Characterization of Enzootic Foci of Venezuelan Equine Encephalitis ...

VECTOR BORNE AND ZOONOTIC DISEASES Volume 1, Number 3, 2001 © Mary Ann Liebert, Inc.

Research Paper Characterization of Enzootic Foci of Venezuelan Equine Encephalitis Virus in Western Venezuela ROBERTO BARRERA,1 NIEVES TORRES,1 JEROME E. FREIER,2 JUAN C. NAVARRO,1 CARMEN Z. GARCÍA,1 ROSALBA SALAS,3 CLOVIS VASQUEZ,3 and SCOTT C. WEAVER4

ABSTRACT The distribution of the sylvatic subtype ID Venezuelan equine encephalitis (VEE) viruses in the lowland tropical forests of western Venezuela was investigated using remote sensing and geographic information system technologies. Landsat 5 Thematic Mapper satellite imagery was used to study the reflectance patterns of VEE endemic foci and to identify other locations with similar reflectance patterns. Enzootic VEE virus variants isolated during this study are the closest genetic relatives of the epizootic viruses that emerged in western Venezuela during 1992–1993. VEE virus surveillance was conducted by exposing sentinel hamsters to mosquito bites and trapping wild vertebrates in seven forests identified and located by means of the satellite image. We isolated VEE viruses from 48 of a total of 1,363 sentinel hamsters in two of the forests on six occasions, in both dry and wet seasons. None of the 12 small vertebrates captured in 8,190 trap-nights showed signs of previous VEE virus infection. The satellite image was classified into 13 validated classes of land use/vegetation using unsupervised and supervised techniques. Data derived from the image consisted of the raw digital values of near- and mid-infrared bands 4, 5, and 7, derived Tasseled Cap indices of wetness, greenness, and brightness, and the Normalized Difference Vegetation Index. Digitized maps provided ancillary data of elevation and soil geomorphology. Image enhancement was applied using Principal Component Analysis. A digital layer of roads together with georeferenced images was used to locate the study sites. A cluster analysis using the above data revealed two main groups of dense forests separated by spectral properties, altitude, and soil geomorphology. Virus was isolated more frequently from the forest type identified on flat flood plains of main rivers rather than the forest type found on the rolling hills of the study area. The spatial analysis suggests that mosquitoes carrying the enzootic viruses would reach 82–97% of the total land area by flying only 1–3 km from forests. We hypothesize that humans within that area are at risk of severe disease caused by enzootic ID VEE viruses. By contrast, equines could actually become naturally vaccinated, thus preventing the local emergence of epizootic IC VEE virus strains and protecting humans indirectly. Key Words: Venezuelan equine encephalitis virus—Remote sensing—Culicidae—Tropical forests— Venezuela. Vector Borne Zoonotic Dis. 1, 219–230.

INTRODUCTION

V

(VEE) is a classical emerging arboviral disease that represents a public health and veterinary threat throughout most of the Americas, including the ENEZUELAN EQUINE ENCEPHALITIS

United States (Walton and Grayson 1989, Murphy 1994, Murphy and Nathanson 1994). VEE virus (Togaviridae: Alphavirus) has caused repeated epidemics and equine epizootics since the 1920s, involving hundreds of thousands of equines and tens of thousands of people with se-

1 Instituto

de Zoología Tropical, Facultad de Ciencias, Universidad Central de Venezuela, Caracas, Venezuela. Center for Animal Disease Information and Analysis, Fort Collins, CO. 3 Instituto Nacional de Higiene, Caracas, Venezuela. 4 Center for Tropical Disease and Department of Pathology, University of Texas Medical Branch, Galveston, TX. 2 USDA

219

220

vere morbidity and mortality (Walton and Grayson 1989, Rivas et al. 1997). The VEE virus complex is one of three major Alphavirus serogroups found in the New World. At least 13 members of the VEE complex have been identified in South, Central, and North America (Young and Johnson 1969, Walton and Grayson 1989). Two particular serotypes, antigenic subtypes IAB and IC, have been responsible for all major VEE epizootics, while the remaining serotypes circulate primarily in enzootic foci. Epizootic viruses have been isolated only during outbreaks involving equines, but never from sylvatic foci. Enzootic viruses do not generally cause outbreaks (Johnson and Martin 1974, Walton and Grayson 1989, Weaver 1998), but some can apparently generate epizootic strains via mutation (Weaver et al. 1992, Powers et al. 1997, Wang et al. 1999). Recent sequencing studies including several of the viruses isolated during the course of this investigation revealed an extremely close genetic relationship between enzootic ID strains from northern South America and epizootic IC viruses from an outbreak of human and equine VEE that occurred in 1992–1993 (Wang et al. 1999). This result implies the existence of a potential for the eventual reemergence of VEE in areas without adequate equine vaccination. Apparently, not all enzootic ID viruses circulating in nature have the potential to produce epizootic viruses. For example, sylvatic ID subtype VEE viruses isolated from the Coastal Cordillera in Venezuela (Moncayo et al. 2001, Salas et al. 2001) are genetically less related to the IC epizootic variants than the ID viruses isolated near the Catatumbo and Magdalena Rivers in western Venezuela and north-central Colombia, respectively. The latter foci are within the same biogeographic region as the areas where the main epizootics have originated in northern South America (Moncayo et al. 2001), with the exception of the 1995 epizootic/epidemic, the origin of which is uncertain (Brault et al. 2001). However, although those epizootic and enzootic VEE areas occupy the same biogeographic region, specific habitats for both virus cycles do not overlap. Enzootic VEE viruses have been isolated regularly from tropical lowland forests in the

BARRERA ET AL.

