Introduction, Spread, and Establishment of West Nile Virus in the Americas

Abstract

The introduction of West Nile virus (WNV) to North America in 1999 and its subsequent rapid spread across the Americas demonstrated the potential impact of arboviral introductions to new regions, and this was reinforced by the subsequent introductions of chikungunya and Zika viruses. Extensive studies of host–pathogen–vector–environment interactions over the past two decades have illuminated many aspects of the ecology and evolution of WNV and other arboviruses, including the potential for pathogen adaptation to hosts and vectors, the influence of climate, land use and host immunity on transmission ecology, and the difficulty in preventing the establishment of a zoonotic pathogen with abundant wildlife reservoirs. Here, we focus on outstanding questions concerning the introduction, spread, and establishment of WNV in the Americas, and what it can teach us about the future of arboviral introductions. Key gaps in our knowledge include the following: viral adaptation and coevolution of hosts, vectors and the virus; the mechanisms and species involved in the large-scale spatial spread of WNV; how weather modulates WNV transmission; the drivers of large-scale variation in enzootic transmission; the ecology of WNV transmission in Latin America; and the relative roles of each component of host–virus–vector interactions in spatial and temporal variation in WNV transmission. Integrative studies that examine multiple factors and mechanisms simultaneously are needed to advance our knowledge of mechanisms driving transmission.

Background

West Nile virus (WNV) is the most widespread flavivirus (Flaviviridae) in the world with evidence of activity on all continents except Antarctica (Karabatsos 1985). It was one of the earliest arboviruses identified to infect humans, first having been isolated in 1937 from the blood of a febrile woman in the West Nile district of Uganda (Smithburn et al. 1940). In the 1950s, it was isolated from apparently healthy children in Egypt (Melnick et al. 1951). Outbreaks were first reported in the 1950s in Israel near Haifa (Bernkopf et al. 1953), followed by Europe in the 1960s (France, Russia), and Eastern Europe in the 1970s (Belarus, Ukraine) (Hayes 1989).

There was little reported WNV activity for the next 20 yr until 1996 when the epidemiology of WNV appeared to change. From 1996 through 2000, human and equine cases were reported in many locations in Europe including Romania, Czechoslovakia, Morocco, Algiers, Tunisia, Italy, Russia, Israel, and France (Murgue et al. 2002). In 1998, WNV was found in migrating storks and domestic geese in Israel (Bin et al. 2001). An outstanding question is whether increased detection of human disease during this period resulted from increases in enzootic transmission of WNV, and if so, whether this contributed to the introduction of the virus to North America. Phylogenetic studies suggested that the WNV strain that emerged in the western hemisphere was most closely related to isolates from Israel (Lanciotti et al. 1999). Here, we synthesize our understanding of the factors that facilitated the introduction, spread and establishment of WNV in the Americas. We discuss the conditions and characteristics of WNV that contributed to its emergence and ongoing success and use previous knowledge to identify outstanding questions that need to be answered to fully understand the future of WNV and other invasive arboviruses.

