The RNA Polymerase of Influenza Virus, Bound to the 5′ End of Virion RNA, Acts in cis To Polyadenylate mRNA

Abstract

We previously demonstrated, by limited mutagenesis, that conserved sequence elements within the 5′ end of influenza virus virion RNA (vRNA) are required for the polyadenylation of mRNA in vitro. To further characterize the nucleotide residues at the 5′ end of vRNA which might be involved in polyadenylation, a complete set of short and long model vRNA-like templates with mutations at nucleotides 1′ to 13′ (prime notation denotes numbering from the 5′ end) of vRNA were synthesized and transcribed in vitro. The products were assayed for mRNA production with both reverse transcription-PCR and [α-³²P]ATP incorporation assays. Results from these independent assays showed that vRNA templates with point mutations at positions 2′, 3′, 7′ to 9′, and 11′ to 13′ synthesized polyadenylated transcripts inefficiently compared with those with mutations at positions 1′, 4′ to 6′, and 10′. Positions 2′, 3′, 7′ to 9′, and 11′ are known to be involved in RNA polymerase binding. Furthermore, residues at positions 11′ to 13′ are known to be involved in base pairing between the 3′ and 5′ ends of vRNA. These findings demonstrate that the RNA polymerase has to bind to the 5′ end of the template vRNA, which must then interact with the 3′ end of the same template for polyadenylation to occur. These results support a model in which a cis-acting RNA polymerase is required for the polyadenylation of influenza virus.

Influenza A virus contains eight segments of single-stranded RNA of negative polarity (19). These RNA segments associate with the nucleoprotein and the three P proteins (PB1, PB2, and PA) to form a ribonucleoprotein (RNP) complex, which is responsible for transcription and replication of the viral genome. During the life cycle of the virus, the virion RNA (vRNA) genome is transcribed into mRNA and replicated into cRNA. cRNA is a full-length copy of vRNA and functions as a template for vRNA synthesis (13). In contrast, mRNA is an incomplete copy of vRNA. Transcription of mRNA is initiated by a capped RNA fragment which is cleaved from host mRNA by an endonuclease activity of the polymerase complex (21). mRNA synthesis is terminated at a track of uridines about 17 nucleotides (nt) from the 5′ end of the vRNA template, and polyadenylation then ensues. The polymerase complex (PB1, PB2, and PA) is responsible for both transcription and replication (7), but the factors controlling the alternate modes of transcription and replication are not well understood. The nucleoprotein may be involved in the “switch” from transcription to replication by inhibiting premature termination of cRNA synthesis (1, 28).

All eight vRNA segments have at their 3′ and 5′ ends, respectively, conserved sequences of 12 and 13 nt which are partially complementary and which were proposed to be regulatory elements for RNA transcription and replication (3, 24). The recent development of in vitro and in vivo systems has allowed the study of RNA signals for the regulation of influenza virus vRNA, cRNA, and mRNA syntheses (8, 10, 11, 16–18, 20, 23, 26). In vivo studies showed that model vRNA templates with the 26 3′-terminal and the 22 5′-terminal nt of influenza A virus contain the signals for transcription and replication and for packaging of RNA into viral particles (16). Initially, the conserved 3′ end was thought to suffice as the promoter for mRNA and cRNA syntheses (20, 26). Further in vitro studies, however, showed that the conserved 5′ end contains a polymerase binding site (4, 30) and that the polymerase requires both the 3′ and the 5′ end for transcription initiation (4, 5) and for endonuclease activity (2, 6).

