Hematophagous Diptera Collected from a Horse and Paired Carbon Dioxide-Baited Suction Trap in Southern California: Relevance to West Nile Virus Epizootiology Open Access

Abstract

Hematophagous Diptera landing on a horse were removed by vacuum, and their numbers were related to a paired carbon dioxide-baited suction trap at three locations in southern California where West Nile virus activity was high during the preceding year. Insects collected from the horse included mosquitoes (nine species), biting midges (Culicoides sonorensis Wirth & Jones), and black flies (Simulium bivittatum Malloch). Mosquitoes were predominantly collected from the head, crest, withers, neck, chest, and shoulders of the horse, whereas biting midges and black flies were predominantly collected from the ventral midline of the horse. Culex erythrothorax Dyar was by far the most abundant mosquito species collected overall. Frequency of engorgement for mosquitoes captured from the horse ranged by species from zero to 58.3%, with Culex pipiens quinquefasciatus Say having the lowest value (16.7% or one of six mosquitoes) of species that fed on the horse. The number of insects captured at the horse and paired CO₂-baited suction trap was not different for Anopheles franciscanus McCracken, Culex tarsalis Coquillett, and S. bivittatum. Cx. p. quinquefasciatus was captured in greater numbers in the CO₂-baited suction trap, whereas Anopheles hermsi Barr & Guptavanji, Cx. erythrothorax, Culiseta inornata (Williston), and Culiseta particeps (Adams) were captured in greater numbers from the horse. The horse biting rate was very low for Cx. p. quinquefasciatus, intermediate for Cx. tarsalis, and very high for Cx. erythrothorax. Both Cx. tarsalis and Cx. erythrothorax should be considered likely epizootic vectors of West Nile virus to horses in rural southern California.

West Nile virus (family Flaviviridae, genus Flavivirus, WNV) is a mosquito-borne virus that causes West Nile fever and West Nile encephalomyelitis in infected humans and horses. WNV was first identified in the United States during 1999 in New York City (Komar 2000), and then it spread rapidly across the country, reaching California in 2003 (Reisen et al. 2004). WNV is now the leading cause of human and equine arboviral encephalitis in the United States.

WNV has been detected in 60 mosquito species in North America (CDC 2007). Most Culex spp. and several Aedes spp. have been shown to be competent vectors of WNV in the laboratory, and they may serve as either enzootic or bridge vectors (Turrell et al. 2005). California mosquito species that have been shown to be highly or moderately competent vectors of WNV under laboratory conditions include Culex tarsalis Coquillett, Culex erythrothorax Dyar, Culex stigmatosoma Dyar, Culex pipiens pipiens L., Aedes dorsalis (Meigen), Aedes melanimon Dyar, Aedes vexans (Meigen), and Culiseta inornata (Williston) (Goddard et al. 2002). Vector competence of Culex pipiens quinquefasciatus Say varied by geographic location, with populations collected from southern California being the least efficient laboratory vectors.

With the exception of Cx. erythrothorax, common Culex species in southern California are considered principally ornithophilic. However, mosquito host-feeding preference, especially within the genus Culex, is often biased according to the predominance of host animals in the vicinity of collection sites (Tempelis 1975), and Cx. tarsalis as well as Cx. p. quinquefasciatus in urban environments will take opportunistic bloodmeals from mammals, including humans and horses (Edman 1974, Reisen and Reeves 1990, Niebylski and Meek 1992, Zinser et al. 2004). Cx. tarsalis is also known to shift feeding preferences seasonally, with a greater proportion of bloodmeals taken from mammals relative to birds during mid- to late summer, perhaps associated with the loss of their usual avian food sources (Tempelis 1975). This shift in host preference during mid- to late summer, at a time when infection rates in many mosquito populations are at a peak, may increase the role of Cx. tarsalis in epidemic/epizootic transmission of WNV to humans, horses, and perhaps other mammals making up their new food source (Kilpatrick et al. 2006).

