H5n1

Cell 368

nections between seemingly unrelated genes and processes. E-MAPs will yield the best results when comprehensive, unbiased strategies are used to define functionally relevant subsets for detailed study. But how do we choose? Large-scale localization data is useful, but functionally related genes are not always spatially restricted, subcellular localization patterns may be transient or ambiguous, and epitope tags may disrupt targeting determinants. The E-MAP strategy will be useful for refining the results of noisy data sets from transcriptional, protein-protein interaction and lowdensity SSL studies. However, the best E-MAPs might come from analyzing the top hits of systematic, quantitative phenotypic screens, which should constitute a comprehensive set of mutants that make the most important contributions to a given cellular process. With the development of genome-wide RNA interference (RNAi) libraries, it is now possible to carry out high-throughput phenotypic screens and map genetic interactions in other organisms (Carpenter and Sabatini, 2004). Flies, worms, and mice have many more genes than yeast, making the determination of all possible pairwise genetic interactions an order of magnitude more difficult. For these organisms, E-MAPs of functionally restricted groups of genes are a viable alternative. However, generating double perturbations in other systems is technically challenging. In C. elegans, RNAi feeding libraries might be used to knock down gene expression in worms with a germline mutation in a second gene. It may also be possible to simultaneously suppress two genes using RNAi, or carry out RNAi screens in the presence of a drug that inhibits the function of a specific gene. Because RNAi reduces protein levels without eliminating protein expression, such strategies may yield effects comparable to hypomorphic alleles and will require appropriate analysis techniques that may not be effectively modeled using yeast knockout data. Phenotype profiling may be more likely than gene expression profiling to identify genes involved in key regulatory mechanisms controlling a biological response. Moreover, genetic interactions are reported to outperform protein interactions in detecting functionally related gene pairs (Wong et al., 2005). Reliably mapping genetically interacting genes in model systems could prove invaluable for identifying genes responsible for diseases that are caused by mutations in multiple genes. The analysis of high-density genetic interaction data helps not only to define the molecular components of complexes and pathways, but also to understand how they are connected and interrelated. E-MAPs may point the way to a better understanding of complex biological systems, and prevent us from getting lost in a sea of large-scale data.

Elizabeth Conibear Centre for Molecular Medicine and Therapeutics Child and Family Research Institute Department of Medical Genetics University of British Columbia Vancouver V5Z 4H4 Canada

Selected Reading Carpenter, A.E., and Sabatini, D.M. (2004). Nat. Rev. Genet. 5, 11–22. Cordell, H.J. (2002). Hum. Mol. Genet. 11, 2463–2468. Davierwala, A.P., Haynes, J., Li, Z., Brost, R.L., Robinson, M.D., Yu, L., Mnaimneh, S., Ding, H., Zhu, H., Chen, Y., et al. (2005). Nat. Genet. 37, 1147–1152. Kaiser, C.A., and Schekman, R. (1990). Cell 61, 723–733. Kelley, R., and Ideker, T. (2005). Nat. Biotechnol. 23, 561–566. Novick, P., Field, C., and Schekman, R. (1980). Cell 21, 205–215. Schuldiner, M., Krogan, N.J., Collins, S.R., Thompson, N.J., Denic, V., Bhamidipati, A., Punna, T., Ihmels, J., Andrews, B., Boone, C., et al. (2005). Cell 123, this issue, 507–519. Segre, D., Deluna, A., Church, G.M., and Kishony, R. (2005). Nat. Genet. 37, 77–83. Tong, A.H., Lesage, G., Bader, G.D., Ding, H., Xu, H., Xin, X., Young, J., Berriz, G.F., Brost, R.L., Chang, M., et al. (2004). Science 303, 808–813. Wong, S.L., Zhang, L.V., and Roth, F.P. (2005). Trends Genet. 21, 424–427. DOI 10.1016/j.cell.2005.10.016

