Sylvatic Mosquito Diversity in Kenya—Considering Enzootic Ecology of Arboviruses in an Era of Deforestation

As new and re-emerging vector-borne diseases are occurring across the world, East Africa represents an interesting location, being the origin of several arboviruses with a history of urbanization and global spread. Rapid expansion of urban populations and alteration of natural habitats creates the opportunity for arboviruses to host-switch from wild, sylvatic hosts or vectors into urban transmission affecting human populations. Although mosquito surveillance regularly takes place in urban areas of Kenya, for example identifying vectors of dengue virus or malaria viruses, little work has been carried out to determine the distribution and abundance of sylvatic vectors. Here, we describe the mosquito vector species and diversity collected at twelve forest habitats of rural Kenya. We conducted arbovirus screening of over 14,082 mosquitoes (47 species, 11 genera) as 1520 pools, and detected seven viruses (six bunyaviruses, and one flavivirus-bunyavirus co-infection) isolated from pools of Aedes dentatus, Anopheles funestus, Culex annulioris, and Cx. vansomereni. Awareness of sylvatic vector species and their location is a critical part of understanding the ecological foci and enzootic cycling of pathogens that may be of concern to public, animal or wildlife health. As natural ecosystems come under anthropogenic pressures, such knowledge can inform us of the One Health potential for spillover or spillback leading to outbreaks, and assist in vector control strategies.


Introduction
The occurrence of vector-borne diseases has been observed in East Africa for several centuries, associated with a variety of etiological agents, host reservoirs and severity of outcome in humans or animals. For example, Crimean-Congo hemorrhagic fever caused by a tick-borne nairovirus is frequently fatal [1]; protozoa of Plasmodium spp., spread by mosquitoes, continue to cause malaria in certain regions [2]; East coast fever, caused by a tick-borne protozoan (Theileria spp.), affects livestock, and has a huge economic impact in cattle-producing regions [3]; and lymphatic filariasis, caused by a vector-borne nematode infection, has a major public health burden in multiple African countries [4]. Focusing specifically on mosquito-borne viruses, medically and veterinary important arboviruses such and we also sought to identify foci for enzootic transmission of arboviruses where spillover events could occur that lead to urban outbreaks.

Mosquitoes
During 2013 and 2014, mosquitoes were collected at 12 forested or thicket sites in Western, Coastal or Central Kenya away from current human residences (Figure 1). Sites were chosen, at elevations below 2000 m (to make detection of mosquitoes more likely, since they are sensitive to cold temperatures), based on the presence of wild animals, in particularly NHPs (the suspected enzootic reservoir of CHIKV and ONNV), and a distance of over 2 km from human activity. They were also chosen according to available permits from Kenya Forestry Service, Kenya Wildlife Service, or local permissible access. Sampling effort varied at each site due to the objectives of a larger study seeking CHIKV isolates; results are therefore weighted accordingly. Mosquitoes were collected at the sites using a combination of trap types: BG Sentinel (baited with BG Lure), CDC light trap (baited with CO 2 from a dry-ice source), and CDC gravid traps (using fermented grass or mango leaf-infused water as bait), as well as resting boxes, and hand-held aspirators.
Traps were set in the early afternoon and overnight on a transect at each site (traps were spaced over 200 m apart, usually at a height of 1.5 m, with some canopy trapping also conducted using BG or Light traps elevated on a rope pulley to approximately 20 m). Geographic coordinates for each collection point were recorded using a hand-held Global Positioning System (GPS) unit (Garmin International, Olathe, KS, USA) in WGS84 datum, and imported into ArcView GIS software (ESRI, Redlands, CA, USA). Trap contents were retrieved at dawn. Captured mosquitoes were separated from other insects, stored at −80 • C and transported to the Kenya Medical Research Institute (KEMRI) in Nairobi, where they were identified to species using a chill table and taxonomic keys [31]. Mosquitoes were pooled (up to 25 individuals per pool) according to collection date, location, and species, for subsequent viral screening.