In southern California, Cx. tarsalis and Cx. stigmatosoma are thought to be the principal enzootic vectors of WNV, with the poorly to moderately efficient peridomestic Cx. p. quinquefasciatus associated with epidemic WNV transmission to humans in urban southern California, especially in areas where American crows, Corvus brachyrhynchus Brehm, and western scrub jays, Aphelocoma californica (Vigors), are concentrated (Reisen et al. 2006). Both bird species develop high viremias needed to infect Cx. p. quinquefasciatus (Komar et al. 2003, Reisen et al. 2005).

Still unclear is the role of Culex species in epizootic WNV transmission to horses in the more rural areas of southern California. Cx. p. quinquefasciatus in rural California may not commonly feed on large mammals (Reeves 1953, Reisen and Reeves 1990, Mullens and Gerry 1998). In contrast, Cx. erythrothorax in southern California feeds readily on both birds and mammals (Tempelis 1975) and may serve as a bridge vector of WNV in some areas. Even strongly mammalophilic mosquito species with laboratory demonstrated vector competence (such as Cs. inornata) may play a role in epizootic WNV transmission after nonviremic virus transfer between mosquitoes cofeeding on the same mammalian host (Higgs et al. 2005).

In 2004 and 2005, the distribution of WNV-infected horses in southern California exhibited distinct clustering in time and space over the length of the WNV transmission season (Fig. 1) (Hullinger, unpublished data). This clustering of cases may have been influenced by the geographic distribution of mosquito species, by variation in the local abundance of WNV epizootic vectors, or by the movement of infected enzootic hosts throughout the region. This study examined the species composition and relative abundance of mosquitoes and other hematophagous Diptera attracted to a horse and paired CO₂-baited suction trap in southern California near locations where and during months when equine cases of WNV were clustered during 2004. Insects feeding on horses in these areas might serve as epizootic vectors of WNV.

Materials and Methods

Insect Collections.

Host-seeking mosquitoes, and other hematophagous Diptera, were collected on 21 d, from 11 July through 26 September 2005, at three inland southern California sites near areas where horses were infected with WNV during 2004. Collection sites were located in the inland southern California cities of Norco (33° 55′ N, 117° 35′ W, 176-m elevation), Nuevo (33° 52′ N, 117° 07′ W, 434-m elevation), and Lake Elsinore (33° 41′ N, 117° 21′ W, 389-m elevation) (Fig. 1). Collection sites were located within 500 m of the Santa Ana River, San Jacinto River, and Lake Elsinore, respectively. Host-seeking insects were collected at Norco on 9 d during July and August (11, 15, 18, 20, 25, and 27 July, and 1, 8, and 16 August), at Nuevo on 8 d during August and September (3, 10, 18, 22, and 29 August, and 13, 19, and 26 September), and at Lake Elsinore on 4 d during August and September (31 August and 7, 14, and 21 September). The collection dates chosen for each site were based on the temporal pattern of equine infections with WNV during 2004 (Fig. 1A).

Host-seeking Diptera were collected by aspiration directly from a 14-yr-old gelding Quarter Horse with a gray hair coat by using a battery-powered mechanical aspirator (DC Insect Vac, BioQuip Products Inc., Rancho Dominquez, CA). Host-seeking Diptera also were collected using standard CDC type suction traps (model 512, JW Hock Inc., Gainesville, FL), without light or rain shields, baited with CO₂ (dry ice). Each trap was operated at an approximate height of 1.25 m, and CO₂ was presented in a 3.7-liter foam-insulated can with four openings on the sides and two openings at the bottom (diameter = 4 mm). Both methods of capture were used simultaneously for comparison. At each collection site, the horse and CO₂-baited suction trap were placed 50 m apart with positions alternated on consecutive trap days to reduce position effects. Insects were aspirated from the horse as they were noted by two human collectors wearing headlamps with red lenses to see insects landing on the horse after sunset. Insects were also collected by periodic aspiration of the ventral midline of the horse where some insect species concentrated.

Human collectors wore protective clothing and applied a 25% Deet insect repellent to exposed areas of their bodies. The horse had been vaccinated with the WNV Innovator vaccine (Fort Dodge Animal Health, Overland Park, KS) during 2004, followed by booster vaccinations with the Recombitek Equine West Nile vaccine (Merial U.S., Duluth, GA) during spring 2005 before the start of this study.