Early sequence analysis of the influenza virus mRNA indicated that the polyadenylation site is a track of uridine residues (five to seven residues long, i.e., U_5–7) near the 5′ end of vRNA (24, 25). Since the track of uridines is next to the predicted base-paired region, this observation led to a model for polyadenylation in which the base-paired region acts as a physical barrier for viral transcription (25). Therefore, instead of transcribing the 5′ end of vRNA, the viral polymerase reiteratively copies the U_5–7 track, thereby adding a poly(A) tail to the 3′ end of mRNA. Later, in vivo studies demonstrated that the U_5–7 track and the adjacent base-paired region were essential for the expression of a model chloramphenicol acetyltransferase (CAT) reporter gene (14, 15). However, the involvement of the 5′ vRNA terminus in transcription initiation and polymerase binding suggested a new model for polyadenylation (4, 30). In this revised model, the polymerase is unable to transcribe the 5′-terminal sequence to which it is bound because of steric hindrance. Instead, reiterative copying of the U_5–7 track results in the addition of a poly(A) tail to the transcript. Our recent study supports this new model by demonstrating that the 5′ end of vRNA is required for polyadenylation (23).

Here, to further understand the molecular mechanism controlling polyadenylation, point mutations of each of the 13 residues within the conserved 5′ terminus of vRNA were constructed and polyadenylation was characterized by two independent in vitro assays (23). We found that residues which are known to be involved in polymerase binding (4) were required for polyadenylation. In addition, residues which are known to be unimportant for polymerase binding but which are known to be needed for base pairing between the 3′ and 5′ ends in the RNA fork structure (4) were also required for polyadenylation. These observations suggested that the influenza virus RNA polymerase complex has to bind to the 5′ end of the template vRNA and further implied that the 5′ end of the template vRNA has to interact with the 3′ end of the same template for polyadenylation to occur.

MATERIALS AND METHODS

Preparation of influenza A virus polymerase.

RNA polymerase was isolated from influenza A virus strain X31, a reassortant of A/HK/8/68 and A/PR/8/34, as described previously (26). Briefly, the virus was disrupted with Triton X-100 and lysolecithin. RNP was then separated by glycerol step gradient centrifugation, followed by micrococcal nuclease (Sigma) digestion to remove endogenous vRNA.

Construction of plasmids.

The 717-nt-long wild-type RNA and its mutants were synthesized from pBXPCAT1 (a gift from P. Palese) and its derivatives. Plasmid pBXPCAT1 encodes a 717-nt-long RNA with an antisense CAT gene flanked by linker sequences and vRNA terminal sequences derived from segment 8 of influenza virus A/PR/8/34 (Fig. 1A) (14). Mutated plasmids of pBXPCAT1 were synthesized by PCR (22). Plasmids encoding the wild-type 49-mer RNA and its mutants (Fig. 1B) were obtained by digesting pBXPCAT or its derivatives with XhoI and BglII, end filling with the Klenow fragment, and religating with T4 DNA ligase. All of the mutated sequences were confirmed by DNA sequencing.

FIG. 1.

RNA templates used in in vitro influenza virus transcription reactions. The proposed base pairs in the RNA fork model (4, 5) are indicated by vertical lines. Prime notation denotes nucleotide numbering from the 5′ end to distinguish from 3′-end nucleotides (4). Point mutations from positions 1′ to 13′ are indicated above the sequences. The U₆ track, the proposed poly(A) site, is in boldface type. (A) Sequence of the 717-nt template. Arrowheads indicate the XhoI and BglII sites in plasmid pBXPCAT1. Underlining and overlining indicate initiation and termination codons of the CAT gene, respectively. (B) Sequence of the 49-mer template.

RNA template preparation.

RNA templates were synthesized with 25 U of T7 RNA polymerase in 20-μl reaction mixtures containing 0.25 μg of BpuAI-linearized plasmid DNA, 10 U of placental RNase inhibitor, 1 mM each nucleoside triphosphate (NTP), 40 mM Tris-HCl (pH 8.0), 8 mM MgCl₂, 50 mM NaCl, 2 mM spermidine, and 10 mM dithiothreitol. Reaction mixtures were incubated at 37°C for 20 min to 2 h, followed by 10 min of incubation with 2 U of RNase-free DNase I at 37°C. The RNA was purified by phenol-chloroform extraction and ethanol precipitation. RNA templates were quantified either by gel electrophoresis followed by ethidium bromide staining (717-nt RNAs) or by gel electrophoresis of RNA, labelled using polynucleotide kinase and [γ-³²P]ATP, followed by PhosphorImager (Molecular Dynamics) analysis (49-mer RNAs). Except for the 2′ (prime notation denotes numbering from the 5′ end) (G→U) vRNA mutant in the [α-³²P]ATP incorporation assay (see Results), the same amounts of each template RNA were used for each influenza virus reaction within an assay.