WNV Introduction to North America

The globalization of trade and travel by humans over the last five centuries (and especially in the last 50 yr) has resulted in the introduction of many arboviruses to new continents, but only a subset of introductions have resulted in establishment and spread. WNV was first detected in New York in 1999, when a cluster of four human cases with unusual symptoms was noted (Briese et al. 1999, Lanciotti et al. 1999). At the same time, birds were observed to be dying (Lanciotti et al. 1999, Steele et al. 2000). After identifying the etiologic agent as WNV, the CDC retested archival serum samples collected in the United States from 1996 to 1999 that were suspected to be due to infection with the closely related St. Louis encephalitis virus (SLEV). However, no evidence of WNV-specific antibody was found, suggesting that WNV was truly a new introduction in 1999 (Gubler et al. 2000). Several other zoonotic arboviruses also have been introduced previously into the United States, including Venezuelan equine encephalitis virus in Texas (Zehmer et al. 1974), Mayaro virus in Louisiana (Calisher et al. 1974), a South American variety of eastern equine encephalitis virus (now named Madariaga virus) in Mississippi (Calisher et al. 1971), and yellow fever virus in California (Prevention 2000). In contrast to WNV, these virus introductions did not lead to establishment and subsequent spread. However, in 2014–2016, a new strain of SLEV, most closely related to a strain isolated in 2005 in Córdoba, Argentina, was introduced and became established in Arizona and California (Diaz et al. 2018). Until then, no evidence of SLEV had been detected in CA following the introduction of WNV (Reisen et al. 2008b). Similarly, Western equine encephalitis virus (WEEV), which was first isolated in CA and often reported in the plains states in the western and central United States, has not been detected in CA since 2007 (Brault et al. 2015). Unlike SLEV and WNV, these two viruses are not known to cross-react immunologically. WEEV also has been rarely detected in Mexico, but in February 2019 it caused an outbreak among equines in that country (ProMED Western equine encephalitis - Mexico: (NA) equine, OIE Archive Number: 20190331.6396453). The factors that facilitate the interactions and re-emergences of these enzootic viruses are still not well understood.

Recently, several other arboviruses for which humans are amplifying hosts, including dengue, chikungunya, and Zika viruses, have been introduced repeatedly into the United States with only limited instances of locally acquired cases. For example, even with thousands of travel-related cases of Zika throughout the contiguous United States since 2015 (https://www.cdc.gov/zika), there has been very limited local transmission of the virus in isolated regions of Texas and Florida (Philip et al. 2019) (CDC https://www.cdc.gov/zika/intheus/texas-update.html). These relatively rare instances of locally acquired cases and/or positive mosquito pools demonstrated that autochthonous transmission following these arbovirus introductions is rare, even though their primary vectors, Aedes aegypti (L.) and Aedes albopictus (Skuse) have been established for centuries and decades, respectively. Given the enormous volume of trade and travel over the last three decades from all corners of the globe, there have almost certainly been other virus introductions that were not detected and did not become established. The lack of establishment of other arboviruses in the United States is likely due to a lack of congruence of biological, environmental, ecological, and socioeconomic drivers of transmission (Smolinski et al. 2003).

In 1999, however, a convergence of intrinsic and extrinsic factors, including many species of naive susceptible avian hosts, moderately competent mosquito vectors (including the previously introduced Culex pipiens complex) that feed disproportionately on a small subset of moderately competent avian hosts (Kilpatrick et al. 2006a, Simpson et al. 2012), and favorable climatological and environmental factors, facilitated the establishment of WNV in the northeastern United States and its subsequent spread. Specifically, the previous introductions of Culex pipiens L. facilitated the successful introduction of WNV to NY (reviewed extensively in Reisen 2013).

Hypotheses on how WNV was introduced into the United States include an imported viremic wild bird, a viremic migratory bird (blown off course; no species regularly migrate between Europe and New York), infected mosquito eggs, larvae, pupae, or adult female mosquitoes hitching a ride on an airplane in the cabin or cargo area, or an infected human (Kilpatrick et al. 2006c). Although prospective and retrospective analyses have been performed for the relative risk of these pathways for the introduction of WNV to Hawaii, Galapagos, and Barbados (Kilpatrick et al. 2004, Kilpatrick et al. 2006c, Douglas et al. 2007), no quantitative analysis of the risk of these pathways has been performed for the 1999 introduction into North America. In addition, the virus has yet to be detected in Hawaii or Galapagos, despite active surveillance and apparently high risk for these locations (estimated number of WNV-infectious mosquitoes arriving each year in Hawaii was 7–70 and in the Galapagos was 0.5–18 (Eastwood et al. 2014, Hofmeister et al. 2015). The lack of detection reflects either lower risk of introduction than previous analyses suggested (e.g., possibly due to lower numbers of live mosquitoes on airplanes currently than measured in the older studies from which calculations were based), or a low probability for establishment in these locations. In Hawaii there are both competent hosts and a mosquito vector. Experimental infection studies suggest that both native (Hawaii amakihi, Chlorodrepanis virens) and non-native (House sparrow, Passer domesticus; house finch, Haemorhous mexicanus; Japanese white-eye, Zosterops japonicus) birds in Hawaii are moderately competent for WNV (~10–32% of mosquitoes biting them over a 5-d viremic period would becoming infectious 14 d later), and local populations of Culex quinquefasciatus Say are moderately competent vectors (Lapointe et al. 2009).

