A major goal of this study was to determine on the basis of a set of representative WNV strains () whether virulence in two avian hosts was directly related to virulence in a mouse model of WNV neuroinvasive disease. In addition, we wished to determine whether the isolation of WNV strains in the Americas with reduced virulence in the mouse model indicated possible changes in the transmission dynamics of WNV that would explain apparent differences in the epidemiology of WNV in the Caribbean, Mexico and Central America compared with that in other areas of North America. Identification of the potential attenuating role of mutations in these recent WNV isolates compared with the NY99 North American prototype may provide information on selective criteria associated with the emergence of WNV avian-attenuated phenotypes in tropical or subtropical ecosystems. The results presented indicate that avian and murine virulence and attenuated phenotypes most likely resulted from complex biological host interactions in disparate vertebrate systems.
Loss of glycosylation of the surface E glycoprotein of natural, laboratory-adapted16,22
and site-directed mutants15
has been associated with decreased replication capacity and neuroinvasive properties in a mouse model. Glycosylation status has also been associated with cell type or species specificity for proper viral maturation and release of WNV lineage I and II viruses,23
pH stability, particle assembly, virion maturation, and infectivity in tissue culture systems.23,24
Natural Kunjin viral isolates lacking E glycosylation have been shown to rapidly incorporate a glycosylation motif after serial passage in vertebrate cell culture.24
Transmission electron microscopy has further indicated that the maturation of glycosylation-deficient WNV particles derived from the Sarafend WNV strain on intracellular membranes could explain reduced titers in culture systems.17
These investigators further demonstrated that providing a glycosylated envelope gene in trans
was sufficient to abrogate this defective phenotype.17
Despite the aforementioned debilitation in viral replication and loss of murine neuroinvasiveness, nonglycosylated WNV isolates have been found repeatedly from the field.12,25,26
With the exception of a previous study that demonstrated reduced virulence of a plaque-purified WNV lacking a functional glycosylation motif in chicks,18
little is known regarding the role of glycosylation on the modulation of virulence of WNV in natural avian hosts for WNV. Viremia production (American crows/house sparrows) and survival time variation of American crows infected with the TM171-03-pp5 and TM171-03-pp1 variants that differ principally in their glycosylation status and in 5′UTR position 85 (G to A), NS5-224 (Arg to Gly) and 3′UTR position 10989 (A/G mixture to G) mutations indicate the potential for additional attenuating mutations in the genome of the variants from Mexico. Poor peripheral replication presumably reduced the efficiency for dose-dependent neuroinvasion, resulting in murine attenuation of the E-gly(−) variant. Variable peripheral replication in avian hosts appeared to modulate virulence phenotype. However, the results derived from the E gly(−) variant from Mexico indicated the potential for differential tissue tropism or virulence phenotypes of some WNV strains for avian hosts. In contrast to the murine model, in which increased viremia has been directly correlated with neurovirulence, there was a poor correlation between avian death and viremia with the variants from Mexico reported herein, which suggests that these viruses may have increased neurotropism.
Both of the TM171-03 variants and the BIRD1153 strain were significantly attenuated for virulence in the American crow model (as measured by differences in daily viremia at 1–5 dpi, mean peak viremia, viremia onset, mortality rate, and average survival times) compared with the NY99 North American prototype and the Texas BIRD1461 strain. The TM171-03-pp1 and -pp5 viruses both encoded amino acid differences compared with NY99 virus at residues prM-141, NS4B-245, and NS5-89812
(), and it is likely that one or more of these residues plays a significant role in the reduced avian virulence of these variants. However, reductions in the magnitude and duration of viremia observed for TM171-03-pp1 compared with TM171-03-pp5 () suggested that E glycosylation may have some modulating effect on the virulence phenotype and/or tropism of these variants. Alternatively, other mutations identified in the 5′ and 3′ UTRs of either TM171-03-pp variant or at NS5-225 of TM171-03-pp5 () may also have contributed to the differences in viremia profiles observed for these two variants. These potentially confounding additional genetic changes and those in the nonglycosylated plaque purified viruses from Mexico could be responsible for the reduced virulence and viremia observed in this study or in chicks with a plaque-purified glycosylation defective mutant.18
Future reverse genetic studies will be required to specifically assess the role of envelope glycosylation and additional mutations identified herein on avian pathogenesis phenotypes.