Neotropics, from both mosquitoes and vertebrate reservoir hosts (Scherer et al. 1963, Scherer and Dickerman 1972, Walder and Suarez 1976, Walder et al. 1984, Dickerman et al. 1986). However, we know very little about the environmental conditions that allow epizootic variants to emerge from enzootic foci to reach susceptible equines and initiate epizootics (Méndez et al. 2001). Characterization of those enzootic VEE foci from which epizootic viruses are most likely to originate may provide some understanding of the processes involved in epizootic emergence. Remote sensing (RS) of the environment is a tool that allows the classification of land and vegetation features on the surface of the Earth, based on the differential emission of electromagnetic energy in selected wavelengths (Jensen 1996). RS, coupled with geographical information systems (GIS), has been successfully applied toward habitat identification of a variety of vector species (Beck et al. 1997, Dister et al. 1997, Dale et al. 1998, Daniel et al. 1998, Laveissière and Meda 1999, Moncayo et al. 2000), and to understand better vector-borne disease transmission dynamics and risk (Washino and Wood 1994, Barrera et al. 1999, Beck et al. 2000, Glass et al. 2000). In this study, we use RS (Landsat 5 imagery) and GIS to characterize the spectral reflectance patterns of habitats that maintain ID VEE virus transmission in southwestern Venezuela, a region where these viruses had previously been isolated (Walder and Suarez 1976, Walder et al. 1984). We searched for sylvatic ID VEE viruses in lowland tropical forests, studied the reflectance patterns of foci as detected by RS imagery, identified other locations with similar reflectance patterns, and used the generated data to describe the location of other potential VEE foci. MATERIALS AND METHODS Study area The study area (8°569550–9°149150 N/ 72°279140–72°459400 W) comprised the Catatumbo, Socuavo, and Tarra River basins in southwestern Zulia State, Venezuela (Fig. 1). This region is part of the Maracaibo Lake de-

221

FOCI OF VEE VIRUS IN WESTERN VENEZUELA

FIG. 1. Geographic location of the sites where VEE virus surveillance was conducted in the Catatumbo region, 1996–1997, Zulia State, Venezuela.

pression, with low terrain elevations (30–80 m), meandering rivers that drain into the lake, and large swamps between rivers with low slopes (,0.05%). There are two main geomorphologic terrains: gentle foothills from the Tertiary and alluvial plains along the major rivers from the Quaternary. Mean annual temperature and precipitation are 27.2°C (1978–1984) and 2,991.1 mm (1978–1996), respectively. There is a short dry season from December to March, but precipitation exceeds evaporation during the rest of the year. High precipitation, low slopes, erosion, accumulation of sediments, and a high water table contribute to frequent floods and swamps. Reticular soil erosion (Moment “II” of Stagno and Steegmayer 1972) produces small canals of varying dimensions on the forest ground that fill with water and become the main preadult mosquito habitats. The original lowland tropical forest vegetation has been cleared extensively for cattle raising. The main economic activities are cattle ranching, African palm (Elaeis guianeensis) cultivation, and oil extraction. The main forest species are tropical hardwoods and palms dominated by Copaifera publiflora, Hymenaeae spp., Ficus spp., Jacaranda copaiba, Protium alicastrum, Attalea butyracea,

and Bactris manaca. Swamps are commonly inhabited by Pterocarpus officinalis, Inga sp., Tabebuia rosea, Erytrina spp., Mimosa pigra, and several species of Heliconiaceae and Cyperaceae. Study sites We visited the lowland tropical forests from which Walder and Suarez (1976) and Walder et al. (1984) reported VEE isolations in the Catatumbo region (from 1973 to 1981). Those and additional forests with similar spectral reflectance patterns were located with the aid of a Landsat 5 satellite image and Global Positioning System (GPS) units, resulting in the following sampling sites: Madre Vieja I, Madre Vieja II, Las Nubes, Rio Claro I, Rio Claro II, Ceres, and Socuavo (Fig. 1 and Table 1). In total, nine field trips were made to the Catatumbo study area in 1996 and 1997 to visit these seven study sites, which resulted in 25 field surveys (Table 1). Virus surveillance Syrian golden hamsters from a colony obtained from the Instituto Nacional de Higiene in Caracas were exposed to mosquito bites in

222

BARRERA ET AL. TABLE 1.

