We sampled mosquitoes in a variety of habitats along Main Park Road from the Everglades Visitor Center to Flamingo (permit EVER-2015-SCI-0054). The sampling locations were divided into 3 regions, upper, middle, and lower, representing natural physiographic regions of south Florida (Table 1; Figure 1). The upper region, from Royal Palm North (25°24′08.3′′N, 80°36′56.7′′W) to Pa-hay-okee South (25°25′56.0′′N, 80°46′38.9′′W), was dominated by large expanses of upland pine or hardwood forest. The middle region, from just south of Pa-hay-okee South extending to Nine Mile Pond (25°15′14.1′′N, 80°47′53.5′′W), was dominated by wet sawgrass prairie (Everglades marsh) with smaller hardwood hammocks (tree islands). The lower region, from Snake Bight Trail to Bear Lake Trail (25°8′55.82′′N, 80°55′23.69′′W), was dominated by extensive mangrove swamp.

We aspirated all mosquitoes from resting shelters (15), except at Mahogany Hammock, where we targeted natural resting sites, such as fallen logs and deep recesses, to abide by permit restrictions. We performed aspirations using a modified handheld battery-powered vacuum (DustBuster BDH1800S; Black and Decker Corporation, https://www.blackanddecker.com) fitted with a funnel and stainless-steel mesh-bottom collection cup (BioQuip model 2846D; BioQuip Products, Inc., https://www.bioquip.com), as described previously (16,17). Resting shelters were equivalent in size and shape to previous models (15) but were constructed of PVC pipe and fittings so that they could be easily disassembled after each collecting trip, in accordance with permit requirements. We sampled a total of 406 resting shelter days in ENP, at a rate of 14–27 resting shelters per sampling period; we placed 1–5 shelters at each site and sampled them for 3–5 consecutive days between 7:00 am and 1:00 pm. We placed resting shelters in areas with maximum shade within 90 m of Main Park Road or 45 m of trails. Sampling occurred within the Everglades dry season (in February and May) and wet season (in June and August) of 2016.

We identified mosquitoes to species using morphological features of the adult (16). Because of well-known difficulties of identifying Culex (Melanoconion) females (17,18), we initially confirmed identifications by morphology of the cibarial armature (17) and molecular assays targeting the 18s mitochondrial gene (18). Narrow decumbent scales of the vertex (19) served as a helpful diagnostic feature to separate Cx. cedecei from other Culex (Melanoconion) spp. in Florida.

We performed blood meal analysis on individual blood-engorged Cx. cedecei females using published PCR-based techniques (20). We amplified extracted DNA using PCR assays targeting cytochrome B and 16s rRNA genes of vertebrate hosts. We initially screened samples to identify blood meals from mammalian and amphibian hosts, using primers L2513/H2714 (21) that target 16S rRNA of the host animal (20). We then screened samples that produced no amplicon using primer pairs L0/H1 and/or L0/H0, both targeting the cytochrome b gene of birds (22), and primer pair 16L1/H3056 used in phylogenetic studies of reptiles (23,24). Primer sequences and cycling conditions are described in Blosser et al. (20). We sent PCR products to Eurofins Scientific (https://www.eurofins.com) for Sanger sequencing (forward direction only). We aligned sequences with published sequences in the National Center of Biotechnology Information sequence database using BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). We considered nucleotide similarity >95% a positive match. We maintained cold chain for all samples from the time of capture until DNA/RNA extraction. To minimize host DNA cross contamination during sorting and processing, we handled specimens by their legs (to avoid puncturing the abdomen) using clean forceps and gloved hands. In general, blood-engorged females from resting shelters were fully intact, with no evidence of ruptured abdomens. We discarded any specimen with ruptured abdomen.

