Picky eaters are rare: DNA-based blood meal analysis of Culicoides (Diptera: Ceratopogonidae) species from the United States

Biting midges in the genus Culicoides (Diptera; Ceratopogonidae) have been implicated in the transmission of a number of parasites and highly pathogenic viruses. In North America, the complete transmission cycles of many of these pathogens need further elucidation. One way to increase our knowledge about the evolution and ecology of Culicoides species and the pathogens they transmit is to document the diversity of vertebrate hosts that Culicoides feed upon. Our objective was to identify the diversity of Culicoides hosts in the United States. We sequenced two vertebrate mitochondrial genes (cytochrome c oxidase subunit 1 and cytochrome b) from blood-engorged Culicoides to identify Culicoides species and their blood meals. We detected the mitochondrial DNA of 12 host species from seven different Culicoides species from three states. The majority of the identified blood meals were from the C. variipennis species complex in California. The hosts included both mammals and birds. We documented new host records for some of the Culicoides species collected. The majority of the mammalian hosts were large ungulate species but we also detected a lagomorph and a carnivore. The bird species that were detected included house finch and emu; the latter is evidence that the species in the C. variipennis species complex are not strictly mammalophilic. These results demonstrate that Culicoides will feed on multiple classes of vertebrates and may be more opportunistic in regards to host choice than previously thought. This knowledge can help with identification of susceptible host species, pathogen reservoirs, and new vector species which, in turn, will improve disease outbreak risk assessments.


Background
Understanding the sylvatic cycles and predicting epizootics of vector-borne pathogens requires knowledge about vector feeding behavior. Elucidating vector and host interactions helps with discovery of unknown vector species, susceptible hosts, reservoir species, and hostparasite-vector co-evolution, e.g. [1,2]. One of the most critical ecological parameters for predicting vector-borne transmission is vector biting rates on susceptible hosts [3][4][5]. Thus, accurate host identification is a requirement for determining the impact these pathogens may have on human, wildlife, and livestock health.
Sylvatic transmission pathways of vector-borne pathogens are often difficult to delineate because they involve multiple host species. Many blood-feeding arthropods that serve as vectors, especially in the Diptera, have feeding behaviors that can be plastic in response to environmental change and host availability [6,7]. For many vector species, we lack knowledge about the diversity of hosts they choose to bite [8][9][10]. Disentangling vectorborne pathogen transmission networks requires full characterization of the breadth of host species upon which vectors feed [11].
Biting midges of the genus Culicoides Latreille, 1809 (Diptera: Ceratopogonidae) are globally ubiquitous, except for New Zealand and Antarctica, and the group includes over 1400 described species [12,13]. Females of most species are hematophagous and transmit a multitude of pathogenic viruses, filarial worms, and blood parasites to humans, livestock, and wildlife [14]. Among the most damaging Culicoides-borne pathogens are bluetongue virus (BTV), epizootic hemorrhagic disease virus (EHDV), vesicular stomatitis virus (VSV), African horse sickness (AHSV), and Schmallenberg virus (SBV) [13,15,16]. Each of these viruses causes outbreaks in wildlife and livestock globally that have been linked to severe economic losses due to both lost production and disease-related trade restrictions [17][18][19][20]. In North America, the most commonly encountered vector species of BTV is C. sonorensis Wirth & Jones, but the evidence for which species is the main vector of EHDV is ambiguous [21][22][23][24].
Multiple methods can be employed to identify the source of arthropod blood meals, but, to date, genetic techniques have proven to be the most accurate when applied to Culicoides, e.g. [25,26]. In North America, three studies used molecular methods to identify the source of Culicoides blood meals [27][28][29]. However, these studies were conducted over limited geographic areas and the sample sizes were very small. Here, we expand our collective knowledge of Culicoides host choice by sequencing two mitochondrial DNA loci from bloodengorged Culicoides collected from three states, including one area where BTV is enzootic in both livestock and wildlife [30,31]. The information gained from this study will help illuminate the diversity of Culicoides hosts, which will help us to better understand pathogen transmission across the livestock/wildlife interface, in addition to advancing our understanding of the natural history of this ubiquitous, abundant, and economically important group of insects.

