74.3 63.6 93.0

76.2 89.7 79.2 88.1

98.8 92.8 87.0 99.4

85.1 98.5 96.5 93.2

(80.8–94.7) (67.5–80.5) (54.7–72.0) (87.4–96.8)

157

(66.7–84.5) (84.5–93.9) (69.2–87.7) (82.5–92.5)

(96.1–99.8) (88.9–95.7) (80.0–92.6) (97.9–99.9)

(76.2–91.9) (96.9–99.4) (92.6–98.6) (88.4–96.4)

Veterinary Parasitology 244 (2017) 154–159

K.K. Patel et al.

Table 3 Apparent sero-prevalence, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and Kappa statistic values (95% confidence interval) and McNemar’s Chi-square statistic for latex agglutination test (LAT) and ELISA at optimised and manufacturer’s cut-off when compared with Western blot as a gold standard. Parameters

LAT

Cut-off

Titre: 1:32 (manufacturer)

Titre: 1:64

S/P:15.5%

S/P:30% (manufacturer)

Apparent sero-prevalence% Sensitivity% Specificity% Positive predictive value (PPV) % Negative predictive value (NPV) % Kappa statistica McNemar’s Chi-square statisticb

47.2 84.4 73.5 63.9 89.5 0.54 14.7

32.5 (26.8–38.3) 72.2 (61.8–81.2) 89.5 (83.7–93.8) 79.3 (70.6–85.9) 85.3 (80.5–89.0) 0.63 (0.53–0.73) 1.5 (p = 0.28)

38.5 91.1 90.7 84.5 94.8 0.80 2.13

29.8 (24.1–35.4) 78.9 (68.8–86.5) 97.5 (93.4–99.2) 94.7 (86.2–98.3) 89.2 (83.5–93.2) 0.79 (0.71–0.87) 9.8 (p = 0.001)

a b

ELISA

(41.1–53.4) (74.9–90.9) (65.8–79.9) (54.5–72.3) (82.7–93.9) (0.44–0.64) (p < 0.001)

(32.5–44.5) (83.2–96.1) (85.2–94.7) (77.1–89.9) (90.5–97.3) (0.73–0.88) (p = 0.21)

Agreement between LAT or ELISA and Western blot. Difference between sero-positive proportion when LAT or ELISA are used compared with sero-positive proportion when Western blot is used.

required, given the decrease in SpELISA was by a lesser 5.7 percentage points. Also, higher SeELISA (98.8%) and NPVELISA (99.4%) at optimum cut-off would ensure fewer false sero-negatives and therefore will help to establish a more accurate assessment of T. gondii exposure in farmed deer. However, higher SpELISA will be required to confirm clinical toxoplasmosis in individual animals, for which a manufacturer’s recommended cut-off could be used. It should be noted, although unlikely, the presence of other parasites in the same family, such as Neospora caninum and Sarcocystis spp., may affect the specificity of this ELISA when used in deer and possible cross-reactivity should be examined further. Moreover, the resulting NPVELISA and PPVELISA are dependent on the true prevalence of T. gondii in the study population and therefore the NPVELISA of 99.7% at the optimum cut-off in this study will not be the same in a test population with different T. gondii sero-prevalence. Given the poor reproductive performance and suspect involvement of T. gondii in a preliminary investigation by Wilson et al. (2012) in deer, the ELISA can be used to detect T. gondii exposure in non-pregnant and aborting deer. The WB has been previously used as a gold standard test to detect T. gondii antibodies in feline(Sohn and Nam, 1999) and ovine (Wastling et al., 1994; Wastling et al., 1995) species. The band range observed on the immunoblot image in this study for deer was comparable with band range obtained in sheep (Wastling et al., 1994), goat (Conde et al., 2001), pig (Al-Adhami and Gajadhar, 2014), and cat (Sohn and Nam, 1999) studies. Whilst few studies on the T. gondii sero-prevalence in the deer have reported sensitivity and specificity for the serological assays using WB as a gold standard, a Mexican study on white-tailed deer (Odocoileus virginainus), employing an in-house ELISA, reported a SeELISA of 73.3% and SpELISA of 100% at cut-off S/P of 30% (OlamendiPortugal et al., 2012). These assay sensitivities were similar to the SeELISA and SpELISA at the manufacturers cut-off S/P of 30% reported in this study. However, when WB was used as a gold standard, the SeELISA and SpELISA estimates differed at the optimized cut-off S/P of 15.5% wherein the SeELISA and SpELISA were 16.8 percentage points higher and 9.3 percentage points lower again raising discussion as to the purpose of the assay whether the better detection of true positives or true negatives is more desirable. This current study also provides the first report of WB sensitivity (95.8%) and specificity (95.1%), using BLC analysis, for the use in deer. While sero-prevalence estimation was not the primary aim of the study, the BLC analysis also provided a true sero-prevalence of 33.0% in this study population. At the manufacturers’ cut-off, the LAT estimated a 14.4 percentage point higher sero-prevalence whereas the ELISA under-estimated sero-prevalence by 3 percentage points. However, when optimised cut-off values were used, the sero-prevalences estimated by LAT and ELISA were within 3 percentage points of the seroprevalence estimated by WB (35.7%) and within 0.5 and 5.5 percentage points, respectively, of the true prevalence.

