Introduction

With human-induced environmental change accelerating, current needs for biodiversity conservation far outweigh the available resources. Biodiversity conservation is reliant on up-to-date information on the abundance, diversity, and distribution of species to guide the implementation and deployment of conservation actions and resources (e.g., the purchasing of land and establishment of protected areas, drafting of legislation, development of species management plants, etc.)1. This critical information is generated by field surveys or biological monitoring, which requires considerable investment of resources. Vertebrates are often considered flagship, umbrella, keystone, or indicator species2,3, and, as such, these species or communities are the targets of field surveys and monitoring. Traditional methods of vertebrate survey and detection are effective, but tend to be time-consuming, expensive, and require extensive field and taxonomic expertise. Given these limitations, it is increasingly important to develop efficient and innovative ways to improve biodiversity survey and detection methods that leverage modern technologies in this critical era of biodiversity loss4.

Advances in molecular ecology and DNA sequencing technology have facilitated the implementation of environmentally derived DNA (eDNA) as a means of detecting vertebrates within complex environments by extracting DNA from water, air, soil, or other environmental samples such as spiderwebs or bulk arthropod samples5,6,7,8,9. Invertebrate-derived DNA (iDNA), i.e., DNA collected from invertebrates that feed on vertebrate blood, tissue, or excrement, builds on the eDNA approach and shows potential for use in biomonitoring applications10 and may enhance efficiency and cost-effectiveness in comparison to conventional surveys (e.g., direct observation, camera traps, drift fence trapping, etc.11,12). Invertebrate-derived DNA may enable researchers to glean insights that would be impossible through other sources of eDNA (e.g., water, air), though iDNA may be logistically more challenging to collect. Often, iDNA from an arthropod represents a noninvasive blood sample derived from an individual vertebrate animal. This nuance makes iDNA useful for tracking disease in vertebrate populations (i.e., xenosurveillance13) or, as a genetic sample that could be useful to conservation genetics, for example, in recognizing individuals or estimating abundance of vertebrate populations14. While other invertebrates (e.g., leeches, carrion flies) have been comparatively well-studied as sources of vertebrate DNA in iDNA-based biomonitoring, relatively few studies have assessed the potential of mosquitoes to detect vertebrate animals10,11,15,16. Their widespread geographic distribution, high abundance (in many terrestrial ecosystems), use of diverse vertebrate hosts, rapid blood meal digestion, and limited post-feeding flight ability make mosquitoes viable candidates for use in iDNA surveys of terrestrial vertebrates17,18,19. Further, mosquito host associations and molecular techniques for mosquito host identification have been studied for more than a century, albeit from a largely epidemiological context20. Despite these factors, mosquitoes have received comparatively little attention as candidate sources of iDNA and methodologies for conducting such surveys have not been fully explored.

Several recent studies investigated the potential of mosquito blood meals as iDNA for vertebrate detection in natural ecosystems. Kocher et al.21 collected hematophagous dipterans (sandflies and mosquitoes) to use as vertebrate samplers in the northern Amazon and concluded that mosquito blood meal analysis could be a helpful tool in biodiversity surveys but needed further investigation. Massey et al.22 collected carrion flies, sandflies, and mosquitoes in Mato Grosso, Brazil across urban and semi-urban habitats and compared the detection of terrestrial vertebrates from iDNA to detections from camera traps. Camera traps detected the highest species richness but were found to be biased towards carnivores and ungulates while the iDNA detections were more varied and included many vertebrate species and taxa (e.g., birds, bats, reptiles, amphibians) that were undetected by camera traps. Mosquitoes detected fewer hosts than other iDNA taxa, and were determined to be biased toward humans22. Danabalan et al.23 compared blood-fed mosquitoes, non-blood-fed mosquitoes and flies (Sarcophagidae, Calliphoridae, Muscidae) as sources of iDNA for detecting mammal diversity in Germany and found non-mosquito flies detected more mammals than mosquitoes, but the mosquito sampling methods (sweep netting) collected few blood-fed females and only 58 total blood meals were analyzed. Vieira et al.24 used 480 mosquito blood meals to detect host diversity and to screen for serological evidence of exposure to Ross River virus. Of these, 346 were identified representing 26 species, but this sample was dominated by human (73%) and cattle (9%) detections, likely reflecting the land-use of the study location, Brisbane, Australia. Additional research involving relatively small samples sizes concluded mosquitoes had potential as sources of iDNA25,26. Together, these studies suggest that mosquitoes can be useful sources of iDNA, albeit with limitations, and that methodologies for collecting and analyzing mosquito iDNA should be optimized.

Among the previous work exploring mosquitoes as a source of iDNA, only Danabalan et al.23 consistently identified collected mosquitoes to species. Host association is an important and idiosyncratic element of the biology of a mosquito species that is relevant to iDNA-based vertebrate detection using mosquitoes. Mosquitoes feed on all terrestrial vertebrate classes: Amphibia, Aves, Mammalia, Reptilia, and some species feed on fish and invertebrate annelids27,28. Among mosquito species, host associations are variable and understanding which mosquito species are associated with which vertebrate groups may be beneficial to interpreting the results of mosquito-based iDNA surveys or to designing and implementing surveys that target particular vertebrate groups or species. Mosquito host associations are nuanced, and go beyond simple classification of species as generalists or specialists. While a few mosquito species are relative generalists (i.e., species that feed on all terrestrial vertebrate classes without a strong association with particular classes), most specialize, to varying extents, on particular subsets of the vertebrate community within their ecosystems. Even among mosquito species that specialize on a particular vertebrate class, there may be especially strong associations with taxa within that class. For example, Culiseta melanura is strongly associated with avian hosts, especially passerine birds, while Culex erraticus, a relative generalist, often feeds from birds, but among birds, is strongly associated with wading birds29,30. Likewise, Culex cedecei feeds predominantly from mammals, but has a strong association with rodents31. Such nuanced host associations would be expected to make individual mosquito species more or less useful to a mosquito-based iDNA survey for vertebrates, depending on the objectives of the surveys (e.g., characterizing the vertebrate diversity of a site, detecting a particular target taxon of conservation or management importance). Thus, recognizing the mosquito species that have the greatest return on investment may be in important step in optimizing the utility and value of a mosquito-based iDNA survey, along with selecting appropriate sampling methods that maximize collection of the most useful species.