Field collection and species identification
Blood-engorged Culicoides individuals were captured opportunistically as part of surveillance efforts, or for other studies, between 2011 and 2014 from three regions in the USA. Insects were captured with CDC mini light traps (Bioquip, USA) or New Jersey light traps baited with ultraviolet light and CO 2 . Insects were stored on dry ice immediately after capture then stored at -80°C, or placed immediately in 99% ethanol. Culicoides were sorted from trap by-catch under stereomicroscopes on a cold plate (Bioquip, Rancho Domingo, California, USA). Species identification was first based on wing patterns following Wirth et al. [32]. For some subgenera and species groups, wing patterns are not species-diagnostic so we used combinations of morphological characters identified as diagnostic by Blanton & Wirth [33], Battle & Turner [34] and Holbrook et al. [35] to identify Culicoides species to the lowest possible taxon. We also included three blood-engorged individuals from the C. sonorensis USDA-ARS captivity colony to validate our blood meal identification assays. These individuals were fed domestic sheep (Ovis aries L.) blood.

Molecular species identification
Morphological identification of the species within the C. variipennis species complex is not always straightforward as there are few, if any, discrete synapomorphies that define the species. In parts of each species' range C. occidentalis, C. sonorensis and C. variipennis are sympatric. The females of the former two species are morphologically indistinguishable whereas the latter two species have differences in maxillary palp morphometrics which are not consistent as intraspecific variation leads to overlapping size ranges [35]. In an attempt to differentiate these sympatric sibling species, we sequenced 650 base pairs of the 5′ region of the mitochondrial cytochrome c oxidase subunit 1 gene (cox1). Currently no comprehensive database of the cox1 gene for North American Culicoides species exists so we developed our own cox1 sequence database that included 5 males of C. occidentalis from Lake County, California and two C. sonorensis females, one from the captive breeding colony maintained by the USDA-ARS in Manhattan, Kansas, and one collected along the front range of Northern Colorado, where C. sonorensis is allopatric. We also included three C. variipennis samples from three states, Indiana, Michigan and New Jersey, where C. variipennis is allopatric. These reference specimens were identified to species by slide mounting and following the key provided in Holbrook et al. [35].
Genomic DNA from individual blood-engorged females and their blood meals were extracted using the DNEasy blood and tissue kit (Qiagen, Hilden, Germany). Insects were extracted in a room dedicated to noninvasive and environmental DNA samples. Each specimen was placed in a sterile Petri dish and we separated the abdomen from the thorax with sterilized tools. The entire insect was placed in 300 μl of Qiagen buffer ATL. We then added 20 μl of proteinase K and incubated the samples overnight at 56°C. The following day the supernatant was transferred to a new 1.7 ml tube, leaving the exoskeleton of the insect in the initial tube. We added absolute ethanol to the exoskeleton and stored it at -20°C for slide-mounting for morphological species identification. We used an automated extraction protocol "Isolation of DNA from forensic casework samples part B (purification)" extraction protocol designed for the QIAmp DNA investigator kit (Qiagen, Germany) on the QIAcube (Qiagen, Germany) to extract DNA from the supernatant. The final elution volume was adjusted to 50 μl. An extraction blank containing only reagents was included with each extraction set to monitor for contamination. DNA extracts were stored at -20°C until further processing.
We used primers from two different studies to sequence 650 bp of cox1, BFculicFm1 [36] and C1-N-2191 [37]. The gene was amplified using PCR in 25 μl reactions with 0.25 mM of each dNTP, 2 mM MgCl 2 , 0.4 uM of each primer, 2 units of Amplitaq Gold polymerase (Thermo Fisher, Waltam, Massachusetts, USA) and 1 μl of DNA extract. The PCR program we used had an initial denaturing step at 95°C for 15 min then 40 cycles of 94°C for 30 s, annealing at 54°C for 30 s, extension at 72°C for 1 min, and then a final extension of 72°C for 5 min. Negative controls were included with each PCR to monitor contamination. PCR success was evaluated with 2% agarose gel stained with ethidium bromide. Successful reactions were cleaned up using ExoSAP-it (Affymetrix, Santa Clara, California, USA). Cycle sequencing was performed using BigDye v3.1 using the manufacturer's recommended thermocycler program but with an annealing temperature of 52°C. The cycle sequence products were purified with sephadex clean up using either PrepEase Sequencing Dye Clean-up Kit (Affymetrix) or 96-well filter plates (Whatman, Maidstone, UK). Sanger sequencing was performed on an Applied Biosystems Genetic Analyzer 3500xl (Life Technologies, USA).
Sequences were edited and aligned using CLUSTALW [38] implemented in GENEIOUS R7 (Biomatters, San Francisco, California, USA). We used a neighbor-joining (N-J) tree constructed in MEGA 6 [39,40] to identify genetic clusters in the C. variipennis species complex that were associated with reference samples. One sequence from C. reevesi collected in California was included as an outgroup for the N-J tree. The tree was built using an uncorrected p-distance and branch support was evaluated using 1000 bootstrap replicates [41,42].

