In the 20th century, three pandemics of human influenza A virus occurred in 1918, 1957 and 1968; the one in 1918 was the most severe as 20 to 40 million people died worldwide. Pandemics are caused by influenza viruses with 'new' antigens. Such viruses can enter humans through the reassortment of genes between viruses that infect humans and those that infect animals or birds. (The 1957 and 1968 viruses acquired their genes from viruses that infected humans and an avian species1,2.) However, human pandemic influenza viruses may also arise in other ways. In 1997 in Hong Kong a lethal avian influenza virus was transmitted directly to humans from chickens and of the 18 known cases 6 were fatal. Despite the severity of the disease, relatively little is known about the viral proteins involved in pathogenicity.

The virulence of a virus is defined as its capacity to produce disease in a host3. For influenza viruses, due to the many factors involved it has proven difficult to elucidate the mechanism of virulence. The best understood determinant of virulence is whether the precursor of the hemagglutinin glycoprotein is cleaved intracellularly or extracellularly4. Other factors are also important; for example, virulence of the 1997 Hong Kong H5N1 influenza virus in mice is associated with the presence of lysine at residue 627 of the polymerase protein PB2 (ref. 5). The abilities of influenza virus to antagonize the host-defense interferon system6 and to either promote or inhibit apoptosis are also likely to contribute to the overall virulence of the virus. In this issue of Nature Medicine, Chen et al.7 describe a newly identified protein, derived from a second open reading frame (ORF) of PB1 and show that the PB1-F2 protein associates with mitochondria and induces cell death.

The discovery of the new influenza virus protein is a classic example of how chance favors the prepared mind. The work began with an investigation of whether alternative reading frames of viral genes could be translated and the polypeptides used to generate antigenic peptides that mediate a cytotoxic T-lymphocyte (CTL) response—a CD8+ phenotype restricted by class I major histocompatibility (MHC) antigen. Infection of cells by influenza virus was chosen as a model system because of the authors' experience with the virus and because the viral genome size is manageable (that is, it consists of eight negative-stranded RNA segments ranging from 2341 to 890 nucleotides in length for a total of about 13,588 nucleotides depending on the subtype8).

The authors performed a systematic search of the viral sequence in all reading frames for putative peptides with motifs containing primary binding sites for the mouse MHC H-2Db peptide-binding groove, and they tested candidate synthetic peptides in CTL-stimulating assays. They identified a novel influenza-virus-specific peptide that bound to H-2Db with high affinity. This novel peptide, PB1-F262–70, was derived from a previously unrecognized second (+1) ORF of PB1 (nucleotides 119–379; 87 amino acids for influenza virus strain A/PR/8/34). The peptide PB1-F262–70 was naturally expressed in influenza virus-infected cells and sensitized target cells against CD8+ T cells.

The PB1-F2 ORF is maintained in 64 out of 75 influenza A virus strains. Not surprisingly, given the lack of conservation, a genetic knockout of the PB1-F2 ORF yielded a viable virus for growth in tissue culture. (Genes that are non-essential for replication in tissue culture for many viruses, often play key roles in virus–host interactions, especially host-range and pathogenicity.)

Expression and mutagenesis studies suggest, but do not prove, that there is a single mRNA that is bicistronic and is used to translate PB1 and PB1-F2. Examination of the subcellular distribution of PB1-F2 in virus-infected cells showed that the protein is detected predominantly in the mitochondria, but is also present in the nucleus and cytosol; however, when PB1-F2 was expressed from a vector it mostly localized to mitochondria. PB1-F2 expression caused alterations to mitochondrial morphology and immunogold-labeling electron microscopy showed that PB1-F2 was almost exclusively associated with both the outer and inner mitochondrial membranes of transfected cells.

In a tour de force of synthetic peptide synthesis, the 87 residue PB1-F2syn protein was made, purified and microinjected into cells. PB1-F2syn localized to mitochondria and caused mitochondrial rounding, nuclear shrinkage and cell death. Remarkably, exogenous addition of PB1-F2syn to cells also induced cell death. Thus, as suggested by the authors7, perhaps PB1-F2 forms membrane pores.

