Use of two commercial caprine arthritis-encephalitis immunoenzymatic assays for screening of arthritic goats

Abstract

Roughly one-fourth of goats infected with small ruminant lentivirus (SRLV) develop caprine arthritis-encephalitis (CAE). We compared the profile of antibody response to surface glycoprotein (SU), and combined transmembrane glycoprotein and capsid protein (TM/CA) in SRLV-infected arthritic and asymptomatic goats, and determined the ability of 2 commercial ELISAs to distinguish between arthritic and asymptomatic goats. We used sera from 312 SRLV-seropositive dairy goats in a whole-virus ELISA; 222 were collected from arthritic goats and 90 from apparently healthy goats. Sera were screened with a competitive inhibition ELISA based on SU antigen (SU-ELISA) and an indirect ELISA based on TM and CA antigens (TM/CA-ELISA). Receiver operating characteristic (ROC) curves were prepared for both ELISAs, and areas under the ROC curves (AUC) were compared. The proportion of goats with antibody response stronger to SU antigen than to TM/CA antigen was significantly higher among arthritic than asymptomatic goats (58.1% vs. 28.9%; p < 0.001). Antibody response to SU antigen was a good predictor of the arthritic form of CAE: AUC for SU-ELISA was 89.7% (95% CI: 85.2%, 94.2%), compared to 59.3% (95% CI: 51.9%, 66.8%) for TM/CA-ELISA (p < 0.001). With the cutoff set at percentage of inhibition of 56%, SU-ELISA had sensitivity of 86.9% (95% CI: 81.9%, 90.7%) and specificity of 84.4% (95% CI: 75.6%, 90.5%) in discriminating between arthritic and asymptomatic goats.

Introduction

Small ruminant lentivirus (SRLV) produces lifelong infection that, in 20–30% of naturally infected goats, leads to the development of emaciation and progressive arthritis affecting mainly the carpal joints.⁵ These are the most typical clinical manifestations of caprine arthritis-encephalitis (CAE), which inevitably result in the culling of diseased goats. SRLV infection induces humoral immune response directed simultaneously to at least 4 main viral antigens: surface (SU, gp135) and transmembrane (TM, gp45) envelope glycoprotein; capsid (CA, p28) and matrix (MA, p16) protein.^2,17 It has been shown that goats reacted more strongly to envelope glycoproteins than to a capsid antigen¹⁰ and that SU glycoprotein played a crucial role in synovial cell infection.⁹ Higher anti-SU and anti-TM antibody titers were found in serum of goats apt to develop clinical arthritis compared to asymptomatic individuals, whereas the level of anti-p28 antibodies in serum did not seem related to arthritis.^11,14 The initial immune response to SU antigen (mediated by Th2 instead of Th1 lymphocytes) was the factor discriminating between goats that developed arthritis compared to those that remained clinically healthy.^3,16,21 Determination of the dominant immunoglobulin (Ig)G1 isotype antibody response to SU antigen in the preclinical phase of CAE proved to be a good predictor of the development of clinical arthritis.²⁰ Nevertheless, this can only be done using Western blot, which is not only commercially unavailable but also not affordable in routine goat medicine.⁶ To date, the mainstays of field diagnosis of CAE are serologic immunoenzymatic assays, based on 1 of 3 different combinations of antigens: whole-virus antigen (mixture of SU, TM, CA, and MA), or combined recombinant TM and CA antigen, or SU antigen.¹⁸

We conducted a study to: 1) compare the antibody response to either SU antigen or TM/CA antigen in SRLV-infected arthritic and asymptomatic goats determined by commercial ELISAs; and 2) determine the ability of commercial ELISAs, which detect the 2 aforementioned antibody responses, to distinguish between arthritic and asymptomatic goats.

Materials and methods

Goats and serum samples

Sera collected from goats during routine diagnostic examinations were used in the study. Sera from 312 dairy goats seropositive in a whole-virus ELISA were used. Goats belonged to Polish White Improved and Polish Fawn Improved breeds and were ≥1 y old at blood sampling. All samples were collected in the summer; goats were in mid-lactation. Blood sample collection was approved by the 3rd Local Ethical Committee in Warsaw (approval 31/2013).

Clinical examination of all goats was performed by 2 experienced veterinarians, diplomates of the European College of Small Ruminant Health Management, and information on goat appetite and daily milk yield was obtained from the owner. Body condition of goats was assessed using lumbar and sternal scorings.¹² Goats that received score 0 or 1 in both scales were considered emaciated, whereas goats with score 2–4 in both scales were regarded as having a normal body condition. Goats were examined for carpal arthritis by observation at rest and in motion, and measuring the carpo-to-metacarpal ratio (circumference of the carpus divided by the circumference of the central part of the metacarpus, measured with a flexible centimeter ruler). Goats that manifested uni- or bilateral forelimb lameness and swelling of one or both carpal joints, defined as the carpo-to-metacarpal ratio of >2, were diagnosed with severe carpal arthritis and concluded to have the arthritic form of CAE. Of the 312 goats, 222 (71.2%) manifested the severe arthritic form of CAE; the remaining 90 goats (28.8%) were apparently healthy.

