A four-year survey (2011–2014) of West Nile virus infection in humans, mosquitoes and birds, including the 2012 meningoencephalitis outbreak in Tunisia

A West Nile virus (WNV) outbreak occurred in Tunisia between mid-July and December 2012. To assess the epidemiological features of the WNV transmission cycle, human cerebrospinal fluid samples from patients with suspected cases (n = 79), Culex pipiens mosquitoes (n = 583) and serum specimens from domestic and migratory birds (n = 70) were collected for 4 years (2011–2014) in the Tunisian Sahel region. Viral testing was performed by polymerase chain reaction (PCR). The WNV genome was detected in 7 patients (8.8%), 4 Culex pipiens pools, and a domestic mallard (Anas platyrhynchos). All PCR-positive samples were from the Monastir region. Phylogenetic analysis revealed that two different WNV strain groups circulated, and isolates from the reservoir (bird), vector (Culex pipiens), and dead-end hosts (humans) were closely related. The Monastir region is a hot-spot for WNV infection, and the reiterative presence of WNV over the years has increased the risk of viral reemergence in Tunisia, which highlights the need for more enhanced and effective WNV surveillance in humans with public awareness campaigns strengthened by monitoring mosquitoes and maintaining avian surveillance for early detection of WNV circulation.


Introduction
The West Nile virus (WNV) is an arthropod-borne virus of the Flaviviridae family, genus Flavivirus, belonging to the Japanese encephalitis serocomplex and has a complex life cycle, involving several bird species as primary hosts, mosquitoes as primary vectors and humans, and horses as incidental or dead-end hosts 1,2 . WNV is maintained in nature via native birds who serve as the local amplifying hosts, as well as infected migratory birds who spread the virus intercontinentally 3 . Most human WNV infections are subclinical or asymptomatic; however, symptomatic persons may experience a mild influenza-like illness that induces fever, headache, myalgia, malaise, vomiting, and diarrhea. Less than 1% of infected patients develop neuroinvasive diseases, including meningitis, encephalitis, or acute flaccid paralysis, with some fatalities occurring in elderly and immunocompromised people [4][5][6] .
In recent decades, WNV has expanded geographically and is now endemic in Southern and Eastern Europe, Africa, North America, West Asia, the Middle East and Australia 7 . WNV circulation in the Mediterranean basin has been confirmed, and large human WNV infection outbreaks continue to be reported 8,9 . In Tunisia, WNV caused three major human epidemics with fatalities and severe central nervous system diseases, such as meningitis and encephalitis, which occurred in 1997 10,11 , 2003 5,12 , and more recently in 2012 6 , with sporadic cases recorded in 2007, 2010, 2011, and 2016 9,13,14 . The WNV strains identified during 1997 and 2003 human outbreaks belonged to lineage 1a and were closely related to the Israeli-American cluster 5,11,15 . WNV is thought to be endemic/enzootic in Tunisia, as shown by several serological studies performed in humans [16][17][18] , wild birds 19 , and other mammals [20][21][22][23] . Furthermore, WNV has been detected in mosquitoes from Central Tunisia 24 .
The Tunisian Sahel is a coastal region of Eastern Tunisia near Malta and the Sicilian Islands. It stretches along the eastern shore of three governorates (Sousse, Monastir, and Mahdia). The Sahel region was affected by the three WNV epidemics, and its governorates are considered hotspots 22 based on avifauna biodiversity and the regional abundance of wetlands and mosquito populations, primarily Culex species, which are potential WNV vectors in Tunisia 25,26 . Moreover, the Sahel region is the resting and wintering place of many migratory and native bird species, which could be a risk factor for WNV infection 25,27 . Thus, monitoring Flavivirus infections in this coastal region is important for public health.
In the present study, we report outcomes of the first molecular WNV characterization from clinical samples from patients who had suspected WNV infections or neurological symptoms during the 2012 outbreak in the Tunisian Sahel region. The study also provides a survey of mosquitoes and birds conducted to compare the WNV strains circulating in wildlife and humans and to obtain epidemiological information about the WNV transmission cycle in the Tunisian Sahel region.
Among the 79 cerebrospinal fluid (CSF) samples, 7 (8.8%) were positive for WNV RNA (5 females and 2 males) originating from 7 localities (Fig. 1): 4 and 1 were obtained in November and December 2012, respectively, and 2 in January 2013 (Table 2). Four patients had neuroinvasive diseases: three had meningitis and one meningoencephalitis with a past medical history of hypertension, one patient experienced febrile seizures and two patients died. All patients received supportive treatment, and the brain computed tomography scan of the meningoencephalitis patient was normal.

