In 2006, we reported 52 avian-human signatures based on a small set of influenza sequence data of 15,785 protein sequences. The selection was based on an entropy threshold value of 0.4 set at position 627 in the PB2 gene (82 Es and 13 Ks from 95 avian PB2 sequences) because that position has been considered associated with host-restriction (8–11). Of the 52 positions, 45 are in the genes PB2, PB1, PA, NP, M1, M2, NS1, and NS2 examined in this work. Today, >100,000 influenza protein sequences are available at NCBI, and a new entropy threshold of 0.33 was set based on the currently available avian sequences of PB2-627, which contain 3,113 Es, 228 Ks, 46 Vs, 2 As, and 2 Gs. This threshold was adopted to update the list of 47 avian-human signatures in Table 1 for the 8 proteins of interest. Consistent with our earlier findings, most signatures are located on the NP gene (15 positions), followed in number by PA (10 positions), PB2 (9 positions), M2 (5 positions), M1 (3 positions), PB1 (2 positions), NS2 (2 positions), and NS1 (1 position). The 20 signatures associated with PB1, NP, and M1 do not differ between the 2 datasets of 2006 and 2009. In PB2, two new signatures are identified at positions 567 and 702. These 2 positions were only just omitted from the 2006 list because their entropy values (0.490 and 0.404, respectively) exceeded the 0.4 threshold, based on the 95 avian sequences examined at that time. New entries in Table 1 also include PA-100, M2-93, and NS1-81. In 2006, although PB2-674 was reported as a signature, it is disqualified here because the entropy of 0.3376 (3,146 As, 88 Es, 87 Ts, 38 Ss, 13 Gs, 4 Vs, and 1 K for avian virus) exceeds the new 0.33 threshold at this position. Similarly, PA-382 (2,421 Ds, 311 Es, 2 Vs, and 1 N in human viruses, with an entropy of 0.3633) and NS2-70 (2,898 Ss, 352 Gs, 28 Rs, and 1 D in avian viruses, with an entropy of 0.3483) were both removed from the 2006 list.

Taubenberger et al. (12) identified 10 polymerase gene positions that separate avian viruses from human influenza A viruses. Table 1 shows 8 of them (PB2 199, 475, 567, 627, and 702; PA 55, 100, and 552), suggesting that the method is robust in finding these signatures. Two other polymerase gene positions that Taubenberger et al. also reported are PB1-375 and PA-382; the latter has already been mentioned above. The other missing position in Table 1, PB1-375, has an entropy value of 0.8865 for human and 0.6338 for avian viruses. This position was also excluded from the 2006 list because of an entropy of 0.698 from avian viruses, which substantially exceeded the 0.4 threshold.

To elucidate the potential adaptive mutations of the pandemic (H1N1) 2009 viruses, we studied the amino acid sequences of pandemic (H1N1) 2009 viruses at the positions that represent the so-called species-specific signatures of avian and human viruses. As shown in the last column of Table 1, 36 of the 47 positions display avian-like signatures in the pandemic (H1N1) 2009 virus. Two positions, PB2-588 and NP-313, exhibit neither avian- nor human-like signatures. Eight human-like signatures were found in pandemic (H1N1) 2009 strains, except for NS1-227, in which all new viruses have an early-terminating NS1 protein and, therefore, contain no residue.

Table 1 presents the updated avian-human signatures for influenza A viruses; Table 2 summarizes the swine-human signatures. Medical literature documents that the swine virus population has distinct evolutionary lineages that originated from the classic 1918 virus referred to as classic or North American swine virus, and the others of post-1979 Eurasian swine virus and subsequent triple reassortants. Because the residue diversity at many positions markedly increased for these swine viruses because of their distinct origins, only 8 swine-human signatures met the 0.33 threshold. Unlike some positions in which human-like signatures of pandemic (H1N1) 2009 were found (Table 1), in this study, all 8 locations of the swine-human signature of this new virus are characteristic of swine. Notably, Table 1 lists all 8 positions in Table 2, with each having the same signature as in the avian virus. Restated, avian and swine viruses contain the same amino acid residue at the 8 human-swine signature positions.

