Flaviviruses represent a group of mosquito-borne viruses in Brazil that are annually involved in a large number of human cases of dengue fever countrywide and sporadic local outbreaks of sylvatic yellow fever [1], [2]. Outbreaks caused by other flaviviruses have also been reported in the country. In the 1970s, the largest Brazilian epidemic of arbovirus encephalitis was caused by Rocio virus (ROCV) in southeast Brazil [3]. More recently, an outbreak of hemorrhagic manifestations was linked to Saint Louis encephalitis virus (SLEV) [4]. Sporadic human cases caused by other sylvatic flaviviruses, including Bussuquara virus (BSQV), Ilheus virus (ILHV) and Cacipacore virus (CPCV) have also been reported in Brazil [5], [6], [7]. Furthermore, yellow fever epizootics in howler monkeys were reported in 2008 and 2009. Approximately 200 carcasses tested positive for Yellow fever virus (YFV) and about 2000 deaths were reported [8].

Thirteen flaviviruses have been reported in Brazil, listed here in chronological order of discovery: YFV, ILHV, BSQV, SLEV, ROCV, CPCV, Dengue virus 1 (DENV-1) and Dengue virus 4 (DENV-4), Dengue virus 2 (DENV-2), Iguape virus (IGUV), Naranjal-like virus (NJLV), Dengue virus 3 (DENV-3) and Culex flavivirus (CXFV) [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. In 2009, serological evidence of West Nile virus (WNV) infection in Brazilian horses, was collected for the first time, in the Pantanal wetland region of Mato Grosso do Sul state (MS) [21].

The Pantanal wetland is a subtropical region of great biodiversity with strong potential for maintenance and evolution of mosquito-borne viruses. Comprising approximately 140,000 km², the Pantanal is a vast sedimentary floodplain characterized by seasonal flooding which determines specific ecosystem processes, with the occurrence of plants and animals that are adapted to the annual shrinking and expansion of habitats due to the seasonal hydrological regime [22]. The region, which covers mainly Brazilian but also Paraguayan and Bolivian territories, is ecologically classified into 11 sub-regions according to vegetation, flooding and physiography. In Brazil, the Pantanal is located within the states of Mato Grosso (MT) and MS in the west-central region of the country [23].

Most of the flaviviruses isolated in Brazil are unknown or understudied in the Pantanal. A small number of investigations in the Nhecolândia Sub-region of the Pantanal, MS, have detected serological evidence for four flaviviruses, including ILHV, SLEV, WNV and CPCV [21], [24], [25]. Recently, three other serological studies reported the detection of neutralizing antibodies for WNV in equines from MT and again in MS, including from outside the boundaries of the Pantanal [26], [27], [28]. Complete surveys for flaviviruses in the Pantanal have not been reported. Therefore, we conducted a serosurvey for WNV and the 11 Brazilian flaviviruses of potential medical importance, utilizing equines, sheep and caimans as indicators.

In October 2009, April, September and October 2010, blood samples were taken from 724 equines, 238 sheep and 61 caimans from 16 beef cattle ranches comprising an area of approximately 2,700 square-km in the Nhecolândia Sub-region of Pantanal, MS, west-central Brazil (Figure 1).