VEE VIRUS SURVEILLANCE IN THE CATATUMBO REGION , ZULIA STATE, VENEZUELA , 1996–1997

Locality

Latitude

Longitude

Las Nubes Ceres Madre Vieja I Madre Vieja II Rio Claro I Rio Claro II Socuavo Total

9°039450 9°059050 9°089020 9°089070 9°019060 9°009470 8°579310

72°379040 72°429180 72°409410 72°409310 72°419330 72°419530 72°399540

“coquito” cages (Dickerman et al. 1986) for 7 days in the forests. Depending on availability and the number of sites to be studied on each field trip, we exposed a mean of 53 hamsters per site per date (30–150 hamsters). Cages were suspended 1.2–1.5 m above the ground and placed in transects at 25-m intervals. Hamsters were inspected and fed early each morning. Blood samples were collected by cardiac puncture from sick hamsters and from those surviving a 1-week exposure and then the hamsters were sacrificed. Heart and spleen samples were dissected from hamsters found dead during the exposure period and preserved in liquid nitrogen. Parallel to the transect where sentinel hamsters were exposed, we placed 40–45 Sherman and 20–30 Tomahawk traps for 7 days. Bait for the Sherman traps was replaced every day and consisted of a mixture of sardines, corn flour, corn grains, bird food, peanut butter, vanilla extract, and vegetable oil. Ripe plantains and cassava were used as bait in the Tomahawk traps. Captured animals were bled by cardiac puncture, and those that could not be readily identified were sacrificed, preserved with formaldehyde, and placed in plastic bags with 80% ethanol. Blood samples and organs were collected as described above. Specimens (blood, heart, and spleen homogenates) were triturated and inoculated into plastic tubes with monolayer cultures of Vero cells. Infected cultures that developed cytopathic effects were harvested, and a suspension of infected cells was placed on 12-well spot slides. After the cells were dried, they were fixed in cold acetone and examined for the

Field visits

Hamsters exposed

VEE isolations

6 1 2 4 5 6 1 25

466 1,365 1,340 103 270 389 1,330 1,363

37 0 0 0 0 10 0 47

presence of arbovirus antigens by IFA using hyperimmune mouse ascitic fluids against the following: Alphaviruses, Flaviviruses, Group C Bunyaviruses, Capin, Guaroa, and Simbú (reference reagents from National Institutes of Health). Cultures IFA-positive for Alphaviruses were subtyped using monoclonal antibodies described previously (Roehrig and Bolin 1997). The hemagglutination inhibition method was used for serum antibody detection in mammals using eastern equine encephalitis and VEE virus antigens (Beaty et al. 1989). Image classification and analysis A Landsat 5 Thematic Mapper (TM) image (path 7, row 54) acquired on September 6, 1996 was used to obtain the lowest cloud interference possible. The image was georeferenced using coordinates gathered at road intersections visible in the image by means of differential GPS (DGPS) methods using a rover unit (Trimble Geoexplorer, Trimble Navigation Ltd.) with data from a commercial base station located in Caracas. Post processing software (Geo-PC, Trimble) was used to correct raw values. Georeferencing was accomplished with TNTmips software (Microimages Inc.) using the Affine method. The Affine transformation corresponds to an orthographic or parallel plane projection from a source xy-plane onto a target xy-plane and uses all control points to perform a least squares fit in order to determine the best overall transformation to apply. We used 15 ground control points scattered throughout the image, obtaining a 30.26 6 14.73 m (mean 6 S.D.) residual.

223


The image was classified using both unsupervised and supervised algorithms. Because low clouds and haze from fog and smoke covered part of the area of interest, we used only the infrared bands (4, 5, and 7) as they are the least influenced by atmospheric conditions. The unsupervised classification was carried out using the ISODATA clustering routine with 40 initial classes or clusters. The ISODATA algorithm analyzes a sample of the input to determine a specified number of initial class centers; then cells are assigned to classes by determining the closest class center (minimum Euclidean distance). After each classification step, the process calculates a new center for each class by finding the mean vector for each class (TNTmips software). Some earth features could readily be identified from the clusters and image, such as water and bare ground. Homogeneous classes were merged, thus reducing the number of classes that needed interpretation. Field visits were conducted to associate classes with features on the ground. A final supervised classification was performed using 13 classes of vegetation/land use (Table 2). Training sites for the 13 classes were identified in the field. A training site is a homogeneous section of each land use/vegetation class identified in the field and georeferenced using a DGPS. Coordinates and codes of training sites were used to identify pixels with similar values in the satellite image. The MAXLIKELIHOOD

algorithm (TNTmips) was used to classify the pixels in the 1996 image. The Maximum Likelihood algorithm interprets the cell values in each training set class as having a Gaussian (normal) distribution described by the mean vector and the covariance matrix. The likelihood of a given cell value belonging to a particular training set class can be determined using these statistics (TNTmips). The classified image was later validated by visiting 30 sites and scoring the land cover at each location by an independent observer (ground proofing). Land use/vegetation predicted by the image was compared with observed ones, resulting in 90% accuracy. Most contradictions were between the managed and unmanaged pasture classes. The classified image provided the location of tall, lowland tropical forests in the study area. Large-format maps of the classified image and color and false-color images were used to locate the forests studied. Digital data were used within TNTmips to generate statistical reports of classified features, such as area, perimeter, and length. Because maps of the area were out of date, we created a road data layer by digitizing roads observed in the image. In addition, image improvement was accomplished through a Principal Component Analysis of the visible and infrared bands (six bands). The first principal component revealed dirt roads, whereas the second principal component clearly delineated paved (asphalt) roads.