Serologic testing

Initial selection of samples was carried out using an indirect monophasic ELISA based on the whole-virus antigen (ID Screen MVV/CAEV indirect screening test, IDVet Innovative Diagnostics, France) because this test proved to have very high specificity (99%) in a previous study.¹⁵ Moreover, all positive results were reevaluated using a biphasic confirmatory ELISA (ID Screen MVV/CAEV indirect confirmation test, IDVet) in order to maximize the positive predictive value of the result. The manufacturer’s cutoff of a sample-to-positive control ratio (S/P) of >60% was used in both tests. A serum sample needed to test positive in both ELISAs to be included in the study.

Sera were then screened with a competitive-inhibition ELISA based on SU antigen (SU-ELISA; Small Ruminant Lentivirus Antibody Test Kit, cELISA, VMRD, Pullman, WA) and an indirect ELISA based on TM and CA antigens (TM/CA-ELISA; IDEXX MVV/CAEV p28 Ab Screening, IDEXX, Westbrook, ME). Both tests had been found highly sensitive and specific in detecting SRLV infection in goats.^4,8

Optical density (OD) of serum samples was read using a scanner with the upper OD limit of 4.0 (Epoch Microplate Spectrophotometer, BioTek Instruments, Winooski, VT). Antibody response was expressed by the following indices: 1) in competitive-inhibition ELISA as a percentage of inhibition (%I) given with the formula: %I = 100% × [1 – (sample OD/negative control OD)]; and 2) in indirect ELISAs as S/P ratio (%) = (sample OD – negative control OD)/(positive control OD – negative control OD) × 100%.

The maximum value of the S/P ratio was limited by the upper OD read by the scanner, which returned the reading “overflow” when the OD of a serum sample was >4.0. Given that the highest measurable S/P ratio observed in our study was 238%, all serum samples with an “overflow” reading were arbitrarily assigned a S/P ratio of ≥240%.

Statistical and epidemiologic analysis

Antibody response was given as a median, interquartile range (lower-to-upper quartile, IQR), and raw data. The antibody response between arthritic and asymptomatic goats was compared with a Mann–Whitney U test, separately for each ELISA. To compare the antibody response to SU and TM/CA antigens, ranks were assigned separately to measurements of each of the 2 ELISAs. The rank transformation was applied because antibody response was measured by the 2 ELISAs in different units and could not be compared without transformation, and because many measurements in the indirect ELISA were ≥240%, which was an ordinal category rather than a numerical measurement. Then, the SU-to-TM/CA rank ratio was calculated by dividing the rank of the SU-ELISA result by the rank of the TM/CA-ELISA result for a given goat, and a Mann–Whitney U test was used to compare rank ratios between arthritic and asymptomatic goats. Moreover, as a SU-to-TM/CA rank ratio >1 would mean that antibody response to SU antigen was stronger than to TM/CA antigens, the proportion of such results was compared between arthritic and asymptomatic goats by calculating the difference between percentages with its 95% confidence interval (95% CI) according to the Newcombe method¹ and by using a chi-square test.

Accuracy of the 2 ELISAs in identifying arthritic goats was assessed by drawing a receiver operating characteristic (ROC) curve for each test and comparing areas under curves (AUCs).⁷ Performance at different cutoffs was expressed as sensitivity (Se), specificity (Sp), and an overall parameter—Youden index (J = Se + Sp − 1); the cutoff at which J index took the highest value was considered as the optimal cutoff for discrimination between arthritic and asymptomatic SRLV-seropositive goats. For the chosen cutoff, likelihood ratios (LRs) of a positive (LR+) and negative (LR–) result as well as positive and negative predictive values (PPVs and NPVs, respectively) of results were calculated.

All statistical tests were 2-sided. A significance level (α) of 0.05 was assumed. Analyses were performed in Statistica 12 (StatSoft, Tulsa, OK), and simulations of ELISA performance in EpiTools (http://epitools.ausvet.com.au).

Results

Antibody response to both antigen sets (SU and TM/CA) was stronger in arthritic than asymptomatic goats (Fig. 1). Asymptomatic goats showed significantly weaker antibody response to SU antigen than to TM/CA antigen (median rank ratio: 0.48, IQR: 0.25–1.20), whereas in arthritic goats the antibody response to SU antigen was stronger than to TM/CA antigen (median rank: 1.11, IQR: 0.78–1.50, p < 0.001). Moreover, the proportion of goats with a SU-to-TM/CA rank ratio >1 in arthritic goats (129 of 222, 58.1%) was significantly higher than in asymptomatic goats (26 of 90, 28.9%)—a difference of 29.2% (95% CI: 17.2%, 39.7%; p < 0.001).