Virus Detection and Identification
Mosquito pools were initially examined using cell culture, since this is more efficient than going straight to molecular analysis when processing a large volume of mosquitoes for which viral prevalence rates are low. Pools were homogenized in 1 mL of Dulbecco's Minimum Essential Medium (Gibco, Gaithersburg, MD, USA) with 10% fetal bovine serum, fungicide and antibiotics added, then centrifuged at 10,000 rpm for 4 min. The resulting supernatant was removed and 50 µL was used to inoculate sub-confluent cell monolayers in 12-well plates (Costar, Corning, NY, USA). Negative (media-only) controls were included. The two cell lines used were Vero E6 (African Green Monkey kidney) and C6/36 (Aedes albopictus), maintained at 37 • C and 28 • C, respectively, with 5% CO 2 . Cells were monitored for the development of cytopathic effects [CPE] for up to 12 days. Samples showing CPE were passaged twice in the same cell line for virus confirmation and amplification, prior to harvest and storage at −80 • C. Viruses were identified via RT-PCR following RNA extraction using a QIAamp Viral RNA mini kit (Qiagen, Germantown, MD, USA) according to the manufacturer's instructions, using generic primers specific for alphaviruses, orthobunyaviruses and flaviviruses (Table 1). Viral RNA from each genus group and water were used as positive and negative controls, respectively. RNA was used as a template for cDNA synthesis with 5X first strand buffer, dNTP (10 mM), DTT (100 mM), random hexamer, RNase inhibitor (40 U/µL), and reverse transcriptase. The cycling conditions were as follows: 70 • C for 10 min, 4 • C for 5 min, 25 • C for 15 min, 42 • C for 50 min, 70 • C for 15 min, and then held at 4 • C. RT-PCR amplification was performed on 2 µL of the resulting cDNA, in a 25 µL volume of Amptaq mastermix, primers and water. Further virus identification, beyond family, was not available due to limited resources for this project.
suspected enzootic reservoir of CHIKV and ONNV), and a distance of over 2 km from human activity. They were also chosen according to available permits from Kenya Forestry Service, Kenya Wildlife Service, or local permissible access. Sampling effort varied at each site due to the objectives of a larger study seeking CHIKV isolates; results are therefore weighted accordingly. Mosquitoes were collected at the sites using a combination of trap types: BG Sentinel (baited with BG Lure), CDC light trap (baited with CO2 from a dry-ice source), and CDC gravid traps (using fermented grass or mango leaf-infused water as bait), as well as resting boxes, and hand-held aspirators.

Statistical Analysis
Species richness (S), the total number of species detected, was calculated for each site. Species diversity and evenness of mosquito abundance was calculated using Shannon's H, as used previously in [32]. Due to uneven sampling effort between sites, collections at each site were standardized using species accumulation curves based on the number of individuals caught. Validity of this method has been verified as a better weighting of effort to compare species communities than number of nights, and also avoids problems from differences in trap efficacy [33]. A modification of a randomized species accumulation curve was made using Estimate S, as per Wilcott (1999), to provide a comparative estimate of the rarefied species richness that was independent of sample size [34,35]. Arboviral infection rate (MIR) in positive mosquito species was computed for cytopathic viruses using maximum likelihood estimates (MLEs) of pooled samples (infection rate per 1000 mosquitoes) as described by Eastwood et al. (2016), via Excel© Add-In software available at http://www.cdc.gov/westnile/resourcepages/mosqSurvSoft.html [32,36].

Mosquito Species
A total of 14,082 mosquitoes (11 genera and 47 species) were collected across 12 sylvatic sites in Kenya. Genera included Aedeomyia, Aedes, Anopheles, Culex, Coquillettidia, and Mansonia, and several of the mosquito species collected are known vectors for arboviruses or for malaria, for example Aedes mcintoshi, Anopheles funestus, Cx. pipiens, and Mansonia africana. Table S1 indicates the species and numbers of all mosquitoes captured during the study. Culex vansomereni (58%) represented the majority of sylvatic mosquitoes captured in total during the study, detected at all sites. Other predominant mosquito species varied according to site ( Table 2). The commonest species collected at the coastal sites were Cx. annulioris (Arabuko-Sokoke forest) and Cx. pipiens (Shimba Hill Reserve). Aedes chaussieri was also a dominant coastal site species, and one which was absent from collections from Western Kenya. In contrast, Coquillettidia fuscopennata and Cx. annulioris were the second most prevalent species after Cx. vansomereni captured at sites in Western Kenya. Aedes cumminsi accounted for the majority of mosquitoes collected near Kitale, followed by Cx. pipiens. Large captures of Cx. vansomereni were made at Rift Valley sites, with Cx. univittatus as the next populous species. At the forest site near Nairobi (Ololua), Cx. vansomereni accounted for 88% of catches, and Aedes tricholabis and Aedes dentatus were next prevalent. The species richness and species diversity (as calculated using Shannon indices) for each site are displayed in Table 2, with evenness illustrated in Figure 2. The highest catch abundance was seen at the Nandi Forest site, near the village of Mwein, with nearly 5000 individuals captured, followed by Kimondi, also part of Nandi forest.