Of the 21 total trap days, crepuscular insect activity was examined on 13 trap days (3 d at Norco, 6 d at Nuevo, and 4 d at Lake Elsinore) and diel insect activity was examined on eight trap days (6 d at Norco and 2 d at Nuevo). To examine crepuscular activity, host-seeking insects were collected for 5 h from 1900 to 2400 hours with each hour divided into a 45-min collection period followed by a 15-min recovery period during which aspirator collection tubes and suction trap collection bags were replaced. On 2 d, 19 and 26 September, collections ended early due to rain at 2100 or 2200, respectively. To examine diel activity, host-seeking insects were collected as described above, except that the 5-h collection time was varied over consecutive trap days to examine insect activity over all 24 h of the day at Norco and from 2400 to 1000 hours at Nuevo. In total, there were 100 separate 45-min collection periods. Collected insects were killed in a cooler with dry ice, and then returned to the laboratory and placed into storage at –80°C before identification, determination of engorgement, and counting. Insects were categorized as engorged if any portion of a bloodmeal could be readily seen. Insects aspirated from the horse by the two collectors were pooled for each collection period before analysis.

WNV Testing of Collected Insects.

Host-seeking insects were pooled into groups of ≤50 individuals of the same species by date of collection, method of capture, and engorgement status. Each pool was placed in a plastic 1.5-ml vial containing a plastic bead and then shipped on dry ice to the Center for Vectorborne Diseases (CVEC) at the University of California, Davis, where pools were tested simultaneously for WNV, St. Louis encephalitis, and western equine encephalomyelitis by multiplex reverse transcription-polymerase chain reaction (RT-PCR), with positive samples confirmed using singleplex RT-PCR with a second primer set (Shi et al. 2001). The CVEC laboratory serves as the official arbovirus diagnostic laboratory for the state of California, and it receives mosquito pools for virus identification from the majority of the mosquito and vector control districts throughout the state.

WNV Testing of Horse.

Blood was drawn from the horse every other week starting 1 wk before the study began through 2 wk after the study ended. Blood samples were submitted to the USDA-APHIS National Veterinary Services Laboratory in Ames, IA, to be tested for the presence of WNV-specific IgM antibodies, which would indicate a recent infection with WNV. A vaccinated horse would be expected to show a transient but detectable increase in IgM antibodies if challenged with live WNV from the bite of an infected insect.

Statistical Analysis.

For each insect species, the number of insects collected by the two methods of capture was compared using Wilcoxon matched pairs signed ranks test (GraphPad InStat version 3, GraphPad Software Inc., San Diego, CA) for each 45-min collection period when at least one insect of the species was collected. Species abundance was examined by collection site for Culex spp. using chi-square analysis (GraphPad InStat version 3). For the more abundant mosquito species, the frequency of capture by trap method was examined at one site (Nuevo) for differences between month of capture (August versus September) and during 1 mo (September) for differences between collection sites (Nuevo versus Lake Elsinore) by using Fisher exact test (GraphPad InStat version 3). Count data were limited to mosquitoes captured from 1900 to 2400 hours and represented a minimum of six similar collection periods over a minimum of four collection days.

Results

Species Composition and Diel Activity.

In total, 7,744 hematophagous Diptera were collected from the horse and paired CO₂-baited suction trap over the study period (Table 1). Collected insects included mosquitoes (10 species), biting midges (Culicoides sonorensis Wirth & Jones), and black flies (Simulium bivittatum Malloch). Species composition varied substantially at the three sites, with differences in collection dates and collection times for each of the three sites likely to be responsible for some of the variation.

