Mutagenic Analysis of the 5′ Arm of the Influenza A Virus Virion RNA Promoter Defines the Sequence Requirements for Endonuclease Activity

Abstract

Short synthetic influenza virus-like RNAs derived from influenza virus promoter sequences were examined for their ability to stimulate the endonuclease activity of recombinant influenza virus polymerase complexes in vitro, an activity that is required for the cap-snatching activity of primers from host pre-mRNA. An extensive set of point mutants of the 5′ arm of the influenza A virus viral RNA (vRNA) was constructed to determine the cis-acting elements which influenced endonuclease activity. Activity was found to be dependent on three features of the conserved vRNA termini: (i) the presence of the 5′ hairpin loop structure, (ii) the identity of residues at positions 5 and 10 bases from the 5′ terminus, and (iii) the presence of base pair interactions between the 5′ and 3′ segment ends. Further experiments discounted a role for the vRNA U track in endonuclease activation. This study represents the first mutagenic analysis of the influenza virus promoter with regard to endonuclease activity.

Influenza A virus is a segmented, negative-stranded RNA virus. Influenza virion RNA (vRNA) serves as a template for the synthesis of two distinct RNA species by a virus-encoded RNA-dependent RNA polymerase. Complementary RNA (cRNA) is a full-length copy of vRNA which serves as a template for the synthesis of new vRNA. mRNA is an incomplete copy of vRNA which is capped at its 5′ end and polyadenylated at its 3′ end. The cap is cleaved from host mRNAs by the endonuclease activity of the RNA polymerase complex (20).

The RNA-dependent RNA polymerase of influenza virus is a complex of three viral proteins (PB1, PB2, and PA), associated with each of the eight nucleoprotein (NP) encapsidated viral gene segments (26). All eight RNA segments have 12 and 13 conserved nucleotides at their 3′ and 5′ termini, respectively (8, 34), which show partial inverted complementarity permitting the formation of an RNA panhandle structure (15). The conserved 5′ and 3′ vRNA termini constitute the promoter and are required for transcription, endonuclease, and polyadenylation activities (12, 14, 25, 27, 29, 32, 33, 35). A detailed mutational analysis of the vRNA promoter sequence in vitro has shown that base pairs within the promoter, rather than the identity of nucleotides themselves, are important for polymerase activity (12). Base-pairing was required between nucleotides 10 and 15 of the 3′ terminus and 11 and 16 of the 5′ terminus, whereas no interaction was detected between nucleotides 1 to 9 of the 3′ terminus and 1 to 10 of the 5′ terminus (12). Further studies suggested that those 5′ terminal nucleotides which were not involved in base pairing between the two vRNA termini formed a 5′ hairpin loop or 5′ hook structure (29, 33). This structural feature was also identified by in vivo reporter gene experiments, which suggested that such hairpin loop structures are present in both the 3′ and 5′ termini of vRNA (11).

Endonuclease activity, resulting in cap snatching from host mRNA, is carried out by the influenza virus RNA polymerase complex binding to the cap structure at the 5′ end of host cell mRNA, which it cleaves 9 to 15 nucleotides from the cap structure (2, 9, 19, 23, 24, 41). The favored cleavage site is usually 3′ of a purine residue (28), and the specificity for mRNA cleavage is probably determined by the recognition of the cap (17). RNAs with either cap 1 structures (possessing both a 7-methyl G and a 2′-O-methyl group), or cap 0 structures (possessing only a 7-methyl G group) can be cleaved efficiently by the influenza virus polymerase complex by a proposed two-metal-ion mechanism (10). The resulting short capped oligonucleotides are then used by the influenza virus polymerase as primers for RNA transcription from the viral template (1, 6, 18). Cap 0 mRNAs are approximately 10% as effective in priming as cap 1-containing globin mRNAs, however, indicating that the presence of a 2′-O-methyl group is important in priming influenza virus mRNA (4, 5). The endonuclease and polymerase activities show differences in metal ion preference (10), and inhibitors selective for either endonuclease or polymerase activities are known (39, 40), showing that the endonuclease active site is separate from that of the polymerase. Cap structures are known to cross-link specifically to the PB2 subunit of the RNA polymerase complex (41), suggesting that the PB2 subunit catalyzes endonuclease activity, although all three polymerase subunits (PB1, PB2, and PA) appear to be required for activity (3, 36).

