INTRODUCTION
West Nile virus (WNV; family Flaviviridae, genus Flavivirus) is a mosquito-borne virus, belonging to the Japanese encephalitis virus (JEV) serogroup. It is amplified through transmission cycles involving birds as main reservoir hosts and mosquitoes of the genus Culex as the principal vectors. WNV, first isolated in 1937 in the West Nile district of Uganda, is now widely distributed throughout different continents [Reference Gubler, Kuno, Markoff and Knipe1–Reference Trock3].
On the basis of phylogenetic analysis, WNV has been classified into at least five lineages. Lineage 1 WNV is widespread and most commonly associated with human and horse neurological disease. This lineage has at least two geographically distinct clades. Clade 1a viruses are circulating between Europe, the Middle East, Africa and America via migratory birds [Reference Lanciotti4, Reference Bondre5].
From the early 1950s to the 1990s, several outbreaks, rarely associated with clinical disease in horses and humans, occurred in the Middle East and the Mediterranean Basin. Concurrent serological studies demonstrated that WNV was endemic in these regions [Reference Bernkopf, Levine and Nerson6–Reference Work, Hurlbut and Taylor8]. Nevertheless, since the 1990s, several outbreaks of neuroinvasive disease occurred in the Middle East and the Mediterranean Basin, extending to Russia and the USA [Reference Bernkopf, Levine and Nerson6, Reference Dauphin9–Reference Pollock11]. Clade 1a viruses were responsible for these outbreaks; one subclade, associated with considerable bird mortality, caused the recent outbreaks in Israel and the American continent. This subclade subdivision suggested that a virus imported from the Middle East may have caused the recent outbreaks in the USA [Reference Lanciotti4, Reference Charrel12].
The first serological studies conducted before the 1990s, demonstrated the presence of WNV antibodies in the Iranian population. In a study reported in 1969, carried out in Khorasan and Khuzestan provinces, WNV antibody was detected in 28·4% of human sera (103/768) [Reference Naficy and Saidi13]. A study conducted in 2010 demonstrated the presence of WNV IgG in 5% of blood donors in Tehran [Reference Sharifi, Mahmoodian Shooshtari and Talebian14]. While these studies clearly demonstrate human exposure to WNV in Iran, similar findings have not been reported in equines.
The study objective was to assess the level of WNV circulation in the equine population in the Islamic Republic of Iran. A large-scale serosurvey was implemented to evaluate the overall seroprevalence rate in horses, as well as any geographic variations in provinces.
Method
Target and study population
There are around 1 547 100 horses distributed throughout Iran that are involved in sport, farming, educational purposes and local transportation. The overall animal density is about 1 animal/km2 and the geographic distribution of the animals is heterogeneous (Fig. 1).
Fig. 1. Geographic distribution of equine population by province and geographic location of collected samples in 27 out of 30 provinces, Iran, 2008–2009.
A total of 1054 healthy equines (844 horses, 9 mules, 201 donkeys) were randomly selected from 260 communities located within 27 of 30 provinces in the country (Fig. 1). The survey was performed between September 2008 and February 2009. A form was completed by a veterinarian for each animal. Along with other data (date of sampling, name of the veterinarian and of the animal's keeper), this form recorded the breed of the animal, its age and sex, the location (easting, northing and elevation), the type of stable (farm or equestrian club) and the group size, i.e. the number of equines living at the same place.
Serological analyses
Enzyme-linked immunosorbent assay (ELISA)
All equine sera were screened for WNV antibodies with a commercial competitive ELISA test (ID Screen West Nile Competition; ID VET, France) according to the manufacturer's instructions.
Reactive serum samples were further tested by IgM-capture ELISA and plaque reduction neutralization test (PRNT). The analysis method for IgM detection was derived from the IgM-capture ELISA method described previously [15] and was performed as described by Murgue et al. [Reference Murgue16]. Plates were read on an ELISA plate reader at the 450 nm wavelength. The following ratios were calculated for each sample:
R1 is the ratio between the absorbance of the test sample in wells containing virus antigen [Ag(+)] and the absorbance of the sample in wells containing control antigen [Ag(−)]. R2 is the ratio between the absorbance of the sample in wells containing virus antigen and the absorbance of negative control serum [C(−)] in wells containing virus antigen. If R1 ⩾2 and R2 ⩾2, the sample was considered IgM positive, if R1 <2 and R2 <2, the sample was considered IgM negative and if R1 ⩾2 and R2 <2, the sample was considered doubtful.