In vitro influenza virus transcription.

Five- to 20-μl reaction mixtures contained 0.1 to 1 μg of RNA template, 1 μg of nuclease-treated RNP (about 5 ng of polymerase protein [27]), 500 μM each NTP, 0.5 mM adenylyl (3′→5′) guanosine (ApG), 50 mM Tris-HCl (pH 7.4), 50 mM KCl, 10 mM NaCl, 5 mM MgCl₂, 5 mM dithiothreitol, and 10 U of placental RNase inhibitor. The reactions were incubated at 30°C for 3 h. Reaction products from the 49-mer RNA were analyzed by gel electrophoresis (see below: [α-³²P]ATP incorporation assay), and reaction products from the 717-nt vRNA were used directly for reverse transcription-PCR (RT-PCR) analysis (see below: RT-PCR assay).

[α-³²P]ATP incorporation assay.

The concentration of ATP in the in vitro transcription mixtures was reduced to 25 μM, and 2 μCi of [α-³²P]ATP (3,000 Ci/mmol) (Amersham) was added to a 5-μl reaction mixture. The reaction mixture was incubated for 3 h at 30°C. Reaction products were phenol-chloroform extracted and precipitated in ethanol with 10 μg of Escherichia coli carrier tRNA and 2.4 M ammonium acetate. RNA pellets were resuspended in formamide loading dyes, heated at 99°C for 3 min, and analyzed on 16% polyacrylamide–7 M urea gels. To quantify the yield of mRNA products, the gels were dried and the high-molecular-weight smear of mRNA products was examined by PhosphorImager analysis.

RT-PCR assay.

Polyadenylated products of 717-nt vRNA from in vitro transcription reactions were reverse transcribed in 10-μl reaction mixtures containing 50 pmol of 5′ GC-clamped T₂₀ primer (5′-GCCCCGGGATCCT₂₀-3′), 200 μM each dNTP, 10 U of placental RNase inhibitor, 100 U of Moloney murine leukemia virus reverse transcriptase (Promega), 10 mM Tris-HCl (pH 9.0), 50 mM KCl, 2.5 mM MgCl₂, and 0.1% Triton X-100 at 40°C for 20 min. Fifty picomoles of CAT-specific primer (5′-CGGTGAAAACCTGGCCTATTTCCCTAAAGGG-3′) and 1.5 U of Taq polymerase were added to the reverse-transcribed products, which were then amplified by PCR (30 s at 94°C, 30 s at 65°C, and 2 min at 72°C) for 33 cycles. PCR products were analyzed by electrophoresis on 1.2% agarose gels in TAE buffer (40 mM Tris base, 20 mM acetic acid, 1 mM EDTA) and visualized by ethidium bromide staining. To observe the cRNA synthesized from the 717-nt RNA templates, 5-μl transcription reaction mixtures were set up to contain 10 μCi of [α-³²P]CTP (800 Ci/mmol) and 50 μM CTP. The cRNA products were analyzed by 4% polyacrylamide gel electrophoresis with 8 M urea. The gels were dried, and the cRNA products were examined by PhosphorImager analysis.

RESULTS

Effect of point mutations of the 5′ vRNA conserved sequence on polyadenylation.

