From March 2007 through June 2008, serum and/or postmortem brain specimens were collected from horses in South Africa with acute fever or neurologic disease; cases were detected by passive surveillance. Specimens were sent to the Department of Medical Virology, University of Pretoria, by the main veterinary diagnostic facilities in South Africa (Onderstepoort Veterinary Institute and the University of Pretoria Faculty of Veterinary Science, Onderstepoort) and by a group of private equine veterinarians from Gauteng and the Northern Cape provinces, who were invited to submit samples from horses with suspected cases. Specimens from horses with fatal and severe neurologic cases mostly came from the University of Pretoria (Pathology Department, Faculty of Veterinary Sciences) and the Onderstepoort Veterinary Institute; most were from horses with neurologic signs for which no alternative diagnoses were made. Specimens from horses with fever were collected by the private veterinarians around Gauteng from horses with less severe disease. Cases that resembled African horse sickness, i.e., pulmonary or cardiac disease, were not included.

RNA was extracted with an RNeasy kit (QIAGEN, Hilden, Germany) according to the manufacturer’s recommendations. A nested real-time reverse transcription–PCR (RT-PCR) specific for the WNV NS5 gene that distinguishes between lineages 1 and 2 with WNV-specific FRET probes on the basis of dissociation curve analysis was used to screen specimens (23). A product of 214 bp could be seen on agarose gel. All positive RT-PCR products were confirmed by sequencing of the NS5 region (genome positions 9091–9191) and analyzed on an ABI 3130 sequencer as recommended by the supplier (Applied Biosystems, Foster City, CA, USA). RT-PCR amplification and sequence analysis of a 255-bp region of the E-protein (genome positions 1402–1656) was conducted as described before (2,24).

Sequences were aligned by using ClustalX version 1.83 (http://bips.u-strasbg.fr/fr/Documentation/ClustalX) with the multiple-sequence alignment option. Maximum-likelihood trees were generated by using PHYML (25). Bootstrap statistics for 1,000 replicates were calculated by neighbor-joining analysis with a maximum composite likelihood model and a gamma parameter of 2, using MEGA version 4 (26). Distances between sequences were calculated by using MEGA version 4 (26) with the P-distance analysis option.

Horse serum samples were tested for flavivirus-specific antibodies first by hemagglutination inhibition (HI) assay and next by a WNV-specific immunoglobulin (Ig) M-capture ELISA on HI-positive specimens as described in (27). IgM-positive specimens were confirmed by neutralization assays as described below.

The IgM-capture ELISA was conducted as described in (28) with virus-specific modifications. In brief, 100 µL/well of goat antihorse IgM µ-chain (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, MD, USA) diluted 1:500 in phosphate-buffered saline (PBS) without magnesium and calcium, pH 7.4, was adsorbed onto ELISA plates (MaxiSorp, Nunc, Denmark) overnight in a humidity chamber at 4°C. Plates were washed 3× with 0.1% Tween-PBS; the same washing procedure followed each subsequent stage of the assay. Plates were blocked with 200 µL/well of 10% skim milk (Merck, Darmstadt, Germany) in PBS and incubated at 37°C for 1 h. After washing, duplicate volumes of 100 µL of each test and control serum diluted 1:400 in 2% skim milk in PBS (diluting buffer) were added to wells in rows A–D:1–12 and to corresponding wells in rows E–G:1–12 and incubated at 37°C for 1 h. After washing, 100 µL/well WNV antigen and mock antigen, diluted 1:400 in diluting buffer, was added to rows A–D:1–12 and E–G:1–12, respectively. After incubation at 37°C for 1 h and washing, 100 µL/well of mouse anti-WNV antibody diluted 1:1,000 was added to each well and incubated at 37°C for 1 h. Production, inactivation, preservation, and safety testing of WNV antigen (strain H442/58), mock antigen, and hyperimmune mouse anti-H442/58 serum were conducted as described (29). After washing, 100 µL/well goat antimouse IgG (H + L chain) HRPO-conjugate (Zymed Laboratories, Inc, San Francisco, CA, USA) diluted 1:2,000 was added to each well and incubated at 37°C for 1 h. Plates were washed, and 100 µL/well ABTS (2,2′-azino diethyl-benzothiazoline-sulfonic acid) peroxidase substrate (Kirkegaard and Perry Laboratories, Inc.) was added to each well and the plate was incubated in the dark for 30 min at room temperature (22°–25°C). The stop reagent, 1% sodium dodecyl sulfate, was added, and optical densities (ODs) were measured at 405 nm. Specific activity of each serum sample (net OD) was calculated by subtracting the nonspecific OD in wells with mock antigen from the specific OD in wells with virus antigen. A threshold value for interpretation of results was determined as mean plus 3 standard deviations of duplicate net OD readings for negative control serum.