To better understand the feasibility of mosquito blood meal-based iDNA surveys of vertebrate diversity, we investigated mosquitoes as a means of detecting terrestrial vertebrate species at the DeLuca Preserve, in Osceola County, Florida and among mosquitoes, we assessed variation in the range of vertebrate hosts detected by individual mosquito species to explore how mosquitoes can be most effectively used in surveys for vertebrate animals. Mosquitoes were collected using methods expected to best target blood-fed female mosquitoes. Collected blood-engorged females were identified to species, and blood meals were preserved and screened for vertebrate DNA using DNA barcoding. The objectives of this work were: (1) to characterize the host associations of mosquito species and range of vertebrate animals detected by mosquito blood meal iDNA sampling at the DeLuca Preserve; and (2) to evaluate the utility of individual mosquito species for iDNA sampling based on the vertebrate “communities” detected by each mosquito species, diversity metrics (species richness, Shannon-Index, Gini-Simpson Index, estimated species richness, and sample completeness) of each community, and a measure of host detection efficiency (i.e., the number of vertebrate host species detected by mosquito species relative to the number of blood meals that were analyzed for that species).

Materials and methods

Study location

The study took place at the DeLuca Preserve, a ~ 10,900 ha conservation area maintained by the University of Florida located in Osceola County, Florida, USA. Natural and modified habitats at the DeLuca Preserve consisted of forests, wetlands, Florida scrub, citrus groves and pastureland. This land provides habitat for native and imperiled Florida wildlife such as the Florida panther (Puma concolor coryi), grasshopper sparrow (Ammodramus savannarum floridanus), red-cockaded woodpecker (Picoides borealis), and gopher tortoise (Gopherus polyphemus32). Additionally, the preserve was within a conservation focus area for the Everglades Headwaters National Wildlife Refuge and Conservation Area, situated between the Kissimmee Prairie Preserve and the Three Lakes Wildlife Management Area33.

Sampling sites

Within the preserve, mosquitoes were collected at sites in scrub, citrus grove, wetland, and forest habitats. Eight sampling sites were selected (Fig. 1), two in each of the habitat types, and mosquitoes were repeatedly sampled at each over an eight-month period. The forest areas sampled at the DeLuca Preserve were hardwood hammocks or mixed forest within mesic flatwood habitat. We sampled mosquitoes in a hardwood and sabal palmetto hammock and in a closed-canopy mixed woodland dominated by bald cypress, hardwoods, and pine, both embedded within a larger swath of mesic flatwoods. Scrub sites consisted of Florida scrub, a habitat found on dry, sandy ridges in Florida. The study habitat we classified as wetland consisted of a type of non-forested wetland habitat, depression marsh, a shallow, rounded, depression area that has sandy soil and herbaceous plants that grow in concentric bands, with the outermost bands being the driest, and the innermost being the wettest34. Citrus grove was the only human-modified habitat that was surveyed at the DeLuca Preserve. The groves consisted of citrus trees planted in rows on raised beds, often with tall grasses and other herbaceous vegetation growing in furrows between the rows, where water often pooled after rainfall35.

Fig. 1
figure 1

Map of the eastern side of the DeLuca Preserve, Osceola County, Florida, USA, indicating the location of field sites where mosquitoes were collected (red squares). Blood-fed mosquitoes were collected from a total of eight sites, which represented four habitats, mesic flatwood forest, citrus grove, scrub, and depression marsh wetland. Two sites were located in each habitat type. At each site, one ten-minute aspiration was performed on each sampling day (n = 45) and five to seven resting shelters were deployed. Hashed black lines indicate the borders of the DeLuca Preserve. Inset (top right) indicates the location of the DeLuca Preserve in the state of Florida. Map created in ArcGIS Pro using the Statewide Land Use Land Cover shapefile hosted by the Florida Department of Environmental Protection.

Full size image

Mosquito collection

Mosquitoes were collected in 2022 from January through August, across three seasons: winter (with collections taking place in January and February), spring (March, April, and May), and summer (June, July, August), to account for potential seasonal variation in mosquito abundance36, host association37,38,39 and vertebrate behavior (e.g., migration, seasonal activity, above ground activity). Mosquitoes were collected over nine sampling periods, three during each season. Each sampling period consisted of five consecutive days of sampling. Sampling took place in the morning to early afternoon (~ 09:00–13:00). The total number of field days of mosquito collection across the eight-month sampling period was 45, each with a duration of approximately four hours.

In order to maximize the number of collected mosquitoes and proportion of blood-engorged female mosquitoes collected, sampling efforts within each site focused on areas where mosquitoes were likely to rest after taking a blood meal from a host40,41. Two sampling methods were used to collect mosquitoes each day: resting shelters and a large-diameter aspirator. At each site, five to seven resting shelters were dispersed on the ground in dark areas shaded from morning sun exposure. The shelters were left in place throughout the duration of the study period. Mosquitoes were collected from each shelter daily during each sampling period. The shelters consisted of either a thick black trash bag fitted over a collapsible wire spiral frame or a similar commercially purchased collapsible black yard waste container. To collect mosquitoes from the shelters, a lid holding a collection cup was placed over the shelter entrance and the shelter was compressed several times, forcing the air inside the shelter, along with any mosquitoes, into the collection cup. Both construction of and collection from the shelters followed the methods described by Burkett-Cadena et al.42. The aspirator was created using an automotive radiator fan attached to a cylindrical wire field fencing frame covered in polyester tarp, based upon a model initially described by Nasci43 and modernized as described in Sloyer et al.44. The aspirator was powered by a 12-V rechargeable battery, and a mesh mosquito head net was used as a collection bag, attached to the inside of the aspirator. For each sampling day, each of the eight sites was aspirated once for approximately ten minutes. Aspirations at each site consisted of moving the running aspirator over likely mosquito resting microhabitats: leaf litter or undergrowth, buttressed tree trunks, recently disturbed areas of the ground, and herbaceous vegetation, depending on the habitat type. All collected mosquitoes were knocked down with carbon dioxide (dry ice), transferred into labeled vials, and transported in a cooler with dry ice to the Florida Medical Entomology Laboratory.