TABLE 2. RESULTS OF THE CLASSIFICATION OF A LANDSAT 5 TM (BANDS 4, 5, AND 7) IMAGE OF THE CATATUMBO AREA IN WESTERN VENEZUELA , SHO WING THE 13 CLASS NAMES , AREA OCCUPIED , AND PERCENTAGE OF TOTAL A REA Class

Area (ha)

Percentage

Water Water with sediments Bare ground/urban Burned grounds Pastures with eroded soils Unmanaged pastures Unmanaged, inundated pastures Managed, inundated pastures Open shrublands Short forests Palms and narrow-leaf tree stands Tall, logged forests Tall, original forests Total

2,800.6 2,983.9 2,080.8 1,694.8 3,955.3 29,878.5 9,599.7 32,161.1 14,995.7 16,067.1 16,901.3 17,164.1 23,453.8 173,736.7

1.6 1.7 1.2 1.0 2.3 17.1 5.6 18.5 8.6 9.3 9.7 9.9 13.5 100.0

224

We were primarily interested in comparing the spectral and environmental properties of forests where VEE viruses were isolated to characterize the enzootic foci. Data contained in the satellite image can be used to derive indices of vegetation and other properties that are useful to compare land use/vegetation classes. Kauth’s Tasseled Cap indices (Jensen 1996) were calculated with TNTmips, producing three raster layers: brightness (an index for electromagnetic radiation, similar to radiation intensity), wetness (surface wetness, including the water suspended in the plant biomass), and greenness (a qualitative estimate of green biomass). We also calculated the Normalized Difference Vegetation Index (NDVI) (Jensen 1996) for the study area. NDVI is another measure of green biomass that results from a normalized ratio of the near-infrared to red band in Landsat TM. Two other spatial variables determining landscape and vegetation utilized in this study were altitude above sea level and soil geomorphology. An elevation layer containing contour lines was digitized on a digitizing tablet (CalComp Inc.) from 1:100,000 maps of the study area. A Digital Elevation Model was calculated using triangulation from the original contour line layer of altitudes. The triangulation method builds a network of triangles meeting the Delaunay criterion: For each triangle, the circle that passes through all three vertices encloses no other input points. The method then fits a planar surface to each triangle so that the overall surface is modeled as a collection of triangular planar facets. Data on geomorphology of the study area were obtained as polygons (1:250,000; Venezuelan Ministry of the Environment), which were then rasterized to conform to pixel and scene sizes. We compared the field-surveyed forests based on their spectral properties and indices (brightness, greenness, and wetness, NDVI, and bands 4, 5, and 7), elevation, and geomorphology by means of a Cluster analysis. This was carried out using the centroid rule and Gower Similarity Coefficient with the MVSP program (Multi-variate Statistical Package, Kovach Computing Services). Sample data on the above variables for each forest were derived by drawing a polygon over the field sampling sites

BARRERA ET AL.

(100–200 pixels; pixel size 25 3 25 m), using the polygon to extract values and then calculating statistics for each raster layer. A map was created to show the spatial distribution of forests (.1 ha) and the likely area mosquito vectors may be able to fly, assuming flying distances of 1, 1.6, and 3 km around lowland tropical forests. Areas were calculated and represented on the map as buffer regions with TNTmips. The map was designed to show the potential spatial distribution of VEE virus in the region. We adopted flight ranges based on reported flight distances of two common mosquito species that are good candidates for exporting VEE virus out of the forests in the study site (Méndez et al. 2001). For example, mean flight distance for Culex nigripalpus is 0.76–1.6 km with a maximum of 3.8 km, whereas mean flight distance of Mansonia titillans is 1.3–1.8 km with a maximum of 3.4 km (Edman and Bidlingmayer 1969, Morris et al. 1991). RESULTS VEE surveillance From 1,363 hamsters exposed in forested sites, 48 isolates of the ID subtype of VEE viruses were made, accounting for a 3.5% isolation success (Table 1). Viruses were isolated in two of the seven locations: Las Nubes (one in December 1996, one in February 1997, 14 in September 1997, and 22 in November 1997) and Rio Claro II (two in February 1997, seven in September 1997, and one in November 1997). Thus, for individual trips to locations where there were virus isolations, the percent isolation success varied from 1.3% in Rio Claro II (November 1997) to 27.5% in Las Nubes (November 1997). VEE virus isolation success in the short dry season (1.2%; four isolates from 322 hamsters exposed) was lower than in the rainy season (8.3%; 44 isolates from 533 hamsters exposed). Vertebrate captures were rather low, with only 12 animals captured over a combined Sherman and Tomahawk trap effort of 8,190 trap-nights. Seven rodents were captured: two forest spiny pocket mice (Heteromys anomalus), three spiny rats (Proechimys poliopus), one agouti (Dasyprocta punctata), and one unidentified rat species. Other species captured

225


included one lesser anteater (Tamandua mexicana), three opossums (Didelphis marsupialis), and one lizard (Tupinambis teguixin). Serological tests showed no signs of previous VEE virus infection in any captured animals. RS VEE virus habitats The supervised image classification yielded 13 classes of land use/vegetation in the Catatumbo area (Table 2 and Fig. 2). Pasture was the most extensive land feature (43.5%) and corresponds with cattle raising, which is the main economic activity. Managed pastures were mainly located on alluvial and flood-prone terrain. The most common cultivated pastures were monospecific, flood-tolerant plants, mainly Brachiaria mutica and Hymenachne amplexicaulis. Unmanaged pas-