The endonuclease activity of the influenza virus polymerase complex is dependent on the presence of influenza virus-like vRNA molecules (14). However influenza virus-like cRNA promoter-containing RNAs do not stimulate endonuclease activity, even though they will both bind influenza virus polymerase proteins and are efficiently transcribed (7, 31). Endonuclease activity takes place without the need for transcription. Therefore, cis-acting elements within the influenza virus promoter which stimulate endonuclease activity can be studied in isolation from those which regulate transcription, giving valuable insight into the RNA sequences and structures required for endonuclease activity.

In this study, recombinant influenza virus polymerase was prepared using vaccinia virus vectors expressing the PB1, PB2, and PA influenza A virus polymerase proteins (14, 37). The recombinant polymerase complex was used to investigate the promoter requirements for viral endonuclease activity through site-directed mutagenesis of short synthetic influenza virus vRNA and cRNAs containing promoter sequences. The effect of point mutations of each residue (1 to 13) within the 5′ vRNA arm was determined, including residues in the conserved 5′ hairpin loop and base-paired panhandle regions. Complementary “rescue” mutations which restored base-pairing previously disrupted by the initial point mutations were studied. Finally, the role of the U-rich track, known to be involved in polyadenylation activity, was determined.

MATERIALS AND METHODS

Construction of influenza virus-like model RNAs from a plasmid.

The influenza virus promoter-like wild-type RNA (Fig. 1A) and its mutants were synthesized from an internally truncated version of pBXPCAT1 (a gift from P. Palese), which was constructed by digesting with XhoI and BglII, end-filling with the Klenow fragment of DNA polymerase I, and religating. The resulting plasmid, pBXP49, no longer possessed the antisense copy of the chloramphenicol acetyltransferase (CAT) gene but contained terminal vRNA sequences derived from segment 8 of influenza virus A/PR/8/34. Mutated versions of pBXB49 were made by an inverse PCR technique using Pfu DNA polymerase (32) and sequenced to confirm mutations. The plasmids encoding influenza virus cRNA-like RNA (Fig. 1B), vRNA-like RNA lacking a U track (Fig. 1C), or cRNA-like RNA with a U track (Fig. 1D) were constructed by digesting pBXP49 with HindIII and XbaI and inserting double-stranded DNA synthesized by annealing two oligonucleotides of the required sequence. Mutations were incorporated by inverse PCR (32). The Thogoto virus (THOV) construct, pBXPTHOV, was made in the same way with oligonucleotides which encoded a 49-mer RNA with the sequence 5′ AGA GAA AUC AAG GCG UUU UUU UCA GAU CUC GAG UAC GCC UGU UUU UGC U 3′. The plasmids pBXP49, pBXP49c (Fig. 1B), and pBXPTHOV were linearized with BpuA1, and influenza virus vRNA-like RNA was synthesized by T7 RNA polymerase run-off transcription reactions (20 μl), typically containing 0.25 μg of linearized plasmid DNA, 25 U of T7 RNA polymerase, 10 U of RNasin (Promega), and 1 mM each of rNTPs in a buffer containing 40 mM Tris-HCl (pH 8.0), 8 mM MgCl₂, 50 mM NaCl, 2 mM spermidine, and 10 mM dithiothreitol. Reaction mixtures were incubated for 2 h at 37°C, extracted with phenol-chloroform, precipitated with ethanol, and redissolved in water. Incomplete transcripts and unincorporated rNTPs were removed by Centri-spin 20 columns (Princetown Separations) before the RNA was added to the endonuclease assays. The quantities of RNA used in the reactions were standardized after electrophoresis on 12% acrylamide (native) gels, followed by ethidium bromide staining.

FIG. 1.

vRNA and cRNA-like constructs (49 nt long) used in the mutagenic analysis of influenza virus endonuclease activity, showing potential RNA secondary structures formed. (A) vRNA; (B) cRNA; (C) vRNA lacking a U track; (D) cRNA with a U track adjacent to the 5′ arm. Nucleotides within the 5′ promoter arm are termed 1′, 2′, 3′, etc., to distinguish them from 3′-terminal nucleotides, which are termed 1, 2, 3, etc.