Cx. erythrothorax was the most abundant mosquito species overall, making up >90% of all mosquitoes collected at Nuevo and Lake Elsinore. Culiseta particeps (Adams) was the second most abundant mosquito species collected. However, the majority was collected at Norco, with 90% collected on the first trapping day in mid-July. Anopheles franciscanus McCracken and Anopheles hermsi Barr & Guptavanji were commonly collected only at Nuevo. The distribution of Cx. p. quinquefasciatus was significantly different (P < 0.0001) from the distribution of Cx. tarsalis and Cx. erythrothorax, with Cx. p. quinquefasciatus more common at Norco and less common at both Nuevo and Lake Elsinore relative to Cx. tarsalis and Cx. erythrothorax.

Of the two hematophagous species collected that were not mosquitoes, C. sonorensis was abundant at Norco and Nuevo, both sites located <3 km from operational dairies containing dairy wastewater ponds, which are typical developmental sites for this species (Mullens and Rodriguez 1988), whereas S. bivittatum was collected only at the Norco site.

Host-seeking mosquitoes were generally most active for the first few hours after sunset, with the more abundant species captured in lower numbers throughout the night (Table 2). Exceptions were An. hermsi, which had a period of peak activity just before sunrise, and Cx. erythrothorax, which had a period of peak activity after sunset and a smaller but still substantial activity peak before sunrise. No mosquitoes were collected during daytime hours. C. sonorensis had periods of peak activity near sunset and sunrise, with a few midges captured during daylight hours, whereas S. bivittatum were only captured during daylight hours before sunset and for up to 3 h after sunrise.

Horse Biting Activity.

Mosquitoes could be readily seen at night by using head lamps with red lenses, and the two human collectors required 2–3 min to visually scan the entire body of the horse, aspirating any mosquitoes noted. Black flies, active only during daylight hours, also were readily seen and rapidly collected. In contrast, due to their small size and crepuscular activity, biting midges were difficult to see on the horse, and they were almost certainly grossly undercollected from the horse.

The majority of mosquitoes landed on the head, crest, withers, neck, chest, and shoulders of the horse. The specific feeding site preference of each species was not evaluated. In contrast, biting midges and black flies landed almost entirely on the ventral midline from the chest to the rear of the belly.

The horse biting rate was temporally high just after sunset. The greatest single period biting rate occurred from 2000 to 2045 on 3 August at the Nuevo site when 265 mosquitoes, predominantly Cx. erythrothorax, were aspirated from the horse. The mean number of mosquitoes aspirated from the horse during the 2000–2045 collection period over all sites and collection dates was 75 mosquitoes.

Engorgement rates for each species aspirated from the horse ranged from zero to 58.3% (Table 3). These values are conservative, because some insects may have been aspirated from the horse too quickly to have initiated feeding. Of mosquito species that did feed on the horse, Cx. p. quinquefasciatus had the lowest engorgement, 16.7% or one of six mosquitoes aspirated from the horse. Although engorgement is also quite low for Cs. particeps (19.5%), the majority of this species was collected near sunset on one evening when this comparatively large mosquito was easily seen and rapidly collected after landing on the horse. Removing this single collection period from analysis, engorgement for the remaining Cs. particeps was 47.8%. Similarly, engorgement in S. bivittatum may be relatively low at 21.2%, because it was a diurnally active insect and therefore readily seen by the human collectors.

Comparison of Horse and Suction Trap Collections.

With the exceptions of Ae. dorsalis and Cs. incidens, collected in very low numbers and known to feed primarily on large mammals, including horses (Bohart and Washino 1978), all species aspirated from the horse also were captured in a paired CO₂-baited suction trap (Table 3). Cx. stigmatosoma, also collected in very low numbers, was the only species captured in the CO₂-baited suction trap and not aspirated from the horse. Two of the most abundant mosquito species from all three sites combined, Cx. erythrothorax and An. franciscanus, showed significant correlation (r = 0.49 and 0.81, respectively; P < 0.05) between the number of insects collected at the horse and paired CO₂-baited suction trap, indicating that both methods of capture may be appropriate to monitor changes in the host-seeking population of these species. Poor correlation between capture methods for the remaining species was likely due to low overall capture or to capture during just a few collection periods making analysis difficult.

