Continued Evolution of West Nile Virus, Houston, Texas, USA, 2002–2012

We investigated the genetics and evolution of West Nile virus (WNV) since initial detection in the United States in 1999 on the basis of continual surveillance studies in the Houston, Texas, USA, metropolitan area (Harris County) as a surrogate model for WNV evolution on a national scale. Full-length genomic sequencing of 14 novel 2010–2012 WNV isolates collected from resident birds in Harris County demonstrates emergence of 4 independent genetic groups distinct from historical strains circulating in the greater Houston region since 2002. Phylogenetic and geospatial analyses of the 2012 WNV isolates indicate closer genetic relationship with 2003–2006 Harris County isolates than more recent 2007–2011 isolates. Inferred monophyletic relationships of these groups with several 2006–2009 northeastern US isolates supports potential introduction of a novel WNV strain in Texas since 2010. These results emphasize the need to maintain WNV surveillance activities to better understand WNV transmission dynamics in the United States.


We investigated the genetics and evolution of West
T he emergence of West Nile virus (WNV) in the Western Hemisphere in 1999 poses an ongoing public health threat in North America as the most common cause of epidemic encephalitis in the United States (1). WNV transmission is maintained in an enzootic cycle between mosquitoes and birds; equids, humans, other mammals, and some bird species act as dead-end hosts (2). Human infections are asymptomatic in 80% of cases, and West Nile fever develops in ≈20% of infected patients, which progresses to neuroinvasive disease in <1% (3).
After introduction of WNV in the United States in 1999 (4), local transmission of the original New York genotype (NY99) in resident Culex spp. mosquito and wild bird populations fueled the geographic expansion of WNV from the northeastern region across the continental United States, north into Canada, and south into Central and South America (5)(6)(7). Subsequent introduction into Texas in 2002 resulted in 105 confirmed human infections, high mortality rates among local corvids, and a 31.2% seroconversion rate among resident birds of Harris County, Texas (Houston metropolitan area) alone (8). Uninterrupted surveillance of the related St. Louis encephalitis virus in local Culex spp. mosquito populations by the Harris County Mosquito Control Division since 1964 provided an ideal infrastructure for the expanded detection of WNV activity in the mosquito vector and the wild bird reservoir on a major bird migratory pathway. Routine collections of WNV-positive birds and mosquito pools to date have provided an outstanding opportunity to investigate WNV diversity and evolution on a fine geographic scale comparable to similar surveillance foci in the midwestern and New England regions of the United States (9)(10)(11). However, because of its geographic location, Harris County represents a different ecosystem, namely a warm year-round climate with unique resident mosquito and avian species. Phylogenetic examination of 2002-2004 Harris County isolates confirmed rapid displacement of the NY99 genotype with the novel North American genotype (NA/WN02) in 2002 (12). Fine-scale geospatial genetic comparisons of these isolates provided further evidence of increased WNV genetic diversification in the greater Houston region relative to the homogenous distribution of the now extinct NY99 genotype (12,13). Subsequently, McMullen et al. identified the emergence of the southwestern genotype (SW/WN03) in the southwestern United States in 2003 and positive selection for the encoded NS4A-A85T and NS5-K314R amino acid substitutions in the WNV nonstructural (NS) proteins (14). To date, the NA/WN02 and SW/WN03 genotypes still appear to co-circulate.
Endemic transmission of WNV in the United States since 2006 has shown a dramatic decrease in the confirmed incidence of clinical WNV disease; <1,100 annual human cases were reported during 2008-2011 (15). Despite identification of regional heterogeneous WNV populations, a relative stasis in WNV evolution has been observed in Harris County, consistent with the logistic molecular clock model and a decreasing viral growth rate proposed on a national scale (16)(17)(18). Notably, the current 2012 WNV transmission season demonstrates major divergence from this status quo; >5,600 human infections have been reported nationwide (15). Incidence of clinical WNV disease in the Texas outbreak alone accounted for >33% of the cases in the United States (1,868 cases, including 844 reports of neuroinvasive disease and 89 deaths) and >994 confirmed cases in the greater Dallas/Fort Worth, Texas, metropolitan area and 101 cases in Harris County (15,19). These changes reflect final US and Texas and WNV cases for 2012 reported by the Centers for Disease Control and Prevention (Atlanta, GA, USA) (www.cdc.gov/media/ releases/2013/a0513-west-nile.html) after submission of this report after review. Therefore, studies concerning the continued evolution of WNV in the central and southern United States remain vital for elucidating the role of dynamic genetic heterogeneity and accumulation of novel mutations in the transmission dynamics and incidence of clinical WNV disease.
We report consensus sequence analyses of 17 novel full-length 2010 (n = 1), 2011 (n = 1), and 2012 (n = 15) WNV isolates collected from WNV-positive birds and Culex spp. mosquito pools in Harris County Texas (n = 14) and the greater Dallas/Fort Worth region (n = 3). Inclusion of these new isolates with 28 additional 2002-2009 Harris County WNV isolates in phylogenetic and geospatial analyses of the greater Houston region provides an ideal model for investigating the role of ongoing WNV evolution relative to environmental and clinical incidence reported over the past decade. Furthermore, isolates from the recent WNV epidemic demonstrate closer phylogenetic relationships with original 2002-2003 Harris County isolates, inconsistent with phylogenetic trends observed until 2011 and supporting evidence for the recent introduction of a novel WNV strain(s) in Texas from another geographic region.

