Materials and Methods

Sample description—Plasma and CSF samples were obtained from 18 horses used in an S neurona challenge experiment to assess the effect of intermittent administration of ponazuril on infection; complete results are reported elsewhere.³⁰ Samples were available from 4 groups of horses. In group 1 (n = 4), 612,500 S neurona sporocysts were administered intragastrically to each horse (day 0); no antiprotozoal treatment was given. In group 2 (n = 5), 612,500 S neurona sporocysts were administered intragastrically to each horse; treatment with the antiprotozoal drug ponazuril (20 mg/kg, PO) was given every 7 days (beginning day 5 after inoculation). In group 3 (n = 5), 612,500 S neurona sporocysts were administered intragastrically to each horse; treatment with ponazuril (20 mg/kg, PO) was given every 14 days (beginning on day 12 after inoculation). In group 4 (n = 4), horses were neither challenged nor treated. Blood and CSF samples obtained on the day immediately prior to sporocyst challenge (start of study; day −1) and at the termination of the experiment at day 84 (end of study) were analyzed for this study. Cerebrospinal fluid samples were obtained from anesthetized horses via puncture of the atlantooccipital space. Although RBC counts were not determined, all CSF samples were transparent and colorless; in the experience of the authors, such samples have RBC counts < 100 RBCs/μL (typically < 10 RBCs/μL^26,31). Albumin concentrations in CSF and blood samples were determined and used to calculate Q_alb as a measure of blood contamination of samples.³²

Culture and enrichment of S neurona—The WSU-1 isolate of S neurona was grown in BT cells,^a as previously described.³³ In brief, BT cells were maintained in Dulbecco modified Eagle medium supplemented with 0.01M HEPES, 1mM sodium pyruvate, penicillin G sodium (100 U/mL), streptomycin (100 μg/mL), 2mM L-glutamine, and 10% heat-inactivated fetal bovine serum; the culture was kept at 37°C in 5% carbon dioxide in air. Sarcocystis neurona merozoites were added to flasks of confluent BT cells. After 7 to 14 days of culture, most BT cells had lysed and released large numbers of merozoites into the medium. Cell debris was removed from culture supernatants by passage through a 24-gauge hypodermic needle followed by low-speed centrifugation (19 × g for 5 minutes). Merozoites were then pelleted via centrifugation at 388 × g for 20 minutes, resuspended in PBSS, and counted. The merozoite suspension was layered over iodixanol with hydrophilic polysaccharide (specific gravity, 1.079)^b and centrifuged at 400 × g for 45 minutes. The pellet, which was highly enriched with S neurona merozoites, was then resuspended in PBSS and washed by 5 cycles of centrifugation (507 × g for 10 minutes) and aspiration. Washed merozoites were used for horse inoculations and the S neurona ELISA.

Preparation of polyclonal anti–S neurona antibody—A 3-year-old mixed-breed gelding was used for the provision of anti–S neurona polyclonal antibody. Merozoites were prepared on the day of each inoculation. Approximately 10⁸ live merozoites/dose were administered IM on days 0, 14, and 28 and IV on day 42. Blood was collected 2 weeks after the final inoculation and was used to prepare immune plasma, which was stored frozen in aliquots at −80°C.

Quantification of IgG in CSF via ELISA—A sandwich ELISA³⁴ was developed to measure IgG in equine CSF. Affinity-purified sheep anti-horse IgG heavy and light chains^c were used as the coating antibody, horse reference serum^c was used as the standard, and wholemolecule horseradish peroxidase–conjugated rabbit anti-horse IgG^d was used as the detecting antibody. The IgG concentration in undiluted reference serum (2,324 mg/dL) was determined via radial immunodiffusion assay.^e Coating antibody (1:8,000 in 100 μL of carbonatebicarbonate buffer) was added to microtitration plates^f that were kept at 4°C for 3 hours and then washed 4 times with PBSST. Next, blocking buffer (3% bovine serum albumin^d in PBSST) was added to wells. After incubation at room temperature (approx 22°C) for 20 hours, blocking buffer was washed off and 100 μL of either equine reference serum or an experimental sample was added to each well for an additional period of 1 hour at room temperature. Standard curves for absorbance versus IgG concentration were formed by use of serial 2-fold dilutions of reference serum ranging from 1:80,000 (290 ng/mL) to 1:2,560,000 (9 ng/mL). Cerebrospinal fluid samples were diluted 1:1,000 and 1:10,000 for assay. After washing wells with PBSST, 100 μL of secondary antibody (1:45,000) was added and plates were incubated for 1 hour at room temperature. Next, plates were washed 4 times and 100 μL of tetramethylbenzidine^g peroxidase substrate was added to each well. Thirty minutes later, the reaction was stopped by addition of hydrochloric acid. Plates were assessed at a wavelength of 450 nm.^h