Fig. 2. Receiver operating characteristic (ROC) curve plotted for ELISA S/P from BLC analysis.

SeLAT at the ≥1:64 cut-off suggests that this cut-off may not be optimum and more research is required in this area. For example, it is possible that the poor concordance of the LAT with the WB seen in this study is due to the detection of IgM positive samples. To the authors knowledge, there have been no studies measuring IgM in deer, thus, the clinical relevance of detecting IgM is unknown. A study in sheep and another in goats has shown that the IgM titre only lasts for approximately 30 days post infection (Lundén, 1995; Conde de Felipe et al., 2007). However, given the age of the hinds ( > 2 years of age) and random sample selection method, it is unlikely that all of the discordant samples were from animals which had been exposed to T. gondii within 30 days of sampling. There have been no studies in deer examining the development of IgG response to T. gondii, but IgG levels are detectable for 10 days post infection in experimental challenge studies in sheep and pigs (Wastling et al., 1994; McColgan et al., 1988; Basso et al., 2017). Whilst most of these studies use the MAT, the IgG ELISA test evaluated here can be modified to determine antibody avidity and hence the timing of the infection, whether the antibodies raised are due to acute or chronic infection, can also be assessed (Syed-Hussain et al., 2013). Additionally, ELISA is less time-consuming and can be semi-automated, hence higher throughput can be achieved with less intra- and inter-laboratory variation. The optimised ELISA S/P cut-off was chosen to achieve higher Se in order to improve the identification of animals at risk of exposure to T. gondii and the potential clinical effects such as in an abortion outbreak investigation. This information could then be used to assess preventative measures, vaccination, or monitoring at either an individual or herd level. The 13.7 percentage point higher SeELISA justifies use of the optimised 15.5% cut-off when the highest sensitivity was 158

Veterinary Parasitology 244 (2017) 154–159

K.K. Patel et al.