Virus Isolates
The World Reference Center for Emerging Viruses and Arboviruses at the University of Texas Medical Branch in Galveston, Texas, supplied the WNV isolates characterized herein provided by the Harris County Mosquito Control Division and Texas Department of State Health Services (Table 1). The fourteen 2010-2012 Harris County isolates were cultured from brain tissue of collected dead WNV-positive birds in Vero cells at the University of Texas Medical Branch in Galveston. Secondary passage of each isolate in Vero cells provided working stocks that were stored at -80°C. These same procedures were used for the 3 Dallas/Fort Worth mosquito isolates.

Genomic Sequencing
Extraction of viral RNA, reverse transcription PCR, and sequencing was conducted according to established protocols (12,14). Resulting sequences were aligned and edited relative to the prototype NY99-flamingo382-99 strain (AF196835) by using ContigExpress in the Vec-torNTI program (Invitrogen, Carlsbad, CA, USA) (4). Sequences were assembled in BioEdit version 7.0.9.0 (20) with 358 full-length North American WNV isolates published in GenBank (as of November 2012) and the IS-98 STD (AF481864) Israeli isolate (21). The encoded 10,299 nt open reading frames (ORFs) for all 372 WNV isolates were aligned by using the MUSCLE algorithm (22); the ORF was used because some viruses published in GenBank include the ORF alone and not the entire genome. Two additional alignments included the ORF of all 42 2002-2012 Harris County WNV isolates with and without the three 2012 Dallas/Fort Worth isolates and the prototype NY99-flamingo382-99 and IS-98 STD reference strains. Screening for potential site-specific positive selection was performed by using the Datamonkey server (23,24) and the single-likelihood ancestor counting, fixed effect likelihood (FEL), and internal branches FEL methods (25,26). Positive selection was defined as d N >d S and a p value <0.05 in >1 method.

Phylogenetic Analyses
Neighbor-joining (NJ) phylogenetic analyses of all alignments were processed in Seaview version 4.3.0 by using the Hasegawa-Kishino-Yano 85 substitution model and 10,000 bootstrap replicates (27). Maximum-likelihood (ML) analyses used RAxML-HPC Blackbox version 7.3.2 (28,29) on the CIPRES Science Gateway version 3.1 server (30) with the generalized time reversible substitution model with invariable sites, a gamma distribution, and 1,000 bootstrap replicates. Bayesian-inferred coalescent phylogenies were produced in BEAST version 1.6.2 (http://beast.bio. ed.ac.uk) by using the generalized time reversible substitution model with invariable sites and a gamma distribution with applied taxa dates, an uncorrelated lognormal relaxed clock model, and Bayesian Skyline prior constraints (31). Resulting BEAST.log and TRE files were down-sampled from triplicate 50,000,000 state runs by using LogCombiner version 1.7.4 and validation in Tracer version 1.5 (31). Inferred phylogenetic trees were edited in FigTree version 1.3.1 (www.mybiosoftware.com/phylogeneticanalysis/2407). In all NJ and ML tree topologies, the IS-98 STD isolate was used as a common phylogenetic outgroup.

Statistical Analyses
The Fischer exact test and post hoc analyses were performed at α = 0.05 (IBM SPSS statistics version 20; IBM, Armonk, NY, USA) to test the association between year of collection and determined phylogenetic groupings.
Adjusted standardized residuals (z scores) at α = 0.05 were compared against the critical z value (± 1.96) with a Bonferroni correction for multiple comparisons.