Spread Throughout the Americas

WNV spread across the contiguous United States within the 4 yr following its introduction, and at the same time moved north into Canada and south into Mexico, Central America, and South America (Fig. 1). Phylogenetic analysis of viral isolates detected by public health departments from 2000 to 2012 confirmed that the introduced virus was able to disseminate rapidly across North America with few barriers to spatial spread, which was suggested by the speed of dispersal and lack of genetic structure (Fig. 2) (Di Giallonardo et al. 2016). The pattern of spatial spread was not consistent with movement based on a simple model of bird migration, but rather, seemed related to the nondirectional dispersal of resident species (Rappole et al. 2006). However, the fit of both models to the data was less than ideal and the virus movement models via birds were overly simplistic. Further, there is now abundant evidence that juvenile dispersing birds play a key role in transmission (Hamer et al. 2008) and that migratory birds are also infected with virus (Dusek et al. 2009), suggesting that both dispersing and migratory birds likely contributed to the initial spread. A study that captured the variation in infectiousness among species (Komar et al. 2003, Reisen et al. 2005, Kilpatrick et al. 2007a), temporal patterns of infection prevalence in birds over time, and differential movement of different species would substantially improve our understanding of the spreading process. In addition, a study that combined detection of the virus by public health agencies in each county with phylogenetic analysis of viral isolates would leverage the power of both datasets.

Fig. 1.

Global distribution of WNV in the Americas. Color of the country, state, or province indicates the first year that WNV was detected. For several countries in the Americas no reports on the presence of WNV could be found, but the virus may be present (e.g., Peru, Chile, etc.).

Fig. 2.

Genetic diversification of WNV in the United States, 1999–2013. A maximum likelihood phylogeny of all available WNV sequences from U.S. isolates. The tree is rooted to a 1998 WNV strain from Israel (GenBank Acession# AF481864). Arrows indicate inclusion of variants from corresponding years. Alignments and maximum likelihood analyses were completed with MAFFT and tree editing was completed with FigTree v1.4.3.

Establishment

The establishment and continued transmission of WNV required the virus to overwinter and genetic data strongly suggests the capacity for local maintenance throughout the Americas (Ebel et al. 2004; Amore et al. 2010; Mann et al. 2013a,c; Ehrbar et al. 2017). Substantial research has examined possible mechanisms of persistence including low level transmission, annual reintroduction, overwintering by infected mosquitoes, and chronic infections in avian hosts. Multiple mechanisms likely contribute to viral persistence (Duggal et al. 2019, Reisen and Wheeler 2019).

The establishment of the enzootic transmission of WNV was determined by relationships among the virus, vector and vertebrate hosts, and the environment, including land use and climate (Kramer et al. 2008, Kilpatrick 2011). WNV infects an enormous range of vertebrates including most birds, mammals, and several amphibians and reptiles (Klenk and Komar 2003, Komar et al. 2003, Kilpatrick et al. 2007a, Gómez et al. 2008). However, species differ enormously in their host competence for WNV. There is substantial variation among birds (with more variability among families than within them) (Kilpatrick et al. 2006a), ranging from two species being refractory to infection (budgerigars and Japanese quail) to species having sustained high viremias that result in an average of 63% of mosquitoes (tufted titmouse) becoming infectious after feeding on birds over a 5-d period (Komar et al. 2003; Reisen et al. 2005; Kilpatrick et al. 2007a, 2013). Most mammals, reptiles, and amphibians are either incompetent or weakly competent, with a few interesting exceptions (Kilpatrick et al. 2007a). Although the role played by different host species in North America has received significant study, the key hosts for WNV amplification in Latin America are almost completely unknown.