Blood meal identification
To identify the origins of Culicoides blood meals, we followed Pettersson et al. [26] and used the general primers COI_short_F/COI_short_R and L14841/H15149 to amplify around 350 bp of vertebrate cytochrome c oxidase subunit 1 (cox1) and cytochrome b (cytb), respectively [43,44]. We first attempted to amplify cox1, but if this failed for samples we tried to amplify cytb. For cox1, the total reaction volume was 25 μl and contained 2.5 μl 10× buffer II, 1 mM of MgCl 2 , 0.1 mM of each dNTP, 0.4 mM of each primer, 2 units of Amplitaq Gold, and 2 μl of DNA extract. The thermocycler profile included an initial denaturing step at 95°C for 15 min then 45 cycles of 94°C for 30 s, annealing at 50°C for 45 s, extension at 72°C for 30 s and a final extension at 72°C for 10 min. The cytb gene was amplified in a 25 μl reaction that contained 2.5 μl 10× buffer II, 1.5 mM of MgCl 2 , 0.15 mM of each dNTP, 0.4 mM of each primer, 2 units of Amplitaq Gold and 2 μl of DNA extract. The cytb thermocycler profile was identical to the one used for cox1 but had an annealing temperature of 52°C. Negative controls were included with each PCR to monitor contamination. PCR success and sequencing methods were identical to those used for Culicoides species identification.
DNA sequences were edited and compared to GenBank via the BLAST function in GENEIOUS R7 (Biomatters, New Zealand), or to the Barcode of Life Database (BOLD; [45]) to determine species identification. We used a threshold of 98% to determine the species identification of vertebrate DNA sequences. If the sequences did not match to a species within this threshold on GenBank or BOLD then we identified it as the genus with the highest match. In some cases, known geographical distributions of hosts were used to assign the host to species (e.g. a sample identified as Odocoileus from California must be blacktailed deer (Odocoileus hemionus Rafinesque)). We tested for differences in blood meal sources between two groups identified as C. occidentalis and C. variipennis species complex using a χ 2 test.

Results
We collected a total of 366 blood-engorged specimens from three states, New York, South Carolina, and California (Table 1; Additional file 1: Table S1). We were able to identify the sources of the blood meals from 199 blood engorged Culicoides from three states, which represented six morphologically identified species and one species complex ( Table 1). The C. variipennis complex accounted for 183 of the 199 individuals (92%).

Culicoides species identification
We successfully sequenced cox1 for 139 of the 183 blood-engorged C. variipennis complex individuals. There were 54 unique haplotypes (including reference samples) that fell into three well supported genetic clusters (Fig. 1). Cluster A is the largest (Fig. 1), with branch support of 99%, contained 43 haplotypes that accounted for 120 individuals and this group also included morphologically confirmed C. occidentalis reference samples. The other two well-supported clusters, B and C, (Fig. 1) contained 8 haplotypes and 3 haplotypes, respectively. The 16 individuals in cluster B were from California and South Carolina, and the three individuals in cluster C were from California. The C. sonorensis and C. variipennis reference samples grouped into cluster B while no reference samples grouped into cluster C. All the individuals in clusters B and C were