PRNT
PRNT90 was performed to confirm the presence of WNV-specific antibodies and was derived from the OIE PRNT90 method [15], with threefold serially diluted sera and WNV lineage 1 IS98-ST1 strain [Reference Lucas17]. Titres were expressed as the reciprocal of the highest serum dilution yielding >90% reduction in the number of viral plaques (PRNT90).
Statistical analyses
Seropositivity risk factor analysis
All the statistical analyses were conducted using R 2·10·1 [18], and maps were produced using a geographic information system (ESRI1, ArcGISTM 9·0). Some values were missing in questionnaires for sex, breed, age and group size. After verifying that the pattern of missing data was compatible with random missing [Reference Rubin19], missing data were imputed using the covariates as predictors (northing, easting, elevation, sex, breed, age, group size, type of stable). Twenty datasets were thus generated using the ‘aregImpute’ function of the Hmisc package [Reference Horton and Kleinman20]. All the statistical analyses described hereafter were separately conducted for each of these 20 datasets, the results being combined as described previously [Reference Rubin19].
A logistic model was performed to analyse the relationship between seropositivity and the covariates. Northing, easting, the corresponding quadratic terms and the northing–easting interaction were included in the model as a proxy for horses' location. Other covariates were grouped by gender (male or female), breed (Arab, Crossbreed, Thoroughbred, others), age (Δ=5 years), group size (Δ=30 animals), elevation (Δ=1000 m), and the type of stable (farm or club). The performance of the model as a classifier was assessed using receiver-operating characteristic (ROC) analysis and computing the area under the ROC curve (AUC). A cross-validation [Reference Stone21] was conducted: half of the tested horses were randomly chosen for fitting the model, which was then used to predict the serological status of the remaining animals.
The prediction error was evaluated as the usual proportion of incorrect predictions.
RESULTS
Out of 1054 sampled animals 249 [23·6%, 95% confidence interval (CI) 21·1–26·3) were positive by PRNT for WNV infection. Three sera could not be tested by PRNT owing to insufficient sample volume. Seropositive equines were observed in all of the provinces except one, where only two animals had been tested. Seroprevalence appeared to be highly heterogeneous and ranged from 1% to 88%. The most affected provinces were located in the western and southern parts of the country, with seroprevalence rates >30% (Fig. 2). Most of the animals possessing WNV neutralizing antibodies had low PRNT titres as only 73 had a titre >90, corresponding to 29·3% of tested animals (95% CI 23·6–35). The majority of these animals were located in provinces with seroprevalence rates >30%. Two sera were not tested by IgM ELISA because there was insufficient sample volume and one serum was removed because it gave an indeterminate result. Nine sera were positive for IgM antibody. The corresponding four provinces had a PRNT seroprevalence rate >30%.
Fig. 2. Anti-WNV seroprevalence determined by plaque reduction neutralization test in equines in 27 out of 30 provinces, Iran, 2008–2009.
A logistic regression model showed an association between elevation and seropositivity, with a 60% decrease of seropositivity odds when elevation is increased by 1000 m (Table 1). The Arab breed was a seropositive risk factor, as was being housed in a farm (reference: club). The seropositivity risk also increased with the age of animals, a difference of 5 years corresponding to a change of seropositivity odds by a factor of 1·3.
Table 1. Logistic model of anti-WNV seropositivity determined by plaque reduction neutralization test in equines, Iran, 2008–2009
OR, Odds ratio; CI, confidence interval.
* Absolute value.
† Quadratic term.
No significant association was observed between sex and seropositivity, or between group size and seropositivity (Table 1). The ROC analysis resulted in an AUC of 0·887, indicating the model had relatively good discrimination ability. The AUC ranged from 0·884 to 0·889 according to the imputed dataset (n=20). Cross-validation showed an individual prediction error of 13·5%: for 86·5% of animals, the predicted PRNT status was thus correct. This prediction error ranged between 12·2% and 14·5% according to the imputed dataset (n=20).
DISCUSSION
The results of this study, the first large-scale WNV serosurvey in equines conducted in Iran, show that antibody response to WNV is widespread in this country as seropositive animals were found in 26 out of 27 provinces. The overall serological prevalence rate of WNV antibodies is high (24%) but varies widely according to the geographic location, with southern and western parts of the country being the most affected areas. Nine provinces had a prevalence ⩾30%, five provinces a prevalence ⩾50%, and two provinces a prevalence ⩾80%. We could analyse a double gradient of prevalence between north and south, and between east and west. The greater part of the provinces with high prevalence is located in a dry area (south) rather than a rainy area (north).
