Assessment of West Nile Virus Lineage 2 Dynamics in Greece and Future Implications

Introduction

West Nile Virus (WNV) is a positive-sense single-stranded RNA-enveloped virus in the family Flaviviridae within the Japanese Encephalitis serogroup. WNV is transmitted through a mosquito-bird-mosquito transmission cycle with the mosquito genus Culex as the vector and birds as the amplifying hosts. Humans and equine act as dead-end hosts when exposed to mosquito bites that transmit WNV. Incubation of WNV in humans is on average 3–14 days and while most people are asymptomatic, 20% develop West Nile Fever and <1% of total cases develop encephalitis and/or meningitis (Petersen and Marfin 2002). Haussig et al. (2018) have reported that in humans with WNV infection, the median age was 66 years with twice as many infections in males compared to females.

There are currently five geographic WNV lineages that are recognized: lineage 1 (1A: Europe, Africa, Middle East, and North America; 1B: Australia; and 1C: India), 2 (sub-Saharan Africa, Madagascar, and Europe), 3 (Czechia and Austrian border, Rabensburg virus), 4 (Caucasus region of Russia), and 5 (India) (McMullen et al. 2013). Before the summer of 2003, West Nile Virus lineage 2 (WNV-2) was endemic only to sub-Saharan Africa and Madagascar. By 2004, infection of a Northern Goshawk in Hungary confirmed the establishment of WNV-2 in Europe (Bakonyi et al. 2006). Outbreaks of WNV-2 in Greece as an epicenter were subsequently reported in 2010–2013 and 2018 (-2019).

In 2010, a major WNV-2 outbreak in Greece resulted in 262 human infections and 35 deaths (Patsoula et al. 2016). In response, and at the suggestion of the European Centre for Disease Prevention and Control (ECDC), Greece implemented a 3-year national Active Mosquito Surveillance System. Since 2010, alongside the national vector surveillance system, the Hellenic Centre for Disease Control and Prevention (HCDCP) has been actively monitoring WNV, implementing a strategy where all probable and confirmed cases (human, equine, and wild birds) are investigated within 24 h of diagnosis in both the clinical and laboratory setting (Gossner et al. 2017). HCDCP in conjunction with the National School of Public Health, local governments, and universities conducts active yearly vector surveillance between June and October to detect circulation of WNV in mosquitoes (Gossner et al. 2017). If WNV is suspected in animal populations, proper veterinary authorities are alerted; this includes year-round passive surveillance in equine, and at a smaller scale in wild avian species (Gossner et al. 2017). In 2014, ECDC began to provide stand-alone Annual Epidemiological Reports on WNV with retrospective analysis back to 2011 (Gossner et al. 2017). The 2017 ECDC Annual Report stated that WNV transmission in the European Union (EU) primarily occurs between May and June, with peak notification of infection between July and September—occasionally leading into November (ECDC 2019).

The complex biological system of WNV transmission makes it challenging to establish a universal risk assessment for potential outbreaks; this is particularly true when an assessment relies solely on publicly available data. While there are surveillance methods in place for vector, avian/animal, and human hosts, the ability to constantly monitor comprehensively (e.g., all avian host species) is not practical or sustainable. The goal is then to sample at an adequate depth. Weather (e.g., temperature and overall precipitation) is an important factor in the context of the biological lifecycle of the vector for WNV, Culex pipiens mosquitoes (Cotar et al. 2016), but it is also important to evaluate the presence of avian species as potential amplifying hosts and consider natural evolution of the virus. Paz et al. (2013) found uncharacteristic elevated temperatures during the 2010 summer correlating with the increase of number of human WNV cases, but noted the migratory patterns of avian species were not clear in Greece.

Multivariable statistical analyses can be used to study the effects of many factors simultaneously. Typically, although, many of the values such as median infection rate, birth/death rates of avian hosts, and Culex mosquitoes are determined by in vitro laboratory data and do not necessarily originate from publicly available data from open-access organizations. To assess WNV-2 in Greece, it was important to focus on easily accessible data such as daily temperature and precipitation, WNV evolution, and migratory areas for avian host species. Vector surveillance programs initiated by Greece and the ECDC can act as a foundation for future studies and risk assessments for emerging vector-borne diseases.