WNV Collection
A total of 14 WNV isolates from Harris County (Table 1) were examined. Two isolates were collected *Strains sequenced in this study are indicated in boldface. Co., county. †Map code designations include B (bird) and M (mosquito) for isolates collected in Harris County, TX, with phylogenetic relationships (groups 1-10) indicated after the hyphen (see Figure 1). -, isolates collected outside Harris County, Texas, that have no known geographic collection information.
‡Zip codes indicated for region of isolate collection within Harris County, TX (see Figure 1). NA, not available. §Mosquito (Culex spp.) isolates were collected from mosquito pools. The Kuritz (also known as TVP8533; GenBank accession no. AY289214) isolate (Southeast Coastal Texas genotype) was included as a common Texas outgroup, and the TX114 (Bird114; GU827998) isolate served as the prototypic member of the NA/WN02 genotype (12,32).

Nucleotide Changes
Comparison of the 14 novel Harris County isolates with the NY99 (NY99-flamingo382-99) prototype strain identified 45-79 nt differences (0.41%-0.72%) per 11,029nt genome (4). Each isolate encodes 8 of the 13 nt changes characteristic of the NA/WN02 genotype (12). In addition, novel UgC transitions at positions 7015 in the NS4B gene and 8811 in the NS5 gene were conserved in 11 of the 12 Harris County isolates from 2012 (except TX8604).
Superimposition of 38 of the 42 Harris County isolates (4 isolates were excluded because of unknown collection location) on a vector-borne WNV incidence map of Harris County, based on known collection information in the 268 mosquito control operational areas, highlights the shift in WNV circulation patterns across the greater Houston region over the past decade ( Figure 2). Mosquito pool isolates demonstrate robust genetic homogeneity and limited geographic distribution (Figure 2   †Indicated substitutions conserved in >1 representative isolate for each of the proposed phylogenetic groups (see Figure 1).
( Figure 3, panel G). In contrast, the novel 2011 TX8349 Harris County isolate retained topologic distribution within the SW/WN03 genotype. Outside the identified group 7-10 monophyletic clusters, all inferred basal node topologies exhibited poor statistical support (<0.70 posterior probabilities) within the NA/WN02 genotype. Principal support for the identified monophyletic lineages is based on limited nucleotide divergence (<0.65%) and retention of unique substitutions between the 2010 and 2012 Harris County isolates and those from the northeastern United States compared with published isolates from Texas and the southwestern United States. In particular, the group 9 monophyletic lineage (Figure 3, panel E) encodes several conserved substitutions: NS2A-T52I, NS2A-L95F, NS3-S334T, and NS4B-S14I with a single NS2A-H119Y substitution shared between the 2008 New York (WNV-1/US/BID/v4097/2008) outgroup and the 2012 TX8546 isolate. Each isolate also encodes the characteristic E-V159A substitution with the conserved absence of the NS4A-A85T and NS5-K314R substitutions supporting monophyletic distribution and ancestral lineage within the NA/WN02 versus NY99 and SW/ WN03 genotypes.

Harris County Paradigm-Model for WNV Evolution
To evaluate application of the proposed Harris County paradigm outside the greater Houston metropolitan region, we collected 3 WNV isolates from Culex spp. mosquito pools in the recent 2012 Dallas/Fort Worth, Texas WNV outbreak (Table 1) were sequenced and included in a Bayesian phylogenetic analysis with all 42 2002-2012 Harris County isolates. Both Collin County isolates (TX AR12-1648 and TX AR12-10674) exhibited limited nucleotide divergence (<0.30%) and robust monophyletic clustering (>0.98 posterior probabilities) with several group 9 and 10 isolates in the Harris County paradigm (Figure 4). In addition, the TX AR12-1468 Denton County isolate demonstrated shared lineage with the nonclustering TX8572 isolate. Overall, our results support potential application of the Harris County paradigm as a relevant model for WNV evolution in Texas as a whole and, possibly, on a national scale.