Dubey, J.P., Dennis, P.M., Verma, S.K., Choudhary, S., Ferreira, L.R., Oliveira, S., Kwok, O.C., Butler, E., Carstensen, M., Su, C., 2014. Epidemiology of toxoplasmosis in white tailed deer (Odocoileus virginianus) Occurrence, congenital transmission, correlates of infection, isolation, and genetic characterization of Toxoplasma gondii. Vet. Parasitol. 3, 00031–00034. Dubey, J.P., 1981. Epizootic toxoplasmosis associated with abortion in dairy goats in Montana. J. Am. Vet. Med. Assoc. 178, 661–670. Freyre, A., Bonino, J., Falcon, J., Castells, D., Correa, O., Casaretto, A., 1997. The incidence and economic significance of ovine toxoplasmosis in Uruguay. Vet. Parasitol. 73, 13–15. Harkins, D., Clements, D.N., Maley, S., Marks, J., Wright, S., Esteban, I., Innes, E.A., Buxton, D., 1998. Western blot analysis of the IgG responses of ruminants infected with Neospora caninum and Toxoplamsa gondii. J. Comp. Path. 119, 45–55. Hartley, W.J., Jebson, J.L., McFarlane, D., 1954. New Zealand type 2 abortions in ewes. Aust. Vet. J. 30, 216–218. Hartley, W.J., 1975. Case of suspected congenital Toxoplasma encephaloomyelitis in a lamb associated with a spinal cord anomaly. Brit. Vet. J. 131 (4), 380–384. Hokmabad, R.V., Khanmohammadi, M., Farhang, H.H., 2011. Seroprevalence of toxoplasma gondii antibodies in sheep by sabin feldman dye test (SFDT) and latex agglutination test (LAT) in northwest Iran. Ann Biol Res 2, 135–139. Jokelainen, P., Näreaho, A., Knaapi, S., Oksanen, A., Rikula, U., Sukura, A., 2010. Toxoplasma gondii in wild cervids and sheep in Finland: north-south gradient in seroprevalence. Vet. Parasitol. 171, 331–336. Joseph, L., Gyorkos, T.W., 1996. Inferences for likelihood ratios in the absence of a gold standard. Med. Decis. Mak. 16, 412–417. Joseph, L., Gyorkos, T.W., Coupal, L., 1995. Bayesian estimation of disease prevalence and the parameters of diagnostic tests in the absence of a gold standard. Am. J. Epidemiol. 141, 263–272. Lundén, A., 1995. Immune responses in sheep after immunization with Toxoplasma gondii antigens incorporated into iscoms. Vet. Parasitol. 56, 23–35. Lunn, D., Thomas, A., Best, N., Spiegelhalter, D., 2000. WinBUGS − A Bayesian modelling framework: concepts, structure, and extensibility. Stat. Comput. 10, 325–337. McColgan, C., Buxton, D., Blewett, D.A., 1988. Titration of Toxoplasma gondii oocysts in non-pregnant sheep and effects of subsequent challenge during pregnancy. Vet. Record 123, 467–470. Nordkvist, C., Rehbinder, C., Mukherjee, S.C., Erne, K., 1984. Pathology of acute and subchronic nitrate poisoning in reindeer (Rangifer tarandus L). Rangifer 4, 9–15. Olamendi-Portugal, M., Caballero-Ortega, H., Correa, D., Sánchez-Alemán, M.A., CruzVázquez, C., Medina-Esparza, L., Ortega-S, J.A., Cantu, A., García-Vázquez, Z., 2012. Serosurvey of antibodies against Toxoplasma gondii and Neospora caninum in whitetailed deer from Northern Mexico. Vet. Parasitol. 189, 369–373. Passler, T., Walz, P.H., Ditchkoff, S.S., Givens, M.D., Maxwell, H.S., Brock, K.V., 2007. Experimental persistent infection with bovine viral diarrhea virus in white-tailed deer. Vet. Microbiol. 