Mosquito Species
A total of 14,082 mosquitoes (11 genera and 47 species) were collected across 12 sylvatic sites in Kenya. Genera included Aedeomyia, Aedes, Anopheles, Culex, Coquillettidia, and Mansonia, and several of the mosquito species collected are known vectors for arboviruses or for malaria, for example Aedes mcintoshi, Anopheles funestus, Cx. pipiens, and Mansonia africana. Table S1 indicates the species and numbers of all mosquitoes captured during the study. Culex vansomereni (58%) represented the majority of sylvatic mosquitoes captured in total during the study, detected at all sites. Other predominant mosquito species varied according to site ( Table 2). The commonest species collected at the coastal sites were Cx. annulioris (Arabuko-Sokoke forest) and Cx. pipiens (Shimba Hill Reserve). Aedes chaussieri was also a dominant coastal site species, and one which was absent from collections from Western Kenya. In contrast, Coquillettidia fuscopennata and Cx. annulioris were the second most prevalent species after Cx. vansomereni captured at sites in Western Kenya. Aedes cumminsi accounted for the majority of mosquitoes collected near Kitale, followed by Cx. pipiens. Large captures of Cx. vansomereni were made at Rift Valley sites, with Cx. univittatus as the next populous species. At the forest site near Nairobi (Ololua), Cx. vansomereni accounted for 88% of catches, and Aedes tricholabis and Aedes dentatus were next prevalent. The species richness and species diversity (as calculated using Shannon indices) for each site are displayed in Table 2, with evenness illustrated in Figure 2. The highest catch abundance was seen at the Nandi Forest site, near the village of Mwein, with nearly 5000 individuals captured, followed by Kimondi, also part of Nandi forest.

Virus Detection
Collected mosquito specimens were screened as 1520 pools on both mammalian (Vero) and mosquito (C6/36) cell lines. One flavivirus (family Flaviviridae) and five orthobunyaviruses (family Bunyaviridae) were isolated from pools representing four different species of mosquitoes, using Vero cells ( Table 3). One of the positive orthobunyavirus samples came from a single male Anopheles spp. mosquito. A further orthobunyavirus was isolated from a Cx. vansomereni pool using C6/36 cells. All positive mosquito pools were captured in the central region of Kenya located near Nairobi. The flavivirus was from the same pool of Cx. vansomereni mosquitoes from which an orthobunyavirus was detected. The pool size was 25 individual mosquitoes and we are unable to state whether this co-infected status came from the same mosquito. Infection prevalence (per 1000 mosquitoes) for arbovirus among the total number of that species screened were MLE = 1. Cx. annulioris). Since all virus-positive pools were detected from the Central Kenya site, the above rates only considered the mosquito pools captured there, rather than from all sites, to limit heterogeneity.

