Association between host species choice and morphological characters of main sensory structures of Culicoides in the Palaeartic region

Culicoides (Diptera: Ceratopogonidae) serve as vectors of several mammalian and avian diseases, including bluetongue, Schmallenberg, African horse sickness, avian malaria and Oropouche. Host preference investigations are necessary to assess the transmission routes of vector-borne diseases and to inform mitigation strategies. A recent study examining the main sensory structures (palps and antennae) of Culicoides species suggests that they be classified as ornithophilic or mammalophilic according to their feeding habits. We analyzed Culicoides host preferences according to the literature and carried out a multiple correspondence analysis linking these preferences with morphological data. Seven out of 12 variables were found to be reliable predictors of host preference in Culicoides species: Antenna Flagellomer-Sensilla Coeloconica-Number: (7–10) and (11–13); Antenna Flagellomer-Sensilla Coeloconica IV–X: presence; Palpus-size: wide and/or narrow opening and shallow pit; Palpus-Shape: strongly swollen; Antenna-Short sensilla trichodea-distal part segment IV to X-Number: 2 seta each. Our results demonstrate that the presence of sensilla coeloconica and the maxillary palpus can be used to separate ornithophilic and mammalophilic or ornithophilic/mammalophilic species.


INTRODUCTION
Vector-borne diseases are health problems for humans, livestock, and wild animals and are transmitted by a variety of arthropods. Culicoides species constitute a diverse and widespread genus with more than 1,400 species world-wide (Borkent, 2014). Biting midges transmit multiple avian and mammalian diseases, including Bluetongue virus (BTV), Schmallenberg virus (SV), epizootic haemorrhagic disease virus (EHDV), African horse sickness virus (AHSV) and avian Haemoproteus (subgenus Parahaemoproteus) parasites.
Direct collection from animals has been considered the most reliable method to study the vector/host ratio (Silver & Service, 2008), which is an essential parameter for the modeling of vectorial capacity and virus transmission (Garrett-Jones, 1964). Various factors, such as habitat type, season, and bait species, contribute to the capture success of engorged females when using light traps. Engorged biting midges can be either fully engorged or have partially digested blood meals. Only fully engorged females were considered for blood meal identification; and the percentage of the engorged females using UV traps was low (Martínez-de la Puente, Figuerola &Soriguer, 2015). The percentage varied from 0.97% to 27.7% with three studies presenting a percentage of engorged females greater than 10% and seven studies presenting a percentage of less than 5% (Bartsch et al., 2009;Votypka, Synek & Svobodova, 2009;Lassen et al., 2011;Lassen, Nielsen & Kristensen, 2012;Garros et al., 2011;Ninio et al., 2011;Martínez-de la Puente et al., 2012;Santiago-Alarcon et al., 2012;Pettersson et al., 2013;Slama et al., 2015;Hadj-Henni et al., 2015;Bobeva et al., 2015).
Culicoides species are mainly mammalophilic and/or ornithophilic but females have also been found to occasionally feed on engorged insects (Ma et al., 2013). Some species of Forcipomyia and Leptoconops (Ceratopogonidae) feed on reptiles and frogs (Borkent, 2005). Hematophagous insects have highly developed olfactory systems and mainly use their antennae and, in some cases, maxillary palps, to detect semiochemicals. Semiochemicals can provide information about the location, suitability, or physiological state of conspecifics, hosts, or breeding sites (Logan & Birkett, 2007). Moreover, several studies carried out on feeding patterns of biting midges found variation in host attractiveness to be correlated with exhaled carbon dioxide (CO 2 ), 1-octen-3-ol, lactic acid, acetone (Zimmer et al., 2015), specific phenolic compounds emitted from urine, (Bhasin, Luntz & Mordue, 2001) or hair fragrance (Mands, Kline & Blackwell, 2004).
Consequently, the morphological characterization of the Culicoides sensory structures can serve as an indirect method to assess host preference (Jamnback, 1965;Braverman & Hulley, 1979;McKeever, Hagan & Wang, 1994;Blackwell, Mordue & Mordue, 1994). Here, we investigate how differences in morphological characters of the sensory structures of female Culicoides may impact host feeding choice. Specific objectives included morphological analysis of the main sensory structures previously explored by Blackwell (2004), Braverman et al. (2012 and Talavera et al. (2015) using host species published in the literature (engorged females and animal baits). We analyzed whether a combination of morphological variables could be used to predict host preference. Table 1 summarizes host species of biting females of species of Culicoides identified using molecular methods and animal baits. In order to exclude variability of morphological characters, we use an Interactive Identification Key for Culicoides (Mathieu et al., 2012). The raw dataset included 12 morphological characters ( : slender or slightly swollen, triangular and moderately swollen, [3]: lender or slightly swollen; (10) Cibarial-Armature (H04); (11) Pharynx posterior-Armature-(H05); (12) Eyes-Inter Ocular-Space-Shape (H02). For the size of the maxillary palpus, only a single sensory pit was used. However, if a Culicoides specimen had multiple irregular pits then we classified them as a small opening. All specimens in this study present a sensilla coeloconica in segment III. Therefore, a new group (Segment-IV-X) has been drawn according to Talavera et al. (2015). Finally, Culicoides species were classified into ornithophilic (O) and mammalophilic (M) or ornithophilic/mammalophilic (O, M) according to their host species (Table 1).