For An. franciscanus, Cx. tarsalis, and S. bivittatum, the number of insects aspirated from the horse was not statistically different than the number captured at a paired CO₂-baited suction trap (W = 110, 222, 12; and P = 0.054, 0.42, 0.25, respectively) (Table 3). Five mosquito species had abundance values that were significantly influenced by the method of capture; Cx. p. quinquefasciatus (W = 4, P = 0.0002) was collected in greater numbers by CO₂-baited suction trap, whereas An. hermsi (W = 6.5, P < 0.0001), Cx. erythrothorax (W = 48.5, P < 0.0001), Cs. inornata (W = 10, P = 0.04), and Cs. particeps (W = 15, P = 0.03) were collected in greater numbers by aspiration from the horse.

Of the four most abundant mosquito species, An. franciscanus and Cx. erythrothorax were captured from the horse at lower frequency during August compared with September (P = 0.0003 and P < 0.0001, respectively) with risk ratios (95% confidence interval) of 0.41 (0.26–0.66) and 0.87 (0.83–0.91), whereas An. hermsi and Cx. tarsalis demonstrated no variation in frequency of capture from the horse by collection month. For mosquitoes collected in September, Cx. tarsalis was captured from the horse at lower frequency at Nuevo relative to Lake Elsinore (P = 0.0016), with a risk ratio of 0.35 (0.17–0.70), whereas no variation in frequency of capture from the horse was noted for the remaining species.

WNV Activity.

Of 361 pools of mosquitoes, biting midges, and black flies tested for the presence of WNV, one pool (n = 12) of Cx. tarsalis contained virus. This pool was captured by CO₂-baited suction trap at the Nuevo site on 3 August, the night of the greatest single collection period. The horse used in this study did not show an increase in WNV specific IgM antibodies that would have indicated challenge with WNV during the study period.

Discussion

Species Composition and Diel Activity.

Of the mosquito species recognized as WNV vectors in southern California, only Cx. tarsalis and Cx. p. quinquefasciatus were captured at all three sites. Cx. stigmatosoma was captured only at Norco. However, Cx. stigmatosoma is poorly collected in ground level CO₂-baited suction traps (Reisen et al. 1990a), and with a strong ornithophilic feeding habit (Bohart and Washino 1978, Reisen et al. 1990a) would not be expected to be captured at the horse. Cx. erythrothorax was the most abundant mosquito collected at each collection site, with the exception of Cs. particeps at Norco, which was predominantly collected on a single evening in mid-July.

The relatively high abundance of Cx. p. quinquefasciatus at Norco is likely due to the presence of dairy wastewater ponds located on the opposite side of the Santa Ana River, within 1–2 km of the collection site, which serve as excellent developmental sites for this species (Reisen et al. 1990b). The Norco site was also in closer proximity to suburban homes relative to the other two collection sites. However, seasonal variation in species abundance cannot be excluded due to the earlier collection dates at Norco relative to collection dates at Nuevo and Lake Elsinore.

An. hermsi and An. franciscanus where commonly collected at Nuevo but not at Norco or Lake Elsinore presumably due to the presence of the San Jacinto River and a nearby smaller tributary both containing abundant vegetation and numerous pools of slow moving water, a preferred developmental site for these species (Bohart and Washino 1978).

Early morning activity peaks for An. hermsi and Cx. erythrothorax are in agreement with a previous study at a southern California wetland (Cope et al. 1986). However, Schreiber et al. (1988) showed only evening activity peaks for An. hermsi and Cx. erythrothorax at a dairy region just a few kilometers from our collection site at Norco. All mosquito species were active only during nighttime hours, whereas black flies and biting midges were active to some extent during daytime hours. Diurnal activity of C. sonorensis has been shown previously (Mullens 1995). Mosquito repellent products applied to horses to reduce biting should be applied in the hour before sunset for maximum protection. To provide protection against S. bivittatum and C. sonorensis, repellent products may need to be reapplied to horses before sunrise to reduce biting during the morning activity period of these species.

Horse Biting Activity.

