Utilization of a zoo for mosquito (Diptera: Culicidae) diversity analysis, arboviral surveillance, and blood feeding patterns

Abstract

Zoos provide a unique opportunity to study mosquito feeding ecology as they represent areas where exotic animals, free-roaming native animals, humans, and mosquito habitats overlap. Therefore, these locations are a concern for arbovirus transmission to both valuable zoo animals and human visitors. We sampled mosquitoes in and around The Nashville Zoo at Grassmere in Tennessee, USA, over 4 months in 2020 using 4 mosquito trap methods and 12 sampling locations. Mosquitoes were identified to species, Culex mosquitoes were analyzed for arboviruses, and all engorged mosquitoes were preserved for host usage analysis. We captured over 9,000 mosquitoes representing 27 different species, including a new species record for Davidson County, TN (Culex nigripalpus Theobald). Minimum infection rates for West Nile virus (WNV) (Flaviviridae: Flavivirus), St. Louis encephalitis virus (Flaviviridae: Flavivirus), and Flanders virus (Hapavirus: Rhabdoviridae) were 0.79, 0, and 4.17, respectively. The collection of 100 engorged mosquitoes was dominated by Culex pipiens pipiens Linnaeus (38%), Culex erraticus Dyar and Knab (23%), and Culex pipiens pipiens–Culex pipiens quinquefasciatus hybrids (10%). Host DNA from 84 engorged mosquitoes was successfully matched to a variety of host species (n = 23), with just 8 species belonging to the zoo. Wild birds were the most frequently fed upon host, in particular northern cardinals (Cardinalis cardinalis L. Passeriformes: Cardinalidae), which are competent WNV reservoirs. Taken together, our results demonstrate the utility of zoos as sentinels for emerging pathogens, for studying wildlife and human risk of zoonotic diseases, and for assessing vector diversity.

Introduction

Mosquitoes are the leading arthropod of medical and veterinary importance in zoos as evidenced by the number of and variety of mosquito-borne disease infections in zoo animals (Adler et al. 2011). Often these infections result in death, which can be detrimental to zoo-based conservation efforts and can have negative publicity. For instance, 2 Mexican gray wolves (Canis lupus baileyi) died from Eastern equine encephalitis virus in a Michigan zoo and a killer whale (Orcinus orca) died from St. Louis encephalitis virus (SLEV) (Flaviviridae: Flavivirus) at a Florida park (Buck et al. 1993, Thompson et al. 2021). These deaths demonstrate the risk of immunologically naïve animal exposure to newly introduced mosquito pathogens, as well as how captivity-related alterations in behavior or prior health issues can make captive animals more susceptible to infection with mosquito-borne diseases (Buck et al. 1993, Thompson et al. 2021). Zoos create a unique environment for disease transmission cycles because they bring together a complicated network of animal interactions including the presence of wild native animals, clustering of exotic species, and movement of captive animals between zoos. Many of the mosquito-borne pathogens affecting animals housed at zoos are zoonotic, hence they are a risk to staff and visitors as well. Despite knowing that medically important arthropods could be abundant in zoos, active surveillance for mosquito vectors and pathogens can be limited by adequate funding and personnel (Adler et al. 2011).

In addition to the importance of zoo locations for monitoring zoonotic vector-borne infections, zoo studies can provide valuable insight into mosquito ecology and feeding patterns. For example, animal enclosures may include a variety of standing water features and water-filled containers that support larval development of numerous species (Beier and Trpis 1981, Derraik 2004a, 2004b, Derraik et al. 2008, Tuten 2011, Heym et al. 2018). Additionally, zoos hold unique aggregations of animals for host-seeking adult mosquitoes. While previous studies on mosquito feeding patterns from zoos generally align with host-class usage trends, novel variations in host usage and reports of mosquito-borne diseases in nontypical captive animals support further investigation (Tuten et al. 2012).

An example of atypical disease transmission at a zoo, was the 2017 death of a bontebok (Damaliscus pygargus) at the Nashville Zoo at Grassmere, hence referred to as “the Nashville Zoo”, from a co-infection of West Nile virus (WNV) (Flaviviridae: Flavivirus) and epizootic hemorrhagic disease virus. The event represents the first WNV infection documented in bontebok (Moncayo et al. 2023). The resulting month-long mosquito surveillance program revealed a higher WNV minimum infection rate (MIR) from zoo collections when compared to collections from the greater Nashville area (Moncayo et al. 2023). Annually, over 1 million guests visit the 76 ha Nashville Zoo, (Nashville Zoo 2020). In 2020, it was home to approximately 3,000 animals representing over 300 different species (Nashville Zoo 2021). The zoo is also home to a variety of native wildlife as it serves as a natural area in the middle of urban neighborhoods and industrial parks. Hence, the zoo is of public health interest due to the potential WNV risk posed to the diversity of captive animals, native wildlife, zoo visitors, and nearby residents.

We sought to further evaluate risk factors that could be contributing to increased arboviral disease transmission within the Nashville Zoo. We tested different methods of mosquito sampling to determine optimal approaches for zoo surveillance of mosquitoes and conducted our surveillance over an extended time frame. We employed these methods to investigate mosquito species diversity (both inside and outside the zoo), host feeding patterns, and viral infection status of key mosquito vectors.

Mosquito species diversity and the WNV infection rates within the zoo were higher than in the surrounding area. Some surveillance methods were more effective than others. In addition, we found that the majority of bloodmeals were from noncaptive animals. Our findings support the utility and importance of zoos as biosurveillance sites for both human and animal health.