FIG. 2.

tures contained perennial herbs (Paspalum sp., Panicum sp.) and sparse shrubs and trees (Vismia sp., Xilopia sp.). Inundated, unmanaged pastures included marshes with a layer of 1.2 m of water and contained a variety of species (M. pigra, Heliotropium sp., Echinocloa sp., Heliconia marginata, Gynerium sagittatum, and Thalia geniculata), including aquatic plants such as Cyperus sp., Eleocharis sp., Phyllantus sp., and Ludwigia sp. Shrublands were common (8.6%) on abandoned farms and mainly comprised shrubs (e.g., M. pigra), young trees (J. copaiba, Heliotropium sp.), and herbs (Echinochloa spectabilis, Panicum sp.). Forests covered ,32.7% of the land in 1996, although they were highly fragmented (Fig. 2). Tall, native forests (13.5%) were evergreen, lowland tropical forests with a closed canopy and emergent trees exceeding 20 m in height, such as

Detail of the northern part of the study area represented by a classified Landsat 5 TM image.

226

Licania arborea and J. copaia. Forests were readily recognized using false composite visualizations (red, green, blue) with bands 4, 5 (near-infrared), and 3 (visible). Logging activity resulted in thinner, tall logged forests (9.9%) usually associated with, and in the periphery of tall, original forests. Dominant species were Attalea maracaibensis, Coriaiana pyriformis, C. publifora, Pseudolmedia laevigata, with an understory represented by Calathea lutea, Rinorea sp., and B. macana. Short forests (9.3%) were either secondary growth stands or forests located on swamps (5–9 m). Common species were Xilopia sp., Vismia sp. and Cassia aculeata. Palms and stands of narrow-leaf trees (9.7%) could be distinguished as a different class, particularly the extensive groves of African palms. Stands of native palms and narrow-leaf trees were associated with gaps and disturbed areas within forests. Other classes in the classified image represented standing water (3.3%; rivers, canals, swamps, small lagoons), bare ground (1.2%; land conditioning, urban, oil-well operations), and burned ground (1%), a cultural practice to eliminate weeds and stimulate growth of unmanaged pastures. Water could easily be identified by the low reflectance values in the three infrared channels of Landsat TM imagery. Bare ground gave high reflectance values on bands 5 and 7, whereas dense vegetation was best discriminated in band 4. Pastures showed higher values on band 5 than forests, and inundated managed and unmanaged pastures (including marshes) had high values on band 4, much in the same way forests did (data not shown). Tall, original forests showed shady canopies, probably resulting from emergent trees. The cluster analysis showed two groups of tall forests, one including Las Nubes and Socuavo, and another one containing the five other forests studied (Fig. 3). Within the latter, it can be observed that forests located in a same area clustered together or had a high similarity index (e.g., Rio Claro I and II). The main spatial differences between the two major clusters were that Las Nubes and Socuavo forests were at lower elevations (39 m in altitude) on alluvial plains, whereas the other forests were located on the Tertiary gentle rolling hills (79–83 m). Las Nubes and Socuavo forests also had slightly higher values in bands 4 and 7, as

BARRERA ET AL.

FIG. 3. Cluster analysis of the forests studied in the Catatumbo region, 1996–1997, Zulia State, Venezuela, based on the similarity of their spectral properties (Landsat 5 TM Bands 4, 5, and 7, Tasseled Cap indices, and NDVI index) and ancillary variables (elevation above sea level, geomorphology).

well as higher values of brightness and wetness than forests in the other cluster. VEE viruses were isolated from forests belonging to both clusters (37 viruses in Las Nubes and 10 viruses in Rio Claro II). Thus, it seems that the differences in spectral and terrain characteristics among dense forests of the study area do not affect virus dispersal. For that reason, the remotely sensed layer of tall forests may represent the extent and location of the likely foci of enzootic VEE viruses in the study area. Lowland tropical forests, roads, and 1-km buffers generated around forests (.1 ha) were mapped to show the extent of the hypothesized biting range of mosquitoes (Fig. 4). Forest fragmentation is evident with many small pieces of forests dispersed throughout the region. It is apparent that most of the study area is within 1 km of any forest, covering 81.9% of the land. The remaining area is mainly located in the upper part close to the Catatumbo River (Fig. 4) and is likely to be out of flying distance for forest mosquitoes. Increasing the buffer to simulate mosquito-flying distances of 1.6 and 3 km from forests resulted in 88.4 and 97.3% coverage of the study area, respectively.