The Genesis of a Pandemic Influenza Virus Pandemic influenza viruses pose a significant threat to public health worldwide. In a recent Nature paper, Taubenberger et al. (2005) now report remarkable similarities between the polymerase genes of the influenza virus that caused the 1918 Spanish influenza pandemic and those of avian influenza viruses. Meanwhile, Tumpey et al. (2005) reporting in Science show that the reconstructed 1918 Spanish influenza virus kills mice faster than any other influenza virus so far tested. In the twentieth century, three influenza viruses emerged that caused major pandemics: the 1918 Spanish flu virus, the 1957 Asian flu virus, and the 1968 Hong Kong flu virus. The 1918 Spanish influenza virus killed an estimated 20 to 50 million people worldwide; the 1957 and 1968 viruses killed between 0.5 and 1 million people in the United States alone. Human pandemic influenza viruses emerge when avian influenza virus genes, previously unseen by the majority of humans, are incorporated into human influenza viruses in a way that allows for efficient spread of these viruses between humans. Influenza viruses have several proteins that are implicated in virulence: the surface proteins hemagglutinin (HA) and neuraminidase (NA), the polymerase complex (including the PB1, PB2, and PA proteins), and the nonstructural (NS) proteins. The 1957 Asian influenza virus acquired avian PB1, HA, and NA genes, and the 1968 Hong Kong influenza virus acquired avian HA and PB1 genes through the process of gene reassortment. During gene reassortment, a new influenza virus is generated through the mixing of the eight gene segments of two different parent influenza viruses. These viral strains gained human transmissibility, in part, by altering the binding preference of their HA proteins for hu-

Previews 369

Figure 1. The Genesis of an H5N1 Pandemic Virus The H5N1 strain of avian influenza virus could acquire the potential to cause pandemic flu in humans through two mechanisms. Coinfection of a human host with an avian H5N1 virus and a human influenza virus could lead to reassortment of genes producing an H5 avian virus that can be transmitted from human to human. Alternatively, an avian H5N1 virus could mutate directly into an H5N1 virus capable of human-tohuman transmission. Currently circulating avian H5N1 viruses are antigenically new, highly pathogenic in humans, and spread systemically in animal models. They exhibit trypsin-independent cleavage of the hemagglutinin (HA) protein and a deletion in the nonstructural (NS) gene, features that are associated with enhanced pathogenicity. As few as one or two mutations in the HA protein and ten mutations in the polymerase complex could result in an avian influenza virus capable of causing a human pandemic. (NS, nonstructural genes).

man host cell receptors bearing sialic acid residues of the α2,6 form. Influenza A viruses are classified into antigenic subtypes based on their HA and NA proteins, with HA proteins falling into H1 to H16 classes and NA proteins falling into N1 to N9 classes. The avian influenza viruses currently circulating that are considered to have the highest pandemic potential include the H2, H5, H7, and H9 subtypes. To induce a pandemic, these viruses need to acquire the ability to be efficiently transmitted from human to human. Here, we consider the threat of the H5N1 strain of avian influenza virus emerging as a pandemic virus in light of new insights into the 1918 Spanish influenza pandemic virus provided by two exciting recent studies by Taubenberger et al. (2005) and Tumpey et al. (2005). The sequence and phylogenetic analyses of Taubenberger et al. (2005) confirm that the 1918 Spanish influenza virus was not a reassortant virus (like the 1957 and 1968 viruses) but rather was an entirely avian-like virus that became adapted to humans through mutation. This means that currently circulating, highly pathogenic avian influenza viruses could evolve into human pandemic viruses either through gene reassortment or through direct mutation of viral genes (see Figure 1). Using their reconstructed 1918 pandemic virus, Tumpey et al. (2005) demonstrate in mice that it is both the polymerase genes and the HA and NA genes that are responsible for the extreme virulence of this virus. In their tour de force study, Taubenberger and colleagues have assembled the sequences of the eight gene segments of the 1918 pandemic virus from RNA fragments obtained from paraffin block and frozen tis-

sue of several victims who perished in the 1918 pandemic. They now report the complete genome sequence of the prototypic pandemic virus (Taubenberger et al., 2005). Unexpectedly, the sequences of the polymerase proteins (PA, PB1, and PB2) of the 1918 virus and subsequent human viruses differ by only ten amino acids from the avian influenza virus consensus sequence. Many or all of these residues must account for the contribution of the polymerase complex to the acquisition of human transmissibility by an avian influenza virus. The human forms of seven of the ten polymerase residues have already been observed individually in currently circulating H5N1 influenza viruses recovered from birds and humans. Under the selective pressure of a suboptimal growth rate in humans, the polymerase genes of an avian H5N1 virus that is currently circulating could potentially mutate such that these ten residues are converted to the “human” forms. As a result the virus may become better suited for efficient human-to-human transmission. The emergence of a pandemic influenza virus from an avian progenitor also appears to involve a switch in preferential binding of the HA protein from α2,3 sialic acid (the major form in the avian enteric tract) to α2,6 sialic acid (the major form in the human respiratory tract). The HA proteins of avian influenza virus species contain Gln226 and Gly228 residues, which form a narrow receptor binding pocket that favors binding of α2,3 sialic acid (Ha et al., 2001). On the other hand, human species usually contain Leu226 and Ser228, which form a broad pocket that prefers α2,6 sialic acid (Skehel and Wiley, 2000). High-resolution structures of the reassembled HA of the 1918 virus show that its avian-like