Discussion
Surveillance of WNV transmission in Harris County, Texas, provides an excellent model for elucidating the dynamics of endemic and epidemic WNV evolution on a fine geographic scale. The southeastern coastal region of Texas serves as a temporary roosting site on a major flyway for migratory birds in transit between more temperate and tropical regions of the Americas. Consequently, in addition to resident bird populations, this region hosts more avian species (and possible WNV reservoir hosts) than anywhere else in the United States. Prior applications of the proposed Harris County paradigm confirmed emergence of NA/WN02 (12) and SW/WN03 (14) genotypes  Figure 2) supports a heterogeneous transmission model with limited correlation between fine-scale geographic dispersion and sequence divergence over time (9,10). However, regional phylogenetic foci among group 2, 4, and 5 mosquito pools and group 8 avian isolates supports potential homogeneous and trans-seasonal WNV transmission in localized vector populations as documented in other studies (11,13,33).
Before the 2010-2012 phylogenetic analyses, a relative stasis in WNV evolution had been observed in Harris County (16) and the northeastern United States (11,13,17,33) after confirmed emergence of the NA/ WN02 and SW/WN03 genotypes. Fine-scale geographic phylogenetic analyses in suburban Chicago and Illinois have identified maintenance and active evolution of heterogeneous WNV populations in local mosquito and avian vectors from the initial introduction into the United States in 1999 until 2008 but not more recently (9,10,18). Applied genomic and phylogenetic comparisons highlight the major divergence of the fourteen 2010-2012 Harris County isolates from historical dynamics of local WNV evolution and emergence of 3 distinct monophyletic lineages (groups 8-10) in 2012 (Figure 1). However, the 2010 and 2012 Harris County isolates exhibit major sequence divergence (>0.42%) relative to group 1-6 isolates despite shared geographic distribution and environmental conditions. In addition, co-circulation of these novel genetic signatures was confirmed in 3 Culex spp. mosquito pool isolates obtained from the 2012 WNV outbreak in the greater Dallas/Fort Worth region. On the basis of these observations, introduction of a novel or existing strain from the United States into the circulating greater Houston WNV populations, and possibly Texas as a whole, since 2010 offers an alternative explanation for  these divergent genetic signatures. However, our results do not exclude possible emergence of related existent viral populations as dominant regional strains in the recent epidemic WNV transmission season.
To test these hypotheses, we included fourteen 2010-2012 Harris County isolates in a comprehensive Bayesian phylogenetic analysis with all 358 published North American WNV isolates. Shared lineage of group 7-10 isolates with several 2006-2009 strains form the northeastern United States within the NA/WN02 genotype was identified with robust monophyletic support (>0.90 posterior probabilities) (Figure 3, panels B-G). Principal evidence for these monophyletic lineages is highlighted on the basis of limited divergence (<0.65%) and conserved expression of unique substitutions relative to observed genetic diversity (0.8%-1.0%) between published group 1-6 strains from the southwestern United States (14).
Our results support inferred lineage of 2010 and 2012 Harris County isolates from ancestral strains in the northeastern United States, consistent with a single or multiple introduction event(s) during 2010-2012 in the greater Houston region. However, poor statistical confidence (<0.70 posterior probabilities) for all inferred basal node topologies within the NA/WN02 genotype limits direct comparison of independent evolution between Harris County isolates in different monophyletic lineages. Furthermore, consistent phylogenetic grouping of the 2011 TX8349 isolate within the SW/ WN03 genotype indicates co-circulation of this genotype in Harris County until the 2012 transmission season. However, limited clustering of sequenced 2010-2012 Harris County isolates within the SW/WN03 genotype may be an artifact of inherent bias in sample collection.
Sampling bias is a recognized constraint in phylogenetic and paired geospatial analyses. Unfortunately, the predominance of WNV surveillance remains restricted to limited regional foci. Of the 372 full-length WNV isolates in our analyses, 31.5% (n = 117) were collected in Connecticut, 22.0% (n = 82) in New York, 16.4% (n = 61) in Texas, and 8.1% (n = 30) in Illinois during 1999-2012. In contrast, a single isolate has been characterized for each of the midwestern/central states of Michigan, North Dakota, and South Dakota, all of which reported a major increase in the number of clinical WNV cases (n>89) in the 2012 epidemic season not seen since 2007 (15). Increased WNV surveillance in mosquito and avian vector populations across the entire United States is needed to provide critical, unbiased insight into the underlying dynamics of WNV evolution and transmission in US host populations.
The Harris County Public Health and Environmental Services Mosquito Control Division, which was founded in response to a St. Louis encephalitis virus outbreak in 1964, provides a model infrastructure for WNV surveillance in resident mosquito and avian populations in the greater Houston region (8). Uninterrupted collection and processing of WNV-positive bird and mosquito pools since the 2002 introduction of WNV into Harris County has provided a conduit for the scientific investigation of real-time disease outbreaks with direct translation of findings towards optimized vector-borne disease control and prevention. Incorporation of this paradigm in public health directives across the United States would provide a proactive approach towards detection and response to clinical outbreak scenarios of endemic and exotic pathogens in the United States.
In conclusion, retrospective analysis of WNV evolution in Harris County over the past decade indicates a recent shift in the genetic and phylogenetic signature of circulating WNV populations, designated the Harris County paradigm. Further evidence supports introduction of a strain into Texas from the northeastern United States since 2010. Continued WNV surveillance is needed to confirm the effect of this genetic shift in the transmission dynamics and incidence of clinical WNV disease in the greater Houston and surrounding US regions.