Arbovirus Prevalence in Mosquitoes, Kenya

Few studies have investigated the many mosquito species that harbor arboviruses in Kenya. During the 2006–2007 Rift Valley fever outbreak in North Eastern Province, Kenya, exophilic mosquitoes were collected from homesteads within 2 affected areas: Gumarey (rural) and Sogan-Godud (urban). Mosquitoes (n = 920) were pooled by trap location and tested for Rift Valley fever virus and West Nile virus. The most common mosquitoes trapped belonged to the genus Culex (75%). Of 105 mosquito pools tested, 22% were positive for Rift Valley fever virus, 18% were positive for West Nile virus, and 3% were positive for both. Estimated mosquito minimum infection rates did not differ between locations. Our data demonstrate the local abundance of mosquitoes that could propagate arboviral infections in Kenya and the high prevalence of vector arbovirus positivity during a Rift Valley fever outbreak.


Sampling
To evaluate the temporal profi le of vector mosquitoes in North Eastern Province, Kenya, trapping was performed during the dry season (August 2006) and during the rainy season (December 2006-January 2007). Mosquitoes collected during December 2006 and January 2007 were trapped during an epizootic/epidemic of RVF. Homestead trapping locations adjacent to homesteads in the regions were randomly selected from previously prepared census lists and were restricted to only those homes where animals (cows, goats, or sheep) were housed alongside human habitats. Each household had only 1 CDC light trap (John W. Hock Company, Gainesville, FL, USA) located next to animal structures; trap was set 1 time for 12 hours, 6:00 PM-6:00 AM.
Mosquito sampling was conducted in 2 areas within Masalani Division, Ijara District, North Eastern Province, where human surveillance had taken place 8 months before the RVF outbreak (2) (Figure 1). Traps were located in the rural village of Gumarey (1°40′12′′S, 40°10′48′′E) and the town of Sogan-Godud (1°41′24′′S, 40°10′12′′E). The population of Gumarey consists of seminomadic herders who live in traditional grass huts near their livestock. Sogan-Godud is more urban with a marketplace and contains a greater proportion of tin-roofed permanent dwellings. The centroids of these 2 locations are 5 km apart, and the borders are within 500 m of each other. Both locations had persistent local fl ooding during the extensive El Niño/Southern Oscillation associated heavy rains during 2006-2007, and both are within 10 km of the Tana River. Persons seropositive for RVFV from both locations were documented in early 2006; seroprevalence rates were greater in rural Gumarey (20% vs. 6%) (2). During that initial study, all homesteads were identifi ed and their locations identifi ed by Global Positioning Satellite. Spatially referenced data on individual residence and homestead exposure features were maintained and analyzed by using ArcGIS version 9.2 (ESRI, Redlands, CA, USA).

Mosquito Preparation
Mosquito genera were identifi ed in Kenya by local entomologists on the basis of microscopic morphologic appearance. Only female mosquitoes were included in this study; male mosquitoes were not further tested. Single leg specimens were preserved in RNAlater (Ambion, Austin, TX, USA) and transported to Case Western Reserve University (Cleveland, OH, USA) for processing. DNA and RNA were extracted from mosquito legs by using a column purifi cation kit (QIAGEN, Valencia, CA, USA) with the following modifi cations: each mosquito leg was placed into a microcentrifuge tube containing 150 μL of RNeasy lysis buffer and fi nely ground with a disposable RNase/DNasefree pestle. After homogenization, samples were processed according to established protocols through either individual QIAGEN RNeasy columns or 96-well plates, washed, and eluted in RNase-free water. The DNase step was omitted so that DNA and RNA could be collected from samples. Individual RNA samples were combined in pools of <12 mosquitoes (median 10, mean 8.7), based on homestead trap for cDNA synthesis and PCR or quantitative reverse transcription-PCR (qRT-PCR).