DISCUSSION The main objectives of this research were to characterize and map the foci of enzootic VEE


227

FIG. 4. Spatial location and extent of lowland, tropical forests in the Catatumbo region in 1996 and 1-km buffers around them, representing the hypothesized risk of exposure to mosquitoes infected with ID VEE virus.

virus in the Catatumbo area of Venezuela. The results showed that the virus exists in several types of tall lowland tropical forests of the study area. Other types of land cover/vegetation, such as short forests and palm groves, have since been monitored and have not yielded enzootic VEE viruses (author’s unpublished data). Our attempt to make a finer classification of dense forests, using spectral and terrain characteristics, did not help to characterize further the enzootic foci. VEE virus was not found in dense forests that were classified as ecologically similar to and in the vicinity of

the forests from which isolates were obtained. For example, we found VEE virus in Rio Claro II but not in Rio Claro I (Fig. 1), where we have isolated enzootic VEE viruses in a previous study (Moncayo et al. 2001). Our results are consistent with previous research in other countries showing that the primary enzootic habitats of VEE viruses are lowland and riparian tropical forests (Grayson and Galindo 1969, Franck and Johnson 1970). Overall, the frequency of enzootic VEE virus isolation has been relatively low in the Catatumbo area. Walder et al. (1984) monitored the

228

region over a span of 9 years, obtaining only 23 VEE virus isolates from sentinel hamsters. Our research showed a higher frequency of virus recovery with a similar surveillance system (47 isolations in ,2 years; 3.5% of hamsters exposed). Virus recovery in another enzootic area located in northern Venezuela gave a 1.25% success (Salas et al. 2001). Results also showed that the likelihood of detecting VEE virus circulation increased in the rainy season, but in no forest were we successful at isolating the virus on every field trip. It may be that virus foci move within the forest (since all collections were made on the same transects within each forest) or the virus may go locally extinct and eventually recolonize the forest from another patch. Local rodent (host) populations seemed to be low in the forests investigated, which in turn may determine that the VEE virus may not be maintained in a given place owing to a lack of nonimmune hosts. It is likely that humans and equines in and around enzootic foci are at greater risk of being bitten by mosquitoes carrying the VEE virus than in areas further away. Our results show that if mosquitoes disperse within 1–3 km from forests, then 82–97% of the total region would be exposed (Fig. 4) to enzootic ID VEE viruses. Several species are candidates for enzootic VEE virus export in this area, particularly Cx. nigripalpus and Ma. titillans (Méndez et al. 2001). These mosquito species complete their preadult development in breeding places of open areas, but visit the forests as adults. We have captured adults of these species inside the forest and in open areas at night 200 m from the forests. Also, enzootic and epizootic VEE viruses have been isolated from these two species (Aitken 1972, Chamberlain 1972, Scherer and Díaz 1972), demonstrating that adult females become infected in nature by feeding on enzootic, viremic hosts. Their competence to transmit VEE viruses, however, seems to be limited since Cx. nigripalpus has a high threshold of infection with epizootic viruses (Sudia et al. 1971), whereas Ma. titillans shows an intermediate capacity to become infected and transmit epizootic viruses (Kissling and Chamberlain 1967, Turell 1999, Turell et al. 2000). Frequent infection of these mosquito species with enzootic viruses may provide op-

BARRERA ET AL.

portunities for the appearance of new genetic variants during replication within the mosquito and perhaps opportunities for the selection of epizootic variants. On the other hand, it is expected that forest fragmentation will increase the contact between mosquitoes from open areas and enzootic hosts carrying the VEE viruses, because of an increased forest ecotone or augmentation of total forest perimeter. For example, had all the study area been continuous forest it would have occupied 1,867 km2 with a perimeter of 180 km. Current forest patches (only those .1 ha) cover 259 km2 with a perimeter of 5,194 km, which is seven times less area and 28 times more perimeter. If the predominant virus being transported out of the forests is subtype ID, then equines located near enzootic foci are likely to be unharmed because ID viruses do not cause serious clinical illness in equids (Walton et al. 1973). Moreover, equines exposed to infection with ID viruses may become naturally vaccinated, because it has been shown that infections with ID virus protected horses from challenge with the epizootic strain P676 (Wang et al. 2001). Those results add confirmatory evidence to the suggested hypothesis that epizootics bypass regions of enzootic transmission owing to natural immunization of equines by enzootic VEE viruses. Indeed, no epizootic VEE outbreaks have been documented in the Catatumbo area, in spite of VEE viruses actively circulating in enzootic foci (Walder et al. 1984, this study). Although equids are not affected clinically by ID strain viruses, humans are known to suffer severe and occasionally fatal disease (Johnson et al. 1968, Franck and Johnson 1970, Zarate et al. 1970). Natural vaccination of equines may indirectly protect humans around enzootic foci if epizootic VEE virus emergence takes place via equines. If epizootic emergence occurs far from the natural immunization belts, hypothesized to exist around enzootic foci, then a relevant question is how epizootic variants originating from enzootic ones are transported and achieve fitness away from the enzootic foci.

ACKNOWLEDGMENTS We thank personnel of the Ministry of Health, Zulia State, Venezuela, for their contri-


bution to the field work and support (Pedro Morel, Ciro Moreno, Exeario Marquez, Yovanni Marquez, Lucio Bustamante, Loreto Sayago, Osmer Paz, Vidal Paz, and Rafael Paz). Susan A. Maroney and Priscilla FitzMaurice contributed to the illustrations. The Ministry of the Environment provided climatologic and soil data. We thank Byron Wood and Louisa Beck, Center for Health Applications of Aerospace Related Technology, National Aeronautics and Space Administration (NASA), for their support and guidance. This research was supported by grant AI39800 from the National Institutes of Health and by NASA.

ABBREVIATIONS DGPS, differential Global Positioning System; GIS, geographical information systems; GPS, Global Positioning System; NDVI, Normalized Difference Vegetation Index; RS, remote sensing; TM, Thematic Mapper; VEE, Venezuelan equine encephalitis.

