Blood meal survey reveals insights into mosquito-borne diseases on the island of Santiago, Cape Verde

Introduction The transmission of pathogens by blood-sucking insects to humans and other animals depends on vector-host interactions. As yet unexplored in Cape Verde, mosquito feeding behavior plays a crucial role in pathogen transmission. Herein, we aim to explore, through blood meal analysis, the relationship between mosquito species and common hosts in Santiago Island, Cape Verde. Methods Engorged female mosquitoes were collected through mechanical aspiration from May 2016 to December 2017 in three municipalities of Santiago Island (Praia, Santa Cruz, and Santa Catarina). Blood-feeding behavior in each municipality was assessed through blood meal analysis using an enzyme-linked immunosorbent assay (ELISA). Results We were able to determine that single-host blood meals were common in Aedes aegypti, Anopheles arabiensis, and Culex pipiens sensu lato (s.l.). In general, the mosquitoes preferred to feed on humans, dogs, and chickens, and on multiple hosts, mainly two hosts. The human blood index (HBI) was highest (i.e., 1.00) in Ae. aegypti, with the lowest value (0.40) observed in An. arabiensis. It was observed that, among single-host blood meals, the likelihood of Cx. pipiens s.l. feeding on humans was significantly high, whereas the likelihood of An. arabiensis feeding on humans was significantly low (log-odds ratio (LOR) = 0.85 and –2.44, respectively). In addition, a high likelihood of Ae. aegypti feeding on humans was observed, but this was not statistically significant (LOR = 0.85). Discussion Overall, our findings demonstrate a lack of feeding preference in Culex pipiens s.l. compared with Ae. aegypti and An. arabiensis. These results provide insights into possible parasite transmission and pathogen spillover/spillback, which threaten human/animal health and the economy in Cape Verde.


Introduction
Vector-host interactions play a key role in pathogen transmission. More than half of the world's population lives in areas at risk of one or more major vector-borne diseases, with mosquito-borne diseases being a major contributor (1). Mosquitoes are known to be vectors of diseases that are devastating to humans, domestic animals, and wildlife (2).
The ability to precisely identify the source of blood meal in mosquitoes and to distinguish anthropophilic and zoophilic species is of prime importance for deciphering host choice and how parasites manipulate the host to promote infection and transmission, and for understanding disease epidemiology for improved entomological surveillance (2,(26)(27)(28)(29).
Nevertheless, mosquito bites and diseases may be preventable by decreasing contact with humans through zooprophylaxis, by the use of repellant air-sprays and house improvements, and, at an individual level, by covering the skin and using insecticide-treated bed nets (1,6,23,(41)(42)(43)(44)(45)(46)(47). Thus, integrated vector management (IVM) with multiple strategies, included as part of the One Health Initiative, has greater potential to improve vector-borne disease control and facilitate elimination than isolated or routine approaches (48,49). Understanding the temporal and spatial dynamics of the vectorhost relationships is a key element in planning IVM.
Entomological studies in Cape Verde have focused more on mosquito density, distribution, and genetic and phenotypic composition than on feeding behaviors, although these are crucial for disease transmission. In our study, we investigated the blood feeding patterns of mosquitoes in three municipalities in Santiago Island: Praia, Santa Cruz, and Santa Catarina.

Ethics statement
All mosquito collections were carried out in private spaces (i.e., homes and their vicinity). We sought consent from owners before collecting the mosquitoes on their properties. The field collections did not involve any endangered or protected species. Materials used in the study posed no health risk to researchers or owners, and no vertebrate animals were harmed.