Preparation of recombinant influenza A virus polymerase/endonuclease.

Recombinant vaccinia virus vectors which express influenza A virus PB1, PB2, or PA proteins were grown in HeLa cells, and titer was determined in Vero cells (37). HeLa cells (non-Swiss) were coinfected with recombinant vaccinia viruses encoding the three influenza virus polymerase proteins (PB1, PB2, and PA) at a multiplicity of infection of 5 for each virus. After 18 h, cells were washed with phosphate-buffered saline (PBS) and then scraped from the walls of the culture flask with a rubber policeman. Cells were washed three times with PBS, and nuclear extracts were isolated by ammonium sulfate precipitation (14).

Synthesis of cap donors.

Cap 0 donors were prepared by a modification of a method described previously (14). Briefly, RNA transcripts were synthesized by SP6 RNA polymerase run-off transcription of SmaI-digested pGEM-7Zf(+) plasmid. The 67-nucleotide (nt) product was then extracted using phenol-chloroform, precipitated with ethanol, and resuspended in water. Unincorporated rNTPs were removed with nucleotide removal columns (Qiagen). The resulting RNA was capped and labeled in a concurrent capping and methylation reaction (25 μl) containing 2 U of guanylyltransferase-(guanine-7-)-methyltransferase-5′-triphosphate enzyme complex (Gibco-BRL) in 25 mM HEPES (pH 7.5)–12 mM MgCl₂–8 mM dithiothreitol–0.1 mM S-adenosylmethionine–5 μg of RNase-free carrier tRNA–10 U of RNasin (Promega)–20 μCi of [α-³²P]GTP (3,000 Ci/mmol). After 1 h of incubation at 37°C, the RNAs were extracted using phenol-chloroform, precipitated with ethanol, and resuspended in water. Unincorporated GTP and incomplete transcripts were removed with Centri-spin 20 spin columns (Princetown Separations).

Capped RNA endonuclease assays.

Typically, 20 to 30 nM capped ³²P-end-labeled 67-nt-long pGEM-7Zf(+)-derived RNA probe was used in a reaction including nuclear extract (usually 3 μl in a 10-μl total reaction volume) containing recombinant influenza virus polymerase in a buffer consisting of 50 mM HEPES (pH 7.5), 5 mM MgCl₂, 100 mM NaCl, 2 mM dithiothreitol, 100 μg of Escherichia coli tRNA, and 10 U of RNasin (Promega). This mixture was incubated at 30°C for 30 min to 1 h with about 1 pmol of synthetic influenza virus RNAs. Reactions were terminated by adding an equal volume of formamide stop solution (80% formamide, 10 mM EDTA, 0.1% xylene cyanol, 0.1% bromophenol blue). Products were analyzed by 15% polyacrylamide gel electrophoresis (PAGE) in 7 M urea and quantified by phosphor-image analysis. The yield of substrate and product was measured, and endonuclease activity was expressed as a percentage of the substrate cleaved.

RESULTS

vRNA but not cRNA catalyzes endonuclease activity in vitro.

To confirm that activation of the influenza virus endonuclease function was specific for vRNA, five independent endonuclease reactions were set up with and without the short synthetic influenza virus 49-nt-long model vRNA and cRNA molecules (Fig. 1, constructs A and B). Endonuclease activity, as detected by the presence of a low-molecular-weight band (P) in the expected position of a 12 nt-long capped product (14), was only detected when the model influenza virus vRNA containing both 5′ and 3′ sequences was added (Fig. 2, lane 5). The influenza virus cRNA promoter and the vRNA promoter of a closely related orthomyxovirus, Thogoto virus (22), failed to stimulate cleavage (Fig. 2, lanes 6 and 9). When a 14-nt-long chemically synthesized RNA identical in sequence to the wild-type 3′ arm of the vRNA promoter was added (5′ ACC CUG CUU UUG CU 3′), endonuclease activity was not detectable (Fig. 2, lane 3). Likewise, there was no cleavage activity in the presence of only the 15-nt-long 5′ arm of the vRNA promoter (5′ AGU AGA AAC AAG GGU 3′) (Fig. 2, lane 4). When PB2-expressing vaccinia virus was omitted from the preparation of recombinant polymerase (see Materials and Methods), no cleavage was evident even in the presence of vRNA promoter-like structures (data not shown). Overall, the results observed are in agreement with earlier studies (7, 14), confirming that vRNA but not cRNA is required for endonuclease activity.