REFERENCES Aitken, THG. Habits of some mosquito hosts of VEE (Mucambo) virus from northeastern South America, including Trinidad. In: Science Publication 243: Venezuelan Encephalitis. Washington, DC: Pan American Health Organization; 1972:254–256. Barrera, R, Grillet, ME, Rangel, Y, Berti, J, et al. Temporal and spatial patterns of malaria reinfection in northeastern Venezuela. Am J Trop Med Hyg 1999; 61:784–790. Beaty, BJ, Calisher, CH, Shope, RE. Arboviruses. In: Schmidt NJ, Emmons RW, eds. Diagnostic Procedures for Viral, Rickettsial and Chlamydial Infections, 6th edit. Washington, DC: American Public Health Association; 1989:797–855. Beck, LR, Rodriguez, MH, Dister, SW, Rodriguez, AD, et al. Assessment of a remote sensing-based model for predicting malaria transmission risk in villages of Chiapas, Mexico. Am J Trop Med Hyg 1997; 56:99–106. Beck, LR, Lobitz, BM, Wood, BL. Remote sensing and human health: new sensors and new opportunities. Emerg Infect Dis 2000; 6:217–227. Brault, AC, Powers, AM, Medina, G, Wang, E, et al. Potential sources of the 1995 Venezuelan equine encephalitis subtype IC epidemic. J Virol 2001; 75:5823–5832. Chamberlain, RW. Venezuelan equine encephalitis infection of mosquitoes and transmission. In: Science Publi-

229 cation 243: Venezuelan Encephalitis. Washington, DC: Pan American Health Organization; 1972:144–147. Dale, PE, Ritchie, SA, Territo, BM, Morris, CD, et al. An overview of remote sensing and GIS for surveillance of mosquito vector habitats and risk assessment. J Vector Ecol 1998; 23:54–61. Daniel, M, Kolár, J, Zeman, P, Pavelka, K, et al. Predictive map of Ixodes ricinus high-incidence habitats and a tick-borne encephalitis risk assessment using satellite data. Exp Appl Acarol 1998; 22:417–433. Dickerman, RW, Cupp, EW, Groot, H, Alarcon, AM, et al. Venezuelan equine encephalomyelitis virus activity in northern Colombia during April and May 1983. Bull Pan Am Health Org 1986; 20:276–283. Dister, SW, Fish, D, Bros, SM, Frank, DH, et al. Landscape characterization of peridomestic risk for Lyme disease using satellite imagery. Am J Trop Med Hyg 1997; 57:687–692. Edman, JD, Bidlingmayer, WL. Flight capacity of bloodengorged mosquitoes. Mosq News 1969; 29:386–392. Franck, PT, Johnson, KM. An outbreak of Venezuelan encephalitis in man in the Panama Canal Zone. Am J Trop Med Hyg 1970; 19:860–865. Glass, GE, Cheek, JE, Patz, JA, Shields, TM, et al. Using remotely sensed data to identify areas at risk for hantavirus pulmonary syndrome. Emerg Infect Dis 2000; 6:238–247. Grayson, MA, Galindo, P. Ecology of Venezuelan equine encephalitis virus in Panama. J Am Vet Med Assoc 1969; 155:2141–2145. Jensen, JR. Introductory Digital Image Processing: A Remote Sensing Perspective, 2nd edit. Upper Saddle River, NJ: Prentice Hall; 1996:179–187. Johnson, KM, Martin, DH. Venezuelan equine encephalitis. Adv Vet Sci Comp Med 1974; 18:79–116. Johnson, KM, Shelokov, A, Peralta, PH, Dammin, GJ, et al. Recovery of Venezuelan equine encephalomyelitis virus in Panama. A fatal case in man. Am J Trop Med Hyg 1968; 17:432–440. Kissling, RE, Chamberlain, RW. Venezuelan equine encephalitis. Adv Vet Sci 1967; 2:65–84. Laveissière, C, Meda, AH. Incidence of African trypanosomiasis and the density of settlements in the forests of Côte d’Ivoire: the possibility of predicting high risk areas for surveillance of disease reservoirs. Trop Med Int Health 1999; 4:199–206. Méndez, W, Liria, J, Navarro, JC, García, CZ, et al. Spatial dispersion of adult mosquitoes (Diptera: Culicidae) in a sylvatic focus of Venezuelan equine encephalitis virus. J Med Entomol 2001; 38:813–821. Moncayo, AC, Edman, JD, Finn, JT. Application of geographic information technology in determining risk of eastern encephalomyelitis virus transmission. J Am Mosq Control Assoc 2000; 16:28–35. Moncayo, AC, Medina, GM, Kalvatchev, Z, Brault, A, et al. Genetic diversity and relationships among Venezuelan equine encephalitis virus field isolates from Colombia and Venezuela. Am J Trop Med Hyg 2001;65:738–746. Morris, CD, Larson, VL, Lounibos, LP. Measuring mos-