Materials and Methods

Results

Selection of locations in Greece

Greece has been the epicenter of multiple WNV-2 outbreaks since the emergence of WNV-2 (Fig. 1). The 2019 outbreak occurred between July and November in central Macedonia, near the epicenter of the 2010 epidemic in the Thessaloniki region; however, cases have been reported as far south as Athens. To represent the geographic range of WNV, four major locations were selected for examination (central Macedonia, Larisa, Lamia, and Athens), and grouped into the Northern (central Macedonia and Larisa) or Central (Lamia and Athens) regions. The four locations span 300 km and represent the various climates of Greece.

OPEN IN VIEWER

FIG. 1.

Transmission of WNV in Greece between 2010 and 2019. Monitored cases of WNV in Greece as reported in the end of season reports by ECDC (blue). Greece was generally divided into Northern, Central, and Southern areas for data analysis. The Northern Greece (purple) was central Macedonia (includes Chalkidiki, Imathia, Kilkis, Pella, Serres, and Thessaloniki regions) and Larisa (Larisa; Thessaly). The Central Greece (teal) was Lamia (Phthiotis) and Athens (Attica). ECDC, European Centre for Disease Prevention and Control; WNV, West Nile virus.

Historical climate in Greece

There were no significant differences in temperature and precipitation across the four locations (Fig. 2A–C). Temperatures exhibited annual bell-curve temperatures shifts reaching a maximum between June-August with Athens having an overall higher average recorded temperature than the other cities (Fig. 2A). Total average precipitation appears to be sinusoid with peaks between September and October and December and February (Fig. 2B). On average, there is >3 cm of total precipitation/year for 2–3 years, followed by a decline of total precipitation for 1–2 years (e.g., 2012, 2013, and 2017) based on the data between 2010 and 2019 (Fig. 2C). Despite an increase in total precipitation starting in 2014 for central Macedonia, it has not been associated with a noticeable shift in temperature either monthly or yearly (Fig. 2C, D).

OPEN IN VIEWER

FIG. 2.

Weather data across four regions of Greece. (A) Average monthly temperature from 2010 to 2019. (B) Average monthly total precipitation from 2010 to 2019. (C) Average yearly total precipitation from 2010 to 2019. Each Greek city is distinguished by color and has applied standard deviation. Northern regional locations are central Macedonia and Larisa, and Central regional locations are Lamia and Athens. (D) The yearly average monthly temperature in central Macedonia from 2010 to 2019.

Climate during major WNV outbreaks

The first major outbreak of WNV-2 in Greece occurred in 2010. In all four locations we analyzed, the peak temperatures were recorded in August and then sharply decreased (Fig. 3A). Temperature patterns were more consistent in Northern Greece with less fluctuations each month compared to Central Greece. In contrast, for the 2019 outbreak, while the peak in temperature arrived in August, there was a plateau characteristic that occurred between June and September (Fig. 3B). When comparing the average monthly temperature during 2010 for the two regions, the Central cities had higher recorded temperatures corresponding to their proximity to the Equator (Fig. 3C). The average yearly total precipitation for the two regions appears to invert one another with less total precipitation in the Northern region linking to the two major outbreaks (2010–2013 and 2018–2019) (Fig. 3D). Statistical significance in the yearly precipitation data could not be ascertained between Northern and Central Greece.

OPEN IN VIEWER

FIG. 3.

Trends in WNV outbreak temperature and precipitation. (A) Average monthly temperature during the 2010 WNV outbreak in Greece. (B) Average monthly temperature during the 2019 WNV outbreak. (C) Comparison of the average regional temperature of Northern and Central Greece during the 2010 WNV outbreak. (D) Comparison of the average yearly total precipitation of Northern and Central Greece between 2010 and 2019 with standard deviation reported. (A, B) Each Greek city is distinguished by color. (C, D) The Northern (central Macedonia and Larisa) and Central (Lamia and Athens) regions are distinguished by color.