Sampling sites and collection
As part of vector surveillance activities, led by the National Institute of Public Health, Cape Verde, mosquitoes were collected from May 2016 to December 2017 in randomly selected areas in three municipalities of Santiago Island [Praia (12 areas), Santa Cruz (four areas), and Santa Catarina (two areas)] ( Figure 1). The selection was based on the history of mosquito-borne diseases, and mosquito and vertebrate composition. Both outdoor and indoor collections were carried out in the morning (7 to 11 a.m.). At each sampling site, mosquitoes were collected through 15 minutes of mechanical aspiration using a modified Centers for Disease Control and Prevention (CDC) backpack aspirator (model 1412; John W. Hock Company, Gainesville, FL, USA). All collected mosquitoes were transported alive in cups to the Medical Entomology Laboratory at the National Institute of Public Health, Cape Verde, and then killed by freezing.

Mosquito species identification
Mosquito specimens were separated according to blood digestion status (i.e., blood fed and unfed) following Reeves et al.'s recommendations (50). Only blood-fed mosquitoes were morphologically identified using dichotomous keys (39, 51). Therefore, mosquitoes with fresh blood and with an intermediary blood digestion status were used for the blood meal analysis. Each mosquito was separated into head, thorax, and abdomen. The abdomens were crushed on Whatman grade 1 qualitative filter paper (GE HealthCare Technologies Inc., Chicago, IL, USA) and preserved at -20°C for enzyme-linked immunosorbent assay (ELISA) analysis. The heads and thoraxes were individually preserved in silica gel (Merck KGaA, Darmstadt, Germany) for DNA extraction. Mosquitoes morphologically identified as Anopheles gambiae sensu lato (s.l.) were submitted to a single total DNA extraction using the NZY Tissue gDNA Isolation Kit (NZYTech genes & enzymes, Lisbon, Portugal). Sibling species of the An. gambiae complex were identified using PCR following the protocol described by Scott et al. (52).

Blood meal identification
Blood meal sources were identified through a direct ELISA adapted from Lardeaux et al. (53). Immunoglobulin G (IgG) from five vertebrate hosts of interest (human, pig, dog, goat, and chicken) was assessed. The choice of alternatives to human hosts was based on observations made during field collections at the sites. The absorbance values were obtained using an absorbance microplate reader (Stat Fax ® 4200) at 450 nm. The cut-off values for each plate were calculated from the mean of three negative controls chosen randomly, plus three times their standard deviation (SD) [i.e., cut-off = mean + (3×SD)].

Statistical analysis
The human blood index (HBI) was estimated for each group of mosquitoes from each municipality in accordance with the Garrett-Jones formula (54). The probabilities of single-host blood meals on humans for each mosquito species, by municipality, and their corresponding 95% confidence intervals (CIs), were calculated. In addition, as proposed elsewhere (55), log-odds ratios (LORs) were calculated from hypergeometrically distributed data on a 2 × 2 contingency table (95% confidence interval) between each mosquito species and host (where positive and negative LORs indicate, respectively, a positive and negative feeding association, meaning that the likelihood that a blood meal of that mosquito species would originate from a given host is, respectively, higher and a lower than random chance). The greater the LOR, the stronger the feeding association. LORs close to 0 suggest no association between the mosquito species and host. In addition, RStudio version 1.4.1717 (packages gglot2, ggdendrogram, and reshape2) was used to generate a heatmap of the LORs in hierarchical clustering, which assembled mosquito species with similar feeding patterns. The percentages of blood meals by host according to season, municipality, and land use were represented graphically using JMP Pro 16.1.0.
Only An. arabiensis took single-host blood meals from all five hosts (pigs, humans, goats, dogs, and chickens) especially in urban areas like Praia during the rainy season ( Figure 2). Cx. pipiens s.l. had a higher preference for human blood, followed by chicken, and this preference did not differ across seasons, municipalities, or land uses (Figure 3). Ae. aegypti also preferred to feed on humans but this preference did not differ across seasons and was not different in Praia and Santa Cruz, the two municipalities where it was possible to capture this mosquito species (Figure 4).
The LOR calculated only for single-host blood meals clustered together Cx. pipiens s.l. and Ae. aegypti with similar feeding patterns. These estimations revealed positive LORs for all the mosquito-host relationships; however, the Cx. pipiens s.l.-dogs and Cx. pipiens s.l.goats relationships show negative LORs, i.e., negative associations. This is similar to An. arabiensis-human host ( Figure 5).   Human + dog + chicken -- Human + dog + pig Human + dog + pig + chicken 3 Human + chicken 4 -4 1 -1 9 1 4 1 24 Human + pig + chicken Human + goat + dog + pig + chicken 2 Human + goat + pig -- Significant positive or negative associations were found for two out of the five hosts (humans and goats). There was a significant clustering positive LOR (LOR = 0.85, p = 0.01) for Cx. pipiens s.l. for human hosts and a strong negative association for goats (LOR = -4.38, p < 0.0001), meaning a higher and a lower likelihood of this mosquito species taking blood from humans and goats, respectively. Similarly, positive associations between Ae. aegypti and humans (LOR = 0.85, p = 0.10) and between Ae. aegypti and pigs (LOR = 2.44, p = 0.16) were observed, but these were not statistically significant. For An. arabiensis-humans a significant strong negative correlation (LOR = -2.44, p < 0.0001) was observed, which points to a lower likelihood of this mosquito species biting humans. A strong positive association was observed between An. arabiensis and goats (LOR = 5.30, p = 7.80) and between An. arabiensis and pigs (LOR = 3.51, p = 0.06), but these associations were not statistically significant. An LOR value closer to 0 was found only for Ae. Aegypti-dogs, which suggests no relationship (LOR = 0.02, p = 0.64) ( Figure 5).