Cell 370

Gln226 and Gly228 residues create a narrow avian-like binding pocket that still allows for high-affinity binding of α2,6 sialic acid (Gamblin et al., 2004; Stevens et al., 2004). Residue 190 alone appears to determine the preference of the 1918 pandemic virus for α2,6 sialic acid versus an avian H5N1 virus for α2,3 sialic acid (Gamblin et al., 2004; Ha et al., 2001; Stevens et al., 2004). In fact, a D190E mutation in the HA of the 1918 virus switches its receptor binding preference to α2,3 sialic acid (Glaser et al., 2005). Conversely, just a single E190D mutation in the HA of the H5N1 strain could potentially switch its binding preference to α2,6 sialic acid, an expected requirement for its evolution into a pandemic virus. Just as transmissibility is a polygenic property, so too is high pathogenicity. The NS gene contributes to pathogenesis by disarming the interferon-based defense system of the host (Garcia-Sastre, 2001). A role for the polymerase genes in pathogenesis is indicated both by the association of the avian PB1 gene with the HA gene in both the 1957 and 1968 pandemic viruses and by the association of a lysine residue at 627 in PB2 with human infection by both H5N1 and H7N7 avian influenza viruses (Hatta et al., 2001; Subbarao et al., 1993). Another hallmark of high pathogenicity is the spread of influenza virus infection to tissues lacking the enzyme trypsin. Multicycle replication by less pathogenic influenza viruses is restricted to the upper respiratory tract where trypsin is expressed. Highly pathogenic influenza viruses usually contain an HA protein with a multibase cleavage site that is recognized by more ubiquitously expressed host cell proteases such as furin. As a result, these viruses can spread throughout the lungs and in some cases throughout the body. The only way for researchers to elucidate how the 1918 pandemic virus caused unparalleled pathogenesis in healthy adults is to reconstruct its genes and study their properties. The first clues to the pathogenicity of the 1918 virus were obtained when researchers inserted its HA and NA genes into the backbone of a less pathogenic virus (Kobasa et al., 2004). They found that the hybrid viruses spread more broadly and caused increased inflammation in the lungs of mice by releasing a storm of cytokines. However, it has not been clear whether other genes, such as the NS or polymerase genes, contributed to the lethality of the 1918 influenza virus. The only way to answer this burning question is to reconstruct the entire virus from oligonucleotide DNA. This is what Tumpey et al. (2005) have done in their seminal study. For the first time, an extinct virus—the influenza virus that caused the 1918 Spanish pandemic—has been resurrected. Tumpey et al. (2005) used reverse genetics to reconstruct the 1918 pandemic virus in cultured cells and then infected mice with the virus to study its pathogenicity. They show that the reconstructed 1918 virus kills mice faster than any previously characterized influenza virus. Based on previous work (Kobasa et al., 2004), it is known that part of the pathogenicity of the 1918 virus lies in its HA and NA surface proteins. Tumpey et al. (2005) now demonstrate that another part of the pathogenicity of this lethal virus is due to its avian-like polymerase genes. Like other highly pathogenic influenza viruses, such as H5N1, the 1918 virus has a HA protein

