Introduction
Many viruses including influenza, herpes, adeno‐ and retroviruses replicate their genomes in the nucleus. Therefore, regulated mechanisms must exist by which the nucleic acids of these viruses enter and leave the nucleus to begin and to complete the virus replicative cycle, respectively. The genome of influenza A virus consists of eight segments of negative‐sense RNA, which are bound by the viral polymerase at the termini and are coated with the viral nucleoprotein (NP) to form ribonucleoprotein (RNP) complexes. Upon entry, virus is first uncoated and delivers into the cytosol its RNP, which is then transported into the nucleus (
O'Neill et al., 1995;
Whittaker et al., 1996b). Influenza virus assembly occurs at the plasma membrane at late times after infection; therefore, a switch in the direction of viral RNA transport from the nucleus to the cytoplasm must occur for the virus replicative cycle to be completed.
The major gateway for nucleocytoplasmic transport of macromolecules is the nuclear pore complex (NPC) (
Davis, 1995;
Nigg, 1997). We have shown that influenza virus NP is required for transport of viral RNA into the nucleus to initiate virus replication and that the viral NP uses the protein nuclear import pathway, including the four soluble import factors karyopherin α, karyopherin β, Ran and p10 for viral RNA import (
O'Neill et al., 1995). The present work addresses the factors involved in the export of viral RNA. Since most RNAs are protein associated, it is believed that protein factors mediate the export of specific RNA classes (reviewed by
Izaurralde and Mattaj, 1995). Indeed, direct involvement of several proteins in RNA transport pathways has already been demonstrated. First, the human immunodeficiency vius type 1 (HIV‐1) Rev protein mediates the nuclear export of unspliced or partially spliced viral RNAs (
Felber et al., 1989;
Malim et al., 1989b;
Fischer et al., 1994,
1995;
Stutz and Rosbash, 1994). Second, cellular mRNAs as RNP particles have been observed in mid‐transport through the NPC (
Mehlin et al., 1992). Because the hnRNP A1 protein shuttles between the nucleus and the cytoplasm and has mRNA‐binding properties, it and other hnRNP proteins are believed to play a direct role as a carrier for the export of cellular mRNAs (
Piñol‐Roma and Dreyfuss, 1992;
Michael et al., 1995). In support of this, the hnRNP A1‐like protein hrp36 of
Chironomus tentans associates with pre‐mRNA of the Balbiani ring complex and remains complexed with the RNA during mRNA processing and transport to the cytoplasm (
Visa et al., 1996). Third, TFIIIA and L5 proteins of
Xenopus have been shown to be involved in the transport of 5S rRNA (
Guddat et al., 1990).
Protein export requires a specific nuclear export signal (NES), which functions independently of surrounding peptide sequences and can confer the ability to be transported out of the nucleus on another protein to which it is fused. Recently, several NES‐containing proteins have been identified, including the lentivirus proteins Rev and Rex (
Fischer et al., 1995;
Palmieri and Malim, 1996), the cAMP‐dependent protein kinase inhibitor (PKI) (
Wen et al., 1995), hnRNP A1 (
Michael et al., 1995), amphibian TFIIIA (
Fridell et al., 1996b) and yeast Gle1 (
Murphy and Wente, 1996). In the case of the HIV‐1 Rev protein, the function of its NES correlated with the ability to bind to the FG‐repeat elements of nucleoporins (
Bogerd et al., 1995;
Fritz et al., 1995;
Stutz et al., 1995). Human (Rab/hRip1) and yeast (yRip1) nucleoporins were identified using the yeast two‐hybrid system as Rev‐binding proteins, which interact with the Rev NES/effector domain. Mutations which disrupt the Rev NES also abolish the ability of Rev to interact with Rab/hRip1 and yRIP1. Nuclear export signals within the proteins Gle1 and PKI also interact with Rab, further strengthening the connection between NESs and nucleoporin function in nucleocytoplasmic transport (
Fridell et al., 1996a;
Murphy and Wente, 1996).
We hypothesized that influenza virus promotes the export of its own RNA by producing a factor(s) which interact(s) with nucleoporins in much the same way that Rev does. By screening influenza virus‐encoded proteins for the ability to bind nucleoporins we hoped to identify the protein(s) which mediate viral RNA export. In the yeast two‐hybrid system, only the viral NS2 protein was able to interact with different nucleoporins. Moreover, the NS2 or a small NS2 amino‐terminal peptide could functionally replace the Rev effector/NES domain as a Rev–NS2 fusion protein. In addition, an amino‐terminal peptide of NS2 promoted the nuclear export of a heterologous protein to which it was cross‐linked. These experiments suggest that NS2 is a Rev‐like, NES‐containing protein. Significantly, antibodies directed against the NS2 protein, when injected directly into the nucleus, inhibited the cytoplasmic accumulation of viral NP and RNP at late times after infection. These experiments suggest a role for the NS2 protein in transport of newly synthesized viral RNA from the nucleus to the cytoplasm. Since NS2 (non‐structural protein 2) is misnamed, i.e. it has been demonstrated to be a structural protein (
Lamb et al., 1978;
Richardson and Akkina, 1991;
Yasuda et al., 1993), we renamed this protein in accordance with the function we have identified—the viral
nuclear
export
protein or NEP.