All caiman and sheep serum samples were negative for flaviviruses-specific antibodies using the 6B6C-1 monoclonal antibody by blocking ELISA. When tested by PRNT₉₀ for WNV, all caiman samples confirmed negative. Sixty caiman samples had PRNT₉₀ titer <10 and one <40 for WNV (low sample volume of the latter required testing at a lowest dilution of 1∶40). From 238 sheep samples tested, 235 had PRNT₉₀ titers <10, two showed low neutralizing antibodies titers of 20 and 10 and one <40. From 760 equine samples initially screened for flavivirus-reactive antibodies by blocking ELISA, 396 (52.1%) were positive. When this subset was tested using PRNT₉₀, 332 (43.7%) had neutralizing reactivity (PRNT₉₀ titer ≥10) for ILHV, 251 (33%) for SLEV, 172 (22.6%) for WNV, 139 (18.3%) for CPCV, 130 (17.1%) for ROCV, 62 (8.2%) for IGUV, 14 (1.8%) for YFV, 12 (1.6%) for BSQV, four (0.5%) for DENV-1, three (0.4%) for DENV-2, one (0.1%) for DENV-4 and none for DENV-3 (Table 1). Employing the criterion of 4-fold greater PRNT₉₀ titer, 77 (10.1%) equines were seropositive for ILHV, 59 (7.8%) for SLEV, 24 (3.2%) for WNV (see Table 2), two (0.3%) for CPCV and one (0.1%) for ROCV. From the remaining 232 blocking ELISA positive samples, 227 (29.9%) were considered seropositive for undifferentiated flavivirus due to cross-reactivity and five (0.6%) were considered seronegative with neutralizing antibodies titers <20 or <10 for all of the flaviviruses that were assayed by PRNT₉₀. These five may represent infections by an unknown flavivirus.

There were no seropositive equines using the four-fold antibody titer difference criterion for BSQV, YFV, IGUV, DENV-1, DENV-2, DENV-3 and DENV-4. Monotypic neutralization reactions (i.e. a sample that reacted with only one of the 12 flaviviruses in the panel) were detected for ILHV (n = 27), SLEV (n = 11) and WNV (n = 6).

The only horse from Nhecolândia Sub-region that presented neurological signs was seropositive for undifferentiated flavivirus with the following PRNT₉₀ titers: 80 for SLEV, 40 for ILHV, 20 for YFV, 10 for ROCV and <10 for all other flaviviruses tested.

Neutralizing antibodies for WNV were detected in 172 equines from all 15 ranches that equines were tested. Considering the four-fold higher titer criterion as confirmation, the 24 confirmed WNV-seropositive equines were detected in ten ranches comprising an area of approximately 2,500 square-km in the Nhecolândia and Nabileque sub-regions of the Pantanal, located in MS (Figure 2). Besides the 24 WNV-seropositive equines, samples from 37 other equines presented PRNT₉₀ titers for WNV ≥40 and among them 26 showed equal or higher titers for one of the other flaviviruses tested. For these 37 equines, the serology profile could not be interpreted to determine whether WNV reactivity was due to infection with WNV or cross-reactivity from other flavivirus infections. In fact, six of them were considered seropositive for one of the 11 other flavivirus tested (Table 3). Low level of neutralizing antibodies was detected for YFV, BSQV, IGUV, DENV-1, DENV-2 and DENV-4 even though none of the equines were considered infected with these viruses. In some cases, the PRNT₉₀ titers for YFV, BSQV and IGUV exceeded 20.

Both ILHV- and SLEV-seropositive equines were detected in 14 of the 15 ranches that equines were tested, comprising roughly a 3,100-square-km area. The only ranch that had no ILHV- and SLEV-seropositive equines was ranch PS (Figure 2).

From 760 equines tested, 441 (58%) had date of birth available. Stratifying the serology results by age demonstrated that risk of exposure to flaviviruses increased with age for flaviviruses in general and for ILHV (1-tailed p<0.001 for both), but not for SLEV (p = 0.17) or WNV (p = 0.46) (Figure 3). SLEV and WNV appeared to increase in prevalence with age for equines <9 years of age, but not for older animals.

Equines seropositive for CPCV and ROCV were sampled in October 2009. The two CPCV-seropositive horses were five and seven years of age and were from ranch PQ, which comprises an area of approximately 140 square-km. The only horse that was seropositive for ROCV had no record of birth date and was from ranch PM that encompasses a 360-square-km area.