We attempted to further elucidate the transition of the amino acid residue on the pandemic (H1N1) 2009 virus that have human signatures by sampling 18 recent ancestral swine viruses (Appendix Table). Doing so enables us to examine more closely the prevalence of amino acid residues specifically with pandemic (H1N1) 2009 viruses. Table 3 summarizes the amino acid statistics of these recent ancestral swine viruses together with avian, human, and pandemic (H1N1) 2009 sequences at the 8 positions containing human residues for pandemic (H1N1) 2009 virus in Table 1. Consider PB2-271, for example, avian viruses have signature T, whereas human viruses have signature A. Although pandemic (H1N1) 2009 viruses also have the human signature A, their predecessors, i.e., the recent ancestral swine viruses, have already acquired the human signature A at this position. PB1-336, along with PA-409, NP-33, -100, -305, and -357, follows the same residue transition, all showing human-characteristic residues in both recent ancestral swine and pandemic (H1N1) 2009 viruses. PA-356 is the only exception, where the residue in recent ancestral swine viruses still maintains an avian-characteristic K before changing to a human residue R in pandemic (H1N1) 2009 viruses. Of particular interest is whether the transition from K to R at position PA-356 is responsible for the ability of pandemic (H1N1) 2009 viruses to replicate and transmit efficiently in humans.

After all 8 human residue-containing positions of pandemic (H1N1) 2009 viruses were found to be within PB2, PB1, PA, and NP protein genes, all amino acid positions of these 4 proteins were scanned for their residue transitions among the 4 virus populations shown in Table 3. The change in the amino acid that may be associated with the transformation of pandemic (H1N1) 2009 virus is summarized in Table 4. As well as PA-356, already shown in Table 2, two additional positions, PB2-684 and PA-204, showed the same dominant amino acid residue in avian and recent ancestral swine viruses, but a different dominant residue in pandemic (H1N1) 2009 viruses and human viruses. Dominance is defined here as 1 residue containing the largest sequence count compared with other residues at a particular aligned position. The previously used entropy measurement in Tables 1, 2, and 3 does not apply to the positions listed in Table 4, in which we emphasize the amino acid transition of dominant residues instead of highly conserved ones subject to the prescribed entropy threshold 0.33. Other than those 3 positions, PB1-216 was found to contain a human residue G in 8 of 9 recent ancestral swine viruses that are closer to pandemic (H1N1) 2009 viruses in the phylogenetic tree published in a study by Smith et al. (6). However, for the other 7 recent ancestors that are more distant from pandemic (H1N1) 2009 viruses, PB1-216 maintains an avian-residue S in 6 of 7 viruses. Our results show that the position-specific transition may serve as a molecular marker for monitoring such adaptive mutations in the future.

Although most studies confer that the death rate associated with pandemic (H1N1) 2009 infection is more moderate than that of subtype H5N1 infection, its virulence may vary with adaptive mutations in viral genes, subsequently increasing the likelihood that the new virus alters its virulence in the new host species. Many of the previously identified virulence factors are apparently not involved. For instance, no E to K mutation at position 627 of PB2 is observed, which has been considered an important factor for avian virus to efficiently replicate in mammalian systems (8–11). Previous studies have indicated that PB1-F2 contributes viral pathogenesis in the mammalian system (13,14). No PB1-F2, however, is predicted in pandemic (H1N1) 2009 viruses because it terminates prematurely at position 12. Its NS1 protein is truncated at position 220 and, therefore, lacks a PDZ ligand interacting domain. As suggested recently, the presence of this PDZ ligand domain increases the pathogenicity of avian influenza A viruses (15). Regardless of whether these known factors are missed, a previous study has demonstrated that the virulence of pandemic (H1N1) 2009 virus is higher than that of seasonal influenza A viruses (16). Although a virulence marker and a host range factor may not be necessarily linked tightly, recent investigations have also demonstrated that altering PB2-627 from E to K in the avian viruses increases its virulence in the mammalian experimental system (9–11). For example, avian influenza virus subtype H7N7 reportedly infects humans (17). A human isolate from a fatal case had its PB2-627 changed from avian-characteristic E to K. Correspondingly, the species-associated signatures identified in this study may serve as potential molecular targets for further evaluating how they impact the virulence of pandemic (H1N1) 2009 viruses in humans.

