A total of 441 equine serum samples from 14 states of Mexico were tested (Figure 1). WNV-specific antibodies were detected in 97 (22%) of the samples. These data probably overestimate the true equine seropositivity rate because sampling focused on herds with a history of clinical encephalitis. Representative data from 22 of the WNV-positive samples obtained in five different states are presented in Table 1. No evidence was obtained of SLEV infection in equines, but VEEV-specific antibodies were detected in serum samples from the states of Veracruz and Yucatán. (Horses are vaccinated against VEEV in the states of Chiapas and Oaxaca, so samples from these locations were not tested for VEEV antibodies.) The positive samples, several of which also contained WNV IgG, represent natural VEEV circulation and infection of horses, presumably with enzootic subtype IE strains (3,4).

Figure 2

Figure 2. Phylogenetic tree, derived by maximum likelihood (ML) by using prM-E sequences for West Nile virus (WNV) isolates including the 2003 Mexican Raven isolate. Strains are indicated by country or abbreviated state...

On May 5, 2003, the CPA-SAGARPA received a report of a dead Common Raven from the El Yumka wildlife preserve in the city of Villahermosa, state of Tabasco. Although this species is native to Tabasco and other regions of Mexico, this bird was one of two ravens imported from the United States in 1999. A necropsy was performed, and virus isolation was attempted on tissue samples at the CPA-SAGARPA biosafety level 3 facility in Palo Alto, Mexico City. On May 16, 2003, cytopathic effects were detected in Vero cells injected with brain suspension. Viral RNA from the isolate was genetically characterized at the National Institute for Epidemiology and Diagnostics (InDre) in Mexico City and at the University of Texas Medical Branch in Galveston. A 2,004-nt genome portion, including the prM-E protein region, was amplified by using a reverse transcription–polymerase chain reaction as described previously (5); the resulting amplicons were sequenced directly with the Big Dye sequencing kit and model 3100 sequencer (Applied Biosystems, Foster City, CA). The sequence of this WNV isolate (TM171-03, submitted to GenBank under accession no. AY371271) was aligned with all homologous WNV sequences of the same length available from the GenBank library (homologous to nt 466–2,469 in the Flamingo382 strain, GenBank accession no. AF196835), and phylogenetic trees were constructed by using maximum parsimony, maximum likelihood (incorporating empirical base frequencies, a general time-reversible substitution model with the following frequencies: A→C 1.34263; A→G 4.18575; A→U 1.55497; C→G 0.044980; C→T 15.08737; G→U 1.00000, a γ shape parameter of 0.228), and neighbor joining programs implemented in the PAUP 4.0 software package (6). All trees had nearly identical branching orders; the maximum parsimony tree is presented in Figure 2. All trees placed the Mexican raven isolate as a sister to a clade that comprises most WNV strains isolated in Texas during 2002. Two other WNV strains from the Bolivar Peninsula, Texas (362, 476), were positioned basally to a 1998 Israeli stork isolate that grouped with all other North American isolates, suggesting that the North American strains may not all have originated from a point source introduction into the New York area in 1999. This topologic finding was the result of a single synonymous, third codon position (genomic position 969 by using the numbering of the Flamingo382 isolate, GenBank accession no. AF196835) U nucleotide (synapomorphy) shared by the 1998 Israeli and all North American strains except the Bolivar Peninsula isolates. However, the relatively poor support, represented by bootstrap values (1,000 replicates) of only 56–59 by using the three different phylogenetic methods, suggests that this topologic finding is not necessarily correct. More robust phylogenetic analyses using complete genomic sequences are needed to clarify these relationships.

Comparison of nucleotide sequences indicated mutations at 9 nt (0.45%) of the Mexican TM171-03 sequence compared to the prototype NY99 strain (Flamingo382, GenBank accession no. AF196835). Some of these nucleotide positions vary among other strains sampled worldwide, suggesting that they are not under strong purifying selection. Comparison with sequences of Year 2002 Texas strains showed only one shared mutation (genomic position 2466 C→U). WNV has remained genetically conserved in the New World; however, based on our limited sequencing, the Mexican strain appears to be the most divergent WNV isolate identified to date in the Americas (Table 2). Two of the mutations resulted in amino acid changes at prM141 (Ile→Thr) and E156 (Ser→Pro), which have not been reported to date in North American isolates. The Ser→Pro amino acid substitution at residue E156 is of interest, as it abolishes a potential glycosylation site that is a putative WNV virulence determinant (7–9). This report is the first of a New World WNV isolate with probable altered E protein glycosylation. Further studies are in progress to assess the possible phenotypic effects of this mutation.

Two recent publications reported serologic evidence of WNV infection among equines in the states of Yucatán and Coahuila from serum samples collected in October and December 2002, respectively (10,11). We obtained serologic evidence of earlier and more widespread circulation of WNV in five other Mexican states, dating back to July 2002. We also report the first isolation of WNV from Mexico from a dead Common Raven that resided in a wildlife preserve in Tabasco.

Genetic studies indicated that the Mexican WNV strain was likely introduced from the central United States. The level of genetic divergence (9 nt) of the Mexican isolate and the unique amino acid substitutions in the prM and E proteins when compared to all other North American WNV isolates suggest that the Mexican strain has been evolving independently for some time and did not simply enter Mexico recently from Texas. We speculate that this strain descended from a WNV strain introduced into the Yucatán peninsula by migrating birds. Nucleotide sequences of viruses isolated from Mexican states close to the U.S. border, once obtained, may more closely resemble strains isolated in Texas during 2002.

Of particular interest is the overlap in distribution of WNV and VEEV (serologic data for VEEV not shown) in the southern Mexican states of Veracruz and Yucatán; the presence of other flaviviruses like SLEV in these states is also likely. Both WVN and VEEV produce clinically similar neurologic disease in horses, and past, presumptive diagnoses of VEEV may have been inaccurate. Steps are now in place at the Mexico City headquarters of the Animal Health Division of SAGARPA to implement appropriate laboratory diagnosis for flaviviruses and alphaviruses. Additionally, field personnel are instructed to investigate epidemiologic signs of possible WNV infection including avian death and unusual human neurologic syndromes.

The biologic and epidemiologic consequences of mosquito-borne encephalitis viruses (12) cocirculating in the same ecosystem should be examined. The impact of WNV on human health in regions (such as Mexico) where inhabitants may have extensive prior exposure to other flaviviruses such as dengue, SLEV, Ilheus, Bussuquara, Jutiapa, and Yellow fever viruses may differ from that in regions (e.g., the United States and Canada) where human exposure to flaviviruses is very limited. WNV infection in persons with previous flavivirus immunity, which could either attenuate disease because of cross-protective antibodies (13) or potentially worsen disease because of immune enhancement (14), should be studied. Our ongoing VEEV surveillance in southern Mexico may identify differences in transmission habitats for VEEV and WNV and assist with optimizing virus containment efforts.