Blood meal identification
Our approach to blood meal identification was successful in that we identified the blood source of the three captive C. sonorensis as domestic sheep, the known host species from which the blood was collected. We also successfully identified the sources of blood meals from199 wild-caught individuals (Table 1; Additional file 1: Table S1). Of the 199 successfully identified blood meals, 183 (92%) came from the C. variipennis species complex. Individuals identified as C. occidentalis using cox1 accounted for 120 of 183 C. variipennis complex blood meals. The other 63 blood meals were from C. variipennis complex individuals that were part of cluster B and C (n = 19), or for which we were unable to obtain molecular species ID (n = 44). From here on we refer to these individuals as the C. variipennis species complex. There were no statistical differences between the blood meal sources identified from C. occidentalis and C. variipennis species complex (χ 2 = 11.2, df = 8, P = 0.192).
Most of the C. occidentalis and C. variipennis species complex blood meals were from large ungulates and these included both wild species, black-tailed deer (n = 72; 39%), 5 domestic mammals (n = 80; 44%), and swine (Sus scrofa L.; n = 3; 2%) which could be either feral or domestic ( Table 1). Eleven of the 183 (6%) were from black-tailed jackrabbit (Lepus californicus Gray) and 11 (6%) from domestic dog (Canis familiaris L.). The remaining six individuals (3%) fed upon emu (Dromaius novaehollandiae Latham). Sixteen of the 199 (8%) identified blood meals were from Culicoides species not in the C. variipennis species complex: C. bigutattus (Coquillett), C. crepuscularis Malloch, C. reevesi Wirth, C. stellifer (Coquillett), and C. utahensis Fox. All of the non-variipennis blood meals were from deer (Odocoileus spp.), except one C. crepuscularis that fed on a house finch (Haemorhous mexicanus (Müller)). We also detected human DNA in some of the samples (not included in the sample size of 199). These samples were excluded from further analysis as determining whether it was a blood meal taken from a human or contamination was not possible. Despite using both genetic loci, we only had a 54% success rate which was due to improper storage conditions for one batch of insects that resulted in DNA degradation in the majority of samples.

Discussion
The main finding from this study was that North American Culicoides host choice was broader than previously understood. Through the use of genetic approaches, we identified a diversity of known hosts, including some that are highly susceptible to infection with BTV and EHDV, and expanded the catalog of hosts through detection of previously undocumented vertebrate host species. Our results represent the most comprehensive dataset to date regarding Nearctic Culicoides host choice. The value of these data cannot be understated as one of the most important shortcomings in our knowledge of Culicoides biology is the lack of data regarding trophic behavior and host selection across diverse habitats [23,46].

Culicoides species identification
The results from molecular identification of the three species in the C. variipennis species complex suggested that identifying species based on sampling location and breeding habitat is not as straight-forward as previously suggested, e.g. [35]. Females collected for this study from California were initially identified as C. sonorensis or C. occidentalis based on sampling location (BMR, unpublished data). However, the molecular results identified almost all of the samples as C. occidentalis (n = 120). Thus, the approach suggested in Holbrook et al. [35] using location or breeding habitat did not work in this case but it does suggest that C. sonorensis is less abundant in Lake County, California than previously thought. Considering that this region has a high rate of BTV and EHDV transmission [31], and given the high occurrence of C. occidentalis, perhaps C. occidentalis should be investigated further as a potential vector. One other interesting point is that haplotypes from California and South Carolina clustered closely together and with the C. variipennis and C. sonorensis reference samples (Fig. 1,  Cluster B). This lack of genetic differentiation between C. sonorensis and C. variipennis brings into question the utility of cox1 to discriminate between these two species, the diagnostic morphological traits, the range estimations, or the current taxonomy. To clarify the molecular taxonomy a multi-locus phylogenetic study is required to help refine the evolutionary relationships, distribution, and ecology of the C. variipennis species complex [24] as this study was not designed to answer such questions.
The high occurrence of C. occidentalis in our California sampling area was surprising as C. sonorensis is thought to be more broadly distributed and abundant [35,47]. There are explanations for the discrepancy between our data and those collected previously. The first is that perhaps cox1 is not a good genetic locus for species identification. However, a more thorough sampling across the US has demonstrated that suggests cox1 can easily discriminate between C. occidentalis and the two other species in the complex [24]. Another explanation could be that C. sonorensis and C. occidentalis hybridize and the results herein are due to mitochondrial DNA introgression. Velten & Mullens [48] were able to successfully hybridize these two species in the laboratory and the cross of C. sonorensis male and C.
occidentalis female had high reproductive success rates. Thus, if these two species hybridize in the wild then C. occidentalis mtDNA haplotypes could be present in C. sonorensis populations due to maternal inheritance. However, previous allozyme studies have not found natural hybrids in zones of sympatry in California [35,47]. Despite some ambiguous species identifications, this study, along with a similar study of Palearctic Culicoides spp. Pettersson et al. [26], demonstrated the importance of combining molecular species identification of pathogen vectors to improve accuracy of associating blood meals with the correct species.