Discussion
The transmission potential of a vector-borne pathogen by a competent vector is influenced to a large extent by blood feeding behavior, vector population density, and longevity (56). In our study, data gathered from multiple sites indicate that the major pathogen vectors (i.e., An. arabiensis, Ae. aegypti, and Cx. pipiens s.l.) share up to three hosts, although this may be influenced by host availability rather than by species-specific host choices. A limitation at this point is the fact that our data could be more accurate if PCR-based molecular approaches were used. Despite Praia being more urbanized, multiple-host blood meals (from four or five hosts) were also observed (Table 1). This observation is not new in mosquitoes, as shown in studies in Brazil, Honduras, and Peru (57-60).
A pertinent result that we found was the generic feeding behavior among An. arabiensis and Cx. pipiens s.l., in contrast to Ae. aegypti, which showed a specific feeding behavior with a strong predilection for human hosts (HBI single-host + multiple-host blood meals = 0.98; HBI single-TABLE 2 Probability of Aedes aegypti, Anopheles arabiensis, and Culex pipiens s.l. from each municipality of Santiago Island taking a single-host blood meal from a human host.

Municipality
No. of single-host blood meals w/human blood Single-host blood meals by Anopheles arabiensis according to season, municipality, and land use. *no mosquitoes captured.
host blood meals = 0.95). Work carried out in Grenada, the USA, Thailand, Puerto Rico, and Australia also found similar feeding behaviors in Ae. aegypti (61-67), as humans provide reproductive and metabolic advantages (68). This feeding behavior could be one of various factors that would explain the large number of cases of dengue fever in 2009 (34) and of Zika in 2015 (32) in Cape Verde, and this is also a point of consideration, as Ae. aegypti from Cape Verde is also able to transmit the Chikungunya virus (69). Few (n = 45) blood-fed An. arabiensis were caught, which limits conclusions on any blood meal analyses. These low numbers could be due to the collection technique employed (70), as we collected only mosquitoes that were within range of the suction effect of the aspirator. It is likely that the use of a collecting tool with wider coverage, such as pyrethrum spray catches, would have increased the numbers collected. In any case, in our study, An. arabiensis showed low anthropophily [HBI single-host + multiple-host blood meals = 0.55; HBI single-host blood meals = 0.41 (LOR = -2.44, p < 0.0001)], similar to the findings of studies in Burkina Faso, Kenya, and Ethiopia (71)(72)(73). We verified that An. Arabiensis fed from multiple hosts in the same gonotrophic cycle (from two to four hosts), which agrees with the gonotrophic discordance phenomenon previously described in other species of this genus in Ethiopia, Brazil, Kenya, Mexico, and Sri Lanka (73)(74)(75)(76)(77)(78). The low anthropophily observed in this species could also explain the relatively low prevalence of malaria in Cape Verde, in contrast to other countries in Africa, where the main malaria vector, An. gambiae s.s., is, essentially, anthropophilic (79).
The mosquito feeding behavior observed in our study demonstrates that it is important to consider the closeness of dogs to humans and their role in parasite transmission. Although humans are dead-end hosts for D. immitis and D. repens (both found in dogs in Cape Verde), earlier stages can cause dirofilariasis and inflammatory response when parasites die in human tissues (99). Ocular and pulmonary dirofilariasis with benign pulmonary and subcutaneous nodules that may be confused with cancer can also be found in humans (99)(100)(101)(102). In addition to Cx. pipiens s.l., Ae. aegypti has also been implicated in the transmission of both parasites, although