FIG. 2.

Effect of various RNA templates on endonuclease activity analyzed by 15% PAGE in 7 M urea. Lane 1, minus enzyme; lane 2, no added RNA template; lane 3, 3′ vRNA sequence (5′ ACC CUG CUU UUG CU 3′); lane 4, 5′ vRNA (5′ AGU AGA AAC AAG GGU 3′); lane 5, vRNA (Fig. 1A); lane 6, cRNA (Fig. 1B); lane 7, vRNA lacking U track (Fig. 1C); lane 8, cRNA-like template with U track (Fig. 1D); lane 9, THOV vRNA (see Materials and Methods); lane 10, chemically synthesized 5′ ³²P-end-labeled 14-nt RNA size marker (labeled M) (12) (see lane 3). Minor bands in lanes 1 to 9 are probably nonspecific RNase degradation products. Minor bands in lane 10 represent contaminating RNA. S, 67-nt-long substrate; P, 12-nt-long cleavage product.

U₆ track does not influence endonuclease activity.

Influenza virus vRNA molecules contain a series of five to seven uridine residues adjacent to the 5′ arm of the promoter, which are termed the U track or U-rich track, known to be required for polyadenylation (30). A U track is not present in cRNA molecules, and so it could also function to identify the template as vRNA-like. To test this hypothesis, we constructed synthetic vRNA molecules lacking the U track and cRNA molecules with a U₆ track adjacent to the 5′ arm of the promoter (Fig. 1C and D, respectively). When cleavage assays were performed using the vRNA promoter lacking the U track, endonuclease activity was comparable to that in the wild type (Fig. 2, lane 7). In contrast, the cRNA molecule with the U₆ insertion was, like the cRNA constructs, inactive (Fig. 2, lane 8), demonstrating that the U track does not influence endonuclease activity.

Systematic study of endonuclease activity using vRNA with point mutations in the 5′ promoter arm.

In order to determine which nucleotides or secondary structures are important for endonuclease function, a set of point mutations at each position of the 5′ arm of the promoter were constructed, involving a transversion at each of the positions 1′ to 13′ as used previously in transcription assays (29). Cap donor endonuclease assays were conducted with each of the constructs, and the cleavage activity was compared with wild-type activity. The results of five independent sets of reactions, quantified by phosphor-image analysis, are summarized in Fig. 3. It can be seen that significant endonuclease activity was still evident in RNA with mutations at residues 4′, 6′, and 7′. However, cleavage was essentially at background levels in the presence of RNA with mutations in positions 1′, 2′, 3′, 8′, 9′, 11′, and 12′. Interestingly, intermediate levels of about 20% activity were observed when residue 5′ or 10′ was mutated. When cleavage reactions included RNA with mutations in both positions 5′ and 10′, no cleavage was observed (data not shown). Mutations at position 13′ resulted in detectable but low cleavage activity.

FIG. 3.

Quantitation of endonuclease cleavage by phosphor-image analysis. Standard deviations of the mean were calculated from five independent experiments. Results are corrected for minor nonspecific background cleavage in the absence of added RNA. Results for each mutant are expressed as a percentage of the cleavage of wild-type vRNA, which was set at 100%. The percent substrate cleaved by the wild type (WT) varied from 28 to 38% in different experiments.

Base-pairing between the 5′ and 3′ arm and the panhandle structure is required for endonuclease activity.