Host orientation by hematophagous Diptera is guided by olfactory and visual cues (Logan and Birkett 2007). For nocturnally active mosquito species, host location is likely dominated by olfactory stimuli (Gillies and Wilkes 1972), which also may play a significant role in the selection of biting sites on the host (De Jong and Knols 1995). Knols et al. (1994) suggested that opportunistic mosquito species may use CO₂ as a reliable kairomone for host location and selection of a biting site, resulting in feeding site selection near the head. With Cx. erythrothorax making up >86% of the mosquitoes collected in this study, this opportunistic mosquito species (Tempelis 1975) seems to use CO₂ or other breath components in selecting feeding sites near the head of the horse. In contrast, S. bivittatum and C. sonorensis were collected predominantly from the ventral midline of the horse, indicating that feeding site selection was determined by physical characteristics of the host body or by olfactory stimuli not associated with the breath. The ventral midline is also known to be the preferred feeding site of C. sonorensis on cattle in southern California (Mullens et al. 2000).

The horse biting rate was very high during the first hour after sunset, with nearly all mosquito species exhibiting peak biting activity. The greatest horse biting rate noted in this study, 265 mosquitoes aspirated from the horse during a single 45-min collection period, would result in an average of nearly six mosquitoes landing on the horse during each minute, assuming an even distribution of the landing rate over the entire 45-min period. Given a mosquito feeding time of several minutes and the likelihood that mosquito biting was not distributed evenly over the collection period, a large number of mosquitoes may have been feeding on the horse at the same time. High biting rates, concurrent biting activity by many mosquito species, and the selection of feeding sites on limited areas of the horse body may provide opportunity for WNV transmission between cofeeding mosquitoes (Higgs et al. 2005).

There was considerable variation by species in engorgement success before aspiration from the horse. Because the insects were collected as they were noted on the horse, engorgement in this study is a measure of speed at which each species initiated blood feeding after landing rather than actual success at obtaining a bloodmeal. Of the mosquito species that did feed on the horse, the species known to feed predominantly on mammals in southern California typically exhibited the greatest engorgement success, whereas the more ornithophilic Culex spp. generally had the lowest engorgement success.

Culex p. quinquefasciatus had the lowest engorgement success of all species that fed on the horse in this study. This finding is in agreement with an earlier study where Cx. p. quinquefasciatus collected in horse-baited traps in Louisiana had very low engorgement rates (0–9.7%) relative to other mosquito species (Samui et al. 2003). Similarly, Mullens and Gerry (1998) showed that Cx. p. quinquefasciatus captured near the Norco site were readily attracted to young cattle, but that they did not feed on them. In contrast, Reisen et al. (1990b) demonstrated by bloodmeal analysis that Cx. p. quinquefasciatus in the Norco area will feed on cattle.

Comparison of Horse and Suction Trap Collections.

Traps are useful tools for monitoring species activity and relative abundance, but they often do not provide an accurate measure of host biting rate (Service 1993). An accurate estimate of biting rate is especially important when determining risk of pathogen transmission by a particular insect species.

With the exception of Cx. tarsalis and An. franciscanus, the CO₂ trap provided a reasonable indication of species presence but a poor estimate of the horse biting rate for mosquito species captured. The horse biting rate was overestimated by CO₂ trap captures for Cx. p. quinquefasciatus, with a correction factor of 0.14 times the CO₂ trap capture relating this value to the horse biting rate. In contrast, the horse biting rate was underestimated by CO₂ trap captures for An. hermsi, Cx. erythrothorax, Cs. inornata, and Cs. particeps. For Cx. erythrothorax, a correction factor of 2.7 times the CO₂ trap capture related this value to the horse biting rate.

The capture ratio for most mosquito species collected in this study corresponds with known feeding behavior (Bohart and Washino 1978), with species known to be strongly mammalophilic having the greatest horse to CO₂ trap capture ratio. Exceptions to this generalization were An. franciscanus and Cx. erythrothorax. An. franciscanus feeds predominantly on rabbits and other small mammals (Bohart and Washino 1978), which would have low CO₂ output relative to a horse. The high capture ratio of Cx. erythrothorax indicates an increased preference for mammal feeding relative to other Culex species captured. In southeastern California, Cx. erythrothorax has been previously reported to feed equally on both birds and mammals (Tempelis 1975).