At the end of 2009, an outbreak of neurologic disease among horses was reported from ranch PV located in the Nabileque Sub-region. From 36 equines sampled, 23 (63.8%) were seropositive for flavivirus infection by blocking ELISA. When this subset was tested by PRNT₉₀, 11 (30.5%) confirmed seropositive for SLEV, nine (25.0%) for undifferentiated flavivirus(es), two (5.5%) for WNV, and one (2.8%) for ILHV (Table 4). Of the nine undifferentiated flavivirus infections, five had SLEV titers >20 and five had higher SLEV titers than other flaviviruses in the panel, indicating that a majority of these could have had a history of SLEV infection. In addition to the relatively low seroprevalence for WNV of just 5.5% in this sample of 36 equines, four equines from this group that recovered from the recent neurologic syndrome were serologically negative for WNV (Table 4). In spite of recent vaccination against equine encephalitis viruses, most equines had low titers for equine encephalitis viruses. These results will be described separately.

In the absence of direct viral detection, diagnosis of arbovirus infections is performed by indirect serological tests. However, cross-reactive antibody responses to closely related flaviviruses can complicate interpretation of serologic test results [36]. In particular, secondary flavivirus infections generate high levels of cross-reactive antibodies, which further complicate the interpretation of flavivirus serology. The cross-reacting antibodies of primary infections typically react with other members of the Japanese encephalitis virus serogroup comprised by WNV, CPCV, SLEV and others principally in Africa and Asia [37], [38], but antibodies generated by secondary flavivirus infections can also bind and neutralize viruses of other serogroups as well. Moreover, some individuals that are sequentially infected by a heterologous flavivirus species can boost antibody levels against the original virus, resulting in a phenomenon known as ‘original antigenic sin’ which can lead to incorrect diagnoses [39]. Finally, the potential circulation of unknown flaviviruses in the region could generate cross-reacting neutralizing antibodies, and lead to misinterpretation. Therefore, we used a conservative threshold for detection of neutralizing antibodies (90%) and we considered homotypic serologic responses to be the most reliable, as these samples reacted with just one of the 12 viruses employed in the tests, with no indication of cross-reaction.

In the present study, serological evidence of ILHV, SLEV, WNV, CPCV and ROCV circulation was detected in the Nhecolândia and Nabileque sub-regions of Pantanal. Nhecolândia is the second largest sub-region of Pantanal comprising roughly 27,000 square-km or 19% of the total Pantanal area and it is the world's largest and most diverse area of subtropical lakes [40], [41]. Nabileque is the fifth largest sub-region comprising roughly 14,000 square-km or 10% of the total Pantanal area. This sub-region is formed by fluvial plains of the Paraguay River and remains flooded for nine months or more during the year [42].

The seroprevalence of ILHV and all flaviviruses combined increased with age in equines (Figure 3). This temporal pattern implies low-level enzootic transmission, where risk of infection increases with time. A drop in seroprevalence in the oldest age category that was observed consistently for all the flaviviruses evaluated may be explained by either waning of detectable antibodies over time, and/or weakening of the humoral immune response in older animals. In the case of SLEV and WNV, seroprevalence appeared to increase with age in the youngest three age categories, through age 9 years old, but the sample sizes were insufficient to detect a statistically significant trend. These increases in the youngest age groups may be explained by one or more introductions with establishment or not of these viruses after 2000 (i.e. within 9 years of the sampling time point in 2009–2010). This finding corroborates the negative results found in a serosurvey conducted with local equines in the 1990s, when no neutralizing antibodies for SLEV were detected [24]. These data also corroborate the results of two other equine serosurveys conducted in the Nhecolândia Sub-region in February 2007 and February 2009, when neutralizing antibodies to SLEV were detected in three and four year old equines, respectively, which suggests that SLEV circulated between 2004–2007 [25] and 2005–2009 [21], coinciding with two human outbreaks caused by SLEV in Argentina in 2005 [43] and southeast Brazil in 2006 [4].

In the present study, the detection of SLEV- and ILHV-seropositive geldings both born in September 26, 2009 and sampled on September 30, 2010 with no history of travel outside of Pantanal indicates that the transmission of SLEV and ILHV had occurred as recently as late 2009 or 2010. The enzootic status of ILHV is corroborated by the concurrent isolation of ILHV from Aedes scapularis specimens captured in the Nhecolândia Sub-region in 2010 [44].