All recent influenza virus promoter models involve base-pairing between the 5′ residues 11′ to 16′ and the 3′ residues 10 to 15 of vRNA, which results in a base-paired region between the two ends of the RNA (11, 12). Residues nearer the 3′ and 5′ termini either remain single-stranded or adopt local secondary structure (Fig. 1A). Interarm base pairs between the 5′ and 3′ residues would be disrupted by mutations to residues 11′, 12′, and 13′, explaining the effect of such mutations at these positions on endonuclease activity (Fig. 3). To test this hypothesis, complementary mutations were inserted at 3′ positions 10, 11, and 12, restoring the base pairs (Fig. 4A). The results of a set of endonuclease assays using these constructs, as analyzed by 15% PAGE, are shown in Fig. 4B. It can be seen that endonuclease activity can be partially rescued by complementary mutations which restore base-pairing between the 3′ and 5′ promoter arms (Fig. 4B, lanes 2, 4, and 6; 48, 62, and 42%, respectively [average of five observations] and Fig. 3). As controls, we also constructed point mutations at positions 10, 11, and 12 of the 3′ arm, which, as expected, in all cases lacked any detectable endonuclease activity (Fig. 4B, lanes 1, 3, and 5). We conclude that base-pairing between the vRNA 5′ and 3′ termini is a major requirement for cap donor endonuclease activity. However, we cannot exclude that the nucleotide sequence per se is unimportant, since rescue was not 100%.

FIG. 4.

(A) RNA constructs used to investigate the effects of mutations in the panhandle region on endonuclease activity. The mutations are underlined. a, 10 U→A; b, 10 U→A + 11′ A→U; c, 11 C→A; d, 11 C→A + 12′ G→U; e, 12 C→A; f, 12 C→A + 13′ G→U. (B) Rescue of endonuclease activity by restoring base pairs in the panhandle, analyzed by 15% PAGE in 7 M urea. Lane 1, 10 U→A; lane 2, 10 U→A + 11′ A→U; lane 3, 11 C→A; lane 4, 11 C→A + 12′ G→U; lane 5, 12 C→A; lane 6, 12 C→A + 13′ G→U; lane 7, wild-type vRNA; lane 8, 14-nt-long marker (M) (see Fig. 2). S, 67-nt-long substrate; P, 12-nt-long cleavage product.

Hairpin loop in the 5′ promoter arm is required for endonuclease activity.

Nucleotide substitutions in positions 2′, 3′, 8′, and 9′ reduced endonuclease activity to undetectable levels (Fig. 3). To test whether this effect was due to the specific identity of these residues or to a hairpin loop RNA structure formed by the base-pairing of residues 2′ and 3′ with 9′ and 8′, two rescue mutations were constructed (Fig. 5A). It was found that a mutation from G to C at position 2′, which reduced endonuclease activity to below the limit of detection, could be rescued to 42% of wild-type activity (average of five experiments) by making the complementary mutation (C→G) at position 9′ (Fig. 5B, lanes 4 and 6). Likewise, a mutation from U to A in position 3′, which reduced endonuclease activity to below the level of detection, was rescued to 56% of wild-type activity (average of five experiments) by introducing the complementary mutation (A→U) at position 8′ (Fig. 5B, lanes 1 and 3). These results suggest that a 5′ hairpin loop structure involving base pairs between residues 2′ and 3′ and residues 8′ and 9′ is required for endonuclease activity.

FIG. 5.

(A) RNA constructs used to investigate the effects of mutations in the stem of the 5′ hairpin loop on endonuclease activity. The mutations are underlined. a, 3′ U→A; b, 8′ A→U; c, 3′ + 8′ rescue mutation; d, 2′ G→C; e, 9′ C→G; f, 2′ + 9′ rescue mutation. (B) Effect of substitution mutations which disrupt and potentially reform base-pairing analyzed by 15% PAGE in 7 M urea. Lane 1, 3′ U→A; lane 2, 8′ A→U; lane 3, 3′ + 8′ rescue mutation; lane 4, 2′ G→C; lane 5, 9′ C→G; lane 6, 2′ + 9′ rescue mutation; lane 7, no added RNA template; lane 8, minus enzyme; lane 9, wild-type vRNA. S, 67-nt-long substrate; P, 12-nt-long cleavage product.

Insertions of A residues into the hairpin loop destroy endonuclease activity.

