In the tropical Americas, where multiple flaviviruses are endemic, it is critical to evaluate the efficacy of WNV diagnostic assays to differentially detect infecting flavivirus agents. While other studies have looked at cross-reactivity with a limited number of tests and single-time-point samples (17
), our sequential infection studies provided a complete set of serum samples collected every 3 days for over 2 months from animals that were known to be free of any previous flavivirus exposure. Furthermore, our study was geographically relevant for our objective and incorporated all the widely available and used equine diagnostic assays. In many countries in Central and South America, horses are both surveillance tools for and susceptible to WNV, making the information generated from this controlled study useful for surveillance programs and for understanding the course of illness and recovery in a biologically important system.
Using the most commonly performed diagnostic assays, we determined that all tests had both advantages and disadvantages; knowing the properties of each particular system provides each user the ability to decide which assay or combination of assays best suits their needs. While we did consider alternative assays (e.g., cross-reactive reduced antigen ELISA or microsphere immunoassay [MIA] bead formats [20
]), we found that the results were no better than the more broadly available assays or that reagents were not available for equine testing (data not shown).
Real-time RT-PCR has proven to be an effective method for detection of WNV, SLEV, and DENV nucleic acid from a variety of sample types (22
). In this report, positive results were obtained only from sera of WNV exposed horses. This corresponds with previous studies that have shown that a small percentage (<10%) of horses develop clinical disease and low but detectable viremia (5
). Therefore, this test is an important tool for diagnosing early infection, especially with viruses that cause minimal viremia. For comparing acute-phase-sample assays, virus isolation is not likely to be the most fruitful approach, as it is more expensive than other techniques and requires both specialized facilities and training. However, if the objective is to obtain virus stocks for future studies, virus isolation is essential.
Serological assays are the most commonly used diagnostic assays due to their simplicity, comparably low cost, and requirements for few specialized apparatus or facilities. Our tests fell into two distinct methods categories, ELISA and PRNT, with each having both advantages and disadvantages. ELISA formats are inexpensive, are safe to perform without specialized containment, as no live virus is used, and can rapidly screen large numbers of samples but can have variable sensitivity and specificity. PRNT assays require containment facilities along with well-trained technical staff, are more time-consuming and expensive, and may provide the most conservative interpretation of etiology. PRNT assays may also yield different titer values, depending upon the cutoff value used. In this study, we report the titers using an 80% cutoff but also calculated the titers for both 50% and 90% (see the supplemental table) and, as expected, noted a change in the reported antibody titer. However, in virtually all samples, the interpretation of results and determination of infecting etiology were not affected.
The WNV IgM ELISA is the test of choice for diagnosis of recent infection in humans and was modified for horses in this study (30
). Curiously, while this is typically considered to be an early appearing antibody, in this study, it was not detected any earlier than IgG antibody. Furthermore, although WNV IgM was detected in most instances where WNV Nt-antibody titers were present, there were cases of IgM presence in the absence of Nt-antibody titers. This could be a false-positive IgM detection, or more likely, the IgM generated early in infection was not neutralizing, as has been shown in humans (6
). There were other instances where IgM was not observed until after Nt antibodies were detected, making the IgM ELISA less sensitive than some of the other assays.
As previously noted, the WNV and SLEV IgM assays had significant cross-reactivity in this study (31
). For example, in cohort 3 horses, the results for the SLEV IgM assay were positive for most WNV IgM-positive samples, even though these animals were never exposed to SLEV. The IgM ELISA does have the advantage of being the only assay able to state that a WNV infection was recent. In one animal, WNV IgM persisted for as few as 7 days (equine no. 12), conclusively indicating a recent WNV infection. Furthermore, the WNV-specific IgM ELISA never produced a positive result prior to WNV exposure, indicating that this assay is indeed WNV specific.