Mosquito identification and blood meal preservation

Field-collected mosquitoes were kept in a − 20 °C or − 80 °C freezer until processing. Mosquito specimens were examined under a stereoscope, separated by species, and counted. Species were morphologically identified using a dichotomous key to the mosquitoes of North America45,46. All blood-engorged females were separated from all other mosquitoes, counted, and given an identifying number. To preserve the host DNA, blood-engorged females were individually placed on Whatman Flinders Technology Associates (FTA) Classic Cards, and a sterile pipette tip was used to roll out the blood meal onto the card47,48. The blood meal was characterized on a scale of one to three based on the level of digestion48, either as blood-fed (BF) 1 (fresh, in early stages of digestion, and red in color), BF2 (partially digested, brown with some red coloration persisting), or BF3 (mostly digested and entirely brown or black in color). Cards with preserved blood meals were stored at room temperature.

DNA barcoding and sequencing

From each preserved blood meal, two 1 mm punches were taken from the associated Whatman FTA Card. Using the Hot Sodium Hydroxide and Tris (HotSHOT) method49,50, the DNA was extracted, then amplified by polymerase chain reaction (PCR) with primers targeting a fragment of the cytochrome c oxidase subunit I (COI) gene of vertebrates51,52. Three primer combinations were used in the hierarchical approach described in Reeves & Burkett-Cadena52: VertCOI_7194 (5′–CGM ATR AAY AAY ATR AGC TTC TGA Y–3′) + Mod_RepCOI_R (5′–TTC DGG RTG NCC RAA RAA TCA–3′), Mod_RepCOI_F (5′–TNT TYT CMA CYA ACC ACA AAG A–3′) + VertCOI_7216_R (5′–CAR AAG CTY ATG TTR TTY ATD CG–3′), and VertCOI_10096_F (5′–CHC AAT ACC AAA CNC CHY TNT TYG–3′) + Mod_RepCOI_R. The PCR products were visualized on an agarose gel and amplicons displaying a band of the expected size were sent to Eurofins Genomics (Louisville, KY) for chain termination sequencing53. The returned DNA sequence files were edited using Geneious Prime 2020.1.2 and submitted to the Barcode of Life Data System (BOLD) Version 4 Identification Engine for identification by comparison to reference sequences54. At the time of analysis, reference COI sequences were not available for Sylvilagus palustris (marsh rabbit) in publicly accessible sequence databases, and there appears to be geographic variation in Anolis carolinensis (green anole) COI sequences51. Thus, Sequences that were close matches (> 85% similarity) to Sylvilagus spp. (cottontail rabbits) and Anolis carolinensis (> 95% similarity) were compared to independently collected COI sequences from those species, derived from locally collected specimens51. Ten high-quality sequences were poor matches (94–97% similarity) to Setophaga and Geothlypis warblers. Because it was unclear what species these represented and how many species were represented, these ten sequences were excluded from the dataset of identified hosts, and those blood meals were considered unidentified. One high-quality sequence was a poor (94%) match to several passerine species that do not occur in Florida. Because this sequence was assumed to be derived from a passerine species not yet present in the BOLD database, and was distinct from all other blood meal sequences, it was included in the dataset as “Unidentified Passeriformes.” In two cases, COI sequences could not distinguish pairs of closely related species. Odocoileus virginianus (white-tailed deer) and Odocoileus hemionus (mule deer) have identical COI sequences and could not be separated. Because Odocoileus hemionus is not native to Florida and not known to occur in the wild, all Odocoileus sequences were attributed to Odocoileus virginianus. Similarly, Ardea herodias (great blue heron) and Ardea cocoi (cocoi heron) could not be separated, but all sequences matching this species pair were attributed to Ardea herodias based on geography.

Statistical analysis

All statistical analyses were run on R Version 4.3.155. We characterized mosquito host associations by compiling all host species detected by each mosquito species and by mosquitoes overall. Each assemblage of vertebrate species detected by a mosquito species was treated as a biological community. For each mosquito species and for all mosquito species combined, we constructed sample completeness profiles using the inext.4steps package56 to determine the upper bound of the proportion of total species in the assemblage (“assemblage” here referring to the range of vertebrate species the mosquito species feeds upon at the DeLuca Preserve) that were detected by the mosquito species, the proportion of individuals in the assemblage that are belong to detected vertebrate species, and the proportion of highly abundant species that were detected. To assess the diversity of the host communities detected by each mosquito species and by mosquitoes overall, vertebrate species count data were analyzed using the iNEXT package57. For each community, diversity metrics were calculated including species richness, Shannon index, Gini-Simpson index, and estimated species richness based on accumulation curves58. Accumulation curves were created using the iNEXT package in R for mosquito species with ten or more identified blood meals and for mosquitoes overall57. The curves were created to enable a robust comparison of the richness detected by each mosquito species based on asymptotic estimates of diversity. Accumulation curves were similarly created to assess the completeness of sampling (all mosquitoes species) of each vertebrate class individually to identify vertebrate classes that were well or poorly detected by the sampling effort.

Because mosquito host associations vary by species, we developed the metric “host detection efficiency” to evaluate and compare the ability of each mosquito species in our sample to contribute toward vertebrate diversity detection in our iDNA-based survey. Host detection efficiency was calculated as the square of host species richness detected by the mosquito species divided by the total number of blood meals analyzed for that species. The host detection efficiency represents a measure of the number of vertebrate host species detected by a mosquito species relative to the number of blood meals that were analyzed for the species. These values provide an indication of the return on investment for each mosquito species. Some mosquito species are narrowly associated with a small range of host species, and thus, may not be worth the expense and effort of including in a vertebrate diversity survey if the detection of as many vertebrate species as possible is the goal. Such species are expected to have host detection efficiency values closer to zero. Other mosquito species feed on a wider range of vertebrate species and are thus more useful to the goal of broad species detection than others. The larger the host detection efficiency value, the more efficient a mosquito species is at detecting the largest range of vertebrate species.

Results

A total of 54,637 mosquitoes were collected throughout the duration of the study, 3,508 (6.4%) of which were blood-fed. Of the blood-fed females, 2,051 (58.4%) resulted in species level vertebrate host identifications. For blood meals with digestion extent estimated as BF1 (n = 1,334), 90.0% resulted in positive identification of a vertebrate host species. For blood meals estimated as BF2 (n = 1,011), 64.8% resulted in positive host identifications, and for those categorized as BF3 (n = 1,163), 19.9% of blood meals resulted in a positive host identification.