we observed a low probability of this species biting dogs (Table 1, Figure 5) (87,103,104). Single-host blood meals by Aedes aegypti according to season, municipality, and land use. *no mosquito captured. Single-host blood meals by Culex pipiens s.l. according to season, municipality, and land use. It is possible that our results could be a reflection more of host availability than of mosquito host preference: the high numbers of Cx. pipiens s.l. and Ae. aegypti individuals collected and their feeding behaviors (on both humans and domestic animals), as reported in this study, call for further investigations into parasites circulating in these mosquitoes and also in dogs and humans in Cape Verde. For situations where humans would behave as amplifying hosts of other dog parasites not mentioned here, from our results of simultaneous blood meals in dogs and humans, we may have an even greater amplification event due to interaction augmentation between these two hosts, which increases as their abundance also increases (105). However, as pointed out elsewhere (18, 106, 107), it is necessary to consider the specific attributes to such interactions.
Future directions from this study should therefore consider on combining data from mosquito and animal densities and richness to explore transmission risk according to season and land use. In addition, considering a larger panel of hosts, including non-domestic animals, and increasing sampling efforts with proper sampling techniques, will ensure enough representation of samples for biological relevance of data generated. Another limitation is that we did not record the total number of collected mosquitoes. Therefore, we cannot provide the percentage of blood-fed mosquitoes for our field study.
In conclusion, our study provides the first characterization of blood sources utilized by medically important mosquitoes in Santiago Island, Cape Verde. Importantly, this work has shown that feeding behavior among Cx. pipiens s.l. is more random than that demonstrated by Ae. aegypti and An. arabiensis. Our findings also provide interesting information on disease transmission between hosts and pathogens spillover or spillback phenomena. These phenomena, despite the fact that they have not yet been thoroughly investigated in the country, could threaten livestock, poultry farming, wildlife health, and the economy, and this may be even more important in view of climate change. Thus, inclusion of the One Health Initiative (involving animals) in disease surveillance could help track diseases of humans with animal ancestry and zoonosis for planning interventions. Interventions such as zooprophylaxis, in which animals are treated with mosquito-killing drugs, and mosquito bites are redirected can greatly reduce mosquito survival and lower their potential to transmit pathogens.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further enquiries can be directed to the corresponding author.

Author contributions
AG and SL conceived and directed the study. AG, DM, IV, and AD performed species identification. AG, DM, IV, and AD undertook sample processing, DNA extraction, and molecular assays. AG and AD performed data analyses. AG and AD interpreted results. AG wrote the first draft of the manuscript with inputs from SL and AD. All authors commented on and edited the draft manuscript and approved the final manuscript. AG and SL compiled the final manuscript. All authors contributed to the article and approved the submitted version.

Funding
This work was supported by the National Institute of Public Health, Cape Verde.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. Single-host blood meals associations between hosts (humans, dogs, chickens, goats, and pigs) and mosquito species. Mosquito species were arranged by hierarchal clustering according to the likeness.