A standard curve for optical density (at 450 nm) versus total IgG concentration was generated for each plate, and the best-fit 4-factor sigmoidal curve was determined via nonlinear regression analysis. The IgG concentrations in CSF samples were then quantified by interpolation from the standard curve.

Quantification of IgG concentration in plasma via radial immunodiffusion assay—For determination of total IgG concentration in plasma samples, radial immunodiffusion assays were performed at an external laboratory.ⁱ

ELISA for S neurona–specific IgG—For use in the S neurona ELISA, merozoites were prepared as described and mixed with a protein extraction reagent^j for 20 minutes. Remaining cell debris was removed via centrifugation (13,000 × g for 10 minutes), and the protein concentration of the lysate was determined by use of a microbicinchoninic acid method.^k

A direct ELISA was developed for the detection of S neurona–specific IgG. A stock solution of S neurona lysate (protein concentration, 5.6 mg/mL) was used as the coating antigen, immune plasma from the S neurona–inoculated horse was used for generation of a standard curve, and horseradish peroxidase–labeled rabbit anti-horse IgG^c was used as the detecting antibody.

Lysate (100 μL [0.5 μg of protein/mL]) in coating buffer (carbonate-bicarbonate buffer; pH, 9.6) was added to the wells of microtitration plates.^g After incubation for 20 hours at 4°C, coated plates were washed 4 times with PBSST. After a 2-hour blocking and washing step, serial 2-fold dilutions of 100 μL of immune plasma (1:800 to 1:25,600 in PBSST) were added for 1 hour at room temperature. Plates were again washed, and detecting antibody (1:2,000 in PBSST) was added to wells; incubation was performed at room temperature for 1 hour. After a final washing procedure, 100 μL of tetramethylbenzidine was added to wells. Plates were shielded from light at room temperature for 15 minutes, and the reaction was then stopped by the addition of hydrochloric acid. Plates were assessed at a wavelength of 450 nm.

A standard curve for log optical density (at 450 nm) versus S neurona IgG titer was generated for each plate, and the best-fit 4-factor sigmoidal curve was determined via nonlinear regression analysis. The S neurona–specific IgG titers of samples on that plate were then determined by interpolation into the standard curve. These values were expressed as EUs, arbitrary units that were based on a nominal titer of 100,000 EUs/mL for undiluted immune plasma.

Western blot analysis—Cerebrospinal fluid samples collected at the start (day −1) and end (day 84) of the study from all 18 study horses underwent western blot analysis. All western blot analyses were performed at a single commercial laboratory.ⁱ Results were reported as positive, nonspecific positive, or negative depending on the band pattern revealed by samples.²⁶

Calculation of Q_SN, Q_IgG, and AI—The Q_SN, Q_IgG, and specific AI were calculated by use of equations as follow:

View Expanded

Statistical analysis—Because results for groups 1 through 4 did not meet assumptions for normal distribution (Shapiro-Wilk W test³⁵) or homogeneity of variance (Levene statistic³⁶), group data were analyzed by use of nonparametric tests. Effects of group on Q_alb, CSF albumin concentration, SN_plasma, SN_CSF, Q_SN, IgG_plasma, IgG_CSF, Q_IgG, and AI at each of the 2 time points (ie, start and end of the study) were examined by use of Kruskal-Wallis tests. When a significant effect of group was identified, the difference between values for the pair of groups was explored by use of a Mann-Whitney U test. Subsequent analyses involved comparisons between group 4 (unchallenged control horses; n = 4) and the combined group of challenged horses (groups 1, 2, and 3; 14). Because these reclassified data sets did not meet the homogeneity-of-variance assumption for ANOVA, they were transformed as natural log values. Transformed data were examined via a repeated-measures ANOVA involving the general linear model univariate procedure, with time (ie, start of study vs end of study) as the within-subjects factor and challenge as a between-subjects factor. Effect of challenge was determined as P values for time × challenge. For data grouped by western blot analysis, log-transformed values were analyzed for effect of western blot classification via a 1-factor ANOVA involving the general linear model univariate procedure. When a significant effect of western blot result was identified, the difference between values for a pair of groups was examined post hoc by use of a Tukey honest significant difference test. In all instances, significance was ascribed to a value of P ≤ 0.05. A commercial statistical software package^l was used to perform analyses.

Results

By day 42 after inoculation, CSF samples from all 4 horses in group 1 (those challenged with S neurona sporocysts but not treated with ponazuril) and 7 of the other 10 challenged horses in groups 2 and 3 yielded positive western blot results. All Q_alb values (median, 1.38; range, 1.05 to 1.96) were less than the published³² upper limit for adult horses (2.35). Albumin concentrations in CSF samples ranged from 21.8 to 51.6 mg/dL (median, 32.4 mg/dL). Significant effect of treatment or time after S neurona challenge on CSF albumin concentration and Q_alb was not identified. Consistent abnormal clinical signs or histologic CNS lesions attributable to S neurona challenge were not observed in the horses of any group.³⁰

Sarcocystis neurona–specific IgG titers in plasma and CSF and total IgG concentration in CSF samples were quantified via an ELISA that was developed for the study. The lower limits of detection for total IgG and S neurona–specific IgG were approximately 10 ng/mL and 3.8 EUs/mL, respectively. For the total IgG ELISA, the mean ± SEM coefficient of variation for repeated assay of samples was 9.1 ± 1.4% within assays and 10.7 ± 1.9% between assays. The coefficient of variation for repeated assay of samples for the S neurona–specific IgG ELISA was 9.3 ± 2.3% within assays and 16.0 ± 4.0% between assays. Although an effort was made to adapt the ELISA for measurement of plasma total IgG concentration, IgG measurements had poor parallelism³⁴; thus, plasma IgG concentration was determined via radial immunodiffusion assay.

Initially, group data were analyzed for effects of time and treatment via nonparametric tests. At the end of the study time point (ie, after sporocyst challenge), significant differences in SN_plasma and SN_CSF between each of the challenge groups 1, 2, and 3 and control group 4 were detected; however, no effect of ponazuril treatment was identified among groups 1 through 3 for any of the variables examined. Because significant differences were not detected among the S neurona challenge groups, data from these 3 groups were combined for subsequent analyses. When results were log transformed, all data sets relating to comparisons between the group of unchallenged control horses (n = 4) and the combined group of challenged horses (n = 14) met assumptions for normal distribution and equal variance and were evaluated via univariate or repeated-measures general linear model procedures.

Prior to sporocyst challenge at the start of the study, there were no significant differences in any of the variables examined between the challenge and control groups (Table 1). Significant effects of challenge on SN_plasma and SN_CSF were detected. The CSF-to-plasma ratio of these titers (ie, Q_SN) did not change significantly after challenge, reflecting the fact that mean plasma and CSF anti–S neurona antibody titers increased proportionally (25- and 31-fold increases, respectively) in horses inoculated with S neurona sporocysts. A significant effect of challenge was not identified for Q_IgG or AI.