Blood meals were collected from 21 mosquito species (Table 1) and included species from eight genera: Aedes, Anopheles, Coquillettidia, Culex, Culiseta, Mansonia, Psorophora, and Uranotaenia. Blood-fed Uranotaenia sapphirina were collected but excluded from analysis because this species is a specialist of annelid hosts in Florida28. Identifications for blood-fed females that were initially identified as Aedes atlanticus based on morphology and distribution45 were revised to Aedes atlanticus/tormentor because DNA barcoding analysis59 of specimens collected at the DeLuca Preserve for a separate project indicated that both morphologically similar species were present at the preserve (Table 3). Culex nigripalpus was the most abundant mosquito species in the blood meal sample, and was the species with the largest number (n = 1,024) of vertebrate host detections, representing 49.9% of all vertebrate detections from mosquito blood meals. Aedes infirmatus produced the second largest number of vertebrate detections, with 267 vertebrate detections (14.3%), followed by Culex pilosus with 262 detections (14.1%).

Table 1 Vertebrate taxa detected by mosquito blood meal-based iDNA surveys at the DeLuca Preserve, Osceola Co., Florida, USA. For each vertebrate class, values represent the number of orders, families, and species detected by screening mosquito blood meals for vertebrate DNA, and, for each order, the number of mosquito species from which DNA from that class or order was detected.
Full size table

Host species representing four classes of terrestrial vertebrates were detected in mosquitoes collected at the DeLuca Preserve (Table 1). From the 2,051 total detections, 3.0% were amphibians (n = 61), 32.6% were birds (n = 668), 48.8% (n = 1,042) were mammals, and 15.7% (n = 322) were reptiles. In total, 86 vertebrate host species were detected. The full list of detected vertebrate species is available in Supplemental Table 1. Of these, seven were amphibians, 57 were birds, 14 were mammals, and eight were reptiles. All detected amphibian species were anurans (frogs). Detected bird species represented 12 orders, with Passeriformes (songbirds) being the most species rich (28 species). Detected mammal species represented eight orders and detected reptile species represented three, with squamates (lizards and snakes) the most species rich.

The five most-collected blood-fed mosquito species (Cx. nigripalpus, Cx. pilosus, Ae. infirmatus, Psorophora columbiae, and Cs. melanura) contributed 91.0% of all vertebrate detections. Among the other 16 mosquito species, fewer than 40 blood-fed individuals were collected from each, and of these, fewer than 10 blood-fed individuals were collected from 11 species throughout the duration of the study. In the overall sample of detected vertebrate hosts, a few host species were common while many were rare. Four host species were detected more than 100 times: Odocoileus virginianus (n = 788), Anolis carolinensis (n = 273), Meleagris gallopavo (wild turkey; n = 152), and Strix varia (barred owl; n = 113). Sixty-four of the 86 detected host species were detected fewer than ten times, and 24 were detected only once.

Mosquito host associations

Most mosquito species (Ae. atlanticus/tormentor, Ae. infirmatus, Aedes taeniorhynchus, Coquillettidia perturbans, Mansonia titillans, Psorophora ciliata, Ps. columbiae, Psorophora ferox) fed from a higher proportion of mammalian hosts than other vertebrate classes (Fig. 2). Culiseta melanura was the only mosquito species that fed primarily on birds (> 70% of identified bloodmeals), though the single identified Culex interrogator blood meal was derived from a bird. Culex pilosus fed primarily on reptiles, and both Culex territans and Uranotaenia lowii took more blood meals from amphibians than other host classes (Fig. 2). While no individual mosquito species fed evenly among all vertebrate classes, Cx. nigripalpus fed approximately evenly between birds and mammals, with 46% of blood meals from mammals and 48% from birds. Additionally, 6% of Cx. nigripalpus blood meals were derived from reptiles, and 0.1% from amphibians.

Fig. 2
figure 2

Proportion of blood meals from each vertebrate class (amphibian, bird, mammal, reptile) for each mosquito species collected at the DeLuca Preserve, Florida, USA. Colors indicate vertebrate class (yellow = Amphibia, orange = Aves, red = Mammalia, dark purple = Reptilia). The sample size of identified blood meals for each mosquito is represented in parentheses after the species name.

Full size image

For some mosquito species, a large proportion of the blood meal sample was derived from only one or two host species. For Ps. columbiae (n = 160), 91.3% of blood meals were derived from Odocoileus virginianus. Culex pilosus (n = 292) fed primarily on lizards of the genus Anolis, with 85% of blood meals derived from native Anolis carolinensis and non-native Anolis sagrei (brown anole).

Sample completeness

To compare the ability of mosquito species to contribute to detection of a broad range of host species in an iDNA survey, the assemblage of vertebrate host species detected by each mosquito species was treated as a biological community and sample completeness profiles and diversity metrics were calculated for each species and for all mosquito combined. Here, use of the word “community” refers to the assemblage of vertebrate species that are bitten by and detected or undetected by each mosquito species (or by all mosquito species combined) at the DeLuca Preserve during the sampling period. In the sample completeness profiles estimated for each vertebrate community sampled by a mosquito species, sample completeness increases with diversity order (q) for all mosquito species combined and for each mosquito species individually except for Ps. columbiae (Fig. 3). Thus, for all but Ps. columbiae, it is likely that there were undetected vertebrate species within each community. For the community sampled by all mosquito species combined, sample completeness for q = 0, 1 and 2 was 81.8%, 98.8%, and 100%, respectively (Table 2), indicating that at most 81.8% of the total vertebrate community fed upon by mosquitoes was detected, the species that were detected make up about 98.8% of the individuals in the community, and the detected species make up 100% of the individuals of the abundant species of the community. For all mosquito species combined, at least 1 – 81.8% = 18.2% of vertebrate species in the community were not detected in the sample. The undetected species make up an estimated 1 – 98.8% = 1.2% of the individuals in the community. Based on the q = 1 and q = 2 estimates of undetected diversity, essentially all abundant and highly abundant species in the community were detected.

Fig. 3
figure 3

Profiles of sample completeness as a function of order q between 0 and 1 for vertebrate species detected by individual mosquito species and all mosquito species combined at the DeLuca Preserve, Osceola County, Florida, USA. Sample size of each species is indicated below species names. Mosquito species for which sample size was fewer than ten or for which number of host species detected was fewer than two are excluded.