WNV detection in mosquitoes
A total of 583 Culex (Cx.) pipiens female mosquitoes were collected and grouped into 25 pools (Table 3).
Among the 25 mosquito pools tested for WNV RNA, 4 pools sampled in 4 localities (Khniss, Agba, Moôtmar-Sahline, Oued Khniss) in the Monastir region were positive by PCR (Fig. 1). Mosquitoes from 1 and 3 of the WNV-positive pools were captured in October 2011 and November 2012, respectively.

WNV detection in migratory and domestic birds
Among the 70 bird serum samples, only one bird tested positive for WNV RNA: a domestic mallard (Anas platyrhynchos) sampled in November 2013 from a livestock farm in Ouardanine-Oued El Guelta in the Monastir region ( Fig. 1; Table 4).

Phylogenetic analysis
A phylogenetic tree was constructed from a 283-bp section of the NS5 gene region amplified by the WNV primers (WNV2+/WNV2−). The tree including the 12 WNV isolates from this study and their relationship with other WNV strains is presented in Fig. 2. As shown in the figure, only strains from lineage 1 were detected. The WNV-positive PCR amplicons obtained from three patients, the bird and the 4 mosquito pools formed a monophyletic cluster and were closely related to a Tunisian WNV strain isolated in 1997 (PAH001) as well as the Israeli strain, IS-98. The 4 remaining human sequences