Methods

Field Site and Collections

The Nashville Zoo, an Association of Zoos and Aquariums-accredited institution, is located approximately 10 km south of downtown Nashville. Throughout the zoo, deciduous forest has been modified to mimic habitats such as savannas, Indonesian forest, Peruvian forest, and bamboo forest. Mosquitoes were collected from the Nashville Zoo (36°05ʹ21.1″ N, 86°44ʹ37.4″ W) from 10 June to 1 October 2020. Local weather data was collected from the Nashville International Airport weather station approximately 6 km from the zoo. The average temperature during the collection period in 2020 was as follows: June 25.3 °C (range 11.7–35 °C); July 28.3 °C (range 19.4–36.7 °C); August 26.4 °C (range 18.9–36.1 °C); September 22.6 °C (range 8.9–33.3 °C); October 16.9 °C (range 3.3–30 °C). Total rainfall during the study period was as follows: June 84.58 mm, July 112.27 mm, August 149.10 mm, September 96.52 mm, and October 89.15 mm (NOAA 2020).

Ten sites were selected using prior knowledge of mosquito activity noted by zoo colleagues and with the consideration of minimizing direct viewing by the public (Fig. 1). Two additional sites outside of the zoo were selected for collections based on environmental factors and proximity to the zoo (Fig. 1). Residence A was 2 km from the zoo, had a small creek flowing through the yard, and was adjacent to a wooded area. Residence B was a half-kilometer from the closest zoo border and was adjacent to a drainage ditch. Mosquito trapping areas and enclosure exhibits within the zoo of possible vertebrate host species were georeferenced. Google Maps base layer was used in QGIS 3.28.0 (QGIS Association. http://www.qgis.org) software to delineate the polygons of the exhibits (Fig. 1).

Fig. 1.

Maps showing the locations of A) 10 mosquito sampling sites inside the zoo and approximate outlines of public-facing enclosures, and B) the 10 mosquito sampling sites inside the zoo and the 2 sites outside the zoo in residential sites. Colors represent sites of positive virus pools.

We divided mosquito collection sites within the zoo into 2 groups of 5. We alternated the collection group each week so that collections were made at each site every other week. The residential sites were sampled every week from 10 June to 1 October 2020. At each site, we placed a CO₂-baited miniature CDC light trap, BG sentinel trap baited with a BG lure, and a CDC gravid trap baited with a grass infusion (BioQuip Products, Rancho Dominguez, CA, USA). In addition, we placed custom-built black wooden resting boxes (45.72 cm × 31.75 cm × 24.13 cm) (Crates & Pallets, Atlanta, GA, USA) made from modified crates, lined with felt and painted black at each site (Supplementary Fig. F1A). After preliminary sampling, 5 of the most productive trapping sites were selected for placement of a larger black custom-built plywood resting box with an expanded opening (45.72 cm × 30.48 cm × 76.2 cm) (Supplementary Fig. F1B). Mosquitoes were aspirated from resting boxes using a Prokopack Aspirator (John W. Hock Company, Gainesville, FL, USA).

Traps were set in the morning between 07:00 and 10:00 h and left in place for 48 h each week, weather permitting. Approximately 1.15 kg of dry ice was added to the cooler of the CDC light traps between 14:00 and 15:30 h on the day of set up to last approximately 16 h. An additional 4 kg of dry ice was added when replacing mosquito collection nets on the second day to last for the following 24 h. Adding dry ice in the afternoon was not feasible due to Nashville Zoo scheduling. Mosquito samples were collected on the following 2 mornings from 07:00 to 10:00 h. Collections were placed in plastic bags, labeled by trap location and date, and transported to the Tennessee Department of Health laboratory in a cooler on ice. They were stored at −20 °C for sorting and identification.

The Metro Public Health Department conducted mosquito surveillance over the same time period using gravid traps at approximately 40 sites across Davidson County in 2020. Trapping was performed once a week unless a site was deemed unproductive due to low collection numbers. Gravid traps were operated for 1 trap night. None of these sites were in the zoo.

Mosquito Sorting and Identification

Mosquitoes were sorted on a chill table under a stereo microscope and identified according to published keys (Darsie and Ward 2005, Burkett-Cadena 2013, Harrison et al. 2016). Individuals resembling the Culex pipiens complex and Culex restuans Theobald were combined due to the geographic range of Cx. pipiens complex hybrids and a lack of reliable external morphological features (Harrington and Poulson 2008, Kothera et al. 2009). A small number of damaged mosquitoes could only be identified to genus. Males were discarded, while females were examined for physiological status and considered blood engorged if a bloodmeal was visible in the abdomen. Engorged mosquitoes were placed individually in sterile labeled microcentrifuge tubes and stored at −80 °C before shipment on dry ice to Cornell University for bloodmeal identification. The remaining female mosquitoes were placed in pools of up to 50 individuals by trap date, location, trap type, and species in microcentrifuge tubes. Pools were stored at −80 °C for virus testing. Mosquito data were recorded by species, engorgement status, collection date, trap type, and location in a master spreadsheet. The Metro Public Health Department identified females to the Culex genus and placed them in pools of up to 50 individuals by location and date before being sent to the Tennessee Department of Health for testing.

Virus Testing

Culex mosquitoes from all trap types (including the gravid traps used by the Metro Public Health Department) were tested for WNV, Flanders virus (FLAV) (Rhabdoviridae: Hapavirus), and SLEV according to the Tennessee Department of Health protocols (Westby et al. 2015), with modifications. Briefly, 3 copper BBs (Crosman, Fairport, NY, USA) were added to each microcentrifuge tube containing pooled mosquitoes, followed by the addition of 1 ml of Eagle’s Minimum Essential Medium (Mediatech, Inc., Manassas, VA, USA) with 2% Fetal Bovine Serum (Life Technologies Corporation, Grand Island, NY, USA), 0.5% Sodium Bicarbonate (Mediatech, Inc., Manassas, VA, USA), and 1% Antibiotic-Antimycotic Solution (Mediatech, Inc., Manassas, VA, USA). Samples were homogenized using a MM300 Mixer Mill (Retsch, Haan, Germany) at 30/s for 90 s. The microcentrifuge tubes were removed from the Mixer Mill and left to rest for at least a minute. The microcentrifuge tubes were centrifuged at room temperature for 7 min at 12,000 rpm. Afterward, 140 ml of mosquito supernatant was transferred from each tube into the corresponding well of a 96-well S-block. The S-block was then placed in a BioRobot Universal System 9604 (Qiagen Sciences, Germantown, MD, USA) for RNA extraction using the QIAmp Viral Isolation 96-well protocol. Tubes containing the remaining supernatant were stored at −80 °C.