Of all the species collected in this study, only Cx. p. quinquefasciatus was collected significantly more often in the CO₂ trap relative to collection at the horse. Host orientation by Cx. p. quinquefasciatus seems to be principally a function of CO₂ detection without the need for other host cues (Mullens and Gerry 1998). Response of this species is known to decrease at higher CO₂ concentrations typical of horses or cattle relative to the much lower CO₂ concentrations of birds or smaller mammals (Reeves 1953, Mullens and Gerry 1998). Although the CO₂ output of a standard dry ice baited suction trap is high, 1,100–1,500 ml/min (Mullens 1995), it is still considerably lower than the 1,786 ml/min CO₂ output expected for a resting 5-yr-old Thoroughbred weighing 470 kg (Van Erck et al. 2005), which would be similar in size to the horse used in this study.

C. sonorensis also seemed to be captured in nearly equal numbers at the horse and the CO₂-baited suction trap, but it was not tested for significance due to presumed undercollection at the horse. Previous studies have shown that C. sonorensis were 7.2 times more likely to be captured from a calf than from a paired CO₂-baited suction trap (Mullens and Gerry 1998).

Whereas Cx. tarsalis is known to shift feeding preferences to increase mammal feeding after midsummer (Tempelis 1975), there was no difference in the frequency of capture of this species at the horse relative to the CO₂ trap from August to September. In contrast, Cx. erythrothorax and An. franciscanus were collected at the horse in lower frequency during August relative to September, perhaps indicative of an increased preference for large mammals between mid- to late summer. Cx. tarsalis was collected from the horse in significantly lower frequency at Nuevo relative to Lake Elsinore, perhaps indicating a greater preference for large mammals within the Cx. tarsalis population at Lake Elsinore.

Relevance to WNV Epizootiology.

In 2004, WNV-infected mosquitoes collected in western Riverside County by local mosquito abatement programs included Cx. tarsalis (19 pools), Cx. erythrothorax (eight pools), Cx. p. quinquefasciatus (three pools), and Cx. stigmatosoma (two pools) (Hom et al. 2005). The majority of these WNV-infected mosquitoes were collected along the Santa Ana River from Norco to Riverside during June or July (Wisniewska-Rosales and Williams 2005). Similarly, horses infected with WNV during 2004 were also concentrated along the Santa Ana River during June and July (Fig. 1A). The high level of WNV activity along the Santa Ana River may have been associated with an abundance of American crow roosts along this river (Reisen et al. 2006). A second area of high virus activity in horses occurred throughout the semirural area south of the San Jacinto River. However, relative to the Santa Ana River area, WNV activity south of the San Jacinto River occurred substantially later in the summer with horses infected from late July through October and WNV infected Cx. tarsalis and Cx. erythrothorax collected in August and September (Riverside County Department of Environmental Health [RCDEH], unpublished data).

During 2005, virus activity along the Santa Ana River was greatly reduced with just three pools of WNV-infected mosquitoes collected (one pool each of Cx. p. quinquefasciatus, Cx. tarsalis, and Cx. thriambus) during June and August, even though the trapping effort was greater relative to 2004 (Wisniewska-Rosales, unpublished data). Similarly, few horses were infected with WNV along the Santa Ana River in 2005 and infections were limited to July and August (Fig. 1B). The substantial decrease in virus activity along the Santa Ana River during 2005 relative to 2004 may have been due in part to dramatic mortality within the American crow population in this area (T. Scott, unpublished data), and perhaps other avian hosts as well, after widespread exposure to WNV in 2004. Increased vaccination or previous exposure to WNV also may have been responsible for some of the reduction in horse infections.

In contrast to the Santa Ana River area, WNV activity south of the San Jacinto River was similar between years, though virus activity occurred somewhat earlier in 2005 relative to 2004. Virus activity in both horses and mosquitoes began in late June and continued through September. WNV-infected mosquitoes collected in western Riverside County (excluding those collected along the Santa Ana River) included Cx. erythrothorax (11 pools), Cx. tarsalis (eight pools), Cx. stigmatosoma (two pools), and Cx. p. quinquefasciatus (one pool) (J. W., unpublished data; RCDEH, unpublished data).