Discussion
During the summer-autumn of 2012, a WNV infection outbreak occurred in Tunisia after an apparent silent period following the 2003 epidemic, although  Phylogenetic analysis revealed that our WNV isolates belonged to two distinct clusters: human, avian and mosquito isolates closely related to the virulent WNV strains isolated in Tunisia (1997) and in Israel (1998), suggesting the presence of an active and local WNV transmission cycle in the Monastir area. The second cluster comprised four isolates from human samples possibly related to the Kunjin virus. These findings revealed the emergence of new pathogenic WNV strains different from those circulating in the Monastir area during the 2003 outbreak 5 . Two of the four patients whose CSF samples tested positive for Kunjin virus RNA presented with neuromeningeal diseases: the elderly and adult patient had meningitis and meningoencephalitis with a past medical history of hypertension, respectively. Kunjin virus detection in patients with neurological manifestations raises the question of whether risk factors have a role in the disease severity since the Kunjin virus is considered less virulent than other WNV types 28 . The contributing factors of age and chronic diseases may relate to a decreased blood-brain barrier integrity, which facilitates access of such a neurotropic virus to the central nervous system, and therefore, predisposes infected patients to neurological complications 29 .
Per the National Observatory of New and Emerging Diseases (Tunisia), the 2012 WNV epidemic occurred earlier in Tunisia from mid-July and persisted beyond the usual occurrence period to December with a wider geographical spread in addition to an increasing number of reported neuroinvasive disease cases and deaths compared to the two previous epidemics of 1997 and 2003, which likely indicates a stronger WNV dynamic in Tunisia in 2012 13,30 . This epidemic coincided with WNV emergence in several other European countries [31][32][33][34] and  13 . The presence of a local and active WNV transmission cycle in this region suggests that Monastir is a high-risk area for WNV infection as demonstrated by previous epidemiological studies 22,38 .
In the current study, WNV-positive cases included three patients with non-specific non-neurological clinical manifestations and four patients with laboratoryconfirmed central nervous system infections. However, even without neurological manifestations, WNV infection can cause significant public health problems, as patients such as children and workers who sought medical attention could not attend school or work because of the illness 39 . Only hospitalized patients were included in this study. Thus, WNV can cause mild self-limited febrile manifestations that may not require hospitalization; therefore, the overall disease incidence may be underestimated.
The roles of migratory birds as introductory hosts, as well as indigenous wild and domestic birds as potential WNV-circulating reservoirs, have previously been discussed 40,41 , and several attempts to isolate the virus from birds have been documented [42][43][44] . A domestic mallard (Anas platyrhynchos) sampled in November 2013 from a livestock farm in Ouardanine-Oued El Guelta in the Monastir region, tested positive for WNV RNA. This sampling site was chosen based primarily on a serological survey conducted at the same farm that showed evidence of WNV antibodies in domestic ducks and highlighted WNV circulation in this area; however, no WNV genomic RNA was detected 45 . Asian studies showed the receptivity and susceptibility of ducks to WNV infection as well as the birds' involvement in WNV transmission 46,47 . Our results suggest that this WNV-positive domestic duck was viremic at the time of sampling, and could therefore act as a WNV reservoir; thus, it may have been involved in the local WNV transmission cycle. However, per Hofmeister et al. 48 , it is unlikely that ducks are amplifying hosts or that they play significant roles in WNV transmission. The homology between the sequences obtained from this bird and the mosquitoes suggests that the virus is overwintered locally or reintroduced seasonally, and the environment favors viral amplification. In a similar study conducted in Italy, WNV persistence was also confirmed through viral circulation in birds and mosquitoes 43 . Furthermore, the studied area is situated near wadi El Guelta with approximately 5 km of wetlands, namely Sahline-Sabkha-Airport and Saline-Sahline, with high concentrations of migratory birds, some of which were sampled and included in this study. Although WNV was undetected in the 47 migratory bird samples, the likely contribution of viremic migratory birds in repeatedly introducing WNV in Tunisia, followed by local viral transmission via amplifying avian hosts, has been clearly assessed in many Tunisian studies 10,12,19 .
Cx. pipiens mosquitoes are considered the most important vectors in transmitting WNV 43,49,50 . Entomological  24 . More studies are needed regarding vector competence and overwintering cycles in mosquitoes, local birds and potential non-avian reservoirs 23 as an endemization mechanism in Tunisia.
Isolate clustering from human and mosquito WNVpositive samples in this study confirms the potential role of Cx. pipiens as a bridge species between birds and mammals, and its involvement in the WNV transmission cycle in Tunisia has been described in several Tunisian entomological studies 24,26,52 . WNV detection in both human and Cx. pipiens specimens sampled in the Monastir region during the 2012 outbreak suggests that after WNV amplification via an enzootic transmission cycle, human exposure to infected mosquitoes and the viral shift to humans may have occurred under ecological and environmental conditions including the close proximity of human habitats to wetlands that host migratory bird settlements and are suitable areas for mosquitoes 55 . All patient and Cx. pipiens mosquito specimens in this study sampled during the 2013 and 2014 transmission seasons were WNV-negative. However, WNV detection in birds in 2013 (this study) and in mosquitoes in 2014 24 suggest that the risk of WNV reemergence in Tunisia should be considered high.
As no vaccine or specific treatment exists for WNV infections, only preventive measures, such as wearing long-sleeve shirts and long pants, applying skin repellents and avoiding being outdoors at dusk could reduce the risk of WNV infection by half 56 . Although vector control measures including adulticidal and larvicidal treatments were implemented early in Tunisia in spring 2012, it appears that the population's low background immunity to WNV and the emergence of new pathogenic strains may partly explain the reoccurrence of WNV human cases and the spread of the virus to new areas 13,16 . Thus, human surveillance should be enhanced by updating risk area maps and public awareness campaigns among the population and health care professionals throughout the country. WNV persistence may lead to future outbreaks in Tunisia and along bird migratory routes between Africa and Europe, highlighting the need for entomological studies on persistence mechanisms and identification of mosquito species acting as competent bridge vectors in WNV transmission 24 . Setting up avian surveillance based on serological surveys of domestic birds cohabiting with humans and/or sentinel chickens to assess WNV enzootic transmission might provide useful epidemiological information such as low-noise WNV circulation, which can therefore be used as a warning system to detect viral circulation early and implement prevention and control measures 57,58 .