Reported infections in respect to the change in temperature

Transmission of WNV primarily occurs from May to June with notifications of human infections between July and September. The peak temperature in each Greek location occurs around August; however, the rate of temperature change from May to August appears to have a greater slope (Fig. 2A). By taking the average temperature between May and August of each Greek location, this established the ARTC (°C/month). A trend was observed when comparing the number of WNV cases to the ARTC value (Fig. 4 and Supplementary Fig. S1). When there was a < 2°C/month ARTC value, the number of WNV infections tended to decrease the following year (2013–2015, 2018). However, a shift from >2°C/month ARTC to ≥2.5°C/month resulted in observing an outbreak the following year (2018–2019). During the 2010–2013 outbreak, the ARTC was >2.52°C/month until 2013, whereby a decrease in WNV cases was reported. In 2017, the ARTC was 2.53°C/month with 48 WNV cases and then grew to 311 the following year of 2018.

OPEN IN VIEWER

FIG. 4.

Comparison of reported WNV cases and the change in temperature during peak transmission and reporting between 2010 and 2019 in Greece. Number of reported WNV infection cases as reported by ECDC (red bars) is tracked where no infections were reported between 2015 and 2016. Individual Greek locations (blue closed circle lines) represent the change in temperature between May and August as a rate of °C/month. The average of the four regions (black open circle line) represents the four Greek regions. The threshold of 2.5°C/month is set as the dotted black line.

Range of avian hosts

There are five major avian hosts that have been associated with WNV transmission cycle in Europe: Eurasian magpies, Northern Goshawk, Hooded crows, Sparrows, and Eurasian collared doves (Valiakos et al. 2019). These hosts can be found throughout Europe (Fig. 5); however, the Northern Goshawk and Eurasian magpies do not have as much coverage in Greece compared to their counterparts. The first WNV-2 sequence was obtained from a Northern Goshawk in Hungary, whereas the Eurasian magpie demonstrated local genetic shift in WNV-2 leading to the 2010 outbreak. While an assortment of avian hosts is involved in the WNV transmission lifecycle, it appears that particular species at one given time might support spread of the vector.

OPEN IN VIEWER

FIG. 5.

Migratory spread of common avian hosts of WNV. (A–E) The common (and scientific names) of five WNV avian hosts. (A–E.i) Images of the avian hosts. (A–E.ii) European range of the avian hosts at 100 km grids. (A–E.iii) Lower Italy- and Greece-specific ranges of the avian hosts at 20 km grids. The frequency that the avian species is observed by a violet gradient.

Phylogenetic and phylogeography analysis

A Continuous-Time Markov Chain Bayesian analysis was carried out to examine WNV-2 evolutionary dynamics (Supplementary Tables S1 and S2). Root-to-tip analysis supported molecular clock enforcement (Supplementary Fig. S3). Bayesian inference suggested all WNV-2 sequences share a common ancestor (2004 Hungarian Goshawk). However, Greek-derived sequences were not monophyletic. Interestingly, there was divergence separating the Austrian/Italian and Hungarian/Greek sublineages with posterior probability (PP) of 0.728 (Fig. 6A). One WNV-2 sequence from a 2012 Goshawk in Greece (KC407673) did not group in the Hungarian/Greek clade, but was weakly supported in this position (PP = 0.406).

OPEN IN VIEWER

FIG. 6.

Bayesian inference and Phylogeography Analysis. (A) Reconstructed time-calibrated Continuous-time Markov chain Bayesian MCC phylogeny inferred for the temporal phylogeography of WNV-2. The data are based on a nucleotide Continuous-time Diffusion Model. Taxon labels include the following: Accession Number, Geographic location of isolation, Year of isolation, Strain of WNV, Host isolated from, and Latitude/Longitude. Samples are colored by country origin with the color key represented. Supported clades are indicated by colored branches of the posterior probabilities. Horizontal bars indicate 95% highest posterior density [interval] for the height of the node in the MCC tree. Dotted arrows indicate WNV-2 sequences from Greece that do not directly group with the major clusters of 2010–2013 or 2018. (B) Visualization of the significant paths of WNV-2 overlaid on a geographical map from Google Earth where the colored areas indicate the posterior distributions smoothed using two-dimension kernel density estimation. MCC, maximum clade credibility; WNV-2, West Nile Virus lineage 2.