A 20 µM reaction mixture was used containing 6.25 µl of 4X TaqMan Fast Virus 1 Step Master Mix (Thermo Fisher Scientific Baltics UAB, Vilnius, Lithuania), 0.4 µl of WNV_#3 1160 forward primer (25 µM), 0.4 µl of WNV_#3 1229c reverse primer (25 µM), 0.1 µl of probe WNV_#3 p1186-FAM/BHQ1 (25 µM), 0.4 µl of FLD f_16 forward primer (25 µM), 0.4 µl of FLD r_94 reverse primer (25 µM), 0.1 µl of probe FLD p41 JOE/BHQ1 (25 µM), 0.4 µl of SLE_834 forward primer (25 µM), 0.4 µl of SLE_905c reverse primer (25 µM), 0.1 µl of probe SLE p857 CY-5/ZEN/ (25 µM), and 6.25 µl of PCR-grade water (Supplementary Table S1). For each multiplex RT-PCR, PCR grade water was used as a negative control; WNV, FLAV, and SLEV RNA were used as positive controls. Positive controls were obtained from the Centers for Disease Control and Prevention’s Arbovirus Reference Collection. Pools were considered positive if the control threshold score was less than or equal to 37. The following cycling conditions were used: 50 °C for 5 min, 95 °C for 20 s, followed by 40 cycles of 95 °C for 3 s, and 60 °C for 30 s.

Bloodmeal Analysis

Mosquitoes were maintained on ice while abdomens from engorged specimens were removed and transferred to a sterile microcentrifuge tube. The head and thorax were returned to the original vial for later molecular identification of Culex species. The degree of blood engorgement was scored using the Sella score (Sella 1920). Forceps were sterilized between samples by dipping in 80% ethanol and passing through a flame. DNA extractions were performed using the Qiagen Puregene kit according to the manufacturer’s protocol (Qiagen Sciences, Germantown, MD, USA). To identify bloodmeals, medium size primer sets were used to amplify a 395 base pair vertebrate-specific region of cytochrome c oxidase subunit I (Supplementary Table S2) (Reeves et al. 2018). DNA templates that failed to amplify were used in a second reaction with the small sized cytochrome c oxidase 1 (COI) primers designed by Reeves et al. (2018) (Supplementary Table S2). If amplification failed a second time, then the DNA template was used in a third nested reaction that consisted of the usage of the COI long primers followed by the short COI primers designed by Townzen et at. (2008) (Supplementary Table S2).

For the medium size Reeves primers, reactions were performed in a total volume of 20 µl that consisted of 10 µl of 2.0X Apex Taq RED Master Mix (Genesee Scientific Corp., San Diego, CA, USA), 0.75 µl of VertCOI_7194_F forward primer (10 µM), 0.75 µl of Mod_RepCOI_R reverse primer (10 µM), 7.5 µl sterile nuclease-free water, and 1 µl of extracted DNA. The following thermocycler conditions were used: 94 °C for 3 min, followed by 40 cycles of 94 °C for 40 s, 53.5 °C for 30 s, and 72 °C for 60 s, and a final extension step at 72 °C for 7 min. Except for the substitution of primers, the same conditions were followed for the small size Reeves primers. Townzen primer reactions were performed in a total volume of 20 µl that consisted of 10 µl of 2.0X Apex Taq RED Master Mix (Genesee Scientific Corp., San Diego, CA, USA), 0.75 µl of forward primer (10 µM), 0.75 µl of reverse primer (10 µM), 7.0 µl sterile nuclease-free water, and 1.5 µl of extracted DNA. The cycling conditions for the Townzen primers were as follows: 95 °C for 5 min, followed by 40 cycles of 94 °C for 30 s, 50 °C for 50 s, and 72 °C for 60 s, and a final extension step at 72 °C for 5 min. All reactions included a positive animal fed control and a negative water control. PCR products were loaded onto a 1% agarose gel stained with gelRED, electrophoresed for 45 min, and visualized with BioRad Gel Doc XRS. PCR products with positive bands after gel electrophoresis were cleaned with ExoSap-IT then submitted to Cornell Biotechnology Resources Center for sequencing. Sequences were compared to BOLD and NCBI databases and identified to a host if the matches were greater than 98% (except for a Carolina chickadee (Parus carolinensis) sequence, which had a 97.9% match).

The heads and thorax of unidentified engorged Culex spp. individuals were extracted using Qiagen Puregene kits (Qiagen Sciences, Germantown, MD, USA). The identity of Culexpipiens complex individuals was confirmed through species-specific primers designed by Aspen and Savage (2003). Reactions were performed in a total volume of 25 µl that consisted of 10 µl of 2.0X Apex Taq RED Master Mix (Genesee Scientific Corp., San Diego, CA, USA), 0.80 µl of forward primer (10 µM), 0.20 µl of reverse primer (10 µM), 13.0 µl sterile nuclease-free water, and 1.0 µl of extracted DNA. The following cycling conditions were used: 93 °C for 5 min, followed by 27 cycles of 91 °C for 1 min, 72 °C for 90 s, and a final extension step at 72 °C for 10 min. PCR products were loaded onto a 1% agarose gel stained with gelRED, electrophoresed for 45 min, and visualized with BioRad Gel Doc XRS. Cx. p. pipiens were identified by the presence of a 487 bp amplicon when using PACEF290 as the forward primer and the absence of a band when using QACEF290. Culex p. quinquefasciatus were identified by the presence of a 487 bp amplicon when using QACEF290 as the forward primer and the absence of a band when using PACEF290. Culex p. pipiens–Cx. p. quinquefasciatus hybrids were identified by having a 487 bp amplicon from both reactions (Aspen and Savage 2003). Culex spp. too damaged for morphological identification were identified using the COI primers designed by Hebert et al. (2004). Reactions and cycling conditions were the same as described by Reeves et al. (2021).