Blood meal identification
Documenting the diversity of hosts in a particular area, including those hosts that have been catalogued previously, can help understand the influence of host availability and density on Culicoides host choice and pathogen transmission [5]. We identified the source of blood meals from 199 Culicoides that represented 12 different host species (Table 1, Fig. 1). Four of these mammalian species are susceptible to Culicoides-borne pathogens and have been previously identified as hosts for the C. variipennis species complex including: horse (Equus caballus L.), cattle (Bos taurus L.), domestic sheep and white-tailed deer (O. virginianus (Zimmerman)) [21,[49][50][51]. The other six mammalian species detected: domestic dog, domestic goat (Capra hircus L.), donkey (E. asinus L.), black-tailed jackrabbit, swine, and black-tailed deer, have not been documented as hosts of the C. variipennis complex until this study. The detection of blood meals from hosts that are susceptible to infection with BTV and EHDV is further evidence of the role that the C. variipennis species complex plays in pathogen transmission across the US and demonstrates the presence of potential reservoir species.
The results presented here, combined with previous findings, confirmed that the C. variipennis species complex will feed upon mammals other than large ungulates, indicating that they are opportunistic when choosing host species for blood meals. Tempelis & Nelson [29] used bait animals and immunological approaches to identify blood meals of the C. variipennis species complex in California, but could only distinguish hosts to taxonomic family (Bovidae and Leporidae). Our data substantiated their findings as we found 31 individuals had fed upon cattle (Bovidae), and we documented, for the first time, that this species complex feeds on blacktailed jackrabbits (Leporidae; Table 1). We were also able to document the first occurrence of the C. variipennis species complex feeding on domestic dog. Alexander et al. [52] reported that African carnivores can be infected with BTV. Our confirmation of Culicoides feeding upon a canid species and a wild leporid suggest that carnivores and other small mammals should be evaluated as potential contributors to the transmission cycles of BTV and EHDV.
The data we collected for this study provide the first evidence that the C. variipennis complex will take blood meals from black-tailed deer. Epidemiological data suggest that C. variipennis will bite black-tailed deer as Roug et al. [31] detected considerable numbers of blacktailed deer in northern California that were seropositive for both BTV and EHDV. Free movement of BTV between deer and livestock in the same area has been documented when large numbers of C. variipennis species complex individuals are present [30]. One remarkable point is that black-tailed deer are often asymptomatic when infected with BTV and EHDV such that detecting outbreaks in this host species can be challenging [53]. Asymptomatic hosts can facilitate movement of pathogens across the landscape, and may even facilitate overwintering of the viruses by maintaining viral infections through the months when vectors are not present. The combination of blood meal and epidemiological data provides additional evidence that black-tailed deer can contribute to maintenance of the BTV enzootic cycle in northern California.
The first evidence of the C. variipennis species complex feeding on a bird was discovered in this study (Table 1). Emus are not native to North America but there are commercial farms and hobbyists, including in northern California, that raise the species for meat or other purposes (California Department of Agriculture, personal communication). Our detection of Culicoides blood meals from emus was unexpected and these results counter the assumption that the species in the C. variipennis complex feed exclusively on mammals. In fact, considering the size of emus, one could hypothesize that host choice in the C. variipennis species complex is not necessarily driven by phylogeny but rather by factors associated with body size, such as the larger emissions of CO 2 or volatile compounds. This hypothesis has been tested with mosquitoes and other Culicoides species and the strict mammal versus bird hypothesis was rejected in favor of body size-driven host choice [5,54,55]. One important point from these findings is that emus can be infected with avian blood parasites [56] and, when emus are imported, they should be screened for infection to limit the chances of introduction and potential transmission of foreign blood parasite species to North American poultry and wild birds through Culicoides bites [57].
We obtained blood meal identification for six other Culicoides species not in the C. variipennis complex (Table 1). Culicoides biguttatus, a common species east of the Mississippi River [33], had fed on white-tailed deer, which confirms previous findings [28,[58][59][60]. In fact, Smith et al. [60] documented C. biguttatus feeding on white-tailed deer in an area that is enzootic for both BTV and EHDV pointing to a potential role of this species as a vector. Culicoides stellifer is a common species across continental North America [33] that we also found had fed on white-tailed deer. This species has been documented as feeding on large ungulates and is suspected as playing a role in the transmission of EHDV [23,58,60,61]. We collected one C. crepuscularis, a species that ranges across the continental US, and this individual had fed upon a house finch. This Culicoides species is known to prefer avian species and is also a vector for nematodes and avian blood parasites [62][63][64]. Culicoides utahensis is found only in the western US and is known to use deer as a blood meal source [65], which we confirmed. Culicoides reevesi is distributed in the southwestern US and, to date, has only been recorded feeding on humans [66]. We have added to the knowledge of this species by presenting the first documented case of it feeding upon black-tailed deer.