During the 2010–2013 Greece outbreak, sequences primarily grouped, respectively, to either central Macedonia or Thrace. In addition, the two major sublineages were either derived from the Kavala-Thrace/Greece-2012 (PP = 1.0) or Nea Santa-central Macedonia/Greece-2010 (PP = 0.913) sequences, thus supporting potential multiple introductions (Fig. 6A). Conversely, the MC139-central Macedonia/Greece-2018 sequence grouped with an unknown-Hungary-2017 (PP = 1.0) sequence, whereby the descendants originate. Given most Austrian sequences in the major clade with Italian sequences (PP = 0.996), it would support at least two introductions of WNV-2 into Europe.

The age of the root was dated June 30, 1999 (95% highest posterior density [interval], June 22, 1993, to December 24, 2003) with the upper 95% HPD consistent with the estimated European introduction date of WNV-2 during 2003 (Supplementary Table S3). All relevant statistics yielded acceptable distributions (Supplementary Fig. S2). Lineage-through-time analysis revealed constant diversification of WNV-2 over time (Supplementary Fig. S4A) with two major fluctuations of WNV-2 viral infection around 2005 and 2010 (Supplementary Fig. S4B) based on Gaussian Markov Random Field reconstruction.

A continuous-time diffusion model was selected to examine phylogeography. There were several major significant migratory paths with considerable posterior probabilities (Fig. 6B). Temporal Greek relationships predominately stem from the Hungarian/Austrian region. The distribution demonstrates the diffuse nature of WNV-2 spread in southern Europe.

Discussion

Seasonal outbreaks of WNV in Europe have become an increased concern since the introduction of WNV-2 in 2003. The major WNV outbreaks of 2010–2013 and 2018–2019 have been of lineage 2 descendent. Attempts to model WNV-2 in European nations have revealed a complex system consisting of biological and environmental factors contributing to pathogenesis. Studies by Vogels et al. (2017) have demonstrated temperature, C. pipiens mosquito biotypes, and bird population can affect the reproductive number. Being able to accurately determine future WNV outbreaks is essential, but has been impeded, in part, by the complexity of the biological system.

In this study, four Greek locations spanning Northern and Central Greece, including known hotspots, were examined for potential predictive factors of WNV. Upon examination of temperature and precipitation data, no direct correlations were observed. However, when the average yearly precipitation in Northern Greece was lower than Central Greece, outbreaks were observed. The ARTC appeared to have a potentially predictive behavior to the timing of increased cases. When the ARTC was >2.5°C/month, the current or following year experienced an increase in reported WNV infections. Temperature has been strongly correlated to WNV transmission due to the C. pipiens lifecycle (Groen et al. 2017, Spanoudis et al. 2019); warmer months appear to ignite the transmission of WNV as supported by the ARTC. As the global temperature continues to rise, a variety of vector-borne diseases (e.g., Flavivirus and Alphavirus) are expected to increase in reported cases and location (Negev et al. 2015). While our study used publicly accessible data for analysis, it would be interesting to assess our indicators against in vitro laboratory data and compare outcomes.

Chaintoutis et al. (2019) initially examined the evolutionary dynamics of WNV-2 in Europe and importantly highlighted epidemiological considerations (Chaintoutis et al. 2019). Phylogenetic studies suggest at least three separate introduction events of WNV-2 in Europe. The initial sample was from the Hungary Goshawk in 2004; however, stable transmission pools appear to exist between Hungary/Greece and Austria/Italy—owing to the wetlands and river deltas in Hungary in Austria. WNV-2 phylogenetic relationships in Greece indicate that multiple sublineages are present even within a given outbreak cycle. Two explorable hypotheses are as follows: (1) there are multiple primers of WNV-2 into Greece from neighboring countries and/or (2) a source-sink inversion exists in Greece. The traditional source-sink transmission model would indicate that most or all introductions of WNV in Greece came from surrounding countries (source) where the habitat would be more favorable (e.g., the wetlands in Hungary), and Greece would function as the sink of low-quality habitation (Pulliam 1988). If these were true, ancestral WNV-2 sequences would be basal in the tree topology of the outbreak sequences in Greece. However, a monophyletic cluster is not found during the 2010–2013 WNV-2 outbreak, indicating the virus may have been able to persist without new major introductions. A source-sink inversion model may potentially explain the persistence of WNV-2, where in this theoretical model, the sink environment that would otherwise be of low quality evolves to a more adaptable landscape (Dias 1996). This model can be supported in that the natural habitat of Northern Greece comprises extensive agricultural areas of natural wetland, rice field, and river delta range across Northern Greece, whereby deltas act as a focal point for both native and migratory avian species (Chaskopoulou et al. 2016). This would require a stable population of the vector to exist in Greece. It has been shown that female Culex mosquitoes are able to overwinter (enter reproductive arrest) during colder temperatures in more temperate latitudes, such as Greece (Rudolf et al. 2017). As a result, the combination of overwintered female Culex mosquitoes harboring WNV-2 (and other Arboviruses) and high-quality natural habitat in Northern Greece may allow for persistent outbreaks without requiring the need for outside introductions.