Full size image
Table 2 Sample completeness values as a function of order q = 0, 1, and 2 for vertebrate species detected by individual mosquito species and all mosquito species combined at the DeLuca Preserve, Osceola County, Florida, USA. Sample size of each species is indicated below each species name. Mosquito species for which sample size was fewer than ten or for which number of host species detected was fewer than two are excluded.
Full size table

For individual mosquito species, sample completeness estimates for q = 0, 1 and 2 for individual species varied, especially at q = 0 and q = 1. For q = 2, completeness estimates were near or at 100% (98.7–100%) except for Ae. atlanticus/tormentor (87.2%), indicating that essentially all abundant and highly abundant vertebrate species in the communities fed upon by each mosquito had been detected. For Ae. atlanticus/tormentor, about 12.8% of abundant species were undetected, though for this species, sample size was small (n = 10) and confidence bands were wide. Among individual mosquito species, sample completeness for q = 0 ranged from 26.2% to 100%. Sample completeness values at q = 0 indicate the upper bound of the proportion of species in the community that were detected by sampling. Because values for all species except Ae. atlanticus/tormentor were > 98.7% at q = 2 indicating that all or most of the abundant species were detected, it is likely that the vertebrate species that were undetected by all mosquito species other than Ae. atlanticus/tormentor were species that were rare in each community. The smallest sample completeness value at q = 0 was 26.2% in Cx. erraticus, implying that at least 73.8% of the vertebrate host species fed upon by Cx. erraticus at the DeLuca Preserve were not detected in the sample. At q = 0, the largest sample completeness value was 100% in Ps. columbiae, which fed predominantly from one species, Odocoileus virginianus. For q = 1, sample completeness ranged from 62.1% (Ae. atlanticus/tormentor) to 100% (Ps. columbiae), suggesting that for each species, the majority of individual vertebrate animals in each community belong to species that were detected by our sampling. For Ae. infirmatus, Cx. nigripalpus, Cx. pilosus, Cx. territans, Ps. columbiae, and Ps. ferox, q = 1 sample completeness estimates were > 95%. Culex nigripalpus had sample completeness values of 70.5%, 97.7%, and 100% at q = 0, 1, and 2, respectively. Despite the largest sample size (n = 1,021) and the largest richness of detected vertebrate species among collected mosquito species, sample completeness for Cx. nigripalpus at q = 0 indicated that at most, 70.5% of the species fed upon by Cx. nigripalpus had been detected. This undetected diversity is likely to consist of rare species because at q = 2, sample completeness was estimated at 100%, indicating that no more abundant species were likely to be detected.

Diversity indices

The diversity indices of detected hosts varied by mosquito species. Host species richness detected by individual mosquito species ranged from one to 63 vertebrate species (Table 3). Seven of the 21 collected mosquito species only detected one vertebrate species, though for many of these, sample size was small (n < 10). Culex nigripalpus had the highest species richness of hosts, detecting 63 vertebrate species, followed by Cs. melanura with a richness of 28, Ae. infirmatus with 15, Culex erraticus with 14, and Cx. pilosus with 12. All other mosquito species detected six or fewer vertebrate species. Overall, host detection efficiency for all mosquitoes combined was 3.86. Among mosquito species, host detection efficiency was highest for Cs. melanura (7.47), followed by Cx. erraticus (5.16), and Cx. nigripalpus (3.89; Table 3). Host detection efficiency was < 1.0 in several species (Ae. infirmatus, Aedes triseriatus, Aedes vexans, Anopheles quadrimaculatus s.s., Cq. perturbans, Culex coronator, Cx. pilosus, Ps. ciliata, Ps. columbiae, and Ps. ferox). For most of these, sample size was small, but for Ae. infirmatus (n = 286), Cx, pilosus (n = 292), and Ps. columbiae (n = 160), sample size was large and host detection efficiency was low.

Table 3 Diversity measures for the host “community” detected by each mosquito species based on blood-fed mosquitoes collected at the DeLuca Preserve, Florida, USA. “Host detection efficiency” is calculated as square of host species richness divided by the number of blood meals analyzed. Estimated species richness and sample coverage were not calculated for species with small sample sizes (n < 5) or species that did not detect more than one host species.
Full size table

The Shannon index, an indicator of evenness and diversity, ranged from 0 (only one species represented) to 2.88 (Cs. melanura; Table 3). The Gini-Simpson index, a dominance-based diversity index (ranging from 0 to 1), ranged between 0 (one species detected) to 0.923 (Cs. melanura; diverse community with few dominant species). For both the Shannon and Gini-Simpson indices, Cs. melanura had the highest values, higher than Cx. nigripalpus despite Cx. nigripalpus having a higher observed species richness (Table 3).

Accumulation curves

Accumulation curves were created for the mosquito species for which at least ten blood meals were identified (Fig. 4). Ten of the 21 mosquito species met this criterion. Aedes infirmatus, Cx. pilosus, Cx. territans, Ps. columbiae, and Ps. ferox all reached an asymptote based on the vertebrate species detected by their blood meals (q = 0 sample completeness > 0.90), suggesting that the sampling effort was sufficient to obtain a representative sample of the host communities of these mosquito species at the DeLuca Preserve, and that few additional species would be detected with additional sampling. Despite being the mosquito species with the largest sample size, the curve for Cx. nigripalpus did not reach an asymptote, with q = 0 sample completeness of 70.5% (i.e., at most, 70.5% of the species in the community were detected). Despite relatively large sample sizes, neither Cs. melanura (n = 105; q = 0 sample completeness: 66.3%) nor Cx. erraticus (n = 38; q = 0 sample completeness: 26.2%) reached an asymptote based on the blood meal sample (Fig. 3). Curves for Ae. atlanticus/tormentor and Anopheles crucians s.l. also failed to achieve saturation. The curve for all mosquito species approached an asymptote, but did not reach saturation, suggesting that further sampling would detect additional vertebrate species (q = 0 sample completeness: 81.8%). Based on the 100% sample completeness at q = 2 for all mosquito species combined, vertebrate species that were undetected by mosquitoes in general were likely to be rare rather than abundant species. For all mosquito species, estimated richness of detected vertebrate species at the asymptote was 105.2 species.

Fig. 4
figure 4

Accumulation curves depicting the accumulation of vertebrate species detected by each mosquito species (n > 10) at the DeLuca Preserve, Florida, USA. The solid line to the point on the curve shows where the sample ends and the estimation is represented by the dotted line. Sample size for each species is indicated under each species name. Mosquito species for which sample size was fewer than ten or for which number of host species detected was fewer than two are excluded.