Discussion
We have identified the influenza virus NEP as a virus‐encoded nuclear export factor and present a model suggesting that NEP mediates the nuclear export of the viral RNPs at late times in the infectious cycle. The experimental evidence is 4‐fold. First, we found using the yeast two‐hybrid system that NEP binds to cellular nucleoporins. Second, the NEP can functionally replace the effector domain of HIV‐1 Rev in a CAT‐based reporter assay involving the nuclear export of RNA transcripts. Third, the effector domain was mapped to the N‐terminal amino acids of NEP; specifically, a nuclear export function was found to be associated with amino acids 11–23 since the corresponding peptide promoted the nuclear export of a heterologous reporter protein to which it was linked. Fourth, antibodies directed against NEP when microinjected into the nucleus of infected cells prevented the export of NP and RNP.
In the influenza virus‐infected cell, the NEP localizes to the nucleus (
Greenspan et al., 1985). Previously, the NEP was found in the virion in low quantities, but no function was attributed to it (
Lamb et al., 1978;
Richardson and Akkina, 1991;
Yasuda et al., 1993). In the virion, the NEP is bound to the viral matrix protein M1, and during subfractionation of viral particles the NEP co‐purifies with the M1 (
Yasuda et al., 1993). This interaction has been duplicated in the yeast two‐hybrid system (
Ward et al., 1995; our unpublished data), and in filter binding assays (
Yasuda et al., 1993). Binding of NEP to M1 may be critical for the nucleocytoplasmic transport function of the NEP protein. Participation by M1 in the nuclear export of influenza virus RNP was demonstrated by Helenius and colleagues, and nuclear M1 was shown to be required for viral RNPs to be exported (
Martin and Helenius, 1991). Furthermore, antibodies directed against M1 also inhibited RNP export when microinjected into infected cells (
Whittaker et al., 1996a). Despite this indirect evidence for a role for M1 in nuclear export, we were unable to show that a functional interaction of M1 and NEP is required for this process. M1 is unable to interact with cellular nucleoporins in the yeast two‐hybrid system, suggesting that M1 is not a viral factor which interacts directly with the cellular transport machinery. Although the failure of M1 to interact with FG‐containing nucleoporins in the yeast two‐hybrid system could be due to trivial reasons, such as failure of LexA‐M1 to localize to the yeast nucleus, our other two‐hybrid system data suggest that it is NEP which binds to and acts as the adaptor molecule between the NPC and M1‐containing RNP complexes which form in the nucleus (
Figure 6). Although previous results using the ts51 mutant of influenza virus suggested that the M1 need not be physically associated with the viral RNP during nuclear export (
Rey and Nayak, 1992), our model would favor a direct role for the M1 in the export of the viral RNA. Also, the earlier data could easily be explained if small amounts of M1 (not detected under the assay conditions) were sufficient to allow the NEP to pull viral RNP through the nuclear pore. Finally, it has been shown that although the mutant M1 accumulates in the nucleus, it shuttles between the nucleus and the cytoplasm (
Whittaker et al., 1996a), and thus could participate in nuclear export of viral RNA. On the other hand, our data do not address directly the role of M1 in the nuclear export of RNP, and it is possible that the NEP directly interacts with RNA or (modified) RNP to facilitate export (
Figure 6).
The HIV‐1 Rev and the influenza virus NEP appear to have similar functional domains. Both of these proteins interact with the NPC through FG‐repeat‐containing nucleoporins. In fact, the relative strength of the interaction between NEP and the nucleoporins Rab/hRip1, yRip1 yNup100 and yNup116 is similar to that between Rev and the same nucleoporins. Also, mutations which abolish Rev and NEP NES function abolish their ability to act as nuclear export factors. Despite the limited number of nucleoporins tested, we believe that these results are significant and may suggest that there is a hierarchy of the strength of protein–protein interactions between NESs and the nucleoporin FG‐domains, and that the different binding activities may promote directionality of the export processes.