An early question following WNV establishment had been why WNV, but not SLEV, caused yearly widespread outbreaks in the United States. Initially, SLEV was mistakenly thought to be the etiologic agent of WN disease detected in New York City in 1999. This was understandable, as WNV and SLEV are closely related genetically and cross-react serologically, with both belonging to the Japanese encephalitis antigenic complex. They also have very similar transmission ecology, with Culex mosquitoes serving as the most important vectors and a similar set of avian species serving as amplifying hosts. However, biological differences likely explain the vast difference in transmission between these two viruses, that is, avian SLEV infections result in lower viremia levels that are less infectious to mosquitoes (despite higher susceptibility of mosquitoes) and lower virulence which results in more birds surviving and being immune following infection (Reisen et al. 2005). Although experimental studies have demonstrated that WNV is superior to SLEV in its capacity to retain high levels of host-specific fitness without sacrificing fitness in alternate hosts (Ciota et al. 2007, 2008, 2009), it is still not entirely clear what the mechanistic basis for this is. Understanding a virus’s evolutionary potential and constraints is critical to predicting the epidemiological trajectories of invasive arboviruses and further comparative studies between WNV and SLEV could help identify characteristics of WNV that contribute to its remarkable success as a generalist.

WNV infects a wide range of mosquito species, with approximately 60 species being found infected in North America alone (Kramer et al. 2008). However, as with vertebrate hosts, there is enormous variation in vector competence both among and within species (Turell et al. 2001, 2005; Vaidyanathan and Scott 2007; Kilpatrick et al. 2010; Ciota and Kramer 2013). There has yet to be a quantitative synthesis of all vector competence experiments performed for WNV, and this might offer insight into taxonomic patterns, including which genera are more competent, and whether mosquito species that are more frequently exposed to WNV (and therefore would select on the virus) have greater competence. A synthesis might also address a key outstanding question: what factors determine the enormous variation that occurs in vector competence within a single species (e.g., 0–52% of Cx. pipiens populations transmitting WNV after feeding on identical doses and strains; 1,000-fold range in viremias needed to infect 50% of Culex tarsalis Coquillet and Cx. pipiens) (Reisen et al. 2008a, Kilpatrick et al. 2010). Current research exploring the use of genetically modified mosquitoes and Wolbachia to modulate competence may offer some insight into the role of the genetics and the microbiome in vector competence, yet this work has largely been focused on Aedes (Hoffmann et al. 2011, Kean et al. 2015, Huang et al. 2017). Future studies identifying unique genetic and microbial signatures in phenotypically distinct Culex populations could help clarify the mechanisms of vector competence driving variation.

Mosquito feeding preferences (or more accurately, host utilization since host defense can play an important role in mosquito feeding success) play key roles in spatial and temporal variation in the transmission of WNV. Strong preferences (disproportionate feeding relative to host relative abundance) can increase transmission and focus it on a relatively small subset of the host community, and strong feeding preferences have been found whenever mosquito feeding patterns have been paired with host abundance data (Woolhouse et al. 1997; Hassan et al. 2003; Kilpatrick et al. 2006a; Hamer et al. 2009, 2011; Kent et al. 2009; Simpson et al. 2012; Levine et al. 2016). Because birds are far more infectious than other vertebrates, ornithophilic mosquitoes, primarily in the genus Culex, are far more important in enzootic transmission than mammalophillic mosquitoes in the genera Aedes and Anopheles. A key outstanding question is whether WNV will adapt to specific hosts that are highly preferred by mosquito vectors (e.g., American robins) (Kilpatrick et al. 2006b; Kent et al. 2009) as the virus has to house sparrows (Duggal et al. 2014), or whether costs of specialization outweigh the potential benefits, especially given the variation in the importance of host species contributing to WNV transmission in different regions, years, and seasons.