122, 350–356. Passler, T., Walz, P.H., Ditchkoff, S.S., Brock, K.V., DeYoung, R.W., Foley, A.M., Daniel Givens, M., 2009. Cohabitation of pregnant white-tailed deer and cattle persistently infected with Bovine viral diarrhea virus results in persistently infected fawns. Vet. Microbiol. 134, 362–367. Patel, K.K., 2016. Epidemiological Investigation into Abortion in Farmed Red Deer in New Zealand PhD Thesis. Massey University, Palmerston North. Reichel, M.P., Timbs, D., Ross, G.P., Penrose, M.E., 1999. Seroprevalence of leptospira and toxoplasma in New Zealand farmed deer. Surveillance. Wellington Ministry for Primary Industriespp. 5–6. Schaefer, J.J., Kirchgessner, M.S., Whipps, C.M., Mohammed, H.O., Bunting, E.M., Wade, S.E., 2013. Prevalence of antibodies to Toxoplasma gondii in white-tailed deer (Odocoilues virginianus) in New York state, U. S. A. J. Wild. Dis. 49, 940–945. Sohn, W.-M., Nam, H.-W., 1999. Western blot analysis of stray cat sera against Toxoplasma gondii and the diagnostic availability of monoclonal antibodies in sandwich-ELISA. Korean J. Parasitol. 37, 249–256. Statistics New Zealand, 2016. Agricultural Production Statistics. pp. 4–5 June 2015 (final). Subharat, S., Wilson, P.R., Heuer, C., Collins-Emerson, J.M., 2010. Investigation of localisation of Leptospira spp. in uterine and fetal tissues of non-pregnant and pregnant farmed deer. N. Z. Vet. J. 58, 281–285. Syed-Hussain, S.S., Howe, L., Pomroy, W.E., West, D.M., Smith, S.L., Williamson, N.B., 2013. Detection of Neospora caninum DNA in semen of experimental infected rams with no evidence of horizontal transmission in ewes. Vet. Parasitol. 197, 534–542. Thorne, E.T., Dean, R.E., Hepworth, W.G., 1976. Nutrition during gestation in relation to successful reproduction in elk. J. Wildl. Manage. 40, 330–335. Trainer, D.O., Karstad, L., Hanson, R.P., 1961. Experimental leptospirosis in white-tailed deer. J. Infect. Dis. 108, 278–286. Wanderley, F.S., Porto, W.J.N., Camara, D.R., de Oliveira, V.G., garcia, J.L., de Albuquerque, P.P., de Fonseca Oliverira, A.A., Mota, R.A., 2015. Small Rum Res. 123, 301–305. Wastling, J.M., Harkins, D., Buxton, D., 1994. Western blot analysis of the IgG response of sheep vaccinated with S48 Toxoplasma gondii (Toxovax). Res. Vet. Sci. 57, 384–386. Wastling, J.M., Harkins, D., Maley, S., Innes, E., Panton, W., Thomson, K., Buxton, D., 1995. Kinetics of the local and systemic antibody response to primary and secondary infection with S48 Toxoplasma gondii in sheep. J. Comp. Pathol. 112, 53–62. Wilson, P.R., Patel, K.K., Asher, G.W., Howe, L., Heuer, C., Sinclair, G., 2012. Clinicalinvestigations of foetal loss in farmed deer. In: Annual Conference of the Deer Branch of New Zealand Veterinary Association. Queenstown, New Zealand. pp. 107–111. das Neves, C.G., Mørk, T., Thiry, J., Godfroid, J., Rimstad, E., Thiry, E., Tryland, M., 2009a. Cervid herpesvirus 2 experimentally reactivated in reindeer can produce generalized viremia and abortion. Virus Res. 145, 321–328. das Neves, C.G., Rimstad, E., Tryland, M., 2009b. Cervid herpesvirus 2 causes respiratory and fetal infections in semidomesticated reindeer. J. Clin. Microbiol. 47, 1309–1313.