that is cleaved into an active form in the absence of trypsin. However, unlike any other HA proteins from highly pathogenic influenza viruses that have been characterized so far, the 1918 virus HA does not have a multibasic cleavage site that can be cleaved by furin and furin-like proteases. Instead, its own NA protein is involved in cleavage of HA by a new mechanism that is not yet understood. As a result, low pathogenic influenza viruses could potentially increase their virulence not only through mutations in their HA gene but also through mutations in or reassortment of their NA gene. The apparent similarity between the genesis of the 1918 Spanish influenza virus and the emerging avian H5N1 strain is of growing global concern. Both the H5N1 and 1918 viruses activate their HA proteins in a manner that is independent of trypsin, albeit through different strategies. Both are also highly pathogenic in mice. In the mouse model of Tumpey et al. (2005), the 1918 virus did not spread beyond the lungs, whereas many of the H5N1 viruses become systemic, spreading throughout the body even to the brain. There are reports of multiple aspects of neurological diseases in humans infected with the 1918 pandemic virus, raising the question of whether the mouse is an appropriate model for studying the Spanish influenza virus. The substantial number of human cases caused by the H5N1 strain of avian influenza virus over the past two years (118 reported cases and 61 deaths), some clusters of probable human-to-human transmission, and the similarity to 1918 pandemic influenza virus all suggest that the only remaining trait for H5N1 to acquire to become a pandemic virus is efficient human transmissibility. Now that the H5N1 strain is endemic in wild migratory birds, its potential for global spread is inevitable. The increased geographical range of migratory birds, which has recently been observed, only increases the chances of avian influenza virus acquiring the ten polymerase mutations, one or two HA mutations, and other unidentified mutations required for human-to-human transmission. The construction of poliomyelitis virus from its basic building blocks (Cello et al., 2002) several years ago was controversial, but this work involved generating an existing organism. Tumpey et al. (2005) have resurrected an infectious agent that is currently extinct. We argue that reconstruction of the 1918 pandemic influenza virus is a risk worth pursuing as it promises the reward of learning how a pandemic virus comes into being and why it is so pathogenic. Researchers already have identified new sentinel sequences in the polymerase and HA genes that are implicated in adaptation of avian influenza viruses to humans. They have also discovered a new mechanism for HA cleavage by NA that may provide a target for drug development. However, we still have not learned enough from the 1918 virus to predict or prevent a future pandemic. The role of the NS gene in disarming the interferon-based defense system of its mammalian host remains unresolved yet is open to study in the mouse. Furthermore, the molecular basis of the transmissibility of the 1918 virus is not yet fully understood. The greater reward of understanding this fundamental question brings with it greater risks associated with generating mutant 1918 viruses and performing transmissibility experiments using ani-

Previews 371

mal models such as ferret, minipig, or primates. Although we believe that the potential rewards for understanding the molecular basis of the transmissibility of the 1918 virus outweigh the risks, we certainly should consider at which point the risks of such experiments become great enough to restrict them to biosafety level 4 (BSL-4) containment. Acknowledgments C.J.R. and R.G.W. are supported by the Cancer Center Support Grant (CA 21765) and the American Lebanese Syrian Associated Charities (ALSAC). C.J.R. is also supported by the Children’s Infection Defense Center (CIDC). R.G.W. is also supported by U.S. Public Health Service grant AI95357. We thank Klo Spelshouse in Biomedical Communications for preparation of the figure.

Charles J. Russell and Robert G. Webster Department of Infectious Diseases St. Jude Children’s Research Hospital Memphis, Tennessee 38105

Selected Reading Cello, J., Paul, A.V., and Wimmer, E. (2002). Science 297, 1016– 1018. Gamblin, S.J., Haire, L.F., Russell, R.J., Stevens, D.J., Xiao, B., Ha, Y., Vasisht, N., Steinhauer, D.A., Daniels, R.S., Elliot, A., et al. (2004). Science 303, 1838–1842. Garcia-Sastre, A. (2001). Virology 279, 375–384. Glaser, L., Stevens, J., Zamarin, D., Wilson, I.A., Garcia-Sastre, A., Tumpey, T.M., Basler, C.F., Taubenberger, J.K., and Palese, P. (2005). J. Virol. 79, 11533–11536. Ha, Y., Stevens, D.J., Skehel, J.J., and Wiley, D.C. (2001). Proc. Natl. Acad. Sci. USA 98, 11181–11186. Hatta, M., Gao, P., Halfmann, P., and Kawaoka, Y. (2001). Science 293, 1840–1842. Kobasa, D., Takada, A., Shinya, K., Hatta, M., Halfmann, P., Theriault, S., Suzuki, H., Nishimura, H., Mitamura, K., Sugaya, N., et al. (2004). Nature 431, 703–707. Skehel, J.J., and Wiley, D.C. (2000). Annu. Rev. Biochem. 69, 531– 569. Stevens, J., Corper, A.L., Basler, C.F., Taubenberger, J.K., Palese, P., and Wilson, I.A. (2004). Science 303, 1866–1870. Subbarao, E.K., London, W., and Murphy, B.R. (1993). J. Virol. 67, 1761–1764. Taubenberger, J.K., Reid, A.H., Lourens, R.M., Wang, R., Jin, G., and Fanning, T.G. (2005). Nature 437, 889–893. Tumpey, T.M., Basler, C.F., Aguilar, P.V., Zeng, H., Solorzano, A., Swayne, D.E., Cox, N.J., Katz, J.M., Taubenberger, J.K., Palese, P., and Garcia-Sastre, A. (2005). Science 310, 77–80. DOI 10.1016/j.cell.2005.10.019