Regarding WNV, the youngest seropositive equine was a 2 year-old untamed mare with no history of travel. Because it was sampled on October 23, 2009, the infection occurred between 2007 and 2009.

The confirmed WNV-seropositive equines were detected in ten ranches comprising an area of approximately 2,500 square-km in the Nhecolândia and Nabileque sub-regions of the Pantanal, located in MS. Monotypic neutralizing reactions for WNV were detected in six ranches encompassing roughly 1,300 square-km, which comprise approximately 20% of the Nhecolândia Sub-region area. WNV-seropositive equines were detected in Nhecolândia and Nabileque ranches located roughly 90 and 100 km distant from the city of Corumbá, MS respectively (Figure 2). Considering the detection of arbovirus activity in the Nabileque Sub-region which borders Bolivia and Paraguay, these two countries should consider our findings for their arbovirus surveillance programs.

The two CPCV-seropositive equines were sampled October 20, 2009 at five years of age and no history of travel outside of Pantanal, which indicates that infection had occurred between 2004 and 2009.

We report here for the first time the detection of ROCV-neutralizing antibodies in the Pantanal region. ROCV, which is an indigenous Brazilian flavivirus, was the causative agent of an extensive encephalitis epidemic in southeastern Brazil in the 1970s [3]. The detection of neutralizing antibodies for ROCV was unexpected, as current transmission events for ROCV are rare. Considering that the only ROCV-seropositive equine had no birth date available, the timing of ROCV transmission could not be clarified. No ROCV-seropositive equines were detected in two previous serosurveys for flaviviruses conducted in the same region [21], [24], which could indicate low circulation or recent introduction of ROCV in the Pantanal.

No seropositivity for BSQV, IGUV, YFV and DENVs was confirmed among the equines tested in the present study. However, except for DENV-3, neutralizing antibodies were found to all the flaviviruses tested. More investigation is needed to understand the potential circulation of these viruses in the Pantanal region, as they may be restricted to ecological cycles for which equines, sheep and caiman fail to serve as indicators. A serosurvey of non-human primates and/or local human residents would be more instructive for these flaviviruses. Testing only equines may provide a biased view of the relative amounts of flavivirus transmission because equines may not mount immune responses to all viruses equally, or they may not attract all vectors equally. For example, they failed to generate primary immune responses to DENV-2 and SLEV [45] and would not attract large numbers of the yellow fever vector mosquitoes, which are primate specialists [46]. However, they are frequently exposed to mosquitoes and often utilized for arbovirus serosurveys elsewhere.

For future flavivirus serosurveys to be held in the Pantanal, not only ILHV, SLEV, WNV, CPCV and ROCV should be included as differential diagnosis, but also those flaviviruses for which neutralizing antibodies titers ≥20 were detected, including BSQV, IGUV and YFV, if a PRNT with a four-fold greater titer criterion is used.