5. Conclusion This study has determined that the ELISA, using the optimised S/P cut-off of 15.5%, is a sensitive test to detect IgG antibodies against T. gondii in red deer and maintains satisfactory specificity. The sensitivity and specificity were lower and higher, respectively, when the manufacturer’s recommended cut-off was used. The ELISA at the optimised cut-off is therefore the test of choice for use in red deer when high sensitivity is required. ELISA had better sensitivity and specificity than the LAT interpreted at the manufacturer’s recommended cut-off. Seroprevalence data supports that exposure of farmed red deer to T. gondii in NZ is common. Acknowledgements The Authors would like to thank the deer farmers, veterinarians, farm staff, and IVABS technical staff for their participation and assistance in the study. The Authors would also like to thank Dr. Geoff Jones for his assistance with Herd Effect statistical analysis. This study was funded by DEEResearch, Massey University, AgResearch, Agmardt, and MSD Animal Health (NZ Ltd). References Al-Adhami, B.H., Gajadhar, A.A., 2014. A new multi-host species indirect ELISA using protein A/G conjugate for detection of anti-Toxoplasma gondii IgG antibodies with comparison to ELISA-IgG, agglutination assay and Western blot. Vet. Parasitol. 200, 66–73. Asher, G.W., Wilson, P.R., 2011. Reproductive productivity of farmed red deer: a review. In. Annual conference of the Deer Branch of New Zealand Veterinary Association 23–29. Audigé, L., Wilson, P.R., Morris, R.S., 1999. Reproductive performance of farmed red deer (Cervus elaphus) in New Zealand − I. Descriptive data. Anim. Reprod. Sci. 55, 127–141. Basso, W., Hartnack, S., Pardini, L., Maksimov, P., Koudela, B., Venturini, M.C., Schares, G., Sidler, X., Lewis, F.I., Deplazes, P., 2013. Assessment of diagnostic accuracy of a commercial ELISA for the detection of Toxoplasma gondii infection in pigs compared with IFAT, TgSAG1-ELISA and Western blot, using a Bayesian latent class approach. Int. J. Parasitol. 43, 565–570. Basso, W., Grimm, F., Ruetten, M., Djokic, V., Blaga, R., Sidler, X., Deplazes, R., 2017. Experimental Toxoplasma gondii infections in pigs: humoral immune response, estimation of specific IgG avidity and the challenges of reproducing vertical transmission in sows. Vet. Parasitology 236, 76–85. Beverley, J.K.A., Watson, W.A., 1961. Ovine abortion and toxoplasmosis in Yorkshire. Vet. Rec. 73 (6-10), 11. Buxton, D., Maley, S.W., Wright, S.E., Rodger, S., Bartley, P., Innes, E.A., 2007. Toxoplasmma gondii < /it > and ovine toxoplasmosis: new aspects of an old story. Vet. Parasitol. 149, 25–28. Cannon, R.M., Roe, R.T., 1982. Livestock Disease Surveys: a Field Manual for Veterinarians. Australian Government Publishing Service, Canberra. Conde de Felipe, M.M., Molina, J.M., Rodriguez-Ponce, E., Ruiz, A., Gonzalez, J.F., 2007. IgM and IgG response to 29–35 kDa Toxoplamsa gondii protein fractions in experimentally infected goats. J. Parasitol. 93 (3), 701–703. Conde, M., Moline Caballaro, J.M., Rodriguez-Ponce, E., Ruiz, A., Gonzalez, J., 2001. Analysis of IgG response to experimental infection with RH Toxoplasma gondii in goats. Comp. Immun. Micro. Infect Dis. 24 (3), 197–206. Consalter, A., Silva, A.F., Franzao-Teixeira, E., Matos, L.F., De Oliveira, F.C.R., Leite, J.S., Silva, F.B.F., Ferreira, A.M.R., 2017. Toxoplamsa gondii transmission by artificial insemination in sheep with experimentally contaminated frozen semen. Theriogenology 90, 169–174. Division of Clinical Epidemiology, M.U. 2015. BayesContinuousDiagnosticTest: A program for estimation of disease prevalence and the characteristics of a continuous diagnostic test and one or two dichotomous diagnostic tests (Montreal, Quebec, M cGIll University). Dohoo, I.R., Martin, W., Stryhn, H.E., 2003. Veterinary Epidemiologic Research. University of Prince Edward Island, Charlottetown, P.E.I, pp. 706. Dubey, J.P., Beattie, C.P., 1988. Toxoplasmosis of animals and man. CRC Press, Florida. Dubey, J.P., Thulliez, P., Weigel, R.M., Andrews, C.D., Lind, P., Powell, E.C., 1995. Sensitivity and specificity of various serologic tests for detection of Toxoplasma gondii infection in naturally infected sows. Am. J. Vet. Res. 56, 1030–1036. Dubey, J.P., Jenkins, M.C., Kwok, O.C.H., Zink, R.L., Michalski, M.L., Ulrich, V., Gill, J., Carstensen, M., Thullize, P., 2008a. Seroprevalence of Neospora caninum and Toxoplamsa gondii antibodies in white-tailed deer (Odocoileus virginianus) from Iowa and Minnesota using four serologic tests. Vet. Parasitol. 161, 330–334. Dubey, J.P., Velmurugan, G.V., Ulrich, V., Gill, J., Carstensen, M., Sundar, N., Kwok, O.C.H., Thulliez, P., Majumdar, D., Su, C., 2008b. Transplacental toxoplasmosis in naturally-infected white-tailed deer: isolation and genetic characterisation of Toxoplasma gondii from foetuses of different gestational ages. Int. J. Parasitol. 38, 1057–1063.

159