During both 2004 and 2005, Cx. tarsalis and Cx. erythrothorax were the mosquito species most commonly collected infected with WNV in western Riverside County. During 2004, the minimum infection rate (MIR) of Cx. tarsalis (3.0/1,000) was higher than that of Cx. p. quinquefasciatus (1.7/1000) and Cx. erythrothorax (0.4/1000) in western Riverside County with total captures of each species numbering 6,403, 1,749, and 19,803 mosquitoes, respectively (Hom et al. 2005). However, the use of numerous gravid traps that collect primarily older Cx. p. quinquefasciatus is likely to have increased the MIR of this species relative to the other Culex spp., which were primarily collected in CO₂-baited suction traps.

Although Cx. tarsalis was not collected in large numbers from the horse during this study, it was the only species collected at our study sites that was infected with WNV. This species is commonly collected throughout the region in CO₂-baited suction traps and was relatively abundant at both the Nuevo and Lake Elsinore sites from July through September 2005 when WNV activity was high in these areas. The high MIR of Cx. tarsalis resulting in the detection of WNV in 28 pools of this species during 2004 and 2005, and the similar attack rate of this species on a horse relative to a CO₂ trap suggest that Cx. tarsalis may be an important epizootic vector of WNV to horses in western Riverside County.

Although Cx. p. quinquefasciatus seems to be an opportunistic blood feeder in urban environments of southern California (Schreiber et al. 1989, Reisen et al. 1990a) and it is considered the epidemic vector of WNV in urban southern California (Reisen et al. 2006), the horse biting rate of this species was very low at our collection sites relative to other potential WNV vector species. Indeed, Cx. p. quinquefasciatus was rarely collected from the horse, even at Norco, where this species was most abundant. Furthermore, southern California populations of Cx. p. quinquefasciatus have shown limited vector competence and a longer extrinsic incubation period for WNV relative to other species captured during this study (Goddard et al. 2002). Given the inflated MIR, the low horse biting rate, and the limited vector competence, the risk of WNV transmission by Cx. p. quinquefasciatus to horses in rural western Riverside County is substantially overestimated by our current surveillance system relative to other potential vector species. This species may still play a significant role in WNV transmission to horses in more urban areas where perhaps abundance and horse biting rate may be greater.

Cx. erythrothorax was by far the most abundant mosquito species collected at our three study sites when nearby WNV activity was high. The exceptionally high abundance of this mosquito species throughout the region compensated for a low MIR, resulting in the collection of 19 pools of WNV-infected Cx. erythrothorax during 2004 and 2005. With the horse biting rate of this species significantly underestimated by CO₂ trap captures, horses in rural western Riverside County are likely at greater risk of receiving infective bites from Cx. erythrothorax than from Cx. tarsalis.

Cx. erythrothorax commonly develop in vegetated river margins (Walters and Smith 1980) and freshwater impoundments, especially those with stands of tule (Schoenoplectus [=Scirpus] spp.) and cattail (Typha spp.) (Bohart and Washino 1978). This species is known to have limited dispersal ability with host-seeking females rarely collected >1.0 km from a release site (Walton et al. 1999). Much of the rural land in southern California that is either zoned for animal housing or on which equine-based recreation is allowed lies along vegetated rivers and freshwater lakes. The risk of WNV transmission by Cx. erythrothorax to horses in these areas may be especially high.

Acknowledgements

We thank Chris Mullens and Greg Williams (Northwest Mosquito and Vector Control District) for considerable assistance in mosquito identification, Keith Jones of the Riverside County Department of Environmental Health for providing data on WNV infected mosquito pools from portions of western Riverside County, and Shaun Swann of Heavenly Ponies and Critters who provided and handled the horse used for this study. We are also grateful to anonymous reviewers who provided very helpful reviews of an earlier draft of this manuscript. This study was funded by Federal Hatch Funds to A.C.G. and conducted under Animal Use Protocol A-0505012-1 approved 2 May 2005.

References Cited