Logistic modelling confirmed that geographic location was the main seropositivity risk factor. Furthermore, seropositivity risk decreased with increasing elevation: this association can be explained by variations in the abundance of the vectors. Seroprevalence also varied with the equines' individual characteristics. Horses had a higher risk for WNV seropositivity when they were Arab breeds. This may be linked to the geographic location of these horses, found mainly in the southwestern part of the country. The increase in seroprevalence rate with equine age suggests recurrent circulation of WNV, as described in some serosurveys performed in endemic areas [Reference Chevalier22, Reference Cohen23]. Considering the fact that seropositive animals were more than 1 year old, seropositivity may not be related to maternal antibodies.
Besides individual characteristics, seroprevalence varied according to the conditions in which animals were managed. For example, horses housed in farms appeared to have been more exposed to WNV than those in clubs.
The number of IgM-positive sera observed in the present study (n=9) could have been underestimated as only sera positive by competitive ELISA were screened for IgM. The corresponding bias was probably minor as IgG and IgM appear roughly simultaneously after infection [Reference Castillo-Olivares and Wood24, Reference Bunning25]. Despite the small number, the positive IgM results revealed a pattern of viral circulation. The corresponding animals were sampled between September 2008 and January 2009 and could thus have been infected from late summer to autumn 2008. Similar observations were made in Europe and the Mediterranean area, with recurring outbreaks in the same seasons that the mosquito population density was highest [Reference Trock3, Reference Murgue16, Reference Tsai26]. The latest IgM-positive serum was obtained at the end of January 2009, suggesting that the animal was infected during the last days of autumn or during winter. This animal was located in a desert area, Sistan-Baluchestan province. Considering the mild winter in this province, it would not be surprising to detect active, infected mosquitoes at that time of year.
A few sera positive by ELISA were negative for neutralizing antibodies against WNV. The lower analytical sensitivity of the PRNT test can account for this finding. This result was unlikely to be due to assay cross-reactivity as no other related flaviviruses have been reported in Iran.
Circulation and persistence of viruses in the ecosystem depends on environmental, as well as biological, factors. WNV is maintained in nature through a transmission cycle between vector mosquitoes and reservoir birds [Reference Komar27–Reference Jourdain29]. It appears that Khuzestan province has more favourable conditions for propagation and circulation of the virus than the other provinces. Khuzestan has the most important wetlands which host unique flora and fauna. The importance of migratory birds as virus carriers and transfer hosts between different areas has been previously reported [Reference Hubalek30, Reference Reed31]. Iran is located at the crossroads of bird migratory routes, with birds coming from WNV-endemic regions. Although pathways of migratory birds in Iran are not well documented, Iran supports nearly two-fifths of the Middle East's important wetlands. The Iranian wetlands represent wintering areas for millions of migratory waterfowl coming from Siberia, Africa, India and South Asia which may introduce different strains of WNV encountered elsewhere.
Since the first reported serological evidence of WNV circulation in the 1950s, the Middle East has been of interest for renewed studies concerning WNV circulation, especially when it was shown that there was a genetic similarity between the New York strain isolated in 1999 and a strain isolated in the Middle East [Reference Lanciotti4]. Co-circulation of different lineages has been previously reported in the Middle East [Reference Briese32]. However, substantiating the circulation of different lineages in Iran needs further study. Efforts are needed to isolate viruses, to identify vectors and vertebrate hosts, to assess the risk of WNV transmission by migratory birds to free ecosystems and also to determine the environmental indicators involved in WNV transmission in Iran.
ACKNOWLEDGEMENTS
We thank all the veterinarians who were involved in field support for this study. We are grateful to J. Maingault, M. Kamyabi, S. Rahimizadeh, S. Jodairy-Eslami, N. Miandehi, A. Moradi-Joshaghan, H. Amiri, H. Najafi. C. Bahuon, M. Chaumien and E. Mazaheri for help with laboratory analyses and their technical assistance. We thank Philippe Pourquier (ID Vet Company) for kindly providing the ID SCREEN® competitive ELISA kits. We also thank Francis Geiger and Marc Grandadam for their help in the preliminary study and Mark Zuckerman for editing and improving the manuscript. The first author also thanks Fereshteh Firouzi for the management of the visiting periods in France. This work was supported in part by a grant from the Service de Coopération et d'Action Culturelle (SCAC) of the French Embassy in Tehran, Iran and the Center for International Scientific Studies and Collaboration (CISSC) Ministry of Science, Research and Technology of Iran.
DECLARATION OF INTEREST
None.