Host Availability Data

Zoo census of captive species and visitors was determined weekly for the duration of the mosquito collection period (Supplementary Tables S3 and S4). There were at least 500 captive animals at the zoo with outdoor access and potential exposure to blood feeding mosquitoes for various amounts of time. We checked the BOLD (http://www.boldsystems.org/) and NCBI (https://blast.ncbi.nlm.nih.gov/Blast.cgi) databases to ensure that DNA was available for these captive species. With capacity limitations due to COVID-19 restrictions, an average of 15,225 guests visited the zoo per week during the collection period. These restrictions combined with general caution related to COVID-19 decreased typical attendance over the summer. Additionally, the first week we conducted trapping the zoo was closed to visitors. The total number of visitors during the study period was 258,827 (Supplementary Table S4).

Data Analysis and Reporting

Data cleaning, analysis, and visualization were performed in Program R (R Core Team 2023) using RStudio (Version 2023.3.1.) (Posit team 2023) with the following packages: dplyr (Wickam et al. 2021), ggplot2 (Wickham 2016), tidyr (Wickham 2020), vegan (Oksanen et al. 2020), lubridate (Grolemund and Wickham 2011), devtools (Wickham et al. 2020), pairwiseAdonis (Martinez Arbizu 2020), and ggthemes (Arnold 2021). Data collected from the Rhino, Farmhouse, and Spider Monkey sites were not included as they were discontinued due to low mosquito collections. Resting boxes of both sizes were combined into the same category.

Mosquito diversity was determined using the Shannon–Wiener index, which considers species richness and evenness of each site by including mosquitoes collected by all trap types (Barnes et al. 1998). The larger the value calculated, the more uncertainty and hence diversity is present for a sample. Comparison of the Shannon–Wiener indices was performed using a Wilcoxon rank-sum test. Multivariate analysis of variance was used to determine factors contributing to differences in species compositions at each site.

MIRs were determined for WNV, FLAV, and SLEV activity using calculator tools in TennSurv (https://vectorsurv.org/). MIRs were calculated as the number of virus-specific positive pools per number of mosquitoes tested (CDC 2013) and expressed as the number of positive individuals per 1,000 tested. Weekly MIRs were also calculated and analyzed using permutational multivariate analysis of variance using distance matrices.

Results

Adult Mosquito Collections

A total of 9,308 adult mosquitoes comprised of 24 species were collected at the 12 trapping sites (Table 1). Overall, 22% of the mosquitoes (2,066 individuals) were collected at the residential sites, while the remaining 78% (7,242 individuals) were collected from the zoo. Mosquito species richness was higher within the zoo; 16 species were collected at the residential sites, whereas 27 species were collected from the zoo sites. The most abundant species collected, defined as over 100 specimens, were Culex pipiens/quinquefasciatus/restuans, Aedes albopictus Skuse, Aedes vexans Meigen, Culex erraticus Dyar and Knab, and Anopheles punctipennis Say. A total of 100 blood-fed mosquitoes representing 11 species were collected, with the majority (n = 96, 96%) from within the zoo as compared to the residential sites.

Table 1.

Mosquito species collected at the Nashville Zoo and residential sites Numbers represent the total number of each species collected by site grouping from June 10 to October 1

Species	Within zoo total (%)	Residential total (%)
Culex spp.	2,494 (34. 44)	781 (37.80)
Culex pipiens/quinquefasciatus/restuans	2,305 (31.83)	691 (33.44)
Aedes albopictus	1,170 (16.16)	492 (23.81)
Aedes vexans	467 (6.45)	19 (0.92)
Culex erraticus	151 (2.09)	22 (1.06)
Anopheles punctipennis	140 (1.93)	13 (0.63)
Aedes triseriatus	73 (1.01)	10 (0.48)
Aedes spp.	69 (0.95)	15 (0.73)
Anopheles quadrimaculatus s.l.	68 (0.94)	2 (0.10)
Culex territans	59 (0.81)	4 (0.19)
Culex nigripalpus	51 (0.70)	2 (0.10)
Psorophora columbiae	48 (0.66)	0 (0)
Culex pipiens pipiens	38 (0.52)	1 (0.05)
Psorophora ferox	32 (0.44)	4 (0.19)
Uranotaenia sapphirina	16 (0.22)	0 (0)
Aedes japonicus	15 (0.21)	0 (0)
Culex pipiens pipiens–Culex pipiens quinquefasciatus hybrids	10 (0.14)	0 (0)
Aedes trivittatus	6 (0.08)	0 (0)
Aedes atlanticus	5 (0.07)	0 (0)
Orthopodomyia spp.	4 (0.06)	2 (0.10)
Anopheles perplexans	3 (0.04)	0 (0)
Psorophora spp.	3 (0.04)	0 (0)
Psorophora cyanscens	3 (0.04)	0 (0)
Aedes infirmatus	2 (0.03)	0 (0)
Culex pipiens quinquefasciatus	2 (0.03)	1 (0.05)
Psorophora mathesoni	2 (0.03)	1 (0.05)
Aedes fulvus pallens	1 (0.01)	0 (0)
Anopheles crucians s.l.	1 (0.01)	0 (0)
Culex salinarius	1 (0.01)	0 (0)
Orthopodomyia signifera	1 (0.01)	3 (0.15)
Psorophora howardii	1 (0.01)	1 (0.05)
Toxorhynchites rutilus	1 (0.01)	2 (0.10)