Full size image

Among mosquito species for which at least ten blood meals were collected and identified, estimated species richness of host communities based on the asymptote ranged from 3.8 (Ma. titillans) to 84.03 (Cx. nigripalpus) vertebrate species (Table 3). Estimated species richness was generally low for mosquito species that took blood meals exclusively or predominantly from one vertebrate class (Fig. 1), especially mammals (e.g., Ae. atlanticus/tormentor, Ae. infirmatus, An. crucians s.l., Ma. titillans, Ps. columbiae, Ps. ferox) and reptiles (e.g., Cx. pilosus, Cx. territans). Conversely, estimated species richness tended to be larger for mosquitoes that took blood meals from multiple vertebrate classes (e.g., Cx. erraticus, Cx. nigripalpus, Cs. melanura).

Overall, seven amphibian, 57 bird, 14 mammal, and eight reptile species were detected by mosquito blood meals. Accumulation curves created based on the detection of vertebrate species by mosquitoes (all species) for each vertebrate class (Fig. 5) approached an asymptote in Mammalia (n = 1,042; sample coverage: 0.95, i.e., 95% of the species in the community were represented by detected species; estimated species richness: 14.67 species), and reached an asymptote in Amphibia (n = 61; sample coverage: 1.00; estimated species richness: seven species). In Aves (n = 668; sample coverage: 77.9; estimated species richness: 73.17 species), and Reptilia (n = 322; sample coverage: 0.78; estimated species richness: 10.24 species), the curves were still rising based on the reference sample.

Fig. 5
figure 5

Accumulation curves depicting the accumulation of vertebrate species detected by mosquitoes at the DeLuca Preserve, Florida, USA by vertebrate class. The solid line to the point on the curve shows where the sample ends. Estimated accumulation of species beyond the sample is represented by the dotted line. Sample size for each class is indicated above each point.

Full size image

Discussion

We collected mosquitoes from four habitats (citrus grove, scrub, forest, wetland) at the DeLuca Preserve across three seasons to investigate the potential of using mosquito-based iDNA for detecting vertebrates in diversity surveys and to compare the assemblages of vertebrate species detected by individual mosquito species. From a sample of more than 2,000 blood meals representing 21 mosquito species, we detected the presence of 86 vertebrate species, and demonstrate that at this site it is feasible to use mosquito blood meal iDNA to survey and detect a diverse range of terrestrial vertebrate species. Collectively, across all mosquito species, all four terrestrial vertebrate classes were represented in blood meal detections. However, individual mosquito species varied in their associations with particular vertebrate classes and in their efficiency in detecting diverse host species. While almost all mosquito species detected essentially all the abundant species among the subset of vertebrate species they feed upon, most had undetected diversity expected to consist of rare species, in line with the expectation that abundant species are easier to detect than rare species. Mosquitoes were able to detect a variety of vertebrate animals including both large- (e.g., ungulates, carnivores, large birds) and small-bodied species (e.g., frogs, lizards, rodents, songbirds). This result parallels mammal surveys based on iDNA from carrion flies which similarly were able to detect mammal species of a wide range of size classes60. The assemblage of vertebrates detected by mosquitoes was taxonomically diverse with species from 22 vertebrate orders, and included species that are nocturnal, diurnal, migratory, resident, fossorial, arboreal, and semiaquatic, and those that are imperiled, invasive, cryptic and rare (Supplemental Table 1). This suggests that a benefit of mosquito-based iDNA surveys is the ability to detect a wide breadth of species using a single survey method. In comparison, many traditional methods of vertebrate detection are most effective for only a narrow range of species. For example, camera traps, a popular method for vertebrate detection and survey, are best suited to detection of larger vertebrates, particularly terrestrial mammals61.

Despite detecting a large number and diversity of vertebrate host species, sample coverage of mosquito blood meals varied among vertebrate host classes. For amphibians and mammals, accumulation curves suggest that essentially all the host species of the communities of these classes were detected and that continued sampling would be unlikely to detect additional species. Conversely, in birds and reptiles, sample coverage was ~ 0.78, suggesting that further sampling would detect additional species of both groups. This was especially true of birds with 57 detected species and an estimated species richness of 73.17, compared to eight detected reptile species and an estimated species richness of 10.24. The estimated accumulation curve for birds approached saturation only after ~ 2000 detections, indicating that for this group, sample size would need to increase by about four times to achieve complete sample coverage of available host species. Conversely, high sample coverage was achieved with less effort for all other vertebrate classes. It is likely, however, that the estimated species richness of the amphibian, mammal, and reptile communities based on mosquito blood meal detections are underestimates of the true richness of each community as these accumulation curves represent the detection process of the pool of species on which the mosquitoes are feeding rather than the entirety of the available vertebrate community at the site. In Florida, and likely at the DeLuca Preserve, these groups include species that are largely inaccessible to host-seeking mosquitoes. Species that are almost fully aquatic (e.g., amphiumas and sirens or snakes of the genus Farancia), and fossorial species that exhibit limited aboveground activity (e.g., Geomys pocket gophers, Scalopus aquaticus, eastern mole, Rhineura floridana, Florida worm lizard, Scaphiopus holbrooki, eastern spadefoot toad, Micrurus fulvius, eastern coral snake) would be unlikely to be encountered and fed upon by mosquitoes, and thus, unlikely to be detected. Further, no bat species were detected by mosquitoes, and only one species of the order Rodentia (Sigmodon hispidus, hispid cotton rat) was detected. Similarly, birds such as woodpeckers and kingfishers that nest and roost in tree cavities or underground, neither of which were detected in our mosquito sample, may be less likely to be detected by mosquito surveys62. These results suggest that, though mosquitoes were able to detect vertebrate species with a variety of biological traits, there are likely groups that are undersampled by mosquito blood meal iDNA surveys.

While all terrestrial vertebrate classes were detected by mosquitoes collectively, there was variation among individual mosquito species in the range of vertebrate classes and species that were detected and no single mosquito species detected the full range of host species detected by all mosquito species combined. This result was expected as mosquito species are well known to exhibit distinct host associations and feeding patterns27 with various degrees of specialization ranging from species that are relative generalists (e.g., Culex iolambdis63) to those that are narrow specialists (e.g., Culex atratus and Cx. pilosus64). The host associations of the mosquito species in our sample were similar to those previously reported for the same species collected elsewhere in Florida39,62,64,65,66,67,68, indicating the robustness of this relationship and the potential utility of targeting particular mosquito species for detecting species within specific vertebrate groups.