Conclusion
Vector-borne pathogens are significant threats to the health and welfare of both humans and animals worldwide [67]. More investigations into the ecology of all Culicoides species, not just the currently known vectors, will lead to better predictions of disease outbreaks and their attendant negative impacts on animal and human populations. In the face of climate change, the distributions of vector species and pathogens are expected to change [68,69]. The potential exists for Culicoides-borne pathogens restricted to warmer parts of the world to move north or south leading to devastating outbreaks [70][71][72]. With the continuous threat of the introduction of foreign pathogens, we must build a solid foundation of knowledge regarding Culicoides host choice so that we can develop better models to predict transmission and to develop methods to study and manage Culicoides and the pathogens they transmit. Our dataset is the most comprehensive blood meal analysis for North American Culicoides to date. One shortcoming of this study is that we used Sanger sequencing which needs high quality DNA and cannot identify mixed blood meals from multiple sources [4]. Applying next-generation sequencing technology could improve the success and sensitivity of Culicoides blood meal identification [73]. Nevertheless, we have demonstrated that molecular tools are important for gaining a solid knowledge base and developing methods to address some of the most pressing needs for managing Culicoides-borne diseases.

Additional file
Additional file 1: Table S1. Summary of data for blood-engorged Culicoides species collected in North America. Provided are the collection locality information (State, county, latitude and longitude), year collected (year), Culicoides species, sample size at each locality (n), and identification of host using DNA-based blood meal analysis (Host ID). The host species are: black-tailed deer (Odocoileus hemionus); black-tailed jackrabbit (Lepus californicus); cattle (Bos taurus); domestic dog (Canis familiaris); domestic goat (Capra hircus); domestic sheep (Ovis aries); donkey (Equus asinus); horse (Equus. caballus); swine (Sus scofa); white-tailed deer (O. virginianus); emu (Dromaius novaehollandiae); house finch (Haemorhous mexicanus). We also included the cluster in the neighbor-joining tree that contained haplotypes from individuals at each location. If we could not obtain genetic information from individuals at that location this was also included. (XLSX 15 kb) Abbreviations BTV: Bluetongue virus; EHDV: Epizootic hemorrhagic disease virus