“Show me which parasites you carry and I will tell you what you eat”, or how to infer the trophic behavior of hematophagous arthropods feeding on wildlife

Abstract Most emerging infectious diseases are zoonoses originating from wildlife among which vector‐borne diseases constitute a major risk for global human health. Understanding the transmission routes of mosquito‐borne pathogens in wildlife crucially depends on recording mosquito blood‐feeding patterns. During an extensive longitudinal survey to study sylvatic anophelines in two wildlife reserves in Gabon, we collected 2,415 mosquitoes of which only 0.3% were blood‐fed. The molecular analysis of the blood meals contained in guts indicated that all the engorged mosquitoes fed on wild ungulates. This direct approach gave only limited insights into the trophic behavior of the captured mosquitoes. Therefore, we developed a complementary indirect approach that exploits the occurrence of natural infections by host‐specific haemosporidian parasites to infer Anopheles trophic behavior. This method showed that 74 infected individuals carried parasites of great apes (58%), ungulates (30%), rodents (11%) and bats (1%). Accordingly, on the basis of haemosporidian host specificity, we could infer different feeding patterns. Some mosquito species had a restricted host range (An. nili only fed on rodents, whereas An. carnevalei, An. coustani, An. obscurus, and An. paludis only fed on wild ungulates). Other species had a wider host range (An. gabonensis could feed on rodents and wild ungulates, whereas An. moucheti and An. vinckei bit rodents, wild ungulates and great apes). An. marshallii was the species with the largest host range (rodents, wild ungulates, great apes, and bats). The indirect method substantially increased the information that could be extracted from the sample by providing details about host‐feeding patterns of all the mosquito species collected (both fed and unfed). Molecular sequences of hematophagous arthropods and their parasites will be increasingly available in the future; exploitation of such data with the approach we propose here should provide key insights into the feeding patterns of vectors and the ecology of vector‐borne diseases.

. Exploitation of information derived from host-specific parasites was difficult in the past because few nonhuman Plasmodium spp. were known before the second half of the twentieth century (Roeder & Anderson, 1990), including species infecting wild mammals such as rodents, bats, ungulates, monkeys, and great apes (Garnham & Heisch, 1953;Thurber et al., 2013). The recent development of molecular screening techniques of host blood and feces, however, has dramatically increased the number of species described, resulting in a plethora of new Plasmodium spp., among which those infecting the African great apes (Liu et al., 2010;Prugnolle et al., 2010;Rayner, Liu, Peeters, Sharp, & Hahn, 2011).
Recently, we reported the results of a longitudinal survey carried out in the rainforest of Gabon with the aim of identifying mosquitoes involved in the transmission of great ape malaria parasites , as well as of other haemosporidians infecting wildlife . During this survey, more than 2,000 individual anophelines were collected, among which only seven were engorged with blood, and 74 were infected with haemosporidian parasites. Here, we used this dataset to gain insights about mosquito trophic behavior, as a proof of concept of such indirect approach. Specifically, we first analyzed the origin of the blood meals and then inferred the blood-feeding behavior of unfed mosquitoes from the haemosporidians they were carrying.

| METHODS
The sequencing data concerning the haemosporidian parasite screens have already been described in Makanga et al., 2016 andBoundenga et al., 2016; . Therefore, here we provide only some general information about the collection sites and the molecular approach used to obtain such data.

| Mosquito collections and identification
Anopheles mosquitoes were collected in two wildlife reserves in Gabon (the Lopé National Park and the private game reserve of La Lékédi) that host large natural populations of mammals, including great apes (gorilla, chimpanzee), monkeys (e.g. mandrill, several species of Cercopithecus and Colobus), ungulates (e.g. red river hog, buffalo, duiker, sitatunga), rodents, and bats. Mosquitoes were sampled using CO 2 -baited CDC light-traps positioned in several sites of each reserve corresponding to dense equatorial forest or to forest patches in a forest/savanna mosaic. Traps were operated between 5 p.m. and 7 a.m. from October 2012 to December 2013, totaling 1,620 trap-nights (see Makanga et al., 2016 for details). Sampled arthropods were killed at −20°C during 1 hr and were then observed under a Leica M80 binocular to (1) identify and isolate Anopheles specimens using taxonomic keys (Gillies & Coetzee, 1987), and (2) to detect female mosquitoes engorged with blood. All female mosquitoes were then individually stored at −80°C until they were processed for further molecular analyses.

| Blood meal analysis
Host DNA was extracted from the blood contained in female mosquito guts using the Qiagen DNeasy Blood and Tissue kit . PCRs to amplify a 415-bp fragment of the host cytochrome b gene (Cyt-b) were carried out in 50 μL reaction volumes containing 5 μL of 10× reaction buffer (Qiagen, Germany), 3 μL of 25 mmol/L MgCl 2 , 1 μL of 10 μmol/L of each primer [Cyt-b-(f) and Cyt-b-(r) primers] (Townzen et al., 2008), 1 μL of 10 μmol/L dNTPs, 0.3 μL of DNA polymerase (Qiagen, Germany) and 3 μL of template DNA using a GeneAmp 9700 thermal Cycler (Applied Biosystems, USA) under the following cycling conditions: 95°C for 1 min; 35 cycles at 95°C for 30 s, 52°C for 50 s, 72°C for 1 min, and final extension at 72°C for 5 min. PCR-amplified products (10 μL) were run on 1.5% agarose gels in 1× TBE buffer for quality control and then sent to Beckman Coulter Genomics (France) for sequencing both strands after purification. Nucleotide sequences were edited and aligned using BioEdit 7.0.9.0 (Hall, 1999) and compared with homologous host sequences contained in GenBank using the basic local alignment search tool (BLAST; http://www.ncbi.nlm.nih.gov) (Altschul et al., 1997). This allowed determining the vertebrate identity of the blood DNA samples. The closest related sequences (all corresponding to wild ruminants with similarity >98%) (Table S1) were used to construct a phylogenetic tree using maximum likelihood (ML) in PhyML v. 3.0 (Guindon et al., 2010), available at the ATGC bioinformatics platform (http://www.atgc-montpellier.fr/). The maximum-likelihood tree and corresponding bootstrap support values were obtained with PhyML using NNI (nearest neighbor interchange) + SPR (subtree pruning regrafting) branch swapping and 100 bootstrap replicates.