MATERIALS & METHODS
For the statistical analysis, the morphological characteristics and species classification are coded as qualitative variables (see Data S1). Therefore, quantitative methods such as the Principal Coordinates Analysis (PCO) or Nonmetric Multidimensional Scaling (NMDS) are not applicable here. Multiple correspondence analysis (MCA) is a data analysis technique for qualitative variables (Greenacre & Jörg, 2006), to obtain maps showing the distances between the qualitative variables and the observations. MCA was performed with XLSTAT software and used to explore the correlation structure between morphological characteristics and host preference. Within the indicator matrix, the rows represented individuals and the columns represented categories of the variables. Correspondence      Lassen et al. (2011), Lassen, Nielsen & Kristensen (2012, Ninio et al. (2011), Pettersson et al. (2013) Capreolus capreolus COI-Cytb b , Cytb c Lassen et al. (2011), Lassen, Nielsen & Kristensen (2012 Anas platyrhynchos, Columba palumbus COI-Cytb b Lassen et al. (2011) Equus caballus PNOC, Cytb, Cytb c , COI, COI-Cyt d Ninio et al. (2011), Lassen, Nielsen & Kristensen (2012, Santiago-Alarcon et al. (2012), Pettersson et al. (2013), Hadj-Henni et al. (2015) Ovis aries PNOC, COI-Cyt d Ninio et al. (2011), Pettersson et al. (2013) Oryctolagus cuniculus, Sus scrofa   analysis was applied through the symmetric matrix of all two-way cross-tabulations, to present the indicator matrix in a low-dimensional Euclidean space. The first axis was found to be the most important dimension, the second axis the second most important, and so on, in terms of the amount of variance accounted for. All Culicoides species (Data S1) were first coded into a 12, 10 or 7 morphological characteristics indicator matrix and analyzed by MCA to obtain the discriminant factors. All species were then projected in order to indicate the species preferences. The MCA map showed that the inertia for the two first dimensions is >70%. So, 3 groups were obtained into three ellipses, based on the F1 and F2 axis. The ellipses were built with the average of each group ± 1 SD (standard deviation), includes about 68% of the observations. A student test was used to compare the several categories (M vs. M/O and O vs. M/O).

RESULTS
Overall, 53 species were investigated. Five species (C. duddingstoni, C. minutissimus, C. reconditus, C. salinarius and C. truncorum) are ornithophilic, 27 are mammalophilic and 21 are ornithophilic/mammalophilic (Fig. 1). Multiple correspondence analysis locates all the categories in a Euclidean space. The MCA map showed that the inertia for the two first dimensions is ranging to 74% and 90% (Fig. 2). The first dimension explained more of 59% of data variability, and the categories are mainly organized along this axis. In the Fig.  2, each point (red, green, blue) corresponds to a Culicoides species category and several species could be plotted in the same point.

DISCUSSION
Some aspects of the epidemiology of vector-borne diseases are linked to the host preferences and feeding behaviors of vector arthropods. This study investigates whether main sensory structures of female Culicoides are correlated to host species feeding preferences. Our results demonstrate that the presence of sensilla coeloconica and the maxillary palpus can be used to separate ornithophilic and mammalophilic or ornithophilic/mammalophilic species as previously reported on five species by Isberg, Hillbur & Ignell (2013). Talavera et al. (2015), proposed to use only four morphological characters to predict Culicoides host preference based on Blackwell (2004), without statistical analysis. In the present study, seven characters are sufficient to assess host preference including the four parameters of Talavera et al. (2015). Interestingly, our results with four morphological characteristics (Fig. 2C) separate the three groups but a lot of Culicoides species are clustered in the same point compared to seven or 10 parameters (Figs. 2A, 2B).
Previous studies have suggested a relationship between the number of sensilla and host preference (Braverman & Hulley, 1979;Isberg, Hillbur & Ignell, 2013). The number of short blunt-tipped sensilla trichodea, sensilla coeloconica and sensilla basiconica are significantly higher in the ornithophilic Culicoides festivipennis compared with the mammalophilic C. obsoletus and C. scoticus (Isberg, Hillbur & Ignell, 2013). In our study, we are unable to classify species having a higher number of sensilla trichodea and sensilla coeloconica as ornithophilic, mammalophilic or both. The ornithophilic species show a number of sensilla coeloconica ranging from eight to 13. The morphological sensillum types of antenna and host preference were not associated with their phylogenetic relationship (Isberg, Hillbur & Ignell, 2013) but rather with volatile organic compounds, captured by different receptors present on sensillum types (Zimmer et al., 2015).
The morphological characters of the main sensory structures of Culicoides and their host preferences are not linked with their breeding sites for 13, 14 and 34 of Culicoides studied species (Kettle & Lawson, 1952;Zimmer, Haubruge & Francis, 2014;Zimmer et al., 2014). Culicoides larvae develop in a wide range of wet substrates. Each species has its own requirements; therefore, larval micro-habitats are generally species-specific, even if species associations are regularly observed (Zimmer, Haubruge & Francis, 2014;Zimmer et al., 2014). The Culicoides species studied are recorded in the same substrates.
Finally, our study, based on 10 and seven characteristics, confirms the empirical classification of Culicoides into ornithophilic and mammalophilic or ornithophilic/mammalophilic, whereas, Talavera et al. (2015)'s analysis was based only on four morphological characters.

ADDITIONAL INFORMATION AND DECLARATIONS Funding
The authors received no funding for this work.