Primers and Generation of Standard Controls
To verify the quality of the RNA and the integrity of the cDNA products after reverse transcription, mosquito 18S rRNA primers were designed to amplify within a region conserved in many Culicidae spp. mosquitoes (30). These mosquito primers were designed against the 18S rRNA gene sequences for Aedes spp. (GenBank accession no. AB085210) and Culex spp. (GenBank accession no. U48385) mosquitoes to amplify an optimally sized product (124 bp) for qRT-PCR. WNV primers were based on the New York 1999 WNV isolate (GenBank accession no. AF196835.2) described by Lanciotti et al. (11). These primers have been shown to detect Old and New World WNV strains, including a strain isolated in Kenya in 1998 (11,31). RVFV primers, which amplify a conserved region of the large segment (90 bp), were used as described by Bird et al. (32) (Table 1).
An RVFV standard control was generated by amplifying RVFV vaccine strain rMP-12 in Vero E6 cells for 72 h and then extracting viral RNA from supernatant and cell lysate by using the PureLink Total RNA Purifi cation System (Invitrogen, Carlsbad, CA, USA). A WNV standard control was generated by using confi rmed WNV-positive samples received from the Ohio Department of Health. Mosquito 18S rRNA-, WNV-, and RVFV-positive controls were generated by using the primers listed in Table 1 and cloned by using the pCR 8/GW/TOPO TA cloning kit (Invitrogen). All inserts were verifi ed by sequencing of the plasmids.

cDNA Synthesis, PCR, and qRT-PCR Conditions
Two-step qRT-PCR was performed on all pooled samples. First-step total cDNA synthesis was performed on RNA extracted from mosquito leg tissue by using random hexamer primers. The reaction mixture was incubated at 65°C for 5 min, chilled on ice, and combined with 4 μL 5× First-Strand Buffer, 1 μL 0.1M dithiothreitol, 1 μL RNase inhibitor, and 0.5 μL SuperScript III Reverse Transcriptase (Invitrogen). The fi nal reaction mixture was incubated at 25°C for 10 min, 50°C for 50 min, and heat inactivated at 70°C for 15 min.
After cDNA synthesis, 1 μL of total cDNA was added to the qRT-PCR mixture containing 0.2 μmol/L forward primer and 0.2 μmol/L reverse primer (18S and WNV test-ing), 12 μL FastStart Universal SYBR Green Master mix (Roche, Indianapolis, IN, USA), and 12 μL sterile, nuclease-free water. The qRT-PCR was conducted in an Applied Biosystems 7300 instrument (Applied Biosystems, Foster City, CA, USA) with a heating cycle of 50°C for 2 min and 95°C for 10 min; followed by 45 cycles of 95°C for 15 s, 60°C for 1 min, 95°C for 15 s; and an additional dissociation step of 60°C for 1 min. All samples, which registered a cycle threshold value <35 cycles and had a lower cycle threshold value than negative controls, were considered positive for their respective targets. All pools were further PCR tested for RVFV by using 2 μL cDNA, 0.5 μmol/L each forward/reverse RVFV primer, 10.5 μL sterile, nuclease-free water, and 12.5 μL JumpStart ReadyMix Taq (Sigma-Aldrich, St. Louis, MO, USA). PCR cycling parameters were 94°C for 5 min, with 30 cycles of 95°C for 30 s, 60°C for 1 min, 72°C for 1 min, and a 10-min 72°C extension. PCR products were run on a 2% agarose gel with SYBR Safe (Invitrogen) for band visualization ( Figure 2). An initial sampling of RVFV PCR products was cloned by using the above-mentioned methods and sequenced for confi rmation. Mosquito minimum infection rates (MIR) for RVFV and WNV were calculated on the basis of maximum-likelihood estimation by using the PoolScreen 2.0 program (University of Alabama at Birmingham, Birmingham, AL, USA) (33)(34)(35).