Influenza virus infection of monocytes causes apoptosis and one inducer of apoptosis is the viral non-structural protein, NS1 (ref. 9). Infection of a monocyte cell line and freshly isolated human monocytes with wild-type influenza virus or the PB1-F2 ORF knock-out virus showed a 50% increase in apoptosis by the wild-type virus. However, such changes in the degree of apoptosis were dependent on cell type and not observed in standard laboratory epithelial cells. Although more work needs to be done to understand the mechanism by which PB1-F2 mediates apoptosis, the cell-type dependency may be important for pathogenicity. When PB1-F2 acts in cis it may disable virus-infected monocytes or other host innate immune cells whose normal role is to block viral infections. When PB1-F2 acts in trans (that is, when it is released from dying cells) it may inactivate the same types of host cells when they are recruited to the site of infection. Studies in appropriate model animal systems may help resolve some of these issues.

In addition to suggesting a mechanism by which the influenza A virus can inactivate the host immune response, the study by Chen et al.7 offers important insights into viral gene expression. The influenza A virus and the two closely related human viruses, influenza B and C, use a remarkable collection of mechanisms to expand their genome-coding capacity beyond the one-gene-per-genome RNA segment10 (Fig. 1). These strategies include translation of unspliced, spliced and alternatively spliced mRNAs, bicistronic mRNAs, as well as coupled stop/start translation of tandem cistrons. In many of these cases, protein translation occurs using overlapping reading frames. Influenza C virus also utilizes an unusual internal signal peptidase cleavage of a precursor protein to generate the integral membrane protein CM2 (Fig. 1e).

Figure 1: Coding strategies of influenza A, B and C viruses to increase viral proteins.
figure 1

a, The two ORFs in influenza A RNA segment 2 that encodes the polymerase PB1 and the newly described PB1-F2 proteins7. A single bicistronic mRNA is probably used for translation of both proteins. b, The two ORFs in influenza B RNA segment 6 that encodes the integral membrane proteins NB and NA (NA is a neuraminidase; NB function is unknown.) It seems that a bicistronic mRNA is translated to yield NB and NA using two AUG codons separated by four nucleotides. The 2 discrete proteins are translated from overlapping ORFs. TM, transmembrane domain of integral membrane protein. c, The 2 ORFs in influenza B RNA segment 7. A single mRNA contains 2 tandem cistrons, which encode the M1 and BM2 proteins. BM2 translation is thought to occur by a stop/start mechanism. The M1 (matrix) protein acts as a structural scaffold protein for the virion; BM2 function is unknown. t, termination codon for M1; i, initiation codon for BM2. d, Unspliced and alternatively spliced mRNAs derived from influenza A virus RNA segment 7. V-shaped lines indicate introns in the mRNAs. M1 functions as in c; M2 protein is an integral membrane protein with ion-channel activity, which is the target of the antiviral drug amantadine. CYT, cytoplasmic tail. e, Unspliced and spliced mRNAs derived from influenza C RNA segment 6. M1 is translated from a spliced mRNA. The p42 precursor protein is translated from the unspliced linear transcript mRNA and is processed by a signal peptidase at an internal signal sequence to yield p31 (rapidly degraded) and the CM2 integral membrane protein (function unknown). SP, signal peptide; ECT, ectodomain; CYT, cytoplasmic tail. f, Unspliced and spliced mRNAs derived from RNA segment 8 of influenza A and B and RNA segment 7 of influenza C. (Influenza A is shown here.) NS1 is an antagonist of interferon; NEP (NS2) protein is implicated in viral-nucleocapsids12. (Adapted from ref. 10, except for a).

So far neither ribosomal frameshifting nor suppression of termination of translation has been identified for influenza viruses. Furthermore, the process of pseudo-templated addition of nucleotides to generate RNAs of different coding capacity, which was found for the V/P gene of paramyxoviruses11, has not been found for influenza viruses. Nonetheless, influenza viruses have not only increased the number of proteins encoded by genomes of limited size by using diverse coding strategies, but these strategies have also provided a means by which to regulate the expression of these proteins.