Further similarity between the nucleoporin‐binding domains of NEP and Rev was demonstrated in Rev functional assays: the NEP NES can substitute for the Rev effector domain in Rev–NEP fusion proteins, confirming that NEP contains an NES. It should be noted that in this assay system the Rev RNA‐binding domain provided the specificity for transport of RRE‐containing transcripts, and NEP alone was not able to promote the nuclear export of unspliced CAT mRNA. This is in contrast to the herpes simplex virus Us11 protein, which does not require fusion to the RNA‐binding region of Rev in order to interact with and induce transport of unspliced RRE‐containing RNAs (
Diaz et al., 1996). The transport activity of NEP is likely to be specific for influenza virus RNP, and this specificity appears to reside in peptide sequences downstream of the NES. Indeed, an M1‐binding domain has been mapped to the carboxy‐terminal 70 amino acids of NEP (
Ward et al., 1995). Unlike Rev, NEP has no inherent RNA‐binding activity; contact of NEP with influenza virus RNA is likely to be mediated by the protein M1 (
Figure 6). Thus, in this respect, the NEP may be similar to adenovirus E4‐34 kDa protein, which contains an NES but contacts viral mRNA through another virus protein E1B‐55 kDa (
Dobbelstein et al., 1997). However, in contrast to the RNA transporters of lenti‐ and adenoviruses, NEP mediates the export of non‐polyadenylated, non‐messenger viral genomic RNA.
In addition to NEP, the viral NS1 protein also bound to Rab/hRIP1. However, NS1 failed to bind to other nucleoporins tested. NS1 is a viral protein which regulates mRNA transport and inhibits export of viral and cellular mRNAs (
Alonso‐Caplen and Krug, 1991,
1992;
Fortes et al., 1994;
Qiu and Krug, 1994). Whether the NS1–Rab interaction is relevant to mRNA retention by NS1 is not clear but is presently under investigation.
In summary, the influenza virus NEP possesses a Rev‐like nuclear export function. It is an adaptor, which allows for the binding of viral RNP and nucleoporins and most likely for the transport of RNPs through the NPC. Unlike Rev, the NEP does not bind directly to its viral RNA but most probably recognizes the viral RNA as an RNP (
Figure 6).
Materials and methods
Yeast two‐hybrid constructions and screen
Saccharomyces cerevisiae EGY48 (MATα trp1 ura3 his3 LEU::pLEX‐Aop6‐LEU2); pEG202, pRFHM1 and pSH18‐34 were kindly provided by Dr R.Brent (Harvard Medical School) and have been described elsewhere (
Gyuris et al., 1993;
Zervos et al., 1993). Two‐hybrid constructs encoding yRip1 (amino acids 151–275), yNup100 (amino acids 278–539), yNup1 (amino acids 438–737) and yNup116 (amino acids 459–672) in pJG4‐5 were provided by Dr M.Rosbash (Brandeis University) (
Stutz et al., 1995). pVP16/Rab was provided by Dr B.Cullen (Duke University) (
Bogerd et al., 1995). pLexA‐NP and pLexA‐NS1 were described previously (
O'Neill and Palese, 1995;
Wolff et al., 1996). Two‐hybrid bait plasmids containing influenza virus cDNAs derived from the A/PR/8/34 strain (Young
et al., 1983;
Greenspan et al., 1985) were constructed in pEG202. PCR‐amplified influenza virus open reading frames were cloned between
EcoRI and
SalI restriction sites (PA),
EcoRI and
XhoI sites (NEP and M1),
NotI and
XhoI sites (PB1), or
BamHI and
XhoI sites (PB2). Assay of β‐galactosidase expression from pSH18‐34 in yeast cells transformed with various combinations of bait and prey plasmids was performed as previously described (
Gyuris et al., 1993;
Zervos et al., 1993).
Mammalian expression vectors
For tissue culture expression of NEP, Rev, the Rev M10 mutant protein and derivative fusion proteins, cDNAs were subcloned into the mammalian expression vector pcDNA3 (Invitrogen). Rev, Rev M10 and amino acids 1–69 of Rev (Rev
*) were cloned between
KpnI and
EcoRI restriction sites. NEP and the nested set of NEP deletions were subcloned between the
EcoRI and
XhoI sites. pcRev and pcM10, from which these constructs were derived, were kindly provided by B.Cullen (Duke University) (
Malim et al., 1989a). The reporter plasmid pDM128 was kindly provided by Tristram Parslow (University of California, San Francisco) (
Hope et al., 1990).