As shown in Tables 1 and 2, the number of signature positions decreases significantly from 47 (human vs. avian) to 8 (human vs. swine), and the positions of the latter are a subset of those of the former. These observations may have the following implications. First, the 3 host species of interest differ, with each providing a unique environment for infection by the influenza virus. When the avian virus enters humans or swine, its genetic feature is shaped by a particular evolutionary path. The viruses, therefore, have different signatures. Second, some avian-like signatures are preserved in swine viruses, suggesting that both avian species and swine may provide similar conditions for harboring influenza A viruses. The body temperature may be a determinant. As is generally known, many avian species have a body temperature exceeding 40^oC; for most pigs it is variable but still higher than the human body temperature, which is 37^oC. Consequently, the signatures are retained when an avian virus enters the swine population, with similar signature-related viral replication mechanisms in both species. Third, the 39 signature positions shown in Table 1, but absent from Table 2, may be correlated with certain functional domains that interact with host factors unique in humans while differing significantly from those of avian and swine. Finally, the number of signature positions of swine versus humans is substantially lower than those of avian versus humans, suggesting that the species barrier to humans is easier for a swine virus to cross than for an avian virus.

The entropy-based computation depends strongly on a good multiple sequence alignment. The 2 surface proteins HA and NA are excluded from this analysis because both contain sequences that diverged sufficiently from so many subtypes of a given species. Locating conserved residues at particular positions on the basis of these alignments is extremely difficult. The entropy threshold is the other parameter requiring attention to locate a signature position. In this study, the entropy determined from PB2-627 of the aligned residues of all avian viruses is used because PB2-627 is the most laboratory-proved host-restriction marker (8–11). A complete new set of signatures can be reproduced rapidly by using a different entropy threshold based on other factors. The diverse genetic origins of influenza viruses would also have great impact on the reported signatures. The proposed entropy-based method to reach the 8 positions listed in Table 2 was based on all swine viruses of different origins, including North American-(classic 1918) origin strains, Eurasian (post-1979 avian)-origin strains, and recent triple reassortants. A comparison of, for example, all human viruses versus classic 1918-origin swine viruses before 1978 (≈75 strains, or 20% of our swine sequence population) would report 60 signature positions (data not shown). In this work, we included all swine viruses of multiple origins in producing Table 2 to consider only host-specific genomic signatures that have been shaped by the same swine species regardless of origin. For the same reason, we did not subdivide avian or human populations into lineages when reporting avian-human signatures in Table 1.

This study analyzed a complete collection of species-specific influenza A viral sequences, including the long-evolving avian, recent ancestral swine and human viruses, as well as pandemic (H1N1) 2009 viruses, which is still in its infancy. The amino acid sequence transition of pandemic (H1N1) 2009 virus at the signature positions was also elucidated by applying the entropy-based signature analysis to these sequences. They were found mostly to be characteristic of avian species, as presented in Table 1. Notably, 8 of them changed from avian-like signatures to human-like signatures. Close examination of the residue transition at these 8 positions in Table 3 showed that PA-356, unlike the other 7 positions, retained an avian-like signature in the recent ancestral swine population and changed to a human-like signature only in pandemic (H1N1) 2009. This finding suggests that PA-356 may be related to host-restriction factors from swine to human species. Similarly, all ribonucleoprotein positions were scanned for the same transitioning pattern as in PA-356, i.e., a retained avian-like residue in the recent ancestral swine population and a change to the human residue in pandemic (H1N1) 2009 viruses. Table 4 lists them all. Although 1 of the positions, PB1- 216, was not dominated by the residue S as we would have expected, it exhibited a mixture of 2 residues involving a transition from avian to human viruses. In summary, Table 4 provides a list of candidate host-restriction factors that we believe are important to adaptive mutation of influenza A viruses among the 3 host species. Continuous monitoring of these signatures in nonhuman species will help in influenza surveillance and in evaluating the likelihood of further adaptation to humans.