In terms of the general population, by using geographic information systems analysis of spatially referenced census data ( Figure 5 (Figure 3). The temporal distribution of these mosquitoes correlates with previous studies showing that Aedes spp. mosquitoes predominate in the initial weeks after substantial fl ooding and then curtail after the fi rst month of fl ooding, at which time Culex spp. and Anopheles spp. mosquitoes emerge as the predominant species (17,36). It is believed that the dramatic proliferation of transovarially infected Aedes spp. mosquitoes immediately after fl ooding re-introduces virus into an epizootic/epidemic cycle, after which Culex spp. mosquitoes propagate the virus in an epizootic/endemic cycle among humans and animal species (17).

Discussion
A substantial proportion of the mosquito population collected within our study area consisted of RVFV-or WNV-infected potential vectors. The close proximity of these infected mosquitoes to amplifying hosts and susceptible animals and humans during an RVFV epizootic/ epidemic warrants further investigation of transmission dynamics. RVFV RNA in mosquitoes collected within the area was high, and the substantial presence of WNV RNA in these mosquito samples was unexpected. The presence of WNV in mosquitoes from Sogan-Godud and Gumarey in our study corroborates recent documentation of the widespread presence of WNV in Kenya and the ability of mosquito populations, including Cx. quinquefasciatus, to acquire and transmit WNV (23).
The previous isolation of WNV from male Culex spp. mosquitoes in Rift Valley Province suggests a natural transovarial transmission cycle among some mosquito vectors but is unlikely to contribute greatly to virus main-tenance between enzootic periods (23). Additionally, although human epidemics and outbreaks of WNV have not been reported, the presence of the virus in local mosquitoes suggests that the virus is maintained in a natural cycle yet to be elucidated and that the actual incidence of WNV in human populations in the region could be underestimated. Improved fi eld diagnostics are necessary for rapid and accurate diagnosis of circulating arbovirus threats and expedient translation into preventive public health practices.
The isolation of RVFV and WNV RNA from mosquito leg samples confi rms that these viruses were disseminated within the bodies of the mosquitoes tested. These results also confi rm that single mosquito leg samples are suffi cient for PCR/qRT-PCR detection of RVFV and WNV, respectively. Positive results from testing of the mosquito legs also diminish concern about false-positive results from testing whole mosquitoes, which might contain recent bloodmeals with substantial viral content. Our study confi rms that RVFV disseminates to the legs of wild Cx. quinquefas- ciatus mosquitoes and suggests that these mosquitoes, promiscuous feeders, could play a role in the maintenance or transmission of RVFV in disease-endemic regions (20,21). Other vector competence studies have shown that RVFV does disseminate in Cx. quinquefasciatus mosquitoes but have yet to show that they are effi cient vectors for RVFV (20,21,37). Although identifi cation of viral RNA in the legs of Cx. quinquefasciatus as well as the other mosquitoes tested supports dissemination of virus, no conclusions can be made from these results regarding the role of these mosquitoes in maintaining these arboviruses in this environment or their ability to transmit virus. Additional studies are required to determine vector competence of Cx. quinquefasciatus and other mosquito species tested for these 2 viruses.
During this RVFV outbreak, we documented >1 arbovirus circulating in local mosquitoes. During an arbovirus outbreak, other viruses may be circulating concomitantly without recognition and serve as alternative causes of fever. Additional arthropod surveillance studies during RVFV outbreaks in Kenya have found arboviruses in mosquitoes, including fl aviviruses and alphaviruses, which can cause febrile illness in humans (38). Because diseases from arboviral infections can be nonspecifi c in humans and animals, it is necessary, even during large outbreaks, to document the true cause of disease with detailed testing. Cases of other arboviral infections could be missed if suspected cases are attributed to the epidemic arbovirus without accurate diagnosis.