Transfections and CAT assays
Plasmid transfections were performed using the cationic liposomal reagent DOTAP (Boehringer‐Mannheim) according to the manufacturer's instructions. Briefly, duplicate 3.5 cm dishes of 293 cells, which were 60% confluent, were transfected with 2 μg of pDM128 and 3 μg of an effector plasmid expressing Rev, NEP or a Rev–NEP fusion protein as described in the text. Transfected cells were harvested after 48 h. Lysates were prepared by freeze–thawing three times in 100 μl of 0.25 mM Tris, pH 7.5, followed by microcentrifugation for 5 min at 4°C. Aliquots were then assayed for CAT activity, which was then quantitated by beta scanning of phosphorylated and non‐phosphorylated chloramphenicol separated by thin‐layer chromatography.
Cross‐linking of oligopeptides to a reporter protein
GST was expressed from pGEX‐2TK in
Escherichia coli BL21 (Novagen) and was purified from induced cultures on glutathione–Sepharose (Pharmacia) as previously described (
O'Neill and Palese, 1995). Two oligopeptides [NS2 NES (CDILLRMSKMQLES) and NS2mut1 (CDILLR
ASK
AQ
AES)] were synthesized at the Mt. Sinai Oligopeptide Synthesis Facility. Peptides were conjugated to GST through an amino‐terminal cysteine residue, which is not part of the NS2 sequence. Oligopeptide sequences correspond to NS2 amino acids 11–23. In order to activate GST for cross‐linking to peptides, 2.8 mg of GST in 1 ml was mixed with 2 mg of sulfosuccinimidyl 4‐(
N‐maleimidomethyl)cyclohexane‐1‐carboxylate (Pierce Chemical Co.) for 1 h at room temperature. Unincorporated cross‐linker was removed by chromatography on G50–Sepharose. Approximately 1 mg of GST in 600 μl was incubated with 3 mg of oligopeptide for 12 h at 4°C. GST conjugates were dialyzed against phosphate‐buffered saline (PBS) to remove unincorporated oligopeptides, and were stored in small aliquots at −70°C.
Affinity‐purified antibodies
GST was expressed from pGEX‐2TK (Pharmacia) in
E.coli BL21 and affinity purified on glutathione–Sepharose. NEP was expressed as a hexahistidine‐tagged protein in
E.coli BL21 from the vector pET28a (Novagen) and affinity purified on Ni‐NTA–agarose (Qiagen). Expression and purification of proteins were performed according to protocols provided by the manufacturer of the respective protein affinity matrices. Antisera were raised by immunization of rabbits with 300 μg of affinity‐purified protein in complete Freund's adjuvant followed by two booster injections with 150 μg in incomplete Freund's adjuvant at 3 week intervals. Affinity‐purified antibodies were prepared from 5 ml of rabbit immune serum by adsorption to NEP or GST, which had been cross‐linked to CNBr‐activated Sepharose 4B according to the recommendations of the manufacturer (Pharmacia, Inc.). Binding, washing and elution of bound antibodies were performed according to standard protocols (
Harlow and Lane, 1988).
Microinjection of cells
MDBK cells were cultured in reinforced minimal essential medium with 10% fetal bovine serum. Cells were seeded to glass coverslips 24 h prior to injection and 36 h prior to infection. Individual cells were microinjected using a micromanipulator and transjection apparatus from Eppendorf.
Oligopeptide‐conjugated GST and a control monoclonal antibody [HT103 (Dr J.L.Schulman, Mt Sinai)] were mixed, diluted to a final concentration of 1 mg/ml each and injected into the nuclei of MDBK cells. After injection, cells were incubated at 37°C for 30 min, and were then fixed and permeabilized for 20 min in 2.5% formaldehyde/0.25% Triton X‐100/PBS. Injected antigens were detected by indirect immunofluorescence. GST was detected in two steps using a rabbit polyclonal serum directed against GST and an FITC‐linked goat anti‐rabbit Ig serum. HT103 was detected in one step with Texas red‐linked goat anti‐mouse Ig serum. Injected cells were visualized by confocal laser microscopy.
Affinity‐purified anti‐NEP antibodies and protein A‐purified normal rabbit immunoglobulins were dialyzed against PBS (pH 7.4) and diluted to a final concentration of 1.3 mg/ml. Antibodies were microinjected into MDBK cells 24 h after seeding on glass coverslips. At 12 h post‐injection, cell monolayers were infected at 37°C for 30 min at a multiplicity of 0.5 infectious particles per cell. Eight hours after infection, cells were fixed and permeabilized as described above. The influenza virus nucleoprotein was detected in two steps using the monoclonal antibody HT103 and a Texas red‐linked goat anti‐mouse Ig serum. Injected antibodies were detected in one step with an FITC‐linked goat anti‐rabbit Ig serum.