Spatial Analysis of Japanese Encephalitis Virus Transmission Risk Factors in Singapore

Background Singapore was previously endemic for JEV but the threat of JE remains. There is continued JEV transmission in animal hosts despite the banning of pig farming in 1992. Surveillance of mosquitoes, ardeid birds and swine gives public health ocials useful warning signs for the monitoring of JEV. In this study, we aim to develop a JE Risk Map based on the data gathered by spatial and quantitative survey of the JE-associated vectors and animal hosts. Methods For over a year across the island, the distribution and population of herons were visually surveyed and the distribution of wild boars was captured with wildlife camera traps. Mosquito population data was collected from 16 sites with ecological factors suitable for JE transmission. The animal and vector data were then spatially analysed using open-sourced GIS software. Results Four JE high risk clusters were interpolated from the combined risk factors of data points. A JEV positive Culex tritaeniorynchus pool was subsequently detected in the largest of these JE high risk clusters.


Introduction
Japanese encephalitis virus (JEV) is the most important member of the Japanese encephalitis (JE) serogroup of the genus avivirus, family Flaviviridae with up to 68,000 cases and 20,400 deaths annually.
It is prevalent throughout Eastern and Southern Asia and the Paci c Rim. [1][2][3]. Genetic studies of JEV show that the Indonesia-Malaysia region has all the genotypes of JEV circulating. This suggests that JEV originated from its ancestral virus in the Malay Archipelago and evolved into the 5 different genotypes which spread across Asia [2].
In temperate areas, JEV transmission peaks in summers and falls, while in the subtropics and tropics, JEV transmission occurs all year-round [4]. Proximity to rice elds has been shown to be an important environmental risk factor for JEV transmission, with JEV-infected encephalitis patients 2.93 times more likely to be living near rice elds than non-infected patients [5]. Out of the 1.9 billion people living in rural JE-prone areas, 220 million people live nearby rice irrigation schemes. Richards et al. have shown positive links between the proportion of rice elds and mosquito abundance [5].
JEV can been isolated from over 30 species of mosquitoes but the paddy breeding mosquitoes of the Culex vishnui subgroup, speci cally Cx. tritaeniorhynchus, are the major vectors of the virus. Other species such as Cx. gelidus, Cx. fuscocephala and Cx. annulirostris, have also been implicated as important secondary or regional vectors. Other Culex species that are competent vectors in the laboratory include Cx. quinquefasciatus, Cx. pipiens pallens, Cx. pseudovishnui, Cx. gelidus, Cx. fuscocephala, Cx. annulirostris and Cx. sitiens [6]. There has also been reports of JEV isolates from midges, Forcipomyia (Lasiohelea) taiwana and Culicoides spp. but it is not clear what role these arthropods play in the transmission of JEV [7,8].
Ardeid wading birds are the primary enzootic hosts of JEV. Extensive eld and immunologic studies by Buescher et al. have established the role of the black-crowned night heron (Nycticorax nycticorax), plumed egret (Egretta intermedia), little egret (Egretta garzetta) and the Asian cattle egret (Bubulcus ibis coromandus) in the ecology of JEV [2,[9][10][11]. Domestic and wild pigs and boars have been shown to be amplifying hosts of JEV due the following reasons ; i) JE vectors are highly attracted to pigs, ii) Infection rates of up to 98-100% have been observed in pigs, iii) Pigs can produce high levels of viremia 24 h post infection, which can last up to ve days. In addition, all domestic pigs are able to infect mosquitoes, iv) The turnover of pigs is rapid and the older and presumably JE immune population is replaced with younger pigs [12][13][14][15].
Air transport of infected vectors had been implicated in the introduction of JEV to the Paci c islands of Guam and Saipan [16,17]. With the advent of cheap air travel, we may expect increased risk of JEV introduction into new areas. Migratory birds may also play an expansionary role in JEV spread, as shown in the existence of JEV in India, Pakistan and Nepal where there is little pig farming [18,19]. Mani et al. found that JEV seroconversion rates of children in rice eco-agro systems with presence of herons were 50-56% while systems without herons in close proximity were 0-5% [20]. The East Asian-Australasian (EAA) yway is one of the great yways that migratory birds take during their seasonal migrations.
Migratory herons transit through this yway which passes through JE endemic areas [21].
Climatic condition is a major factor in the geographic and temporal distribution of vectors, their life cycle characteristics and the resultant dispersal patterns of JEV, and its transmission e ciency from vector to host. Effects of climate change such as global warming may increase vector population and increase transmission e ciency leading to rapid disease emergence in new areas [22,23].
Surveillance of mosquitoes, migratory birds, swine, and sentinel chickens gives public health o cials useful warning signs for the monitoring of JEV [24]. Singapore was previously endemic for JEV before the complete phasing out of pig farming in 1992. Although only six cases of JE infection in humans were reported from 1991 to 2005, JEV antibody prevalence in working dogs and chicken were reported in 2004.
Further evidence of JEV seropositivity in ardeid birds, raptors, wild boars and sentinel chickens from 2010 till 2017 demonstrated that JEV transmission in animal hosts has continued despite the abolishment of pig farms [25][26][27].
In this study, we aim to develop a JE Risk Map based on the data gathered by the spatial and quantitative survey of JE-associated vectors and animal hosts. Field data collected from cross-sectional studies at 16 sites that have ecological factors suitable for JE transmission were analysed using GIS. The ecological drivers that we identi ed at these sites include the presence of i) heron nesting, roosting and feeding, ii) wild boar, iii) Culex mosquitoes and iv) the primary JEV vector, Culex tritaeniorynchus species.