The requirement for base-pairing between residues 2′, 3′, 8′, and 9′ of the 5′ arm of model vRNA shows that a hairpin loop is required for endonuclease activity. In order to determine how stringently the structure of the hairpin loop must be conserved within the vRNA promoter to maintain endonuclease function, five further mutants were constructed, with insertions between residues 3′ and 4′ or between 6′ and 7′. Specifically, we inserted an A between residues 3′ and 4′. We also inserted one, two, three, or four A residues between residues 6′ and 7′, as shown in Fig. 6A. All mutants were found to be inactive in the cap donor endonuclease assay (Fig. 6B, lanes 1 to 5) compared to the wild-type control (Fig. 1A). We conclude that although point mutations in the hairpin loop can be partly tolerated (Fig. 3), insertions, even of single A residues, abrogate activity.

FIG. 6.

RNA constructs used to investigate the effects of insertions within the 5′ hairpin loop on endonuclease activity. (A) The mutations are underlined. a, an A residue added at position 4′; b, an A residue added in position 7′; c, two A residues added in position 7′; d, three A residues added in position 7′; e, four A residues added in position 7′. (B) Effect of insertion mutations on endonuclease activity, analyzed by 15% PAGE in 7 M urea. Lane 1, one A residue added at position 4′; lane 2, one A residue added in position 7′; lane 3, two A residues added in position 7′; lane 4, three A residues added in position 7′; lane 5, four A residues added in position 7′; lane 6, wild-type vRNA. S, 67-nt-long substrate; P, 12-nt-long cleavage product.

Identity of nucleotides 5′ and 10′ in vRNA molecules plays a role in cleavage.

Because the mutation of nucleotides 5′ and 10′ reduced endonuclease cleavage markedly (Fig. 3) while neither residue is apparently involved in standard Watson-Crick base-pairing (Fig. 1), these nucleotides were mutated to all three alternative nucleotides. Five independent sets of reactions were performed, and a representative result is shown in Fig. 7. It was found that when 5′ G was replaced with either an A, C, or U residue, cleavage activity was reduced to between 5 and 32% of wild-type activity (Fig. 7, lanes 1, 2, and 3). Likewise, when 10′ A was replaced with a G, C, or U residue, cleavage was reduced to between 7 and 24% (Fig. 7, lanes 5, 6, and 7). As the 5′ G and 10′ A residues represent either a sequence or a structural difference between the vRNA promoter and the cRNA promoter (Fig. 1), they may represent one of the features which distinguish vRNA and cRNA.

FIG. 7.

Effect of substitution mutations in positions 5′ and 10′ on endonuclease activity, analyzed by 15% PAGE in 7 M urea. Lane 1, 5′ G→A; lane 2, 5′ G→C; lane 3, 5′ G→U; lane 4, wild-type vRNA; lane 5, 10′ A→C; lane 6, 10′ A→G; lane 7, 10′ A→U; lane 8, 14-nt-long RNA marker (M). S, 67-nt-long substrate; P, 12-nt-long cleavage product.

DISCUSSION

The influenza virus RNA polymerase complex, composed of the PB1, PB2, and PA polymerase component proteins, is involved in the synthesis of all three influenza RNA species—mRNA, vRNA, and cRNA. In addition to RNA polymerase activity, the complex possesses an endonuclease activity which cleaves cap structures complete with 9 to 15 heterologous nucleotides from host mRNA. The endonuclease activity of recombinant influenza virus polymerase complexes devoid of RNA is dependent on the addition of synthetic influenza virus vRNA-like molecules, but is not stimulated by the addition of synthetic cRNA molecules (7), indicating that vRNA is required to interact with the influenza virus polymerase complex when stealing cap structures from cellular mRNA in vivo. The failure of added synthetic cRNA templates to stimulate endonuclease activity demonstrates that, although both species are competent templates for polymerase binding and transcription, differences between the two promoter regions influence endonuclease activity.

Influenza virus vRNA molecules possess a sequence of five to seven uridine residues adjacent to the 5′ component of the promoter, but cRNA molecules do not. The U track is essential for polyadenylation activity and therefore for correct mRNA synthesis (30). Therefore, it is possible that it is also involved in the stimulation of mRNA cleavage and cap-snatching activity by influenza virus vRNA molecules. To test this, wild-type vRNA and cRNA constructs were examined for their ability to stimulate specific endonuclease cleavage by recombinant influenza virus polymerase proteins. The results (Fig. 2) were compared with those for a vRNA construct lacking the U track and with those for a cRNA construct with six U residues next to the 5′ promoter component. Cleavage only occurred when the vRNA promoter, with or without the U track, was added. The cRNA construct with an added U₆ track did not show detectable endonuclease activity, demonstrating that, in contrast to polyadenylation, the U track does not influence mRNA cleavage and cap snatching.