Ethics statement
The study protocol was approved by the Ethics and Research Committee of the Fattouma Bourguiba University Hospital (Monastir, Tunisia, Committee's advice on 20 September 2013), and written informed consent was obtained from the 79 patients.

Patients and clinical sample collections
Seventy-nine CSF samples were obtained from patients admitted to different departments of the Fattouma Bourguiba University Hospital (Monastir, Tunisia) from November 2012 to December 2014. Per routine clinical care, lumbar punctures were performed by the examining physician on patients showing signs of suspected acute meningitis upon admission at the emergency department or within 24 h of hospitalization. CSF specimens were processed for microbiological testing by routine hospital procedures and were bacterial pathogen-free. CSF cytobiochemical examinations provided pleocytosis as well as protein and glucose concentrations.
Based on the examining physician's clinical diagnosis and medical records, patients with febrile symptoms and/ or exhibiting suspicious signs of neuromeningeal infection, such as aseptic meningitis, encephalitis, and meningoencephalitis, were included in the study. Complete demographic characteristics and clinical findings were obtained for all patients.

Study sites and mosquito collections
Mosquito sampling was performed at different sites in the Sahel region's 3 coastal governorates, Monastir, Sousse, and Mahdia, mostly from September to November for 4 years (2011-2014). Two battery-powered CDC light-traps were placed at 14 sampling sites and operated approximately from sunset to sunrise. Epidemiological characteristics were considered for the sampling site selection including occurrence of WNV human cases, proximity to wetlands and migratory bird settlement flight. Mosquito traps were placed in urban (houses and gardens) and rural (riding stables and livestock farms) environments.
Field-collected specimens were transported to the laboratory, and female Cx. pipiens were morphologically identified using identification keys 59 . Female Cx. pipiens were then pooled by date and collection site by handling mosquitoes individually with sterile stainless steel tweezers with a maximum of 50 individuals per pool and stored in cryotubes at −70°C until being assayed for viral detection.
Prior to nucleic acid extraction, mosquito pools were homogenized following a procedure described previously 60 .

Bird collections
Seventy serum samples from 23 domestic and 47 migratory birds were obtained from 2012 to 2014 during the autumn season and active mosquito circulation periods. Migratory bird blood sampling was conducted at several bird wintering and resting sites. Domestic bird blood was collected from livestock farms located on wild bird flyways and/or close to wetlands in the Tunisian Sahel region. For each bird, 5 ml of blood was obtained and centrifuged at 12 000 rpm for 10 min. Serum samples were stored at −20°C until tested.