Materials And Methods
Distribution and population count of herons.
Preliminary survey of all possible Ardeidae roosting, nesting and feeding areas in Singapore was conducted from 5 September 2013 to 21 January 2014. Choice of locations and their priorities were derived from the island wide Heron Watch survey conducted on 31 August 2013, with the addition of potential estuaries, ponds, golf courses and reservoirs not included in the survey. Census walks were conducted in the mornings (7 am -11 am) and evenings (4 pm -6 pm). Birds were located through direct sightings and movements. When a heron, egret or bittern was sighted, surveyors stopped and noted down: 1) species and count; 2) location plotted on a GPS device; 3) if ying, the direction in which the bird was ying to and from; and 4) whether it was a (potential) roosting and/or nesting site. A roosting site was suspected when many birds were found resting on tree(s) at the same spot and con rmed when the birds were found returning to the roosting spot just before sunset (6.30 pm -7 pm). A nesting site was suspected when juveniles were spotted and con rmed only when nests were observed. A suspected roosting or nesting site was revisited just before sunrise (6.30 am -7.30 am) and/or just before sunset (6 pm -7 pm) for veri cation. When it was not possible to visit at dawn or dusk (e.g. in the case of golf courses which entrance was allowed only at 8am), survey was done at the site at the earliest time possible. As black-crowned night herons are nocturnal, the roosting and nesting sites were veri ed during the day (7.30 am -6 pm).

Population size and range determination
For a more accurate count and range determination, concurrent surveys were conducted at every roosting, nesting and feeding location in areas with more than 5 herons. Due to manpower limitations, the island was divided into four sectors (North-East, North-West, South-East and South-West) and concurrent surveys were done sector by sector. For the purpose of data comparison, all identi ed locations were surveyed once in the peak of the migratory (December to January) and non-migratory (June to July) season respectively. Surveys were conducted in the following periods: 9 December 2013-21 January 2014; 30 June 2014-7 July 2014; and 6 January 2015-13 January 2015. Two surveyors were assigned to each location and counts were made concurrently at sunrise. After which, surveyors observed the 1) species and count; and 2) ight direction of birds at their respective locations until 10 am. Surveys are conducted continually over the course of the year to search for potential roosting and nesting sites and coordinated surveys conducted every half a year to monitor the population and ranging behaviour at the roosting and nesting sites.

Distribution of wild boars
From 2013 to 2015, surveys to assess the presence of wild boar at various sites across mainland Singapore were selected based on previous reliable reports of wild boar, and the presence of large tracts of potential habitat (e.g. unmanaged forest patches or secondary scrub), with a focus on sites located close to urban areas. These sites were selected for camera trapping, where 5-6 Reconyx PC900 camera traps were deployed in selected locations during each session. Camera locations were selected based on areas with evidence of recent wild boar activity, as well as the presence of trees of suitable size for the cameras to be attached. Each camera was secured to the trunk of a tree with a Masterlock Python Adjustable Locking Cable 8413DPF, 40 to 50 centimetres above the ground. Upon activation, camera traps were then left in the eld for 14 days to record images of wild boar. Mosquitoes were trapped using the Centre for Disease Control (CDC) light traps baited with dry ice at ten selected locations in each trapping site. Every trapping session comprised of two consecutive nights and each night lasted for 16 hours which includes two hours of dusk and two hours of dawn.
Mosquitoes were collected after each 16-hour trap night and transported back to the laboratory on ice. These mosquitoes were sorted and identi ed to species when possible and to species group for morphologically similar species. The entire process was performed on ice. Samples of the same species were pooled in groups of 3 and kept at -80°c until further experiments were performed. Spatial analysis of JE risk factors QGIS (Open source software) was used to render the spatial and quantitative eld data into a JE Risk Map. Data layers included the following : 1) Singapore map boundaries, 2) Culex mosquitoes catch data, 3) Heron population and distribution data, and 4) Wild boar distribution data. The quantitative data from the vectors and animal hosts was tabulated into a JE risk matrix (Table 2) which assigns JE risk score for every spatial data point. An interpolation method called Inverse Distance Weighting (IDW) was used to to generate a continuous JE risk map that shows colour-coded risk levels of the different locations.      (Figure 4).