In order to further characterize the role of the promoter in endonuclease activity, a synthetic RNA resembling the THOV vRNA promoter was tested for its ability to stimulate endonuclease cleavage reactions. THOV polymerase complexes can transcribe influenza virus vRNA-like templates in vitro (21) and copy influenza A virus templates in vivo (13). Moreover, influenza virus polymerase complexes can transcribe and polyadenylate THOV vRNA-like templates in vitro (data not shown). However, cells which synthesize THOV core proteins from cloned cDNAs fail to express influenza virus-like CAT RNA constructs in vivo (42). This suggests that subtle differences between the two orthomyxovirus promoters affect a stage of the viral life cycle other than transcription or polyadenylation. Obvious contenders are endonuclease activity and packaging. While the influenza virus vRNA promoter is superficially very similar in primary sequence to the THOV vRNA promoter, the THOV vRNA promoter fails to stimulate endonuclease activity (Fig. 2). As there are very few nucleotide differences between the two viral structures, pinpointing likely residues which may be important for cleavage is straightforward. In the THOV vRNA promoter, nucleotide 3′ is an A, whereas there is a U in this position within the influenza virus promoter (Fig. 1A). In addition, nucleotide 8′ is an A in influenza virus but a U in the THOV vRNA promoter. Thus, although the primary sequence may differ between the two promoters, base-pairing within the stem of the 5′ hairpin loop is maintained. In fact, the THOV vRNA promoter closely resembles an influenza virus 3′ + 8′ rescue mutant, which stimulates endonuclease activity to 75% of wild-type levels (Fig. 5Ac). This suggests that the differences in the base-pairing at 3′ and 8′ between THOV and influenza virus vRNA promoters is not solely responsible for the lack of endonuclease cleavage activity by THOV vRNA. Other residues must be involved. For example, nucleotide 5′ is a G in influenza virus but an A in THOV, which may go some way to explain why the THOV promoter fails to stimulate influenza virus polymerase endonuclease activity, since a 5′ G→A mutation in the influenza virus vRNA promoter inhibited endonuclease cleavage (Fig. 7).

A thorough mutagenic analysis of the 5′ arm of the vRNA promoter showed that nucleotides 11′ and 12′ were essential for cleavage activity (Fig. 3). These residues were also important in previous in vitro and in vivo studies of transcription and polyadenylation (11, 12, 30, 31) because they are known to be involved in forming the RNA panhandle structure by base-pairing with residues in the 3′ arm of the promoter. In order to assess the relevance of these base pairs for endonuclease activity, mutations were made to residues 10, 11, and 12, which form the 3′ arm of the base-paired region of the panhandle, and complementary mutations were made to nucleotides 11′, 12′, and 13′ of the 5′ arm, restoring base-pairing (Fig. 4A). In each case, endonuclease activity was rescued, although only to about 50% of wild-type levels (Fig. 4B). When residue 12 of the 3′ arm was mutated, cleavage activity was still just detectable (Fig. 3). As residues 12 and 13′ form a base pair in the middle of a duplex region (Fig. 1), it is possible that base pairs on either side of the mutation may offer a degree of stability to the panhandle structure in this particular case.

In vitro transcription and polyadenylation of orthomyxovirus vRNA templates are dependent on a 5′ hairpin loop (22, 33). In addition, in vivo work has suggested that both 5′ and 3′ hairpin loops are required at some point in the viral life cycle (11). To determine whether the 5′ hook was required for endonuclease activity, mutations were made at positions 2′ and 3′, disrupting the stem of the hairpin loop. Complementary mutations were also made at positions 9′ and 8′ to reform base pairs with the mutants at positions 2′ and 3′. In both cases, cleavage activity was partially rescued (Fig. 5), indicating the importance of a 5′ hairpin loop in endonuclease activity. When the 5′-terminal A residue was mutated, endonuclease activity was abrogated, whereas mutations to residues 5′ and 10′ had a less profound effect on endonuclease cleavage activity (Fig. 3). Although nucleotides 1′, 5′, and 10′ do not appear to be involved in the formation of secondary structures within the promoter, it is likely that residue 1′ A is required for polymerase binding (12).