The most mosquitoes (n = 1,219) were collected at the macaw site, and the least (n = 361) were collected at the pond site (Fig. 1B). The CDC miniature light traps collected the most species (n = 23), followed by the gravid trap (n = 15), the resting boxes (n = 11), and the BG Sentinel (n = 10). The gravid traps consistently yielded more mosquitoes per trap night than other traps. This is likely due to the gravid traps’ attractiveness to Culex sp. such as Cx. pipiens/quinquefasciatus/restuans, which represented the bulk of the collections in gravid traps (Fig. 2). Culex erraticus was most frequently collected in CDC miniature light traps (n = 79 individuals), followed closely by resting boxes (n = 74 individuals) (Fig. 2). Aedes vexans and An. punctipennis were most frequently collected in CDC miniature light traps (n = 479, n = 95, respectively) (Fig. 2). Aedes albopictus was most frequently collected in CDC miniature light traps (n = 782), but they were commonly collected in the BG Sentinel traps (n = 503) and gravid traps (n = 365) as well (Fig. 2). It is worth noting that due to space constraints at several sites, the BG and CDC miniature light traps were placed within close proximity (minimum of 9 m) of each other, which could lead to confounding collections.

Fig. 2.

The average number (and standard error) of the 6 most abundant species of mosquitoes trapped per night by site and trap type: A) Culex spp., B) Cx. pipiens/quinquefasciatus/restuans, C) Ae. albopictus, D) Ae. vexans, E) An. punctipennis, and F) Cx. erraticus.

Mosquito Diversity

Shannon–Wiener indices of mosquito species diversity varied by site ranging from 1.24 to 1.91 (Table 2). The highest diversity of mosquitoes was collected from the dino trek site, with a Shannon–Wiener index of 1.91 (Fig. 1, Table 2). Residence B had the lowest mosquito diversity with a Shannon–Wiener index of 1.24 (Fig. 1, Table 2). Diversity was significantly higher from sites inside the zoo as compared with those outside the zoo (Wilcoxon rank-sum test; P = 0.0303), potentially due to greater larval habitat presence and diversity within the zoo. Collection week contributed to 43% of the variance between sites (df = 16, P = 0.001); while the site location itself contributed to 13% of the variance (df = 11, P = 0.001).

Table 2.

Shannon–Wiener diversity index values for each collection site

Collection site	Shannon–Wiener index
Dino Trek	1.91
Red Lemur	1.86
Porcupine	1.81
Pond	1.74
Lorikeet	1.66
Gibbons	1.61
Train	1.50
Macaw	1.46
Flamingo	1.43
Tapir	1.43
Residence A	1.32
Residence B	1.24

Arbovirus Testing Results

A total of 5,032 Culex mosquitoes were tested in 565 pools from the 10 sites within the zoo. None of the pools were positive for SLEV, while 4 pools were positive for WNV, and 21 pools were positive for FLAV. FLAV was detected at 8 sites within the zoo (porcupine, red lemur, flamingo, pond, train, gibbons, lorikeet, and dino trek), and WNV was detected at 4 sites within the zoo (red lemur, tapir, flamingo, and macaw) (Fig. 1). At 2 of the sites (red lemur and flamingo) WNV and FLAV were both detected (Fig. 1). A total of 1,495 mosquitoes were tested in 137 pools from 2 sites outside the zoo. None of the pools were positive for SLEV or WNV, while 11 pools were positive for FLAV. FLAV was detected at both sites outside of the zoo (Residence A and B) (Fig. 1). The zoo-based WNV, SLEV, and FLAV MIRs were 0.79 (range, 0.02–1.57), 0, and 4.17 (range, 2.39–5.95), respectively. The sites outside the zoo had WNV, SLEV, and FLAV MIRs of 0, 0, and 7.36 (range, 3.03–11.69), respectively. The rest of Davidson County had WNV, SLEV, and FLAV MIRs of 0.39 (range, 0.10–0.68), 0, and 3.99 (range, 3.07–4.91), respectively. The average weekly WNV and FLAV MIR for the Nashville Zoo were 0.52 (SD ± 1.85) and 2.32 (SD ± 4.21), respectively. When analyzed on a weekly basis the average FLAV MIR for the sites outside of the zoo was 5.31 (SD ± 8.34). The average weekly WNV and FLAV MIRs for Davidson County were 0.20 (SD ± 0.40) and 4.36 (SD ± 4.07), respectively. There was no significant difference between WNV (F = 0.99, df = 2, P = 0.45) or FLAV (F = 1.15, df = 2, P = 0.33) MIRs values for the Nashville Zoo, the sites outside of the zoo, and Davidson County. The peak in weekly FLAV MIR (13.40) preceded the peak in weekly WNV MIR (7.61) by 12 weeks (Supplementary Fig. F2).

Mosquito Species Determination with PCR

A total of 61 engorged mosquitoes were identified morphologically to the Culex genus since members of the Culex pipiens complex and Cx. restuans are morphologically indistinguishable or the other mosquito species were too damaged to identify. We employed PCR using species-specific primers for the Culex pipiens complex (Aspen and Savage 2003). Our results revealed 36 Culex pipiens pipiens Linnaeus, 10 Culex pipiens pipiens–Culex pipiens quinquefasciatus hybrids, and 3 Culex pipiens quinquefasciatus Say. A remaining 12 Culex spp. were tested using COI primers to determine species (Hebert et al. 2004, Reeves et al. 2018). With these primers, we identified 3 Cx. p. pipiens mosquitoes (confirmed with the Culex spp. species-specific primers), 1 Culex salinarius Coquillett, and 1 Cx. erraticus. The identity of 4 individuals in the Cx.pipiens spp. complex and 4 other Culex spp. could not be determined using PCR.