Unlike the significant level of attenuation observed in American crows for the BIRD1153 isolate, the E gly(+) strain from Mexico generated mean peak viremias comparable with that of the murine neuroinvasive Texas strain (BIRD1461) in house sparrows (). This finding could indicate the relative selective importance of different avian hosts for emergence of alternative WNV genotypes. Significantly higher viremias were identified at multiple time points in house sparrows inoculated with NY99 virus compared with the BIRD1461 strain, and differences were only observed at 1 dpi in American crows, which highlights the interspecies incongruence in virulence potential for WNV strains in alternative avian hosts.
The BIRD1461 and BIRD1153 strains differed at eight nucleotide positions (). Of these differences, only the NS4B-249 and NS5-619 amino acids and 10,596, 10,774, and 10,799 3′UTR nucleotide substitutions were unique to the BIRD1153 strain (i.e., not present in the NY99 viral genome). Incorporation of the NS4B BIRD1153 mutation in combination with either the NS5 or the three 3′UTR BIRD1153 substitutions into the NY99 virus genetic backbone were required to elicit a temperature-sensitive, small plaque, murine non-neuroinvasive strain.27
Interestingly, none of the potential genetic determinants have been identified in the variants from Mexico, which indicates that there are multiple molecular mechanisms for the attenuation of these variants through independent, polygenic pathways.
Previous studies have demonstrated that small numbers of genetic substitutions can have a dramatic effect on avian virulence phenotypes. For instance, infection with a WNV strain from Kenya (KN-3829)28
that differed from the NY99 WNV strain by only 11 amino acids29
produced a mortality rate of only 20–30% in American crows and has demonstrated significantly reduced serum viremias compared with the NY99 strain.30
None of the amino acid substitutions identified between the WNV strain from Kenya and the NY99 strain29
have been identified in either of the variants from Mexico or the BIRD1153 murine non-neuroinvasive virus investigated in these studies. This finding was not particularly surprising given the fact that the KN-3829 strain has a virulent, neuroinvasive phenotype in mice.14
In this study, strains that were non-neuroinvasive in mice still caused significant mortality in avian species, indicating that variable pathogenic mechanisms of virulence and attenuation are present in these disparate vertebrate models and that the murine model may not accurately predict virulence in avian hosts.
Viremia and mortality appear to be highly correlated in the American crow virulence model of WNV infection. This finding was evident for infection with the BIRD1461, BIRD1153, and the TM171-03-pp5 strains/variants, in which the mean peak viremias of all decedent American crows exceeded 8.4 log10 PFU/mL of serum and the mean peak viremias for survivors never exceeded 6 log10 PFU/mL of serum. This model is consistent with a hypothesis that once peripheral replication in the avian host reaches a particular threshold (> 6 log10 PFU/mL of serum), invasion of the central nervous system or other vital organ systems occurs through a non-specific passive process. However, the E gly(−) variant from Mexico (TM171-03-pp1) did not demonstrate this viremia response/mortality relationship. Interestingly, peak viremias of crows that died because of infection with this variant were in some cases lower than those identified for survivors (). Only one American crow in this infection group showed development of a viremia > 6 log10 PFU/mL of serum. Nonetheless, four of eight crows died because of infection with this plaque-purified virus, and all four decedents had nonglycosylated WNV isolated from brain tissue at the time of death as determined by sequence analyses.
To investigate this viremia magnitude/mortality relationship, peak viremias were plotted as a function of percent mortality and a regression analysis was performed to describe this relationship deriving a best-fit line (y = 0.059X + 3.143, R2 = 0.702). On the basis of this relationship, and assuming that each of the viruses exhibited similar virulence (i.e., survivorship was exclusively a determinant of viremia magnitude), an estimate of a peak viremia of 6.1 log10 PFU/mL of serum would result in a 50% mortality rate within the infection group. Further investigations of differences in increased viremia duration and differential tissue tropisms are needed to further evaluate this theory. However, the finding of a 50% mortality rate for which three of the four decedent American crows infected with the TM171-03-pp1 E gly(−) variant from Mexico showed development of a peak viremia well below the > 6 log10 PFU/mL of serum hypothetical threshold (3.6, 3.4, 4.4 log10 PFU/mL of serum, respectively) suggest that there could be differences in tissue tropism for some of these strains/variants.