Following a period of increased human cases, there is a marked decrease before rising again. The complex relationship between the vector and avian host may contribute to this observed dynamic. Mosquitoes can transmit genetically diverse WNV quasi-species upon infection of avian hosts; however, due to a strong purifying selection in avians (i.e., anatomical bottlenecks), the genetic diversity is significantly reduced in WNV (Grubaugh and Ebel 2016, Grubaugh et al. 2017). This constriction on diversity is ultimately detrimental to the overall evolution in WNV, yet leaves potentially vector-adaptive WNV variants to be transmitted as a result from strong mosquito-based genetic drift (Grubaugh et al. 2017). Another factor in this complex vector-host relationship is WNV antibodies can protect avian species even at low titers; however, most birds develop undetectable titers 2 years postexposure, whereas juveniles experience a higher rate of antibody decay (McKee et al. 2015). As the population of naive avian hosts grows, alongside with mosquito-driven genetic drift to increase diversity in circulating WNV quasi-species, avian hosts may become more susceptible to infection and propagate a more highly selected WNV variant. Due to the intricacies, further examination is warranted. As a result, the birth-death rate in avian hosts is a valuable measure to consider in statistical models to evaluate the impact of seroconversion surrounding WNV dynamics.

In addition, a new zoonotic arbovirus from sub-Saharan Africa, Usutu virus (USUV), has emerged in Europe (Bakonyi et al. 2007). Similar to WNV-2, USUV uses Culex mosquitoes as its vector and presents clinically similar, although USUV is less associated with severe infections (Bakonyi et al. 2007, Nagy et al. 2019). The first report of USUV in an avian host was in 2001 in Austria (Bakonyi et al. 2007). However, analysis of the first identified human case in 2018 suggests possible co-circulation with WNV (Nagy et al. 2019).

In Greece, effective WNV monitoring with continued early and integrated surveillance of mosquitoes and birds has proven effective (Gossner et al. 2017). Nevertheless, phylogenetic differentiation between WNV-1, WNV-2, and WNV-3 is still lacking. Fortunately, enzyme-linked immunosorbent assays have been developed to distinguish between WNV-1 and WNV-2, and USUV (Reusken et al. 2019) for rapid distinction. Yet, consistent epidemiological surveying on the prevalence of the three WNV lineages has not been robustly reported. This generates a potential public health concern because there are multiple circulating Flaviviridae encephalitic viruses in Europe requiring differential diagnosis, despite their similar clinical management. The development of a multiplex PCR-based assay to distinguish between the multiple WNV lineages should be considered as it would be advantageous, particularly during escalation of human WNV cases. In addition, accessible WNV-1 and WNV-3 lineage near-full length sequences from Europe has been inhibitory toward the progression of ideal phylogeographic modeling.

Conclusion

Greece has served as an excellent model in the surveying of and rapid response to WNV transmission. The implementation of real-time WNV surveillance is critical to timely and effective reporting and to ensuring informed guidance for clinicians and public health professionals, particularly in the face of WNV reintroduction (Phillips et al. 2014). Surveillance of WNV in Greece has improved over the years; however, distinguishing between multiple lineages (as well as various Flaviviridae) and the accumulation of sequence data for phylogenetic modelling could be improved. The continued biological complexity of the transmission cycle for WNV generates an ever-problematic scenario of properly modeling outbreaks. The sustained combined effort of EU nations to enhance WNV surveillance will continue to be epidemiologically and clinically imperative, especially as we look to the impact of ARTC as a potentially predictive indicator for WNV and other vector-borne epidemics in Greece and the lager EU.