Bloodmeal Analysis

Of 100 engorged mosquitoes captured, 84 were matched to a host species. Of the 23 host species identified, only 8 were from captive zoo species (cow [Bos taurus], common eland [Tragelaphus oryx], common ostrich [Struthio camelus], trumpeter swan [Cygnus buccinator], clouded leopard [Neofelis nebulosa], white-tailed deer [Odocoileus virginianus], red kangaroo [Osphranter rufus], and yellow-backed duiker [Cephalophus silvicultor]) (Table 3). While it is possible that the bloodmeals from cattle and a white-tailed deer were from non-zoo animals, this is highly unlikely because the mosquitoes were collected at sites near these animals’ enclosures. Additionally, white-tailed deer were physically excluded from the zoo with deer fencing and the properties near the zoo were residential and too small to adequately support cattle. The origin of the bloodmeal from a chicken (Gallus gallus) is unknown, but it could be from chickens kept at one of the residences bordering the zoo. However, we did not survey animals on residential properties surrounding the zoo. Surprisingly, the most bloodmeals identified within the zoo were from northern cardinals (Cardinalis cardinalis) L. (Passeriformes: Cardinalidae), followed by humans and other wild bird species, rather than meals from the approximately 500 zoo animals on exhibit (Table 3). Four host identifications were from residence B (2 human [Homo sapiens], 1 American robin [Turdus migratorius] L. [Passeriformes: Turdidae], and a northern cardinal). Due to a low number of bloodmeals matched to zoo animals and the lack of data regarding wild host availability, we did not calculate host forage ratios.

Table 3.

Number of bloodmeals categorized by host species and mosquito species. A host identification was unsuccessful for the single engorged Ae. trivittatus collected and the single Cx. territans, hence these species have been removed from the table

Mosquito species	Host species	No. of bloodmeals from host (% of bloodmeals from a host species for a mosquito species)
Aedes albopictus	Human (Homo sapiens)	1 (100%)
Aedes vexans	Eastern Cottontail (Sylvilagus floridanus)	1 (100%)
Anopheles punctipennis	Common Eland (Tragelaphus oryx) a	1 (20%)
	Cow (Bos taurus)a	1 (20%)
	Common Ostrich (Struthio camelus)a	1 (20%)
	Red Kangaroo (Osphranter rufus)a	1 (20%)
	White-tailed Deer (Odocoileus virginianus)a	1 (20%)
Anopheles quadrimaculatus s.l.	Clouded Leopard (Neofelis nebulosa)a	2 (28.6%)
	Common Eland (Tragelaphus oryx) a	2 (28.6%)
	Cow (Bos taurus)a	2 (28.6%)
	Common Ostrich (Struthio camelus)a	1 (14.3%)
Culex erraticus	Human (Homo sapiens)	6 (35.3%)
	Northern Cardinal (Cardinalis cardinalis)	3 (17.6%)
	American Robin (Turdus migratorius)	1 (5.9%)
	Chicken (Gallus gallus)	1 (5.9%)
	Common Box Turtle (Terrapene carolina)	1 (5.9%)
	Eastern Gray Squirrel (Sciurus carolienensis)	1 (5.9%)
	Mourning Dove (Zenaida macroura)	1 (5.9%)
	Trumpeter Swan (Cygnus buccinator)a	1 (5.9%)
	White-tailed Deer (Odocoileus virginianus)a	1 (5.9%)
	Yellow-backed Duiker (Cephalophus silvicultor)a	1 (5.9%)
Culex pipiens pipiens	Northern Cardinal (Cardinalis cardinalis)	17 (47.2%)
	Human (Homo sapiens)	6 (16.7%)
	Carolina Wren (Thryothorus ludovicianus)	3 (8.3%)
	European Starling (Sturnus vulgaris)	3 (8.3%)
	Common Grackle (Quiscalus quiscula)	2 (5.6%)
	Mourning Dove (Zenaida macroura)	2 (5.6%)
	American Kestrel (Falco sparverius)	1 (2.8%)
	Carolina Chickadee (Parus carolinensis)	1 (2.8%)
	White-eyed Vireo (Vireo girseus)	1 (2.8%)
Culex pipiens pipiens–Culex pipiens quinquefasciatus hybrids	Northern Cardinal (Cardinalis cardinalis)	5 (55.6%)
	Carolina Wren (Thryothorus ludovicianus)	2 (22.2%)
	Common Grackle (Quiscalus quiscula)	1 (1.11%)
	Human (Homo sapiens)	1 (11.1%)
Culex pipiens quinquefasciatus	Northern Cardinal (Cardinalis cardinalis)	2 (66.7%)
	American Robin (Turdus migratorius)	1 (33.3%)
Culex salinarius	Cow (Bos taurus)a	1 (100%)
Culex spp.	Green Frog (Rana clamitans)	1 (25%)
	Northern Cardinal (Cardinalis cardinalis)	3 (75%)

^aDenotes captive species.

Mosquitoes with 5 or more successful bloodmeal identifications were An. punctipennis, Anopheles quadrimaculatus s.l., Cx. erraticus, Cx. p. pipiens, and Cx. p. pipiens–Cx. p. quinquefasciatus hybrids. The majority of bloodmeals from An. punctipenis and An. quadrimaculatus s.l. were from mammalian hosts, while the majority of bloodmeals from Cx. p. pipiens and Cx. p. pipiens–Cx. p. quinquefasciatus hybrids were from avian hosts (Table 3). The most bloodmeals were collected from the red lemur site (the only location where WNV was detected) and the predominant bloodmeal host from the site was avian (Supplementary Fig. F3).

Discussion

This study represents one of the most detailed assessments of mosquito diversity, host use patterns, and arbovirus infection at a zoo setting in the United States. Our results support the hypothesis that zoos can be sources of high mosquito diversity and may support transmission risk factors as well as threats to valuable and rare captive zoo species. However, we observed lower than expected feeding on exotic animals as compared to native birds.

