Comparing the nonsynonymous/synonymous substi-tution rate ratio (ω = dN/dS) provides a sensitive means for the mechanism study of sequence evolution at the protein level. An ω ratio greater than one means that nonsynonymous mutations offer fitness advantages to the individuals. In this case, they have higher probabi-lities to be fixed in the population than synonymous mutations (20). Positive selection can thus be detected by identifying cases where ω > 1 (2).
Up to date, several methods have been proposed to estimate the selection pressure, one of which was fixed effect likelihood (FEL) method (12). It is practicable to complete a site-by-site positive selection analysis based on a maximum likelihood method. Kosakovsky Pond et al. have also developed internal fixed effect likelihood (IFEL) method to investigate whether sequences sampled from a population had been subject to selective pressure at the population level (13). In addition, they have proposed an approach for identifying whether selection was operating differen-tially on single codons of a gene sampled from two different populations and embedded it in HyPhy as a standard analysis procedure-CompareSelective Pressure. bf (13, 14).
H9N2 avian influenza viruses (H9N2) were detected to be co-circulating with deadly H5N1 viruses in Hong Kong in 1997 and were reckoned as the genetic donors of HongKong/156/97 (7). Although they lacked multiple basic amino acids at the cleavage sites and were regarded as viruses of low-pathogenicity, these viruses demonstrated fast evolution rate as observed in H5N1 avian influenza viruses and some different sublineages and genotypes have been detected (4). In addition, their co-circulation with H5N1, H6N1 and viruses of other subtypes and proved ability to reassort with these viruses still merited our intense attention.
Although migratory waterfowls were regarded as natural hosts of avian influenza viruses, pigs were once believed to be the genetic vectors of influenza viruses of different subtypes. Accordingly, they were taken as potential intermediate hosts of avian in-fluenza viruses to acquire genetic advantages to infect humans (17, 18). However, other research indicated that quails could also act as intermediate hosts for duck viruses to gain adaptations to cross species barrier (16).
Haemagglutinin (HA) gene is one major surface glycoprotein of influenza A viruses (19). Previous studies have proved that positive selection pressures acting on HAs of H1, H3 and H5 subtypes are responsible for the evolution of the antigenic sites (3, 5, 10). Here we present a study regarding the positive selection and differential selective pressure acting on HAs of H9N2 avian influenza viruses from different hosts. Our analysis provides evidence of the genetic factors that favor their extension of host range and adaptations to different hosts of the H9 viruses.
Sequence data, which were updated in March, 2008, were downloaded from Influenza Virus Resource database at GenBank (http://www.ncbi.nlm.nih.gov/genomes/FLU/FLU.html) directly. Some sequences shorter than 940nt were removed. In particular, all but one completely identical sequence were also excluded. Finally, we included 541 H9N2 HA sequences into this analysis in total. These sequences were further divided into several small datasets according to the hosts where the H9N2 viruses were isolated (Table 1).
Table 1. Sites found to be under positive selection
Multiple sequence alignment was performed manually for each dataset. A Maximum Likelihood (ML) tree was built for each data subset using Phyml v2.4.4 for Windows (8). Transition/transversion (Ts/tv) ratio and proportion of invariable sites were both estimated rather than treated as fixed.
Site-by-site positive selection analysis was perfor-med for each data subset using the FEL method in the HyPhy software Package. ML trees obtained in the previous step were used in the analysis. HKY85 model was selected as the best model of sequence evolution for the data. A global dN/dS ratio was estimated by a codon model obtained by crossing MG94 and HKY85. A two rate FEL model was applied allowing dN and dS adjusted across sites. P ≤ 0.05 was considered to be significant. Furthermore, the same parameters were also applied in IFEL analysis.
In addition, test for differential selection at a site between viruses from any two hosts was completed by the standard analysis CompareSelectivePressure.bf in HyPhy. Similarly, the ML trees built by Phyml were used as the input trees. However, p ≤ 0.05 was regarded as the significance level.
Overall, the FEL analysis showed that most of the sites were under neutral or purifying evolution and there were totally six different sites under positive selection (Table 1). In detail, for viruses from chic-kens, positions 180 and 216 were positively selected. However, positions 127 and 148 were under positive selection only for viruses isolated from ducks. Viruses from quails had three positively selected sites, 83, 216 and 311. In addition, position 216 was under positive selection for viruses from pigs. In particular, no position was detected to be positively selected for viruses from humans and turkeys. However, at the population level, only 180, 216 in HAs from chickens, 148 from ducks and 216 from pigs were positively selected.
Furthermore, the comparative analysis of dif-ferential selection showed that there were always some positions under different selection for viruses from any two hosts into analysis (Table 2). Notably, there were overall fourteen sites under differential selection between viruses from ducks and quail (Table 2, Fig. 1). The numbers of positions under different selection were less than fourteen between any other two hosts (Table 2). For H9N2 viruses from chickens, the positions under different selection differed when they were compared to viruses from different hosts. Remarkably, only four positions were detected to be under differential selection between isolates from chickens and pigs. Position 268 was under different selection for strains from humans and other hosts. Meanwhile, except for isolates from pigs, the human isolates differed with viruses from other hosts in position 111. Furthermore, position 243 was under different selection for strains from quails and other avian hosts, but not mammalian hosts. Likewise, position 311 was not under differential selection between viruses from quails and humans. However, it was not the case when 311 was compared for viruses from quails and other hosts except for humans. Position 86 and 115 were differently selected for viruses from pigs and other hosts except for humans. Additionally, position 216 was under different selection for strains from ducks and chickens, quails and pigs. It should also be noted that position 318 was differently selected between strains from some hosts, such as humans and pigs, ducks and turkeys.
Table 2. Sites found to be under differential selection
Figure 1. Sites under differential selection between H9N2 viruses from ducks and quail. The y-axis represented the 1-p values, while the x-axis stood for the corresponding amino acid positions of H9HA1. Those positions with 1-p ≥ 0.95 were differentially selected.
However, there was no positively selected site that was subject to positive selection pressure for viruses from more than one host and at the same time was differentially selected between viruses from these hosts (Table 1, Table 2).