The HI assay was conducted as described previously (27), except that the sucrose-acetone-extracted H442/58 strain of WNV derived from mouse brain tissue (produced as described for the IgM-capture ELISA above) was used as an antigen. A serum sample was considered seropositive if it had a titer >1.3, equivalent to a serum dilution >1:20.

The serum microneutralization procedure using African green monkey kidney (Vero cells) was conducted as described previously (30), except that the SPU 93/01 isolate of WNV recovered in South Africa was used as a source of antigen. The titer was expressed as the reciprocal serum dilution that inhibited 100% of viral cytopathic effect. A serum sample was considered positive when it had a virus neutralization titer >1.0, equivalent to a serum dilution >1:10.

African horse sickness virus (AHSV), equine encephalosis virus (EEV), and equine herpesviruses (EHV) 1 and 4 were identified by using viral culture and antigen detection assays and/or complement fixation tests (31) on serum samples and using RT-PCR to detect AHSV and EHV (32). Rabies virus infections were identified by fluorescent antigen detection tests on brain tissue (33). Immunoperoxidase staining for EEV, EHV, AHSV, and flavivirus was performed, according to the method adapted from (34), on histopathologic sections of brain, spinal cord, spleen, liver, and lung after postmortem investigations.

All specimens were inoculated onto Vero cell monolayers (18 hours old) in 25 cm² tissue culture flasks supplemented with Eagle Minimum Essential Medium containing 2% fetal bovine serum, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 1mg/mL l-glutamine (GIBCO BRL, Invitrogen, Carlsbad, CA, USA). Inoculated cultures were microscopically observed for cytopathic effects for 10 days.

A total of 80 serum or brain specimens from horses with unexplained fever (n = 48) and/or neurologic symptoms (n = 32) were tested for WNV over the 16-month period. Most specimens from horses with neurologic signs came from Onderstepoort Veterinary Institute and the Department of Medical Virology, University of Pretoria; horses were from across the country; the specimens from horses with fever were mostly from horses with less severe disease in Gauteng.

WNV infection was identified for 7 (21.8%) of 32 acute neurologic cases (Table 1). For 5 cases, acute WNV infection could be confirmed by the presence of RNA through RT-PCR (4 brain specimens, 1 serum sample) as well as virus isolation from 1 brain specimen. Two HI-positive cases could be confirmed as probable recent WNV infections by IgM ELISA and WNV-specific antibodies, which were confirmed by neutralization assays. None of the horses with fever had WNV, although EEV was isolated from several. Affected horses ranged from 4 months to 19 years and were thoroughbreds, Arabians, Lipizzaners, Welsh ponies, warmbloods, and mixed breeds. Cases were identified in Gauteng, the Northern Cape, and North-Western provinces and occurred in April 2007 and from March through June 2008. Of the 7 WNV-infected horses, 5 died or were euthanized for humane reasons (Table 2).