Abundance of key mosquito species also plays a role in WNV establishment and modulates the transmission of WNV both among avian hosts and in spillover transmission to humans. Many studies have linked entomological risk (the density of infected mosquitoes) with variation in the number of human WNV cases over space and time (Kwan et al. 2012, Kilpatrick and Pape 2013), and much of the variation in entomological risk is due to variation in mosquito abundance. Mosquito abundance is determined by a combination of land use and land cover, which influences larval habitat availability, mosquito competitors and predators, host availability, and weather, which alters larval habitats and the rate of mosquito life history processes (Reisen 1995, Wekesa et al. 1997, Andreadis et al. 2004, Barker et al. 2010, Ciota et al. 2014, Ruybal et al. 2016). Consequently, WNV transmission is positively correlated with increased agricultural land use in the western U.S. states, and increased urbanization throughout the United States, ostensibly because these land uses increase larval habitat for Cx. tarsalis and Cx. pipiens complex, respectively (Bowden et al. 2011, Kovach and Kilpatrick 2018). A key unanswered question is the relative roles of variation in mosquito abundance, survival and feeding patterns, and avian host abundance (and resulting community competence) in spatial patterns of WNV transmission. The increase in urbanization in the United States (and globally) and shifts in agricultural land use make this question important for predicting future WNV transmission as climate changes.

Many studies have examined the influence of weather on mosquitoes and WNV to predict future outbreaks and the potential impacts of climate change. There is strong evidence for the role of temperature in the seasonality of WNV, with a short 3-mo long transmission season in northern North America that increases to an 8-mo long transmission season in Florida and other southern states (Fig. 3). Strong correlations with temperature exist on a weekly time-scale (Ruiz et al. 2010), and the WNV transmission season is relatively predictable from year-to-year, with the peak of human WNV cases generally occurring during a 2-mo period (Fig. 4; August–September). However, year-to-year variation in WNV epidemics across the United States were less correlated with temperature and more strongly correlated with drought and human immunity from previous WNV infection (Paull et al. 2017).

Fig. 3.

WNV transmission season length as measured by reports of WNV-infected dead birds. Dates of first and last WNV-infected dead birds in states along the east coast of the United States in 2003 by state and county departments of public health as reported to the CDC. For example, the first WNV-infected dead bird in Florida was June 6 and the last WNV-infected dead bird was reported on December 18. Two letter abbreviations indicate the state.

Fig. 4.

Seasonality of average number of human WNV encephalitis cases in six states along a north-south gradient during 2003–2009. Two letter abbreviations indicate the state.

An outstanding question is exactly how climate modulates yearly variation in WNV transmission. Although a large number of studies have correlated many climate variables with human WNV cases (and in some cases with mosquito abundance or mosquito infection prevalence) (Keyel et al. 2019), these correlative studies have not produced a mechanistic understanding. For example, a comprehensive model of the temperature dependence of the reproductive ratio (R₀) of WNV, parameterized with detailed laboratory studies of temperature effects on mosquito survival, biting rate, and extrinsic incubation period, was significantly correlated with yearly variation in human WNV cases in only 11 of 44 states, even after controlling for other climate drivers and human immunity (Paull et al. 2017). Furthermore, drought, the strongest correlate of the climatic drivers examined in this study, was not correlated with mosquito abundance, but was significantly correlated with WNV prevalence in mosquitoes, suggesting that simple explanations for the role of drought in influencing WNV transmission (e.g., by providing organically rich larval habitat) was insufficient (Paz 2015, Paull et al. 2017). Overall, the wealth of research on weather-WNV relationships indicate that more detailed studies that encompass the full transmission ecology of WNV, including hosts, vectors, and the pathogen are necessary to understand climate–WNV relationships. Further, studies that examine variation in the enzootic cycle (i.e., the density of infected mosquitoes, and avian exposure) are needed to separate factors that influence enzootic transmission from those that alter human cases.