230 quito dispersal for control programs. J Am Mosq Control Assoc 1991; 7:608–615. Murphy, FA. New, emerging, and reemerging infectious diseases. Adv Virus Res 1994; 43:1–52. Murphy, FA, Nathanson, N. The emergence of new virus diseases: an overview. Semin Virol 1994; 5:87–102. Powers, AM, Oberste, MS, Brault, AC, Rico-Hesse, R, et al. Repeated emergence of epidemic/epizootic Venezuelan equine encephalitis from a single genotype of enzootic subtype ID virus. J Virol 1997; 71:6697–6705. Rivas, F, Diaz, LA, Cardenas, VM, Daza, E, et al. Epidemic Venezuelan equine encephalitis in La Guajira, Colombia, 1995. J Infect Dis 1997; 175:828–832. Roehrig, JT, Bolin, RA. Monoclonal antibodies capable of distinguishing epizootic from enzootic varieties of subtype I Venezuelan equine encephalitis viruses in a rapid indirect immunofluorescence assay. J Clin Microbiol 1997; 35:1887–1890. Salas, RA, Garcia, CZ, Liria, J, Barrera, R, et al. Ecological studies of enzootic Venezuelan equine encephalitis in north-central Venezuela. Am J Trop Med Hyg 2001; 64:84–92. Scherer, WF, Díaz, A. Discussion. In: Science Publication 243: Venezuelan Encephalitis. Washington, DC: Pan American Health Organization; 1972:261. Scherer, WF, Dickerman, RW. Ecologic studies of Venezuelan encephalitis virus in southeastern Mexico. VIII. Correlations and conclusions. Am J Trop Med Hyg 1972; 21:86–89. Scherer, WF, Dickerman, RW, Chia, CW, Ventura, A, et al. Venezuelan equine encephalitis virus in Veracruz, Mexico, and the use of hamsters as sentinels. Science 1963; 145:274–275. Stagno, P, Steegmayer, P. La erosión reticular en el Sur del Lago de Maracaibo. Agronomia Trop 1972; 22:99–118. Sudia, WD, Newhouse, VF, Henderson, BE. Experimental infection of horses with three strains of Venezuelan equine encephalomyelitis virus. II. Experimental vector studies. Am J Epidemiol 1971; 93:206–211. Turell, MJ. Vector competence of three Venezuelan mosquitoes (Diptera: Culicidae) for an epizootic IC strain of Venezuelan equine encephalitis virus. J Med Entomol 1999; 36:407–409. Turell, MJ, Jones, JW, Sardelis, MR, Dohm, DJ, et al. Vector competence of Peruvian mosquitoes (Diptera: Culicidae) for epizootic and enzootic strains of Venezuelan equine encephalomyelitis virus. J Med Entomol 2000; 37:835–839. Walder, R, Suarez, OM. Studies of arboviruses in southwestern Venezuela: I. Isolations of Venezuelan and eastern equine encephalitis viruses from sentinel hamsters in the Catatumbo region. Int J Epidemiol 1976; 5:375–384.

BARRERA ET AL. Walder, R, Suarez, OM, Calisher, CH. Arbovirus studies in southwestern Venezuela during 1973–1981. II. Isolations and further studies of Venezuelan and eastern equine encephalitis, Una, Itaqui, and Moju viruses. Am J Trop Med Hyg 1984; 33:483–491. Walton, TE, Grayson, MA. Venezuelan equine encephalomyelitis. In: Monath, TP ed. The Arboviruses: Epidemiology and Ecology, Vol. IV. Boca Raton, FL: CRC Press; 1989:203–233. Walton, TE, Alvarez, O, Buckwalter, RM, Johnson, KM. Experimental infection of horses with enzootic and epizootic strains of Venezuelan equine encephalomyelitis virus. J Infect Dis 1973; 128:271–282. Wang E, Barrera R, Boshell J, Ferro, C, et al. Genetic and phenotypic changes accompanying the emergence of epizootic subtype IC Venezuelan equine encephalitis viruses from an enzootic subtype ID progenitor. J Virol 1999; 73:4266–4271. Wang, E, Bowen, RA, Medina, G, Powers, AM, et al. Virulence and viremia characteristics of 1992 epizootic subtype IC Venezuelan equine encephalitis viruses and closely related enzootic subtype ID strains. Am J Trop Med Hyg 2001; 65:64–69. Washino, RK, Wood, BL. Application of remote sensing to arthropod vector surveillance and control. Am J Trop Med Hyg 1994; 50:134–144. Weaver, SC. Recurrent emergence of Venezuelan equine encephalomyelitis. In: Scheld, WM, Highes, J, eds. Emerging Infections I. Washington, DC: ASM Press; 1998:27–42. Weaver, SC, Bellew, LA, Rico-Hesse, R. Phylogenetic analysis of alphaviruses in the Venezuelan equine encephalitis complex and identification of the source of epizootic viruses. Virology 1992; 191:282–290. Young, NA, Johnson, KM. Antigenic variants of Venezuelan equine encephalitis virus: their geographic distribution and epidemiologic significance. Am J Epidemiol 1969; 89:286–307. Zarate, ML, Scherer, WF, Dickerman, RW. Venezuelan equine encephalitis virus as a human infection determinant. Description of a fatal case occurring in Jaltipan, Ver., in 1965. Rev Invest Salud Publ 1970; 30: 296–302.

Address reprint requests to: Dr. Roberto Barrera Instituto de Zoología Tropical Facultad de Ciencias Universidad Central de Venezuela Apartado 47058 Caracas 1041-A, Venezuela E-mail: [email protected]