We screened RNA extracts (QIAamp viral RNA mini kit; QIAGEN, https://www.qiagen.com) from pooled Cx. cedecei females for presence of EVEV RNA using quantitative reverse transcription PCR (qRT-PCR) assay with (5′→3′) primers (forward: CGAGGAGCTGTTTAAGGAGTATAA; reverse: CCTCTATGGCTATTGGGCTATG) and probe (6-FAM/CGTTAGGTGTGCCGTTGGGAGTT/3BHQ1/) targeting EVEV nucleocapsid structural protein (primers/probes design by Integrated DNA Technologies, https://www.idtdna.com). Cycling conditions were 50°C (30 min), 95°C (2 min), then 40 cycles of 95°C (15 s) and 61°C (1 min). For assay validation we used EVEV strain FE3-7C, obtained from the Centers for Disease Control and Prevention, propagated on Vero cells using standard techniques (25,26). Mosquito pools consisted of <32 unfed and gravid Cx. cedecei females aggregated by month and site. We determined samples with C_t values <33 to be positive for EVEV RNA.

Cx. cedecei mosquito abundance varied across season and region (Figure 2, panel A). Females were significantly more abundant in the wet season in upper (AIC = 228.22, z = 6.488; p<0.001) and middle (AIC = 295.24, z = 3.811; p<0.001) regions of the Everglades but significantly more abundant during the dry season in the lower region (AIC = 251.2, z = −3.897; p<0.001). No other mosquito species exhibited a pattern of greater abundance during the dry season in any region (Appendix Table 1). For all other commonly collected mosquito species, abundance was greater during the wet season than dry season in all areas, although differences were not always significant (Appendix Table 1). Even within a single region and sampling period, Cx. cedecei mosquito abundance varied considerably. For example, we collected 13.50 females per resting shelter at Royal Palm North (60 shelter days), whereas we collected 2.43 females per resting shelter at Pinelands (44 shelter days); the distance between the sites was only ≈6.7 km.

The distribution of host species blood meals was significantly different spatially (χ² = 90.90, df = 20; p<0.001) between the lower, middle, and upper regions of ENP (Figure 2, panel B) but not seasonally (χ² = 14.49, df = 10; p = 0.152) (Figure 2, panel B). Rodents accounted for 77%–100% of the 347 total identifiable blood meals (77.0% of 451 samples returning >95% match), depending on season and region (Table 2). Blood meals from EVEV reservoirs (hispid cotton rat and cotton mouse combined) were a large percentage of total blood meals from all regions, constituting 48% of total blood meals from the lower, 84% from the middle, and 50% from the upper ENP region (Table 2). Reservoir host blood meals originated overwhelmingly from hispid cotton rat (Table 2; Figure 2, panel B), regardless of region or season. However, cotton mouse blood meals were relatively more common in the dry season in the lower and middle ENP (Figure 2, panel B). Blood meals from invasive Rattus spp. rodents contributed to the significant difference in distribution of host blood meals by region, such that much higher numbers of Rattus spp. blood meals were encountered in upper (40%) and lower (30%) than middle (2%) regions of ENP (Figure 2, panel B). Blood meals obtained from humans constituted 7.53% of total blood meals and were detected in all 3 regions and seasons. All other hosts constituted <1% of Cx. cedecei host blood meals (Table 2; Appendix Table 2).

Although the distribution of blood meals from different host species did not vary substantially between seasons, the number of blood meals from reservoir hosts by season and region did (Figure 2, panel B). The number of hispid cotton rat–fed females per resting shelter day was greatest in the wet season in the upper region, peaking at 3.20 hispid cotton rat blood meals per shelter-day in August in the upper region. By contrast, we encountered 0.02 hispid cotton rat blood meals per shelter-day in lower ENP that same month (Figure 2, panel B).

In total, we screened 3,673 females (1,326 in dry season, and 2,347 in wet season) for EVEV RNA by RT-PCR (average pool size + SD of 22 + 6 females) (Table 1), including 3,413 females from resting shelters and 260 females aspirated from natural resting sites. We found 4 pools of Cx. cedecei females, all of which were from the wet season (June or August), to be positive for EVEV by RT-PCR. These EVEV-positive pools included 1 pool of 25 females from Pinelands (upper region) in June (C_t = 21.3), 1 pool of 8 females from Bear Lake Trail (lower region) in August (C_t = 21.8), 1 pool of 25 females from Royal Palm North (upper region) in August (C_t = 24.8), and 1 pool of 25 females from Long Pine Key North (upper region) in August (C_t = 24.5). Multiple logistic regression showed no association between EVEV RNA detected in pooled Cx. cedecei and Cx. cedecei abundance, reservoir (hispid cotton rat plus cotton mouse) host use, or proportion of blood meals from reservoir hosts (χ² = 4.08, df = 3; p = 0.252).