Removal of the TM171-03-pp1 data point from the regression analyses generated a best-fit line (y = 0.048X + 4.71) with a much greater R2
= 0.895, which further supports differential tropisms of this variant from Mexico. This second regression line subsequently increased the hypothetical circulating titer required to result in 50% mortality from 6.1 to 7.1 log10
PFU/mL of serum and was consistent with estimates reported for a variety of avian species.31
When viremia parameters of survivors and decedents were compared, viremia magnitude, onset, and daily magnitudes (1–6 dpi) were all significantly different (P
< 0.0001). On the basis of these preliminary findings, further studies investigating the alternative pathogenicity of the variants from Mexico that yield low viremias are warranted.
Numerous reports have provided serologic evidence of WNV transmission in the Western hemisphere south of the United States.10,32–38
Despite this fact, with the exception of sites adjacent to the United States border,39
WNV was isolated only once in Mexico from a diseased raven in a Tabasco zoo.11
This paucity of viral isolations could be the result of a combination of reduced avian mortality, reduced viral loads within avian hosts, or limited mosquito testing. In addition to observation of reduced avian mortality in Mexico, the Caribbean, and Central and South America, serosurveys of equids have demonstrated evidence of spillover transmission32,34,35,40–42
without documented equine disease, until reaching the temperate latitudes of Argentina.43
A number of theories could explain reduced virulence. These theories include 1) circulation of more attenuated strains in Latin America that have been the result of differential selection in local avian or mosquito fauna; 2) founder's effects from rare introduction events from viremic migratory birds; 3) herd immunity afforded by serologic cross-protection by heterologous flaviviruses44
circulating in these geographic locations; or 4) potential dilution effect of the greater avian species diversity in the tropics that could confound the identification of mortality in highly susceptible species, as has been prevalent in temperate regions of North America.45,46
Data in this study indicate that the small plaque variant isolated from southeastern Mexico, in addition to being non-neuroinvasive in mice, was also considerably attenuated in its ability to replicate in two avian species. In particular, seven of eight American crows infected with the E gly(−) variant from Mexico did not have viremia titers greater than 4.4 log10
PFU/mL. On the basis of dose-response data for common WNV mosquito vectors within the United States,47
mosquito oral infection rates would be expected to be quite low at these titers, which would limit viral transmission potential (and overall fitness) of these avian attenuated strains/variants. The glycosylation-competent, large plaque–forming variant from Mexico produced serum viremias greater than 5 log10
PFU/mL in six of the eight American crows, which indicated that this virus would be much more likely to be infective and subsequently to be transmitted by competent mosquito vectors. Specific vector competence data for mosquito species from Latin America will be needed to further assess this hypothesis.
The fact that the TM171-03 isolate constituted a mixture of two plaque variants12
is intriguing and could indicate that differential selective factors were associated with the presence of two distinctive viral populations in the infected raven. Highly susceptible and temperate region prevalent avian hosts (American crows and house sparrows) were used for these virulence studies and could afford a radically different selective environment from that found in sympatric avian hosts in Latin America. Preliminary data have indicated that the TM171-03 isolate (mixed plaque isolate) was capable of eliciting significant viremias in resident house sparrows, clay-colored thrushes, and great-tailed grackles from Mexico (Komar N and Estrada-Franco JG, unpublished data), indicating that viremia potential could be significantly modulated by avian species and/or intraspecies geographic variants or the potential presence of the mixed viral populations. Host competence studies using sympatric avian hosts with additional isolates from areas lacking human, animal, and avian mortality will be needed to fully assess the role of alternative WNV genotypes with altered disease presentation in Latin America.