Western blot results were obtained for all CSF samples. The IgG values were classified on the basis of negative, nonspecific positive, or positive western blot results, and differences among these classification groups were analyzed (Table 2). There were significant effects of western blot classification on SN_plasma and SN_CSF values. Furthermore, within these variables, each post hoc pairwise comparison revealed a significant difference. Significant effects of western blot classification were not identified for Q_SN, IgG_plasma, IgG_CSF, Q_IgG, or AI.

Discussion

In the present study, an ELISA was developed for quantification of S neurona–specific IgG in plasma and CSF and calculation of Q_SN (the CSF-to-plasma ratio). Because there is some reported variation in the specificities of serum anti–S neurona antibody among horses³⁷ and heterogeneity in surface antigen expression among different S neurona isolates,³⁸ whole-cell merozoite detergent lysate was used as the ELISA antigen. Calculation of a specific AI does not require detection of antibody against any particular antigen of an infectious agent; instead, AI is based on a presumption that the quotient for the specific antibody of interest (the CSF-to-plasma ratio for any quantifiable antibody against an infectious agent) equals the quotient for total immunoglobulin unless additional specific antibody is synthesized in the CNS.¹⁵

Total IgG concentration in CSF samples collected from horses also was successfully measured by use of an ELISA developed specifically for this study. Although the assay was useful over a wide range of CSF IgG concentrations, it was not accurate or reproducible when used for measurement of plasma IgG concentration. Plasma IgG concentration was therefore quantified via a radial immunodiffusion assay that used a goat antibody of the same nominal specificity (equine IgG heavy and light chains) as the sheep capture antibody used in the ELISA. One of the precepts involved in generation of AI is that a quotient pair is the product of a single assay technique.¹⁵ Because 2 different IgG assays were used in the present study, it is possible that absolute values for Q_IgG—and therefore values of AI—were less accurate than they would have been if a single assay had been used. Despite this potential problem, the mean AI value of 0.81 calculated for CSF and plasma samples collected before sporocyst challenge at the start of the study was within the reference range of 0.7 to 1.3 used for most human clinical applications.^15,16 Also, because the combination of assays used for generation of Q_IgG values was the same for all samples, there should have been no effect of assay type on comparisons among groups classified by western blot result or by challenge status. To optimize accuracy and minimize variability of Q_IgG in future investigations of the use of AI in EPM diagnosis, it will be important to use a technique for IgG quantification that will be effective for both plasma and CSF samples from horses, as has been done previously for calculation of IgG index.^20,32

A potential confounding factor in any study involving S neurona assays of CSF is blood contamination of samples. In our study, RBC counts were not performed; thus, the extent of blood contamination of CSF samples could not be evaluated. However, for various reasons, it is unlikely that such contamination affected study results and conclusions. All samples were clear and transparent and were obtained from the atlantooccipital site during anesthesia; in the authors' experience, such samples reliably have < 10 RBCs/μL and certainly < 100 RBCs/μL. Any effect of RBC contamination should be distributed randomly across groups and should not cause an artifactual treatment effect. All values for Q_alb (a crude measure of blood contamination) were < 2.0. Because blood contamination would equally increase the plasma IgG contribution to the numerator and denominator of the equation for AI, effects of blood contamination should be slight. For example, for a hypothetic set of CSF and plasma samples with values of SN_plasma = 40,000 EUs/mL and IgG_plasma = 3,000 mg/dL versus SN_CSF = 150 EUs/mL and IgG_CSF = 12 mg/dL (similar to those found in the present study), addition of blood containing 10⁷ RBCs/μL to CSF at a final concentration of 1,000 RBCs/μL would change the AI value from 0.938 to 0.939.

Even with the technical limitations described, the effect of S neurona sporocyst challenge on Q_SN and AI was interesting. Despite the 25-fold increase in mean S neurona–specific IgG titer in CSF following sporocyst challenge, Q_SN and AI did not change significantly. Similarly, significant effects of western blot classification on Q_SN and AI were not detected even though the mean anti– S neurona antibody titer of CSF samples that yielded positive results via western blot analysis (400.4 EUs/mL) was significantly greater than that of samples that yielded negative results via western blot analysis (8.8 EUs/mL). Taken together, these results suggested that the increase in mean intrathecal anti–S neurona antibody titer associated with sporocyst challenge was not attributable to synthesis within the CNS, but simply reflected increased circulating anti–S neurona antibody concentration. Because the AI relates S neurona–specific IgG titer to total IgG concentration, movement of specific antibody into the CSF, whether as a result of passive diffusion across an intact blood-CSF barrier, leakage across a porous barrier, or blood contamination during sample collection, would not be expected to affect AI.