Several previous studies investigated mosquito-based iDNA as a means of vertebrate diversity survey from natural ecosystems. Kocher et al.21, Danabalan et al.23, Saranholi et al.25,Vieira et al.24, and Chivas et al.26 demonstrated that detection of vertebrate animals through mosquito-based iDNA is feasible. However, these surveys involved relatively limited sample sizes of collected mosquitoes, which may have limited the power of their surveys to detect and characterize the diversity of vertebrate communities. In contrast, Massey et al.22 used next generation sequencing to identify hosts from pooled light trap-collected mosquitoes, both unfed and blood-fed, and reported low numbers of vertebrate detections. At the DeLuca Preserve, we implemented mosquito sampling methods (aspirator and resting shelters) that better target blood-fed female mosquitoes and demonstrate a survey strategy with increased power to collect blood-fed mosquitoes and thus to characterize vertebrate diversity. We identified the hosts of a total of 2,051 blood-fed mosquitoes, compared to the sample sizes of 87, 58, 17, and 480 of previous blood-fed mosquito-based iDNA surveys, which detected 16, 5, 24, and 26 vertebrates, respectively (Kocher et al.21; Danabalan et al.23; Saranholi et al.25; Vieira et al.24, respectively). Here, vertebrate species richness detected by mosquitoes was notably higher than those of previous studies, however, collection methods, sampling effort, geographic location, habitat, and associated differences in mosquito species composition and abundance likely influenced the comparatively high numbers of mosquito blood meals we collected and the richness of vertebrate species we detected. These factors should be explored and considered prior to implementation of a mosquito-based vertebrate survey. In particular, those interested in using this approach should use mosquito collection methods that are known to produce large numbers of blood fed female mosquitoes (e.g., aspirators, resting shelters) and be familiar with their effective use. Unfortunately, blood fed females cannot be collected with the similar ease of host-seeking female mosquitoes (i.e., those that are attracted to host cues), which can be collected in sometimes exceptionally large numbers using traps baited with carbon dioxide or other host cues42. Collection methods should be appropriate for the local mosquito community and geographic location where the study takes place, and, as much as is possible, surveyors should have an understanding of the local mosquito community and the host associations of its species.

Large sample size may be a necessary element of successful mosquito-based iDNA surveys, particularly for those with the goal of characterizing the diversity within vertebrate communities. At the DeLuca Preserve, after detection of 86 vertebrate species from > 2,000 blood meals, species accumulation had not yet reached saturation, and further sampling was likely to detect additional vertebrate host species. Based on the sample completeness profile, at q = 0, sample completeness for the total mosquito sample (i.e., all mosquito species combined) was 81.8% indicating that at most, ~ 80% of the species of the vertebrate assemblage fed upon by mosquitoes at the DeLuca Preserve were detected, and at least ~ 20% were undetected. We found that the assemblage of hosts detected by the total mosquito sample was dominated by a small number of common species (i.e., common within our sample). In this “community,” a few species (i.e., Odocoileus virginianus, Anolis carolinensis, Meleagris gallopavo, Strix varia) were common (commonly detected), while most were rare (rarely detected), in line with a universal law of the structure of ecological communities69. Sixty-four (74.4%) of the 86 vertebrate species we detected were detected fewer than ten times, and 24 (27.9%) were detected only once. At q = 1 and q = 2, sample completeness was 98.8% and 100%, respectively, indicating that 98.8% of the individuals in the community belonged to detected vertebrate species, and all abundant species in the community had been detected. This implies that the ~ 20% of the community’s species that were undetected in the sample consist of rare species, and collectively, they represent only 1.2% of the community’s individuals. Undersampling bias wherein common species are over-represented and rare species and singletons are undersampled is a pervasive problem in biodiversity sampling70. Thus, to minimize undersampling bias, a large sample of blood meals is likely needed to maximize detection of rare species when the goal of the survey is characterization of the vertebrate community. When possible, constructing sample completeness profiles56 based on count data as a survey progresses would be helpful to understanding the proportions of detected and undetected species in the vertebrate host community.

Determining the host detection efficiency of mosquito species at a study site could facilitate maximization of detection of rare species. Our results showing variable host detection efficiency values across mosquito species at the DeLuca Preserve suggest that mosquito species do not contribute equally to species detection in mosquito blood meal-based iDNA surveys. Host detection efficiency ranged from 7.47 (Cs. melanura) to 0.06 (Ps. ciliata), indicating that some mosquito species detected a disproportionally large or disproportionally small number of vertebrate species relative to their sample size. In general, species with high host detection efficiency values feed on multiple host groups, but were strongly associated with avian hosts, while those with low host detection efficiency values were primarily associated with one host class, often mammals or reptiles. Species with high host detection efficiency values included Cs. melanura (n = 105, ~ 5% of the blood meal sample; host detection efficiency: 7.47), which detected 28 (31.4%) of the 86 vertebrate species, Cx. nigripalpus (n = 1,021, about 50% of the overall blood meal sample; host detection efficiency: 3.89) which detected 63 (73.2%) of the 86 vertebrate species, and Cx. erraticus (n = 38, < 2% of the blood meal sample; host detection efficiency: 5.16) which detected 14/86 (16.3%) vertebrate species. All three of these fed on multiple host classes, with many blood meals derived from avian hosts. Among these three, sample completeness at q = 0 was ~ 70% for Cs. melanura and Cx. nigripalpus indicating that both species detected at most ~ 70% of their vertebrate “communities,” while Cx. erraticus had a much lower q = 0 sample completeness of only ~ 25%. For all three species, q = 2 sample completeness was essentially 100% suggesting that the > 30% of undetected species for Cx. nigripalpus and Cs. melanura, and the > 75% of undetected species for Cx. erraticus were all relatively rare species. Comparatively, other species, especially those with strong associations to individual vertebrate classes, contributed relatively little to the detection of new vertebrate species in the overall mosquito sample. For example, Ps. columbiae (n = 160; ~ 8% of the blood meal sample; host detection efficiency: 0.10) detected only four vertebrate species, none of which were undetected by other mosquito species. This suggests that when characterization of a vertebrate community is the goal of the survey (i.e., to detect as many of the vertebrate species that occur at the survey site as possible), species with higher host detection efficiency values (e.g., at the DeLuca Preserve, Cx. erraticus, Cx. nigripalpus, and Cs. melanura) should be targeted by field collections and prioritized in lab processing because they provide the greatest species detection return for resources spent.