The dynamic nature of Cx. cedecei mosquito abundance is probably shaped by seasonal fluctuations in water levels through the seasonal filling and depletion of limestone solution holes (pits in karst that formed when sea level was lower than present levels), which constitute the primary habitat of Cx. cedecei larvae (29). The contrasting patterns of Cx. cedecei abundance in lower and upper regions of ENP suggest that Everglades hydrology has a strong influence on the seasonal reproductive biology of Cx. cedecei mosquitoes, which may have consequences for EVEV transmission. In addition, the finding that Cx. cedecei was the only mosquito species more abundant in dry season in the lower region of ENP suggests that the positive association between precipitation and abundance of other mosquito species may not apply consistently for this species. Our study did not quantify precipitation nor the availability of larval habitat, so explanations of the links between season, region, and Cx. cedecei abundance are speculative. One possible explanation for the higher abundance of Cx. cedecei mosquitoes during the dry season in the lower region is seasonal drying in the low-elevation lower Everglades: during times of higher water, aquatic predators, particularly fish, disperse rapidly throughout ENP (30,31), whereas isolated pockets of water without fish may be more abundant during the dry season. Detailed field and laboratory studies focusing on the ecology of immature stages are needed to understand the mechanisms driving these contrasting patterns of Cx. cedecei abundance throughout the Everglades.

Variation in Cx. cedecei host use between regions has important implications for understanding the transmission of EVEV in Florida. Our data demonstrate that Cx. cedecei mosquitoes feed heavily on mammals; rodents make up a substantial portion of hosts, regardless of season or region. The finding that hispid cotton rat and cotton mouse together constituted a large portion (43.0%–86.2%) of Cx. cedecei blood meals confirms a strong association between Cx. cedecei mosquitoes and these EVEV reservoir host species (3,14). Both hispid cotton rat and cotton mouse are common throughout Florida, so their role in EVEV transmission is probably limited by the distribution and abundance of Cx. cedece mosquitoes i. However, the introduction and establishment of Cx. panocossa mosquitoes in Florida may change this dynamic, resulting in more areas at risk for EVEV transmission (8).

The importance of Rattus spp. rodents as hosts of EVEV in lower and upper regions of ENP bears additional investigation, because these rats were relatively common hosts of Cx. cedecei mosquitoes in these regions (20.0%–55.5% of total). We expect the lower and upper regions of the park, because of their proximity to park boundaries, campgrounds, parking lots, and human activity, to host larger populations of invasive rats than the middle region of ENP, which is comparatively undisturbed. Little information is available on the importance of Rattus spp. rodents as hosts of EVEV. Sanmartin et al. (32) made 4 isolations of VEEV from 41 Rattus spp. rodents sampled during an epizootic of VEEV in El Carmelo, Colombia, where subtypes IAB and IC circulate. In Florida, Bigler (33) detected EVEV antibodies in 12.5% (n = 40) R. rattus rats sampled east of ENP. These findings suggest that these widespread, invasive rats might also support the transmission of EVEV in Florida, although laboratory host competence studies are needed to address this hypothesis.

The regions of the park in which human footprint is greatest also had the greatest relative numbers of human blood meals (11.48% lower; 7.80% upper), compared with findings from the middle region (5.95%). These relatively high levels of feeding suggest that Cx. cedecei mosquitoes could serve not only as an enzootic and epizootic vector but also as a potential epidemic vector of EVEV where it comes into contact with humans.