Discussion
Since mosquito community composition plays a major role in the transmission of zoonotic vector-borne pathogens, it is important to consider the sylvatic mosquito species that may be encountered within forests. Changes are rapidly taking place in East African ecosystems, which leads to increased human-wildlife interaction-for example, use of agriculture space, forest resources (wood, hunting) and/or leisure (tourism), and encounters for spillover or spillback between sylvatic and urban-rural cycles of arboviral transmission are possible.
The concept of arboviruses being maintained in sylvatic cycles outside of periodic epidemic transmission affecting humans (for example, YFV maintained by Hemagogus spp. or Sabethes spp. mosquitoes in a cycle independent from urban transmission), together with an outbreak of CHIKV in Kenya from an unknown enzootic source, prompted our investigations of mosquito diversity and the prevalence of arboviruses in sylvatic habitats. The rapid modification of sylvatic habitats, or increased contact with man, reinforces the importance of such a study; Weaver (2006), for example, highlights that many zoonotic arboviruses circulate in tropical forest habitats that are being disturbed for logging and agricultural operations [15].
Recently, we provided evidence of arbovirus exposure in non-human primates in Western Kenya forests [14]. However, no isolates of CHIKV were obtained from mosquitoes of those regions during the current study, conducted at a similar time period. In fact, despite screening over 14,000 mosquitoes from 12 forest sites for evidence of either mammalian or insect-specific viruses, we found little evidence of arboviral infection in specimens from forested areas in Kenya, despite some of the mosquitoes collected being known vector species elsewhere (see below). Only one of the forest sites sampled (nearest to Nairobi) yielded virus-positive mosquitoes, with detection of six Orthobunyavirus isolates, and one same pool of mosquitoes also being flavivirus positive, indicating co-infection status with that pool. The Orthobunyavirus genus has the greatest species diversity of known virus genera, with widely distributed arthropod-borne viruses found in both hemispheres. Bunyaviruses are capable of infecting a wide range of vertebrate hosts, and include some important viral pathogens, which cause diseases of veterinary and human concern [37,38]. Elsewhere in Kenya, numerous bunyaviruses have been detected, with evidence in both mosquitoes and livestock [39][40][41]. Most Orthobunyavirus activity in Kenya has been detected in the Northeast region, which is dry ecosystem, with little forested habitat except alongside the Tana River. Flaviviruses elsewhere have spilled over from their enzootic cycle-for example, WNV has frequently left its bird-mosquito transmission cycle to infect horses and humans (dead-end hosts) via a bridge-vector mosquito, resulting in neurological infections [42]. Regarding the likelihood of false negatives during screening, it is possible that low-level viral RNA exists within mosquitoes. However, the passaging of samples within cell culture was expected to amplify infectious viral particles to a detectable level should they exist.
Species richness of the sylvatic mosquito community in the Kenya region sampled is high; 47 different species were collected, illustrating the high diversity of sylvatic mosquitoes. Sites in Rift Valley were less diverse, and showed unequal evenness, i.e., that key species dominated, rather than an equal spread of species.
To discuss some of the more notable vector species within sylvatic sites of Kenya, we highlight (i) Aedes (stegomyia) spp., e.g., Aedes africanus and Ae. vittatus, which are arboreal mosquitoes found to be competent for CHIKV transmission [43], and associated with a number of other alpha-and flavi-viruses [44], and Ae. aegypti as a key vector of CHIKV, YFV, DENV (although the latter was only detected at one forest site). (ii) Aedes cumminsi is believed to be a vector of Spondweni virus and RVFV [45], and an isolate of Middleburg virus was made from a pool of Ae. cumminsi from forests in Senegal (Robin, 1969). (iii) Culex spp. are often associated with WNV transmission and represented a large proportion of the mosquitoes collected, in particular Cx. vansomereni (>58% of catches in our study). This species was found to be a competent vector of WNV in the laboratory [46]. (iv) In neighboring Uganda, Ae. tarsalis and Coq. fuscopennata (both found in the study here) were sources of Pongola virus, which can cause febrile illness in humans [47]. (v) Anopheles spp. are generally associated with transmission of malaria. However, An. funestus and An. gambiae are also vectors of ONNV, and this genera of mosquito is increasingly detected with arboviruses; transmission capability requires determination [9]. That a bunyavirus was detected here in a male An. funestus, suggests that vertical transmission is occurring as a mechanism for virus persistence.
Mosquito studies of sylvatic habitats with potential for spillover under habitat disturbance elsewhere show a similar incidence of both disease vectors and lesser known species. In Senegal, tree-hole breeding species such as Ae. luteocephalus, Ae. furcifer, and Ae. taylori (all associated with Aedes-borne viruses) predominated at forest sites [48]. Mosquito community studies elsewhere have been used to suggest whether outbreaks of prominent arboviruses could be supported [49,50]. Knowing more about vectorial capacity, including vector competence, of sylvatic species in Kenya would suggest whether the elements exist to support enzootic outbreaks. Other arboviruses, although with little known about behavior or potential risk, have been detected from mosquito species captured during our study-for example, Kamese virus (family Rhabdoviridae) was first detected from Ae. annulioris in forests in Uganda [45,51]. Mosquito diversity in forests of West Africa has been demonstrated to include many Aedes spp., including several detected with medically important arboviruses including CHIKV [10]. Although no vector evidence of CHIKV was detected here, we previously showed the virus to be circulating in the region via vertebrate-host studies [14]. Further mosquito studies, incorporating improved methodology to detect Aedes stegomyia spp., are recommended.