Adult collections revealed high mosquito diversity at the zoo, despite the use of routine mosquito control interventions such as Bti larvicide applications and deployment of In2Care traps (In2Care BV, Wageningen, Netherlands) to control adult container breeding mosquitoes. The CDC miniature light traps collected the greatest number of species (n = 22), supporting their use for diverse mosquito surveillance. The breadth of mosquito species collected in the zoo reflects the presence of a variety of larval habitats and environmental conditions ranging from man-made containers to natural pools within or in close proximity to the zoo. Although only 15% of the variance in diversity was attributed to the site of collection, our analysis did not factor into account quantifiable habitat similarities or dissimilarities for each site. While there was a significantly higher level of species diversity collected inside the zoo compared to the residential sites, additional surveillance for larval habitats, analysis of associated environmental factors, and additional external sites would provide insight into factors driving diversity. It is also possible that including additional residential sites could lead to the collection of previously undetected species, hence reducing the differences in diversity between the zoo and the residential sites.

Many of the mosquito species collected are of medical and veterinary concern. Several of the Culex species we captured are known maintenance vectors of WNV in avian populations, while species like Ae. vexans could serve as bridge vectors, transmitting WNV from birds to mammals (Molaei and Andreadis 2006). One species of concern included Culex nigripalpus Theobald as it is a vector of SLEV and WNV (Vitek et al. 2008). Additionally, Cx. nigripalpus has been associated with Eastern equine encephalitis outbreaks but vector competence is yet to be determined (Scott and Weaver 1989, Day and Stark 1996). Additionally, the collection of Cx. nigripalpus at the Nashville Zoo is significant because it was the first time Cx. nigripalpus has been documented in Davidson County. This species has historically been detected in the far western part of Tennessee (Darsie and Ward 2005). More recently, Cx. nigripalpus has been identified in additional counties in western and eastern Tennessee (Cohen et al. 2009, Haddow et al. 2009, Fryxell et al. 2014). Our results further support the expansion of Cx. nigripalpus into new regions of the state. This finding also contributes to the growing body of knowledge supporting the northern expansion of Cx. nigripalpus across the United States (Akaratovic et al. 2021).

The detection of a FLAV MIR peak 12 weeks before a WNV MIR supports the use of FLAV as an indicator of future WNV activity (Supplementary Fig. F2). This gap in the MIR peaks is similar to a previous study from Tennessee which detected an average 10-week gap between peaks (Lucero et al. 2016). However, the FLAV peak was detected during the first series of collections, so it is possible that a higher weekly MIR occurred earlier in the season. Poh et al. (2018) detected a one- and three-week gap between the detection of FLAV and peaks in WNV infection rates in mosquitoes collected in Chicago, IL, in 2010 and 2012, respectively. This suggests that there are regional differences in timing of FLAV detections and these relationships warrant further study.

The detection of WNV in pools of Culex spp. and Cx. pipiens/quinquefasciatus/restuans is consistent with patterns seen in the southeastern United States (Godsey et al. 2005). All but 1 pool of WNV and FLAV infected mosquitoes was collected by gravid traps. The gravid traps also resulted in more pools of Culex mosquitoes when compared to the other trap types. Gravid traps are ideal tools for WNV surveillance because they capture females after they have blood fed and potentially ingested an infectious meal (CDC 2013). The WNV MIR for the Nashville Zoo in 2017 from September 19 to October 13 was 9.72 (Moncayo et al. 2023). When looking at a similar interval in our collection period (16 September to 1 October 2020) the WNV MIR for the Nashville Zoo is lower (0.64). Nationally, enzootic WNV amplification may have been greater in 2017 and, although not directly related, human case incidence was higher across the United States in 2017 compared to 2020 (CDC 2020).

The continued presence (in 2017 and 2020) of WNV within the Nashville Zoo is concerning and could suggest factors unique to the Nashville Zoo are contributing to recurring transmission cycles. Nashville is considered a metropolitan area with relatively low access to parks (Trust for Public Land. 2023a). In Los Angeles, California, a city with a similar park accessibility score, urban parks serve as refugia for birds (Vasquez and Wood 2022, Trust for Public Land. 2023b). Likewise, the Nashville Zoo could serve as an oasis for wild birds in the middle of an urbanized area. As a result, the Nashville Zoo could be prone to higher WNV MIR when compared to Davidson County due to the increased concentration of a variety of wild bird species. Wild bird host census would be required to substantiate this hypothesis, with special attention to amplification species, which can also be present in highly residential landscapes. If the zoo does serve as an oasis for amplification species, these birds could be introduction points for WNV into the area following migration as well as providing a continual presence of amplification hosts. Most mosquito bloodmeals in this study were from wild birds, which supports the notion that wild birds at the zoo could play an important role in mosquito infection and transmission dynamics. Notably, Levine et al. (2016), found the Atlanta Zoo had lower WNV MIR when compared to nearby forested areas. However, approximately 40% of the bloodmeals of mosquitoes collected at the Atlanta Zoo were from wild birds compared to 58% of the bloodmeals in our study. This difference alone does not explain the differences we found in mosquito infection rates because not all wild birds are equally competent reservoirs of WNV. Additionally, Atlanta has a significantly higher park accessibility score, which could indicate more urban green space (Trust for Public Land 2023c), which could result in a decreased concentration of birds in the Atlanta Zoo when compared to the Nashville Zoo.