The lack of a clear association between EVEV in pooled Cx. cedecei females and vector abundance or metrics of host use is perplexing. Three of 4 EVEV-positive pools were from upper ENP during the wet season (1 in June, 2 in August), when Cx. cedecei mosquitoes were more abundant (Figure 2, panel A) and obtained a large number of blood meals from cotton rats (Figure 2, panel B). Conversely, the EVEV-positive pool from lower ENP was also from the wet season (August), but Cx. cedecei mosquito numbers were relatively low (Figure 2, panel A), and we observed very few blood meals from rodents (Figure 2, panel B). It is possible that wet-season transmission of EVEV is driven largely by the ecology and reproductive biology of cotton mouse and hispid cotton rat; however, a complex picture of rodent breeding and population dynamics emerges from past studies on this topic. Bigler et al. (14) concluded that the preponderance of EVEV amplification occurred between July and October, when dense populations of both cotton mice and cotton rats inhabited hammocks in the Pinecrest area on the ecotone between the Big Cypress Swamp and the Everglades ecosystem, ≈38 km northwest of the nearest sampling sites in our study. Lord et al. (34) screened mammals for EVEV antibodies in both Big Cypress Swamp and the Everglades, including the upper (Royal Palm Hammock) and middle (Mahogany Hammock) regions sampled in our study. Those results indicated that rodent breeding in these Everglades regions peaked in the dry season (January–February), a reversal of breeding patterns found farther north (14). Smith and Vrieze (35), working on hammocks of Taylor Slough (southeast Everglades), stated that all rodent reproduction occurred in the wet season. These contrasting results indicate that S. hispidus rats and P. gossypinus mice are likely to time their breeding to coincide with local conditions, which may be substantially different across locales. Our results are evidence of spatial variation in seasonality of EVEV transmission as observed in our work and that of others.

It was surprising that we did not detect EVEV in Cx. cedecei mosquito samples from the middle region of ENP, despite relatively high Cx. cedecei abundance (Figure 2, panel A), a relatively high biting rate on hispid cotton rats (Figure 2, panel B), and past evidence of EVEV circulation in mosquitoes in that region (2). Most Cx. cedecei females sampled from the middle region (70.5%; 942/1,336; Table 1) were captured at Nine Mile Pond on the ecotone between sawgrass marsh and mangrove swamps (Figure 1), a location for which historical data on EVEV prevalence is not available.

Although results of our work do not provide a complete understanding of EVEV transmission in ENP, they may help to clarify the perplexing heterogeneity observed in previous studies of EVEV in south Florida. The absence of EVEV in mosquitoes at Pa-hay-okee Overlook (2), for example, could be due to a near absence of Cx. cedecei mosquitoes at that site (Table 1). Differences in EVEV transmission at 2 adjacent hammocks observed by Bigler et al. (14) could be a result of differences in vector abundance, contact rates, or both between vectors and reservoir versus nonreservoir hosts, as observed in our study. We did not quantify the actual abundance of reservoir hosts in the field, so it is possible that a given set of mosquitoes all could have fed on a small number (in low population settings) or a large number (greater populations) of rodents. Our measure of host usage could not account for those differences, nor could it account for how many of those rodents were actually susceptible hosts. The number of available susceptible reservoir hosts is an important factor in maintaining enzootic cycling of viruses. Our data do support conclusions of past studies that EVEV transmission occurs seasonally and is heterogeneous across ENP. Relative number of feedings on reservoir hosts differs across regions of the park, indicating the vector infection ratio may be higher in specific habitats where Cx. cedecei mosquitoes have ample access to reservoir hosts, particularly young, virus-susceptible rodents.

Quantifying the spatial and temporal variation in abundance, host use, and virus infection of Cx. cedecei mosquitoes is a step toward understanding the ecology of EVEV transmission in the United States. Findings that Cx. cedecei mosquitoes feed on both reservoir hosts and humans in nature suggests that this insect could serve as both enzootic and epizootic vector. The establishment of the Cx. panocossa mosquito in Florida, and its putative spread, may change the spatial risk for EVEV transmission, if found to be a competent vector. Future work should evaluate the host competence of Rattus spp. rodents for EVEV and other VEEV subtypes, evaluate vector competence of Cx. panocossa mosquitoes for these viruses, and perform longitudinal studies of the focality of transmission in Florida.

Additional information about patterns of abundance, host use, and Everglades virus infection in Culex (Melanoconion) cedecei and other mosquitoes, Florida.