Between-site habitat differences, weather, and longer seasonal fluctuations can influence results. Although efforts were made to sample throughout the year, rainfall is higher during March-June and October-November. Several mosquitoes, especially those whose morphological quality deteriorated in large and wet overnight catches, could only be identified to genus level here, and so, although Cx. vansomereni was suspected for many Culex spp., further distinction of diversity was prevented.
The highest nightly catch abundances were made at sites in Western Kenya, which were mainly sampled during the month of March 2014. However, other sites sampled in that same month (including Ololua) did not reflect such high abundance. Nevertheless, seasonality is an important factor that should be acknowledged in the design of future studies, where the main objective is comparative abundance of these mosquito species. In Senegal, mosquito diversity was higher during months when more water in larval habitat was likely [48]. Here, we combined a variety of sampling opportunities (e.g., alongside other wildlife studies) and therefore exact comparison of seasonal effects upon vector species was not possible.
We focus on forested habitat, but ecosystems still vary considerably. Unequal sampling was accounted for, however some sites noticeably yielded smaller sample sizes, e.g., few mosquitoes were obtained from Shimba Hills Reserve (Kwale County), reflecting the hot dry ecosystem, with fewer areas of forest (a more savannah ecosystem), plus samples were taken from there in the dry season (as an aside to a primate study). In the coastal Malindi region of Kenya, Mwangangi et al. (2012) found Cx. quinquefasciatus to be the vastly predominant species, especially in urban and peri-urban strata. There, most vectors-An. gambiae, An. funestus and An. coustani-were found in the 'rural' stratum, which was defined, however, as agricultural land rather than intact forest [25]. There are challenges comparing species inventories of sites where there is no indication of how complete the inventory is, or when different methods have been used.
In Kenya, it is now difficult to locate primary (unaltered) tropical forest; Kakamega forest being the only existent remnant of the Guinea-Congolion rainforest belt that spanned equatorial Africa [52], and even then Kakamega forest is frequented by local villagers and popular for ecotourism. Braack et al. (2018) highlight that increased land transformation in Africa and disruption of historical ecological processes promote contact between humans and infected wildlife or sylvatic vectors [45]. A framework presented by Faust et al. (2018) suggests that the highest spillover risk occurs at intermediate levels of habitat loss, whereas the largest, but rare, epidemics occur at extremes of land conversion [53]. Tropical forest habitat is generally reported as threatened due to encroachment for new land for agriculture and settlement and overharvesting of forest resources. In other countries, although undisturbed primary forest has peak mosquito diversity, disturbance has led to mosquito communities with a greater abundance of 'pathogen-transmitting' vector species [54]. This highlights that it is the specific mosquito species that is present that matters, not the quantity or richness in general. Elsewhere, clearing patches of tropical forest often appears to create ideal habitat along forest edges for malaria mosquito vectors to breed, and deforestation is associated with increases in disease through ecological mechanisms [55,56]. Indeed, Afrane et al. (2012) highlight how deforestation in the western Kenyan highlands is associated with increased development and survival of some Anopheles vectors [57].
Host-switching of pathogens, from one vector species to another, or one host species to another, is a common event during long-term arbovirus evolution [15,58], and one we should attempt to anticipate. Taxonomic identification of sylvatic mosquitoes can be a challenging process since many entomological specialists are not as familiar with such species; however, continued monitoring of mosquito vectors is recommended, and not just around towns and villages where surveillance and urban transmission most commonly takes place. For example, supplementation of activities conducted by KEMRI or Ministries of Health to include regions where rural workers or tourists may come into contact with sylvatic cycles of disease pathogens may be useful to avert emergent outbreaks occurring via spillover.
Most arboviruses are zoonotic, and understanding sylvatic transmission of vector-borne pathogens is not complete without determination of the host reservoir involved in their enzootic maintenance. Studies of sylvatic vector species should be complemented by understanding of enzootic vertebrate host interaction-for example, by mosquito blood-meal analysis, viral or serological surveys.

Conclusions
In summary, a wide diversity of mosquito species exists outside of urban or peridomestic settlements in East Africa-many of which are known as competent disease vectors elsewhere. As humans encroach on forests and previously undisturbed sylvatic habitats, awareness of arthropod vector species and their pathogen transmission role is vital to limit the potential for disease outbreak. We demonstrate the high species diversity of mosquitoes across sylvatic Kenyan ecosystems, and encourage that their role in pathogen maintenance networks is investigated further. Recognizing the existence of sylvatic mechanisms of arbovirus proliferation is an essential component to One Health, with forest degradation, spillover and spillbacks between domesticated and wild species (arbitrated by non-urban arthropod vectors) relevant for both public, animal and wildlife health.