For 5 horses, 3 of which had fatal infections, no virus other than WNV was identified. Two horses that died (8 months of age and 6 years of age) had co-infections with AHSV. For both horses, WNV was detected by RT-PCR in the brain and AHSV was detected in the spleen or lungs but not in the brain. The 6-year-old horse had documented records of up-to-date AHSV vaccinations. All cases were negative by immunoperoxidase staining, virus isolation, and/or PCR for EHV and EEV.

All confirmed WNV infections were identified in the group of horses with neurologic signs (Tables 1,2). Five horses with WNV died or had to be euthanized after becoming paralyzed. Signs included ataxia in all cases (7/7), weak hindlimbs and/or forelimbs and paresis (4/7), complete paralysis (2/7), seizures (2/7), chewing (1/7), partial blindness (2/7), jaundice and/or hepatitis (2/7), and miosis of the pupils (1/7). One horse (HS101/08) (Table 2) was recumbent from quadriplegia and displayed limb paddling, teeth grinding, and muscle twitching; signs progressed over 3 days to those similar to rabies, i.e., chewing fits, seizures, and coma before death, but fluorescent antigen detection results for rabies were negative. Fever was intermittent and not reported for all horses. The 2 WNV-infected horses that survived showed clinical signs for ≈21 days and had to be rested for several months, but each recovered fully.

Postmortem investigations were performed on 4 horses (HS125/08, SAE 126/08, M123/08, and HS101/08). All 4 had positive WNV results by real-time RT-PCR of brain tissue, and a WNV isolate was obtained from the brain of HS101/08. Virus isolation attempts on frozen specimens from the other 3 horses were not successful. AHSV was also isolated from the lungs and spleen of SAE126/08 and HS125/08. Immunohistochemical staining of organs, to identify AHSV, EEV, and EHV, confirmed the presence of AHSV in SAE126/08, but results were negative for all other horses. Detailed postmortem findings for the 2 horses that had no complications (HS101/08 and M123/08) are shown in Table 2. Immunohistochemical staining for flaviviruses demonstrated antigen in the lumbar spinal cord and in some gray matter axons of HS101/08 (Figure 1).

The 5 specimens that were positive for WNV by RT-PCR were subjected to sequence analysis and confirmed to be WNV by BLAST (www.ncbi.nlm.nih.gov/blast/Blast.cgi) search analysis. Phylogenetic comparison of the 189-bp NS5 gene region with representative sequences of all 5 currently recognized WNV lineages confirmed that all strains clustered with lineage 2 (Figure 2) and were closely related to lineage 2 strains isolated from humans in South Africa. The horse specimens differed by 0%–3% nucleotides from South African lineage 2 strains in the NS5 region. Recent (2008) strains displayed the least variation from each other (0%–1.2%) and were most closely related to strains SPU116/89 and SA93/01, which had been isolated from a person with fatal hepatitis and nonfatal encephalitis, differing by 1.2% and 1.8%, respectively. Specimens from southern Africa had 1.2%–4.8% differences compared with the lineage 2 Hungary isolate of 2004. All lineage 2 strains from southern Africa differed by 18.6%–19.2% from the Madagascar strain and by 19.2%–25.7% with lineage 1 strains. Additional phylogenetic analysis of a more variable region of the E-protein was performed on the 2008 isolate obtained from the brain of HS101/08 (Figure 3). HS101/08 had 0.07%–2% differences in the E-protein region compared with all other southern Africa lineage 2 strains and was the closest to SPU116/89; it differed by 6.9% from the prototype Uganda strain (B956), differed by 17.3% from a strain from Madagascar, and 23% from NY385/99 (lineage 1). The 2004 Hungary lineage 2 strains differed by only 1.6% from the southern Africa strains, which suggests that the Hungary strains may have originated from southern Africa. On the amino acid level, all southern Africa lineage 2 strains’ E-protein regions were identical to each other and to the Hungary strain but differed by 7% from the Madagascar strain and by 23% from the lineage 1 NY385/99 strain.