Figure 1.

Antibody response to 2 different sets of small ruminant lentivirus (SRLV) antigens—surface glycoprotein (SU) and transmembrane glycoprotein/capsid protein (TM/CA)—in arthritic and asymptomatic SRLV-infected goats.

Antibody response to SU antigen was a fairly good predictor of the presence of the arthritic form of CAE with AUC of 89.7% (95% CI: 85.2%, 94.2%), whereas antibody response to TM/CA antigen performed only slightly better than a test based on random selection (AUC: 59.3%; 95% CI: 51.9%, 66.8%; Fig. 2). The difference between the aforementioned AUCs was 30.4% (95% CI: 23.6%, 37.1%) and was significant (p < 0.001). Percent inhibition (%I) of 56% was the optimal cutoff value to discriminate between arthritic (%I > 56%) and asymptomatic (%I < 56%) goats (Table 1). At this cutoff, a positive result raises the odds of a goat having clinical arthritis roughly 6-fold, whereas a negative result reduces it roughly 7-fold. Thus, the SU-ELISA is a moderately useful detection test.¹³ NPVs are high in the wide range of expected prevalences of the clinical form of CAE, including the figures usually encountered in infected herds (NPV > 80% at a prevalence of 0–60%). On the other hand, PPVs were similarly high only when the clinical form of CAE had affected >40% of goats, which is a rare situation (Fig. 3).

Figure 2.

Receiver operating characteristic (ROC) curves of the surface glycoprotein-based ELISA (SU-ELISA) and the transmembrane/capsid protein–based ELISA (TM/CA-ELISA).

Table 1.

Accuracy of caprine arthritis-encephalitis antibody ELISA based on surface (SU) glycoprotein in identifying arthritic and asymptomatic goats.

Cutoff (%)	No of goats according to SU-ELISA		Sensitivity (%)	Specificity (%)	Youden J index (%)	Positive likelihood ratio	Negative likelihood ratio
	Positive	Negative
10	303	9	100 (98.3, 100)	10.0 (5.4, 17.9)	10.0 (6.8, 13.2)	1.1 (1.0, 1.2)	0
20	292	20	99.5 (97.5, 99.9)	21.1 (14.0, 30.6)	20.7 (16.3, 25.0)	1.3 (1.1, 1.4)	<0.01
30	281	31	99.5 (97.5, 99.9)	33.3 (24.5, 43.6)	32.9 (27.9, 37.9)	1.5 (1.3, 1.7)	<0.01
35*	273	39	99.1 (96.8, 99.8)	41.1 (31.5, 51.4)	40.2 (35.0, 45.4)	1.7 (1.4, 2.0)	0.01 (<0.01, 0.04)
40	264	48	98.6 (96.1, 99.5)	50.0 (39.9, 60.1)	48.6 (43.3, 54.0)	2.0 (1.6, 2.4)	0.02 (0.01, 0.06)
45	248	64	97.3 (94.2, 98.8)	64.4 (54.1, 73.6)	61.7 (56.6, 66.9)	2.7 (2.1, 3.6)	0.04 (0.02, 0.08)
50	232	90	92.3 (88.1, 95.2)	70.0 (59.9, 78.5)	62.3 (57.2, 67.5)	3.1 (2.2, 4.2)	0.11 (0.07, 0.18)
56†	207	105	86.9 (81.9, 90.7)	84.4 (75.6, 90.5)	71.4 (66.9, 75.8)	5.6 (3.4, 9.1)	0.15 (0.11, 0.22)
60	185	137	78.4 (72.5, 83.3)	87.8 (79.4, 93.0)	66.2 (61.7, 70.6)	6.4 (3.7, 11.2)	0.25 (0.19, 0.32)
65	152	160	64.4 (57.5, 70.0)	90.0 (82.1, 94.6)	54.4 (49.5, 58.5)	6.4 (3.4, 12.0)	0.40 (0.33, 0.48)
70	126	186	54.1 (47.5, 60.5)	93.3 (86.2, 96.9)	47.4 (43.1, 51.6)	8.1 (3.7, 17.7)	0.49 (0.42, 0.57)
80	72	240	30.2 (24.5, 36.5)	94.4 (87.6, 97.6)	24.6 (20.7, 28.5)	5.4 (2.3, 13.0)	0.74 (0.67, 0.82)
90	15	297	6.8 (4.1, 10.8)	100 (95.9, 100)	6.8 (5.1, 8.4)	–	0.93 (0.90, 0.97)

Numbers in parentheses are 95% confidence intervals. Dash (–) indicates positive likelihood ratio cannot be computed.

^*Percent inhibition of 35% included because it is the manufacturer’s cutoff for discrimination between seropositive and seronegative goats.