Viral genetics also play a role in the establishment and transmission of arboviruses. Although the reproductive ratio of the introduced strain of WNV (NY99) was more than sufficient for establishment and spread, substantial evolution occurred (Ebel et al. 2004, Pybus et al. 2012, Mann et al. 2013a, Di Giallonardo et al. 2016, Bialosuknia et al. 2019) (Fig. 2), and some changes had measurable fitness effects in some populations of hosts, vectors, or both (Moudy et al. 2007, Duggal et al. 2014) that increased with temperature (Kilpatrick et al. 2008). A single, shared amino acid substitution in the envelope protein, V159A, was associated with the displacement of the NY99 genotype by the WNV02 genotype and has been correlated with variation in vector competence in some populations of both Cx. pipiens and Cx. tarsalis (Ebel et al. 2004, Davis et al. 2005), likely contributing to the prolific spread of WNV westward through the United States (Moudy et al. 2007). Although genetic diversification initially declined following peak WNV activity in 2003 (Snapinn et al. 2007), there have been repeated resurgences of WNV activity throughout the United States (www.cdc.gov/westnile) and continued WNV evolution (Mann et al. 2013b, Di Giallonardo et al. 2016, Ehrbar et al. 2017, Bialosuknia et al. 2019). Broadly, there is substantial evidence that genetic variation is repeatedly generated at local scales and that some successful viral genotypes can spread rapidly across large spatial scales, resulting in very low signatures of genetic isolation by distance and the presence of multiple clades in relatively small regions (Amore et al. 2010, Di Giallonardo et al. 2016, Nelson et al. 2018, Bialosuknia et al. 2019). Overall, the most comprehensive and recent analyses suggest that clades and associated mutations often originate far from where they are subsequently detected (e.g., viruses in SW03 clade were thought to originate in the southwestern United States in 2003 but were subsequently detected in New York in 2002) (Bialosuknia et al. 2019). In addition, although previous studies generally have demonstrated unfettered movement of WNV with little adaptive evolution (Di Giallonardo et al. 2016), a recent comprehensive analysis of WNV strains isolated from 1999 to 2015 identified evidence of positive selection at 12 positions in the WNV genome. In addition, novel genotypes have shown evidence of displacement in New York beginning in 2010 (Bialosuknia et al. 2019).

Future studies should assess variability in vector and host populations and genotype-by-genotype interactions using experimental infections of mosquitoes and avian hosts with multiple viral isolates (Duggal et al. 2014) to understand how local interactions are contributing to WNV evolution and transmission. Human WNV cases, as well as infection prevalence in hosts and vectors, have fluctuated enormously over the past two decades across the United States (with a peak in cases in 2018 in the northeastern USA; www.cdc.gov/westnile) and only a portion of this variance has been explained by weather and host immunity (Kwan et al. 2012, Paull et al. 2017). Further, there have been few studies examining drivers of enzootic WNV transmission (as opposed to human cases) across large spatial scales. The importance of ongoing viral evolution in WNV transmission will require more extensive viral sequencing and phenotypic studies in vectors and hosts and analyses which link spatio-temporal patterns of transmission with viral genotypes and phenotypes.

Conclusions

The introduction of WNV has forever changed the ecology of the Americas, with impacts on human, veterinary, and wildlife health and cascading impacts on other viruses and the ecosystem. The enormous quantity of research on WNV over the past 20 yr have identified key issues to anticipate when the next virus is introduced, including the importance of host–vector–virus and climate interactions, the potential for viral evolution, and the potential for rapid spread and establishment of a new virus if competent host and vectors are present. Our aim here was to synthesize knowledge on several aspects of WNV ecology and evolution to highlight key outstanding questions regarding the introduction, spread, establishment, and ongoing transmission of WNV, so we can continue to learn from this natural experiment.