In the present study, horses inoculated with 612,500 S neurona sporocysts did not develop consistent abnormal neurologic signs or have histologic findings indicative of CNS infection; however, CSF samples obtained from all challenged but untreated horses yielded positive western blot results at 42 days after challenge, which is consistent with the range of 19 to 61 days for post challenge immunoconversion in CSF in previously reported studies.^25–27 In light of the AI values determined in challenged horses, which were not considered abnormal on the basis of data from humans, these results raise the possibility that experimental S neurona sporocyst challenge in horses may not reliably result in CNS infection. It is possible that Q_SN and AI would be abnormally high in horses that underwent more aggressive challenge protocols than that used in the present study. For example, horses given 10⁶ biologically active sporocysts and then transported over long distances consistently developed abnormal clinical signs and some had CNS inflammation.²⁵

Although analysis of clinical samples was not part of the present study, it is reasonable to assume that Q_SN and AI values would be high in samples from horses with EPM. Antibody index values in samples from humans with neuroborreliosis are > 2.0 according to approved diagnostic criteria.³⁹ By contrast, samples from humans with nonneurologic borreliosis or other infectious forms of encephalitis (attributable to infection with herpes simplex virus, HIV, or herpes zoster virus) have AI values < 1.4^15,39; reciprocal specificity for AIs for agents involved in these other forms of encephalitis was also determined.

A potential advantage of determination of AI in the diagnosis of EPM in horses is that the technique should accurately identify the CNS-derived fraction of specific antibody and diminish the specificity problems of currently available tests that rely on the presence of antibody in CSF (western blot analysis),⁹ semiquantification of anti–S neurona antibody in CSF (semiquantitative western blot analysis),²⁶ or preset serum antibody titer ranges (indirect fluorescent antibody testing).⁶ The IgG index, a calculation derived from Q_IgG and Q_alb, has been used in an effort to identify CNS-derived anti– S neurona antibody.^20,32 However, because the IgG index does not use specific antibody and is exquisitely sensitive to contamination of CSF samples with blood,¹² it appears to have little usefulness in the diagnosis of EPM. Because AI assumes that blood-origin specific antibody would be in the same proportion in CSF as it is in blood, low-level blood contamination should not affect the accuracy or interpretation of AI.

Most applications of AI in human medicine require measurement of specific antibody of multiple isotypes.^17,40 The diagnostic usefulness of each isotype is difficult to predict on the basis of serum isotype profiles. Immunoglobulin G, IgA, and IgM dominance are characteristic of the intrathecal antibody responses of humans with neurosyphilis-, neurotuberculosis-, or mumps-associated encephalitis, respectively.¹⁵ It is not known which immunoglobulin isotype is most important during the CNS response to S neurona. Although all currently available commercial diagnostic tests detect S neurona–specific IgG, S neurona-specific IgM has been detected by use of western blot in serum and CSF obtained from horses inoculated with S neurona sporocysts.⁴¹ Sarcocystis neurona immunoassays, such as the one used by the authors and previously by others,^37,38,41 are easily adapted to measurement of different isotypes, and generation of contemporaneous AIs for multiple isotypes could provide greater predictive power than that provided by the IgG AI alone.

In the present study, a specific anti–S neurona AI was used to investigate the origins of S neurona–specific IgG in CSF samples obtained from horses inoculated with S neurona sporocysts. At 42 days after challenge, CSF samples collected from most of the challenged horses yielded positive western blot results, but S neurona–specific IgG appeared to be of systemic origin. Additional investigation will be required to determine the usefulness of AI for diagnosis of EPM in horses.