Detection of WNV RNA
Total RNA was extracted from 200-µl samples using the Trizol ® LS Reagent (Sigma-Aldrich, Madrid, Spain) method for CSF samples and mosquito pool supernatants. Sera from domestic and migratory birds were processed using Trizol ® BD Reagent (Sigma-Aldrich, Madrid, Spain). RNA was then purified with chloroform and precipitated with isopropanol (Sigma-Aldrich, Madrid, Spain).
After washing with 70% ethanol, the pellet was dried and eluted in 30 µl of RNase-free water (Qiagen, Barcelona, Spain). A negative control consisting of RNase-free water was included in this step.
Generic RT-PCR was performed using degenerated primers, Flavi1+ and Flavi1−, whereas generic nested PCR was performed using degenerated primers, Flavi2+ and Flavi2−, as previously described 61 . These two primer sets were designed to amplify a conserved region of the flavivirus genome located in the NS5 gene encoding for polymerase. A second generic nested PCR was performed using degenerated primers designed to specifically target the NS5 gene of the WNV genome (WNV2+: 5′ 8485 AARCCYCTNCTYAAYTCWGAYAC3′ 8507 /WNV2 −: 5′ 8813 TCRTTSARNACNWRYTTIRCWCC3′ 8791 ).
The RT-PCR step was performed using the One Step RT-PCR kit (Qiagen, Barcelona, Spain) per the manufacturer's instructions using 10 µl of RNA and 100 pmol of each primer (Flavi1+/Flavi1−) in a 50-µl total reaction volume. Samples underwent an initial cycle at 50°C for 30 min and 95°C for 15 min, followed by 40 PCR cycles at 94°C for 30 s, 40°C for 4 min, 72°C for 1 min, and a final elongation step at 72°C for 10 min.
The nested-PCR reaction was performed using the Taq PCR Core kit (Qiagen, Barcelona, Spain) per the manufacturer's instructions. One microliter of the first amplification product was added to 100 pmol of the second primer set (Flavi2+/Flavi2−) in a final volume of 50 µl. The reaction mixture was then subjected to the following amplification conditions: an initial denaturation step at 94°C for 2 min followed by 40 PCR cycles at 94°C for 30 s, 50°C for 1 min, 72°C for 30 s, and a final elongation step at 72°C for 10 min. The expected PCR product size was 143 bp.
For the second nested WNV-specific PCR reaction, 1 µl of the RT-PCR product was added to 50 µl of the total reaction mixture containing 100 pmol of each primer (WNV2+/WNV2−) with the following cycling conditions: 94°C for 3 min and 40 PCR cycles at 94°C for 1 min, 45°C for 3 min, 72°C for 1 min, and a final elongation step at 72°C for 10 min. The expected amplicon size was 328 bp.
After the amplifications, 10 µl of nested-PCR product was loaded into a 2% GelRed (Biotium) stained agarose electrophoresis gel in TBE buffer and visualized under ultraviolet light. Amplicon sizes were determined by comparing them with a 100-bp DNA ladder (Qiagen, Barcelona, Spain).

Sequencing and phylogenetic analysis
PCR products were purified using the QIAquick PCR purification kit (Qiagen, Barcelona, Spain) and bidirectionally sequenced (Macrogen Inc., Amsterdam, the Netherlands). The phylogenetic analysis (Bayesian analysis) was conducted using the beast package 62 , with the TN93 evolution model, a gamma distribution and invariable sites. The clock parameters were set to uncorrelated lognormal, using a coalescent constant size model. The chain length was set at 10 million iterations to produce an ESS (effective sampling size) superior to 200. The maximum credibility tree with branch length in number of substitutions was defined from ten thousand trees after discarding 10% and was edited using FigTree (http://tree.bio.ed.ac.uk/software/ figtree/).

Acknowledgements
We gratefully thank all hospital physicians of the Neurological, Pediatric, Emergency and Neurosurgery Departments and the laboratory personnel in the Microbiology Department at the Fattouma Bourguiba University Hospital of Monastir (Tunisia) for their help in obtaining patient clinical samples and laboratory data; the Ministry of Agriculture (Tunisia) and the Regional Office of Agriculture Development (Monastir, Tunisia) for providing field study permissions, mosquito traps and avian samples within the framework of national passive epidemiological surveillance of avian influenza and West Nile virus in wild birds in Tunisia; and all technicians and staff at the Genomics Unit, Centros Cientificos y Tecnologicos UB (CCiTUB) at the Parc Cientific de Barcelona (PCB), University of Barcelona, for technical support. Abir Monastiri received a fellowship grant from the Ministry of Higher Education and Scientific Research (Tunisia). Funding for this work was provided by the Ministry of Higher Education and Scientific Research (Tunisia).
Author details 1