^†Percent inhibition of 56% included as the best cutoff for discrimination between arthritic and asymptomatic goats according to the Youden J index.

Figure 3.

Distribution of positive and negative predictive values (PPVs and NPVs, respectively) of the results of the surface glycoprotein–based ELISA used for detecting goats with the clinical form of caprine arthritis-encephalitis (CAE). Broken vertical lines signify typical proportion of small ruminant lentivirus–infected goats that develop the arthritic form of CAE.

A %I cutoff of 56% yields moderate sensitivity (~87%) and specificity (~84%). Given LR+ and LR– of roughly 5.6 and 0.15, respectively, application of a rule of thumb¹³ shows that a positive result of the test increases probability that a goat is affected by the clinical form of CAE by ~30%, whereas a negative result decreases it by ~35%, regardless of the prior probability of infection. However, prior probability is known when the test is used in a herd with a high SRLV seroprevalence because it can be assumed that the clinical form of CAE affects usually 20–30% of infected goats.⁵ A simulation with prior probability of being affected by the clinical form of CAE set at 25% shows that a positive result would raise it to 64%, which is still a low figure, whereas a negative result would result in 95% post-test probability of being clinically healthy. At this cutoff, only 5 goats with the clinical form of CAE would hide among 100 goats that test negative. If this figure still seems too high, simply lowering the cutoff to %I of 45% will yield Se of ≈97% and result in a negative predictive value of 98%. Now, there would only be 2 such goats. Moreover, in both of the aforementioned scenarios, the PPV is still much higher than the expected prevalence of the clinical form of CAE in a herd (64% and 47%, respectively vs. 20–30%). When only a group of goats that tested positive is then examined clinically, the prior probability of the clinical form of CAE at which examiners are about to make their diagnosis is roughly 2-fold higher than the prior probability among all SRLV-positive goats. As a consequence, examiners are more likely to make a correct diagnosis of CAEV-associated arthritis.

Discussion

Our study confirmed that a different antibody response profile can be demonstrated in CAE-affected goats with and without arthritis using commercial ELISAs—arthritic goats had a much more intense antibody response to SU antigen than did asymptomatic goats, as has been well described previously.^11,14 The immune process responsible for disease progression to arthritis is at least partially mediated by a high level of gp135-specific IgG1 antibodies.^19,20 The novel observation from our study is that this reaction is detectable using commercial ELISAs. We believe that this observation has further practical application.

Currently available serologic tests allow the identification of SRLV-infected goats with high accuracy.^6,18 This is an essential step to control disease in a herd, and eradication of SRLV infection may only be achieved through repeated serologic testing of the whole herd and immediate culling of seropositive goats. However, in herds with high seroprevalence, this would mean total depopulation, which is usually unacceptable for the farmers, especially in regions where they must bear all of the costs associated with CAE control. In this situation, elimination of animals with the clinical form of CAE may be the only affordable solution.

The commercial ELISA that detects antibodies to SU antigen could be used to distinguish between arthritic and asymptomatic SRLV-infected goats without the need for herd inspection requiring clinical examination of each goat. Clinical examination performed simultaneously with blood collection considerably extends the time spent with each goat, and automatically raises the cost of the procedure. In our method, only blood need be collected, and the ELISA results interpreted according to one more cutoff—not only %I of 35%, which is recommended by the manufacturer and signifies seropositivity, but also %I of 56%, which indicates arthritis. Furthermore, available tests for the estimation of both between- and within-herd prevalence of the clinical form of CAE are not only very inconvenient (as they are all based on clinical examination) but also poorly characterized in terms of their accuracy and reliability (diagnoses are likely to differ considerably among examiners). Our study offers a tool of known epidemiologic characteristics. Although of limited accuracy, it employs a commercial ELISA, whose repeatability and reproducibility had been assessed before it was registered, and will allow users to estimate and compare the true prevalence of the clinical form of CAE between herds and regions. Repeatability and reproducibility of the SU-ELISA was not determined for the new cutoff value of 56%. However, as these measures depend on the assay manufacturing process (coating plates with an antigen) and compliance with the assay protocol in a laboratory, rather than on the cutoff itself,²² the measures are likely to remain stable.

The simulation that we conducted shows that our diagnostic approach does not allow for definitive identification of CAE-affected arthritic goats to be culled. However, it can be used for an initial selection of a group of goats at risk, which require further clinical examination by a veterinarian. Obviously, in this group, some goats will not have arthritis (as our test does not ensure a 100% predictive value of the positive result), but the actual benefit is that usually only a few arthritic goats will be left beyond this group (given that a predictive value of the negative result is usually high). As this procedure adds virtually no labor to serologic screening of a herd, time and costs would be considerably reduced compared to the procedure involving clinical examination of the whole herd after serologic screening has been performed.