Discussion
The risk map can also be further developed to include spatial and temporal indicators of infection risks. Spatial indicators of infection risks are associated with ecotones which are highly connected locations at the interface of circulating vectors, hosts or pathogens. Temporal indicators deal with biological mutation of pathogens, invasion of vectors, animal host population change, immigration of naïve human hosts, sudden land use and changes, political and economic changes and rapid climatic changes [30]. Detailed and accurate land cover mapping is required to include the role of landscape features in JE risk modelling. With the absence of rice elds as the primary habitat of Culex tritaeniorynchus, we need to locate areas in Singapore where there are similar environmental or landscape features. Waterlogged areas with abundant grass cover can be a potential environmental feature. However more comprehensive ecological surveys are needed to accurately characterise the local habitat of wild Culex tritaeniorynchus.
Remote sensing technology will be essential in the search for such ecologically analogous areas in urban Singapore. Using ecological niche modelling, we can move on to estimate the abundance of Cx. tritaeniorynchus throughout the whole island of Singapore and develop a more accurate JE risk prediction models [31].
Before the last pig farm was shut down in 1992, the main bulk of pig farming activity was located in the north eastern part of Singapore, namely Punggol. The abundance of amplifying hosts and presence of primary vectors would most likely in uence JEV to be previously endemic in this area. More than twenty years later, we are still able to detect active JEV transmission of the Genotype I variant in Coney Island which is located within the vicinity of the historical pig farms [25]. We are unable to ascertain as to whether the detected JEV is related to the strain that was previously endemic in Punggol. However, seroprevalence of local wild pigs in Singapore remains relatively high at 15.2% for a country without recent clinical cases. From our study, it is shown that local birds have JEV seropositivity rate of 3.2% and the JEV antibody is detected mainly in herons and raptors [27]. This indicates possible sylvatic transmission between the vector and both the amplifying and reservoir hosts.
The limited scope of this snapshot study does not account for the seasonality of mosquito abundance. The distribution and abundance pro les can vary temporally, which can be characterised in the scope of a future study. Singapore is a land-scarce country and development often encroach into the natural habitats of the host animals. The animal host distribution and abundance data have to be actively monitored at sites where there is future development. Setting up these active surveillance systems will require substantial resources, as the animals need to be tagged with geolocation devices and mosquitoes are regularly sampled.
It is di cult to assess and predict JE emergence as there are various biotic and abiotic environmental factors that interact synergistically to in uence JEV transmission dynamics. The abolition of pig farming has greatly reduced the impact of the amplifying hosts as a risk factor. However, there are wildlife such as herons and wild boars which continue to be risk factors in JEV transmission. Other factors such as climate change might accelerate JEV transmission [32,33]. The implementation of active JEV surveillance and the modelling of JE risk maps will enable us to predict and mitigate the re-emergence of JE in Singapore.

Conclusion
The spatial distribution of JEV-vectors and animals hosts, and the quantitative data obtained from our cross-sectional studies were integrated and a JE risk map was generated. In the areas where Very High JE Risk clusters were indicated, we were able to locate previous detection of JEV transmission at AS, and a pool of JEV positive mosquitoes was later discovered in Coney Island, a location quite a distance away from AS. We have thus demonstrated the utility of the JE risk map in identifying clusters of JE transmission risk in Singapore.

Declarations
Ethics approval and consent to participate Not applicable

Consent for publication
Not applicable Availability of data and materials The datasets during and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.  Location of wild boars detected by RECONYX PC900 camera traps. All locations re ected on the map present suitable habitat for wild boars. Spatial analysis of JE Risk Factors. The north-eastern part of Singapore represents the largest cluster of highest JE Risk. This is contributed by the abundance of all three Risk factors over a large area. The western part of Singapore carries the highest JE Risk near a wetland reserve in the north and an Avian Sanctuary in the south. The JE hotspot near MR is in the vicinity of the zoo there is a population of freeroaming herons.