Although MIRs for RVFV were similar in the 2 villages studied, rural Gumarey was more likely to have RVFV-positive pools than was Sogan-Godud. This fi nding concurs with previous human seroprevalence studies that found that risk for being RVFV seropositive is 4× greater for those living in Gumarey than for those in Sogan-Godud (2). Gumarey residents were more likely to report greater contact with animals and mosquitoes (2). Continued research to identify village level and landscape factors responsible for increased human transmission is necessary. Although RVFV can be transmitted to humans by the bite of an infected mosquito, alternative forms of human exposure, such as aerosol and direct contact, may be more critical for transmission during epidemics (2,28,36). More research must be conducted to elucidate the most common and most effective routes of RVFV transmission to humans during epidemic and interepidemic periods.
Few research studies have documented the presence of WNV and the vectors responsible for its transmission in Kenya. The identifi cation of WNV in North Eastern Province indicates a greater prevalence of the virus than was expected. WNV has not been previously reported in mosquitoes from these 2 villages, and study results imply regional variance in infection rates. Further studies may elucidate a difference between these 2 villages with regard   to resident reservoirs (birds) or undiscovered amplifying hosts, especially if data are collected during outbreak conditions of fl ooding and mosquito proliferation. The spatial overlap of human population density with mosquito abundance ( Figure 4) and the proximity of humans to infected mosquitoes ( Figure 5), suggest that RVFV and WNV trans-mission during epizootic/epidemic periods could be high in both villages. Additional exposure-modifying factors, including the relative contribution of aerosol transmission of RVFV and the effects of housing construction, sleep and work habits, and the role of personal protective measures need to be further elucidated (2). Our study has several limitations. Mosquito sampling during the outbreak was not stratifi ed, and pooling of collected mosquitoes was not randomized (39). Mosquito sampling was conducted only at homesteads where specifi c animals, those known to be reservoirs of RVFV, were housed closely with humans. This sampling method may have underestimated the WNV MIR detected. This type of targeted sampling, however, can provide earlier detection of arboviruses and greater understanding of transmission and maintenance factors of these viruses (39). Although only 920 mosquitoes were tested for WNV and RVFV, a fraction of the total mosquito population collected, it has been shown that testing of mosquito pools versus testing of all samples can yield suitable results, thereby conserving time and resources (39,40). The choice of screening pools for arboviruses offers many benefi ts, especially during an outbreak. The potentially limiting factors of cost and time are avoided, while mosquito positivity is accurately identifi ed (33,39).
In conclusion, we found high MIR for RVFV and WNV for many mosquitoes, some potentially effi cient vectors, in our study region during the 2006-07 RVF outbreak in northeastern Kenya. MIRs did not differ between villages, although RVFV pool positivity and human seroprevalence (as measured in a previous homestead-based study during an interepidemic period) were higher in the rural village of Gumarey (2). Our data demonstrate the local abundance of mosquitoes infected with arboviruses in Kenya and highlights simultaneous arbovirus circulation. A greater understanding of how these arboviruses are maintained in nature will improve targeted prevention in regions where disease is endemic and curtail introduction to new areas. Our current inability to quickly detect arboviral infections in endemic communities has led to inaccurate risk assessments, underdiagnosis of clinical cases, and ineffective control measures. Better detection methods in vector, animal, and human populations and recognition of arboviral risk zones and circulation may alter current perceptions about these diseases. These methods could also lead to improved surveillance and better estimates of the true impact of arboviral disease on animal and human populations.  support; Megan Ermler for the generation of the RVFV MP-12 cDNA clone; and Krupen Patel and Florence Bockarie for technical assistance.
This study was supported by National Institute of Health grants R01TW008067 and 1KL2RR024990. Additional support was provided by Kenya Fogarty grant D43TW06576 and the Robert E. Shope Fellowship in Infectious Diseases (A.D.L.).
Dr LaBeaud is an assistant scientist and associate physician at the Children's Hospital Oakland Research Institute. Her research interests include infectious disease epidemiology, domestic and international arbovirology, and emerging infections.