Introductions of West Nile Virus Strains to Mexico

Complete genome sequencing of 22 West Nile virus isolates suggested 2 independent introductions into Mexico. A previously identified mouse-attenuated glycosylation variant was introduced into southern Mexico through the southeastern United States, while a common US genotype appears to have been introduced incrementally into northern Mexico through the southwestern United States.

W est Nile virus (WNV), a mosquitoborne flavivirus for which birds serve as reservoir and amplification hosts, was introduced into New York in 1999 (1) and spread across the United States to California by 2003 (2). By 2002, serosurveys demonstrated WNV circulation in >6 eastern Mexican states and along its northern border with the United States (3)(4)(5). This pattern of WNV appearance in Mexico suggested a southwesterly spread across the United States and into northeastern Mexico through Texas. However, in the spring of 2003, the first WNV isolate found in Mexico was obtained from a raven in the southeastern state of Tabasco (3). If WNV reached southern Mexico by incremental spread through northern Mexico from Texas, the index isolate would have been expected sooner and in a northern Mexican state. Phylogenetic analyses showed the raven isolate to be unexpectedly divergent from contemporary Texas strains, but exact relationships and a route of entry could not be determined by using premembrane and envelope glycoprotein (prM-E) sequences (3).
The divergence between the southern Mexican raven and Texas isolates suggested that WNV arrived in southern Mexico by an alternate route, perhaps the Caribbean. After its spread throughout the northeastern United States, WNV appeared abruptly in Florida in 2001, appearing to bypass several mid-Atlantic states. This pattern could be explained by spread of migratory birds (6) (8)(9)(10).
The possibility of a third WNV introduction into Mexico at the California border must also be considered. A 2003 horse isolate from the northern Mexican state of Nuevo Leon was closely related to Texas isolates from 2002 (11), based on its prM-E sequence. We do not know whether WNV reached California from Texas and the Midwest by crossing the Rocky Mountains or by traveling first into northern Mexico and subsequently spreading north from Baja California. The latter route is suggested by the geographic link with the first detection of WNV in southeastern California (2).
The reported incidence of human West Nile encephalitis is much greater on the US (California) than on the Mexican (Baja California and Sonora) side of the common border. Possible explanations for this discrepancy include differences in disease surveillance and reporting. Another possibility is that the WNV strains circulating in Mexico are attenuated compared to US strains, and the identification of a murine-attenuated glycosylation variant in Tabasco State (12) is consistent with this hypothesis.

The Study
To investigate possible routes of WNV entry into Mexico from the United States, 9 isolates from Mexico (all strains available) and 13 strains isolated in the United States from hypothetical points of introduction into Mexico (2 from Florida, 2 from Louisiana, 3 from Arizona, and 6 from California) were compared. Isolates from several northern Mexican states, 1 from Sonora, 1 from Tamaulipas, and 7 from Baja California, were obtained from a variety of birds and from a horse (Table, Figure 1) by injection of Vero cells. RNA was extracted from first or second Vero cell passages by using the QIAamp Viral RNA Mini-kit (Qiagen Inc, Valencia, CA, USA). Reverse transcription-polymerase chain reactions (RT-PCRs) were performed to amplify the complete WNV genome in 6 overlapping amplicons by using primers described previously (12). Amplicons were purified from agarose gels by using the QIAquick gel-extraction kit (Qiagen), and both strands were sequenced directly by using the PCR primers Complete genomic sequences excluding the 5′ and 3′ terminal 20 nucleotides (representing primers incorporated into amplicons) were aligned with all homologous WNV sequences from the GenBank library by using ClustalW. Sequences were analyzed by using maximum parsimony and neighbor-joining programs implemented in the PAUP 4.0 software package (13) as well as Bayesian analysis using MRBAYES v3.0 (14) with 100,000 generations, a general time-reversible model with empirically estimated base frequencies, and either a codon position-specific (for the open reading frame alone) or a gamma distribution of substitution rates among nucleotide sites.
All phylogenetic trees placed the North American WNV isolates into monophyletic groups with strong bootstrap and Bayesian support values; the tree generated using the Bayesian analyses is presented in Figure 2 Surprisingly, despite the greater geographic distances between Tamaulipas and Baja California/Sonora, compared to the distance between Tamaulipas and Texas, the Tamaulipas WNV strains grouped more closely with strains from Baja California and Sonora than with those from Texas.
Compared to the Tabasco strain, the other Mexican isolates differed by 0.55%-0.66% nucleotide sequence divergence across the genome. The gene with the most sequence divergence was prM, with 0.72%-1.4% divergence from the Tabasco strain. However, the 5′ untranslated region was more variable with 3.0%-4.6% divergence. The most conserved gene was NS2B, with 0.0%-0.24% divergence from the Tabasco strain. The E gene, often used for phylogenetic analyses, had 0.46%-0.66% sequence divergence.
Of the Mexican WNV isolates, only the 2003 Tabasco raven isolate had the E-156 Pro residue, which ablates the N-linked glycosylation site found in most North American strains. In addition, 2 other WNV isolates (GenBank accession nos. AY490240 and AF260968) share this E-156 Pro residue despite their geographic diversity (China and Egypt, respectively) and their placement in different lineages. Although the paraphyletic nature of this Pro mutation suggests that it could be selected either during laboratory isolation or passage, its identification in the low-passage Tabasco isolate may indicate its presence in nature.