Although the 5′ arm of influenza virus vRNA constitutes a polymerase-binding site (12, 16, 29, 38), residue 5′ has not previously been implicated in polymerase binding. Moreover, while residue 10′ has been reported by some to be important in binding (16, 38), this is not universally accepted (12). The effect of mutations at 5′ and 10′ was particularly interesting because these nucleotides represent either a sequence or a structural difference between the cRNA promoter and vRNA promoter. In addition, the residue at position 5 was different in THOV vRNA. Thus, because the three promoters are very similar in other respects, it is likely that the sequence identity or the effect on secondary structure of these two nucleotides represents an important cis-acting element in the activation of endonuclease activity by the influenza virus vRNA promoter. Both residues 5′ and 10′ were replaced with all three alternative nucleotides, and in all cases endonuclease cleavage was inhibited (Fig. 7). As the identity of the bases at positions 5′ and 10′ appeared to be important and because no existing model of influenza virus vRNA promoter structure indicates that either nucleotide is involved in base pair interactions, it is likely that the specific identity of these residues is important for mRNA cleavage.

It has been suggested, based on an in vivo CAT reporter assay, that the 5′ G residue interacts with viral polymerase in a locally single-stranded conformation (11). This indicates that the conservation of the 5′ G is crucial to the viral life cycle. Although cleavage requires both 5′ and 3′ termini of the influenza virus vRNA promoter, the 5′ arm has been shown to be sufficient to stimulate capped mRNA binding by the influenza virus polymerase (7). Therefore, it is tempting to speculate that the 5′ G residue may be required for capped mRNA binding preceding the cleavage of host-derived pre-mRNAs. This would explain why mutation of this residue has such a pronounced effect on the viral life cycle when it has no apparent role in polymerase binding or transcription. In vivo experiments have previously found the 10′ A residue to be of intermediate variability, and it has been suggested that it may constitute a flexible joint within an angular structure of the two rigid elements comprising the panhandle region and the 5′ hairpin loop (11).

To further characterize the role of the 5′ hairpin loop, a series of insertion mutants were tested for their ability to stimulate endonuclease activity in vitro. Constructs were made with one or more A residues inserted into 5′ position 4 or 7. All mutations reduced endonuclease activity to below the level of detection. It may be that residue 5′, already shown to be essential for endonuclease activity (see above), must be held in a specific position by other nucleotides in the hairpin loop and that these mutations prevent the correct interactions between the polymerase proteins and the 5′ G residue.

It is important to note that the partial loss of endonuclease activity of the mutant vRNAs studied here could reflect a failure of the mutant RNAs to bind to the RNA polymerase rather than a direct effect on the binding or endonucleolytic cleavage of the capped substrate. Indeed, it is known that many of the mutants of the 5′ strand of vRNA studied in related but not identical vRNA-like compounds bind RNA polymerase less efficiently than their corresponding wild-type RNAs (12, 16, 38). This may suggest that the primary reason for the loss of endonuclease activity in some mutants (e.g., 2′ and 9′) is a failure of RNA binding. However, in other mutants (e.g., 5′ and 10′), binding is apparently unaffected (12). A much more exhaustive and detailed study, beyond the scope of this paper, is therefore required to establish which mutations affect binding of the enzyme and/or substrate and which, if any, directly affect endonuclease cleavage.

In conclusion, our results confirm that the 5′ arm of the influenza A virus vRNA promoter is stringently required for endonuclease activity. Although the influenza virus cRNA promoter and the vRNA promoter of THOV are recognized and transcribed by the influenza virus polymerase complex in vitro, neither stimulates endonuclease activity. The U track present in all influenza virus vRNA molecules, while essential for polyadenylation of influenza virus mRNA, plays no role in stimulating host mRNA cleavage, indicating that the signals which influence cleavage reside wholly in the promoter. The specific identities of nucleotides 5′ and 10′ within the 5′ promoter arm, along with the residues responsible for forming the vRNA panhandle and 5′ hairpin loop structures, are crucial for cap-snatching activity by the influenza virus polymerase complex.