Regarding the 36 equines collected from ranch PV in the Nabileque Sub-region of Pantanal after a neurologic disorder outbreak occurred in 2009 and 2010, no serological evidence was found linking this epizootic to WNV, which is currently the only flavivirus known to be involved in outbreaks of equine neurological disorders in the New World. Three of four survivors of the recent neurologic syndrome had elevated titers for SLEV, and one was the highest SLEV titer detected in the herd, suggesting that the etiologic agent was a flavivirus that triggered a boost of pre-existing SLEV antibody or that SLEV infection may have been associated with the neurologic syndrome as a contributing risk factor perhaps through antibody enhancement. Equines experimentally infected with SLEV have not developed overt disease [47]. However, the possibility that a novel equine-virulent strain of SLEV caused the neurologic disease in horses in Brazil also merits exploration. SLEV genotype VB was recently isolated from the brain of a single horse that died of encephalitis in the state of Minas Gerais, southeast Brazil, in March 2009 [48]. Of the 36 equines tested from ranch PV, 11 (30.6%) were confirmed seropositive for SLEV, the highest seropositivity compared to 13 other ranches where SLEV-seropositive equines were found (and where maximum seroprevalence was 14%) and significantly higher than SLEV seropositivity detected in all Nhecolândia equines combined (48/724, 6.6%) (p<0.001, chi-square test). The true SLEV infection rate is probably underestimated, as several probable SLEV infections could not be differentiated due to masking by cross-reactive antibodies. Ecological differences between Nabileque and Nhecolândia sub-regions, the former formed by fluvial plains of the Paraguay River and the latter mainly comprised of savannah with subtropical lakes, may have contributed to the difference of seroprevalence encountered in these two sub-regions. Also, the Nabileque ranch is the largest ranch sampled encompassing a 900-square-km area, including portions of the plateau bordering the Pantanal. The high seropositivity for SLEV in Nabileque ranch may reflect higher circulation of SLEV in the plateau surrounding Pantanal than in the floodplains of Pantanal.

Amphibians and reptiles may play a larger role in the transmission cycle of arboviruses than previously assumed. Because of their high energy conversion efficiency, ectotherms typically represent a large portion of the vertebrate biomass in many ecosystems, and may therefore greatly influence transmission dynamics, either as virus amplifiers or dilution hosts [49]. Juvenile Alligator mississippiensis has been shown to develop sufficient WNV viremia to infect blood-feeding mosquitoes [50], [51]. A recent study reported the first evidence of WNV exposure in wild Morelet's crocodiles (Crocodylus moreletii) in Mexico [52].

Because of widespread, high prevalence of the yacare caiman (Caiman crocodilus yacare) in the Brazilian Pantanal, about 3.9 million non-hatchling caimans in 1993 [53], and its potential to naturally amplify WNV in the region, in February of 2009 a serosurvey for flaviviruses was conducted with 30 free-ranging caimans and all animals tested negative for flavivirus antibodies. Despite the negative results, in the same study equines temporally and spatially associated with the seronegative caimans showed neutralizing antibodies for ILHV, SLEV, WNV and CPCV [21]. Taking into account these data, in the present study a wider serosurvey for flaviviruses was conducted with 61 free-ranging caimans from the Nhecolândia Sub-region, but all caimans tested negative for flavivirus antibodies by blocking ELISA and also by PRNT₉₀ for WNV. The absence of serological evidence of flavivirus circulation in caimans was unexpected. Possible explanations for these findings are: (i) low circulation of flavivirus in caimans, (ii) caimans may not mount immune responses to flaviviruses and, finally, (iii) caimans may not attract local mosquito vectors of flaviviruses.

In a study conducted also in the Pantanal, but in a different sub-region, 80% of mosquitoes caught blood-feeding on caimans were Culex (Melanoconion) theobaldi and Mansonia spp. [54], species that are not usually reported as flavivirus vectors. However, more recently, ILHV was isolated from Aedes scapularis, which was the most prevalent species caught landing or blood-feeding on caimans from the Nhecolândia Sub-region [44].

The absence of serological evidence of flavivirus circulation in sheep by blocking ELISA and the detection of only two samples with neutralizing antibodies for WNV in very low titers was also an unexpected result, considering that the tested sheep were temporally and spatially associated with equines that were seropositive for various flaviviruses. However, reports of serological evidence of WNV infection in sheep are scarce and have been showing low prevalence [55], [56]. Possible explanations for these findings are the low prevalence of Brazilian flaviviruses in sheep, or that sheep fail to develop antibodies for Brazilian flaviviruses following exposure by mosquito bite, or finally, that sheep may not attract local mosquito flavivirus vectors in the Pantanal region. An entomological study conducted in southeast Brazil in the 1980s showed that sheep was not an attractive host for mosquitoes compared with man, horse and cattle [57].