Conclusions
Our data support the hypothesis that WNV entered Mexico through at least 2 independent introductions. The introduction detected by the first virus isolation in May 2003 from a raven in Tabasco State probably occurred from a migratory bird that flew southward from the southeastern United States to the Gulf of Mexico or the Caribbean and bypassed northern Mexico. The isolation and sequencing of WNV isolates from islands in the Caribbean may shed further light on how WNV reached southern Mexico. However, the extreme genetic conservation of North American WNV strains may preclude identifying the exact routes of introduction. Independently, other WNV strains probably spread incrementally from the southwestern United States into northern Mexico. Both northward and southward movements of WNV between northern Mexico and California or Arizona may also occur.
Available WNV strains from Mexico indicate that the murine-attenuated, E-156 glycosylation-negative variant identified in Tabasco state may be limited in its distribution to southern Mexico, while the glycosylated variant typical of US strains is widespread in northern Mexico. However, our sampling was limited and may also be biased because many WNV isolates were from sick or dying animals; the attenuated E-156 Pro residue phenotype Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 12 could be undersampled because relatively benign infections are rarely identified.
The epidemiology of WNV-associated disease in Mexico is puzzling. According to the Centers for Disease Control and Prevention, 2,470 human cases of WNV infection were confirmed during 2004 in the United States, with >80% of these from areas of California and Arizona bordering the northern Mexico states of Baja California and Sonora where many of our viral isolations were made. In contrast, only 7 human cases of WNV have been confirmed in Mexico. The cases occurred in the border states of Chihuahua (n = 4), Sonora (n = 1), and Nuevo Leon (n = 1) in 2003, and Sonora (n = 1) in 2004 (15). Our results of extremely low sequence divergence between the southwestern United States and northern Mexican WNV isolates indicate that this epidemiologic discrepancy is unlikely to be explained by genetic and phenotypic differences among WNV strains. The possibility that WNV circulating in Mexico has an attenuated phenotype was suggested by the murine-attenuating mutation in the Tabasco raven isolate (12). However, none of our northern Mexico isolates have the E-156-P attenuating mutation, and all appear extremely closely related to isolates made in southwestern areas of the United States with a high disease incidence.
Another possible explanation for the low incidence of WNV disease in Mexico is resistance in the Mexican human population, possibly because cross-protective immunity from other flavivirus infections such as dengue  Nile virus sequences using a Bayesian analysis. Virus strains are labeled by GenBank accession number followed by the state and/or country, year, and host of isolation. Numbers indicate Bayesian probability values followed by neighbor-joining bootstrap values for groups to the right. The tree was rooted by using an outgroup comprised of Old World strains of West Nile virus, including a lineage 2 strain (Table).