Among the species detected by mosquitoes at the DeLuca Preserve were species that are imperiled (e.g., Gopherus polyphemus, gopher tortoise), invasive/nonnative (e.g., Anolis sagrei, Osteopilus septentrionalis, Cuban treefrog), cryptic or secretive (e.g., Scolopax minor, American woodcock, Antrostomus vociferus, eastern whip-poor-will), and rare (e.g., Crotalus adamanteus, eastern diamondback rattlesnake). While mosquito species that are narrowly associated with a small subset of the vertebrate community are unlikely to substantially contribute to detecting and characterizing the full vertebrate community, they may be informative to vertebrate surveys with other objectives, such as those with the intent of informing management decisions for imperiled or invasive species. Data on the host associations, resting habitats41 and efficient collection methods44 of each mosquito species would be indispensable in this scenario. A benefit of detecting imperiled species through mosquito blood meals or other iDNA sources is that the technique is noninvasive, enabling both the detection of such species and the indirect collection of blood or DNA samples10, avoiding the capture, handling and anesthetizing of individuals. Camera trapping is similarly noninvasive, yet it is biased toward larger species61 and does not provide a sample of DNA from vertebrates that can be used to glean additional information such as pathogen infection status71 or population demographics14.

In this study, we used chain termination sequencing53 to sequence vertebrate COI amplicons. Chain termination sequencing was useful here to rapidly and affordably process individual blood meals in a way that maintained the ability to associate an individual host detection with an individual mosquito specimen. Currently, this may be the most cost-effective approach for associating host species with individual mosquito specimens, though nanopore sequencing and other high-throughput approaches can be affordable and better at detecting multiple host species in mixed blood meals72,73. The ability to associate a host species with an individual mosquito specimen is especially relevant to vertebrate surveys that include xenosurveillance or population genetics questions. With the continued development of new sequencing technologies, high throughput sequencing approaches may become more feasible for processing such samples. If a survey or study does not need to attribute a host species to an individual mosquito specimen, or even species, metabarcoding and high throughput sequencing approaches would be appropriate. Under a metabarcoding approach, blood fed mosquitoes collected from a site could be pooled by species, or all together, and the pool processed by high throughput sequencing to identify all the vertebrate hosts fed upon by that pool of mosquitoes. However, the ability of these approaches to detect rare vertebrate species should be assessed, as our sample was dominated by several abundant species which may make it difficult to detect DNA from singleton or other rare species in the blood meal sample.

Over recent decades, eDNA-based approaches for detecting the presence of species have been increasingly applied to biodiversity monitoring. The types of samples that can be screened for eDNA are diverse, and selection of an appropriate source of eDNA depends on the organisms or ecosystem of interest. For example, in aquatic and terrestrial systems, samples of water and air, respectively, are effective sources of eDNA. Other sources of eDNA that have been used include spiderwebs, twigs browsed by herbivores, soil or aquatic sediments, or ice, among others. Each type of environmental sample has advantages and disadvantages, and is appropriate for particular situations or questions. Prominent disadvantages of mosquito blood meal-based eDNA include the sampling effort that is often necessary to collect large numbers of blood-fed mosquitoes and the learning curve required for field personnel to become effective in sampling larger numbers of blood-fed mosquitoes. Comparatively, collection of water or air samples for eDNA screening requires less field effort and specialized knowledge and experience (e.g., knowledge of mosquito resting behaviors, sampling techniques, and identification). Mosquito blood meal-based eDNA differs from water- or air-based eDNA in the nature of the sample that is obtained for analysis, and this difference enables researchers to glean additional information beyond only the identification of a species’ presence at a locality. Though mosquitoes occasionally feed on more than one host per gonotrophic cycle, most mosquito blood meals are derived from an individual host. Thus, each blood-fed mosquito, in general, represents a blood sample from an individual animal, from which it is possible not only to determine the species, but potentially to collect molecular data relevant to conservation or population genetics (e.g., Martinez- de la Puente et al.14), pathogen infection status or xenosurveillance (e.g., Wisely et al.13), or parasite diversity (e.g., Bartlett-Healy et al.74). Such data would be impossible or challenging to obtain from environmental samples that represent collections of free DNA fragments, cells or organelles75. The nuances of mosquito biology also confer advantages and disadvantages to their use as a source of eDNA. Because mosquitoes are not known to disperse large distances after blood-feeding (up to ~ 360 m18,76,77), and because host DNA degrades relatively rapidly within the mosquito midgut (< 48 h post-ingestion17), the ability to infer the presence of a detected species is limited to a relatively narrow spatiotemporal range, which may be an advantage or disadvantage of mosquito blood meal-based eDNA, depending on the scale and the objectives of the study or survey. Likewise, mosquito host associations can be an advantage or disadvantage of the technique. With the exception of only one known species, all blood-feeding mosquito species feed on the blood of vertebrates, with most specializing, to varying extents, on particular types of hosts. These associations could bias results toward particular vertebrate groups, or they could bolster likelihood of detection for target species by focusing field efforts on those mosquito species most likely to interact with targeted vertebrates. The advantages and disadvantages of the possible sources of eDNA should be considered when designing eDNA surveys to select sources of eDNA that are most appropriate to the study or survey objectives and to maximize the return on investment.

Here, we demonstrate the feasibility of using mosquito-based iDNA as a tool to survey vertebrate diversity on a large scale in conservation lands in central Florida, USA. We found that surveys using this single approach can detect a diverse assemblage of species representing all terrestrial vertebrate classes and species with a range of biological characteristics (e.g., species that are nocturnal, diurnal, migratory, resident, fossorial, arboreal, and semiaquatic, and species that are imperiled, invasive, cryptic, secretive, or rare). At this site, three mosquito species (Cs. melanura, Cx. erraticus and Cx. nigripalpus) contributed the majority of vertebrate species detections, highlighting their utility as vertebrate diversity detection tools. Globally, there are 3,727 described mosquito species78, and among this diversity, mosquito species vary in their host associations, habitats, larval microhabitats, abundance, and geographic distribution, among other factors. Each site is expected to have idiosyncratic mosquito assemblages that would be expected to influence the vertebrate assemblages that can be detected by mosquitoes. Thus, understanding the mosquito species in the local community in relation to their presence, abundance, host association, resting habits, and sampling, prior to the implementation of field surveys intended to detect vertebrate species, will be critical in successful deployment of this method.