The number of captive species fed on by mosquitoes (n = 8) and the percentage of bloodmeals from captive animals (19%) was lower than other studies conducted in the United States at the Greenville Zoo (South Carolina), Riverbanks Zoo (South Carolina), and Rio Grande Zoo (New Mexico) (Greenburg et al. 2012, Tuten et al. 2012). These differences could be due to differences in the mosquito species collected. The most bloodmeals were contributed by Cx. erraticus in Tuten et al. (2012), Aedes vexans in Greenberg et al. (2012), and Culex p. pipiens in this study. Culex erraticus (in summer months) and Ae. vexans primarily feed on mammals, while Cx. p. pipiens primarily feeds on avian hosts (Molaei and Andreadis 2006, Burkett-Cadena et al. 2011, Oliveira et al. 2011, Turell 2012). Many of the captive animals housed outdoors in zoos tend to be mammals, hence creating an abundance of mammalian hosts for Cx. erraticus and Ae. vexans. Differences in trap proximity to enclosures could have varied as well since these metrics were not reported in previous studies. These studies also reported bloodmeals from overlapping host species identified in our study, including cottontail rabbits (Sylvilagus audubonii), frogs (Hyla cinerea), cattle, ostriches, northern cardinals, common starlings (Sturnus vulgaris), mourning doves (Zenaida macroura), American robins, and Carolina chickadees.

Previous studies have indicated that American robins could be important WNV amplifying hosts in urban areas (Apperson et al. 2004, Kilpatrick et al. 2006a, Molaei et al. 2006, Savage et al. 2007, Khalil et al. 2021). We documented few American robin bloodmeals at the Nashville Zoo. Instead, northern cardinals were the most frequently detected animal host in mosquitoes collected at the Nashville Zoo. One possible explanation for the lack of American robin bloodmeals could be host switching during our collection period. Kilpatrick et al. (2006b) documented a host switching trend that was defined as a decrease in the occurrence of feeding on American robins and a simultaneous increase in feeding on mammals beginning in July. Levine et al. (2016) also noted that Culex mosquitoes demonstrated host switching patterns away from American robins beginning in early summer (July), but rather than switching to mammals later in the season, Culex switched to feed on northern cardinals. Models suggest the relative abundance of American robins decreases throughout Tennessee from the months of May through October (Fink et al. 2022). Therefore, it is possible there was a higher prevalence of northern cardinal rather than American robin meals at the Nashville Zoo because we did not begin trapping mosquitoes until midsummer and most engorged mosquitoes were collected after June.

Intriguingly, northern cardinals were also a common avian host at both zoos in South Carolina (Tuten et al. 2012). Northern cardinals have been a common host in blood feeding studies on Culex mosquitoes conducted in Georgia, New Jersey, New York, Tennessee, and Virginia as well (Apperson et al. 2004, Patrican et al. 2007, Savage et al. 2007, Levine et al. 2016, Khalil et al. 2021). In Georgia, WNV seroprevalence is consistently higher in northern cardinals than other avian species (Levine et al. 2016). This led to the suggestion of using northern cardinals as a WNV surveillance species (Gibbs et al. 2006). However, the role that northern cardinals play in WNV transmission remains unclear. Northern cardinals could have a “supersuppressor” effect by serving as a highly utilized bloodmeal source while reducing WNV transmission as a moderately competent reservoir rather than as a highly competent reservoir, such as American robins (Levine et al. 2016). It is difficult to determine what role, if any, northern cardinals are playing in the WNV transmission dynamics due to a lack of knowledge of the avian populations in the immediate area. During our study, a wild bird survey was not taken, so it is unclear how common northern cardinals were relative to other bird species during the collection period, precluding the calculation of forage ratios to determine if the proportion of bloodmeals was proportional to the population. The lack of a concurrent or prior bird census of Davidson County or the Nashville Zoo also made it difficult to determine if the population of American robins in the area is migratory, which is considered a factor supporting host switching (Kilpatrick et al. 2006b). Despite these limitations, our results contribute to the growing body of knowledge on the utilization of northern cardinals by WNV vectors and suggest the Nashville Zoo could serve as a future site to investigate avian communities, host seroprevalence, and vector infection rates.

There is still much to be learned about the nuance of blood feeding patterns and the role of these patterns in the transmission of WNV. Further research should be performed to investigate wild bird populations at the Nashville Zoo and provide more insight into the repeated detection of WNV within the zoo. Despite a low number of bloodmeals taken from captive species, the continuous circulation of WNV puts these animals at risk for infection. The variety of mosquito species collected within the zoo could indicate that captive animals are at risk of infection beyond WNV if proper preventative steps are not taken, such as limiting potential larval development sites or incorporating other mosquito control methods. Humans in and near the zoo could also be at risk for WNV or other arbovirus infections, and they should be cautioned to wear repellants, especially given the proportion of human blood meals collected (17% of the collection). Currently, this type of messaging does not exist at the Nashville Zoo. Arboviral threats in the United States continue to change due to invasive and expanding species, climate change, and habitat modification. The detection of WNV and Cx. nigripalpus within the zoo further demonstrates the value of zoological institutions as biosurveillance sites for existing and potential arboviral threats. Biosurveillance at zoos is a valuable means of monitoring components of zoonotic disease transmission and contributing to our growing knowledge of One Health and emerging pathogens.

Author Contributions

Cierra Briggs (Conceptualization [Equal], Data curation [Lead], Formal analysis [Lead], Investigation [Lead], Methodology [Equal], Project administration [Lead], Visualization [Lead], Writing – original draft [Lead], Writing – review & editing [Lead]), Rayan Osman (Investigation [Supporting], Writing – review & editing [Supporting]), Brent Newman (Investigation [Supporting], Writing – review & editing [Supporting]), Kara Fikrig (Investigation [Supporting], Methodology [Supporting], Writing – review & editing [Supporting]), Philip Danziger (Investigation [Supporting], Writing – review & editing [Supporting]), Emily Mader (Visualization [Supporting], Writing – review & editing [Supporting]), Margarita Woc-Colburn (Conceptualization [Equal], Methodology [Supporting], Writing – review & editing [Supporting]), Laura Harrington (Conceptualization [Equal], Funding acquisition [Lead], Methodology [Equal], Project administration [Supporting], Resources [Lead], Supervision [Equal], Writing – review & editing [Supporting]), and Abelardo Moncayo (Conceptualization [Equal], Methodology [Equal], Resources [Supporting], Supervision [Equal], Writing – review & editing [Supporting])