| Haemosporidian parasite screening
The presence of haemosporidian parasites in the anopheline samples was detected as described previously Makanga et al., 2016). Briefly, total DNA was extracted from the mosquito bodies and salivary glands with the Qiagen DNeasy Blood and Tissue kit . A nested PCR procedure (Ollomo et al., 2009) was performed using individual DNA templates to detect the parasites and to amplify a 950-bp portion of their Cyt-b gene.
The PCR products of infected mosquitoes were then sequenced as described above. Gene sequences were edited and aligned to published sequences (Table S3) using BioEdit and assigned to known haemosporidian species using maximum likelihood (ML) in PhyML (using ATGC platform). As above, the maximum-likelihood tree and corresponding bootstrap support values were obtained using NNI (nearest neighbor interchange) + SPR (subtree pruning regrafting) branch swapping and 100 bootstrap replicates. For each individual infection, we deduced the vertebrate host from which the parasite was acquired during a blood meal based on its natural host range.

| Blood meal analysis of wild caught Anopheles
Among the 2,415 female mosquitoes collected, only seven (0.3%) were engorged with blood, indicating that the baited trapping technique biased the sample toward unfed blood-seeking mosquitoes.
The Cyt-b sequence-based phylogenetic analysis showed that all seven blood meals (in red in Figure 1)

| Inference of mosquito host-feeding behavior from haemosporidian infections
Toward this aim, the 2,415 female Anopheles collected during the survey were screened to identify the presence of haemosporidian parasites.
Haemosporida were detected in 74 females (~3% of the sample), and most of them corresponded to parasites known to infect mammalian hosts, in agreement with the mammalophilic feeding behavior typical of the genus Anopheles (Bruce-Chwatt, Garrett-Jones, & Weitz, 1966).
Specifically, they were parasites of African great apes (58%), ungulates (30%), rodents (11%) and bats (1%) (Figure 2a,b, Table S2). Infection rates varied greatly according to the mosquito species and the host group from which the parasites were acquired (Table 1). DNA detection of the parasite infective stages in appropriate tissues (i.e. sporozoites in mosquito salivary glands) was carried out only for a small number of Anopheles-Haemosporida species pairs, thus limiting the scope of inference about the vector role of each mosquito species for these parasites. Nevertheless, the identification of parasite DNA in mosquito tissues necessarily implies that the parasite was acquired from its natural host in the course of a blood meal. The presence of haemosporidians in mosquitoes, therefore, indirectly inform about the trophic behavior of each mosquito species, because most Haemosporida infect a single vertebrate host species or groups of vertebrate species that are taxonomically related. For instance, the rodent Grammomys poensis (formerly known as Thamnonys rutilans) is the only recognized host for both Plasmodium yoelii and P. vinckei (Stephens, Culleton, & Lamb, 2012 Table S2) should have fed on bats because bats are the only known host of parasites of this genus (Duval et al., 2012;Schaer et al., 2013).
The same type of argument can be applied in the case of mosquitoes infected with Haemosporida whose host are wild ungulates or great apes.   (Haddow, Gillett, & Highton, 1947), or even when dispersing over open ground (Gillies & Wilkes, 1974).
These conclusions should be considered with caution because mosquito host-feeding patterns depend on many interacting factors, such as inherent host preferences and their modulation by endogenous physiological or exogenous environmental factors, such as the host relative abundance and accessibility (Takken & Verhulst, 2012

| CONCLUSIONS
This study improves our knowledge of the trophic behavior of Anopheles mosquitoes living in the pristine rainforest of Central Africa and feeding on wildlife, providing valuable information to understand the transmission pathways of nonhuman Haemosporida. Difficulties with sampling blood-fed Anopheles in this ecological context were at least partially overcome by exploiting information about mosquito-infecting parasites to infer vector-host interactions. Compared with the conventional analysis of blood-engorged mosquitoes, this method increased by more than 10-fold the number of informative mosquito specimens.
The method we propose relies on parasite high specificity in the vertebrate host to determine mosquito blood-feeding patterns. At the other end of the scale, parasites that are not host-specific, as it is often the case for enzootic arboviruses (e.g. Flavivirus or Alpha virus), are uninformative. However, characterization of food networks from mosquito blood-feeding patterns can serve to predict cross-species transfer of pathogens showing such lower host specificity. In this context, mosquitoes with opportunistic feeding patterns are of a particular interest because they are able to bridge nonspecific pathogens toward new hosts, including humans. The approach presented here using mosquitoes and haemosporidians as a model can be suitable for other hematophagous arthropods and/or other pathogens infecting wildlife, provided that sufficient information about host specificity is available. Molecular sequences of hematophagous arthropods and their parasites generated from new high-throughput technologies will be increasingly available in the future; exploitation of such data with the approach we propose here should provide key insights into the feeding patterns of vectors and the ecology of vector-borne diseases.