The WNV-specific blocking ELISA was also extremely specific and never generated a false positive, even after repeated flavivirus exposure, making it an excellent option for diagnosing WNV infection in equines with a history of previous flavivirus infection. This assay has proven to be effective in detecting total IgM and IgG from birds and domestic animals (3
) but less effective for humans from regions where multiple flaviviruses cocirculate (28
). In contrast to the virus-specific blocking ELISA, the flavivirus group blocking ELISA had some sensitivity limitations. In cohort 1, the assay was able to detect all of the initial samples that were SLEV Nt antibody positive but it also produced some false positives. In cohort 3, none of the samples receiving only DENV-2 were positive with the flavivirus-specific blocking ELISA, and in cohort 2, many of the samples with SLEV Nt-antibody titers were negative in this test. This result is particularly interesting considering that the monoclonal antibody used was developed using an SLEV antigen.
Because the IgG assay is designed to work more broadly on flaviviruses and has previously been shown to detect DENV-2, SLEV, WNV, Japanese encephalitis virus, Murray Valley encephalitis virus, Powassan virus, and yellow fever virus antibodies in humans (2
), it was unexpected that this assay generated results similar to those of the WNV-specific blocking ELISA. Possible explanations for this deviation from human studies include the options that IgG responses are different in humans and horses and that the equine-adapted WNV assay is more specific for WNV than the human assay.
Traditionally, the PRNT is the gold standard for serological diagnosis and confirmation. In our study, the PRNT was a conservative test, often resulting in a diagnosis of “recent flavivirus infection.” Additionally, cross-reactivity between different serogroups was observed. This was most clearly seen in cohorts 1 and 2 when DENV Nt-antibody titers were detected after sequential infections. Interestingly, it does not appear that SLEV Nt antibodies cross-react with WNV antigen in instances of single infection. Given the low SLEV antibody titers, it is likely that the SLEV antibody response is simply too low to elicit WNV cross-reactivity. Due to concerns that the SLEV viral dose administered was insufficient to elicit an immune response, a higher dose was also administered. The results were similar with the two dosages, suggesting that SLEV is a poor immunogen.
Previous field reports found seroconversion to SLEV in domestic and sentinel horses in Central and South America, reinforcing the idea that undocumented SLEV infection could affect diagnosis (1
). We found only low levels of antibody against SLEV even after 2 sequential injections, but these antibodies persisted to day 39 and through the subsequent heterologous exposure to WNV. Based on the high WNV Nt-antibody titers observed at the later time points for cohort 2, it can be theorized that the WNV antibodies present are cross-reacting with SLEV antigens. This phenomenon was seen even more clearly with horses initially receiving a DENV-2 injection, where the development of antibody was even more robust. The degree of cross-reactivity between WNV and DENV-2 was somewhat unexpected since these viruses are in distinct serogroups. This finding suggests the possibility that humans with febrile illness in areas where dengue is endemic may indeed have WNV infections even when serological assays suggest a dengue virus etiology, particularly when the patient has had previous dengue virus infection. However, while our results clearly demonstrate this possibility for equines, it is important to point out that we cannot be certain how our results for equines will correlate with the data from human infections. Furthermore, it is significant to note the timing of sample collection as it relates to testing outcome. As our data show, antibody levels can rise and fall rapidly, particularly when only low levels exist. Thus, a single sample may give misleading or inaccurate results of the true etiology, underscoring the importance of testing both acute- and convalescent-phase samples. Furthermore, while numerous protocols with minor technical variations exist (30
), testing all the protocol variants published was not a feasible option; rather, our objective was to evaluate each technique overall.
A final question is whether previous exposure to a flavivirus can modulate disease following a subsequent WNV infection. At least partial protection was seen in hamsters immunized with SLEV and subsequently challenged with WNV (16
). Because of the close antigenic relationship between WNV and SLEV, this result may not be unexpected. More intriguing are the reports that hamsters immunized with DENV were protected against lethal WNV infection (35
). While none of the horses in our study developed clinical illness, the antibody responses developed in horses could prevent subsequent disease which would support earlier studies with rodents. This finding may provide one plausible reason for the absence of WNV epidemics in areas where dengue viruses and SLEV are endemic. Further studies will be necessary to examine this phenomenon more fully.