In this study, we investigated the role of individual nucleotides within the 5′ conserved sequence of influenza virus vRNA in polyadenylation by constructing point mutations. Polyadenylated products derived from mutated vRNA-like templates transcribed in vitro were detected with two previously described assays (23). Long (717-nt) and short (49-mer) vRNA-like templates mutated at positions 1′ to 13′ of the 5′ vRNA conserved sequence were made (Fig. 1). The point mutations chosen were the same as those previously studied to assess RNA polymerase binding to the conserved 5′ region of vRNA by photochemical cross-linking (4) and to determine the role of individual 5′ nucleotides in transcription initiation (5). This latter study used 3′ and 5′ conserved sequence RNAs which had complementary mutations introduced at positions 10 and 11′, respectively. Mutation at position 10 of an added 3′ arm abolishes activity in the in vitro transcription assay, presumably by preventing interaction with the polymerase, which is usually mediated by the associated endogenous 5′ arm. The activity can be restored by adding an exogenous 5′ arm carrying a complementary mutation at position 11′. So, the inclusion of added 3′ and 5′ arms with position 10 and 11′ complementary mutations makes transcription dependent on the addition of the exogenous 5′ arm rather than the endogenous 5′ arms present in the RNA polymerase preparations. The endogenous arms allow the RNA polymerase to transcribe any added RNA provided it carries the 3′ conserved sequence, irrespective of the sequence present at the 5′ end. Therefore, in the present study, transcription initiation occurred for all added templates (except for the 49-mer 2′ [G→U] mutant; see below) because all templates had a wild-type 3′ end. However, the various mutant vRNAs differed in their ability to synthesize polyadenylated transcripts.

RT-PCR polyadenylation assay.

Figure 2, lane 1, shows the detection of polyadenylated RNA transcribed from the 717-nt vRNA-like template carrying wild-type sequences at the 3′ and 5′ termini in the RT-PCR assay (see Materials and Methods). The characteristic broad band contained cDNAs of heterogeneous lengths depending on where priming occurred in the poly(A) tail during reverse transcription with a 5′ GC-clamped T₂₀ primer. The broad band was derived from mRNA and has been rigorously characterized elsewhere (23). The results for templates mutated at positions 1′ to 13′ are shown in Fig. 2, lanes 2 to 14. Polyadenylated products were detected for mutants with mutations at positions 1′ (A→U), 4′ (A→U), 5′ (G→U), 6′ (A→U), and 10′ (A→U). Both 8′ (A→U) and 9′ (C→A) mutants showed a faint sharp band at the position consistent with mispriming on the run of six A residues present in cRNA at the polyadenylation junction. This faint band was considered a negative result because the characteristic broad band was absent. All other mutants (2′, 3′, 7′, and 11′ to 13′) were negative in this assay (estimated as having <10% wild-type activity). All the mutant templates were templates for the synthesis of cRNA, as determined by transcription in the presence of [α-³²P]CTP followed by denaturing polyacrylamide gel electrophoresis and autoradiography (data not shown). These results indicate that mutations at positions 2′, 3′, 7′, 8′, 9′, 11′, 12′, and 13′ of the 717-nt vRNA-like template all drastically affected polyadenylation, while mutations at positions 1′, 4′, 5′, 6′ and 10′ could be tolerated without a dramatic effect on polyadenylation.

FIG. 2.

Effect of point mutations in the conserved sequence at the 5′ end of vRNA in the RT-PCR assay. The assay was carried out with a 5′ GC-clamped T₂₀ primer and the 717-nt vRNA. The RT-PCR assay was performed with consistent results in at least three independent influenza virus transcription reactions for each mutant. Lane 1, wild-type (WT) 717-nt vRNA; lanes 2 to 14, points mutants; lane 15, no template; lane 16, DNA size markers.

[α-³²P]ATP incorporation polyadenylation assay.

In the [α-³²P]ATP incorporation assay, both polyadenylated mRNA and cRNA products were transcribed from the wild-type 49-mer vRNA-like template (Fig. 3, lane 1). The transcription products typically ran as a high-molecular-weight mRNA smear and a major cRNA product (Fig. 3). The authenticity of these products was demonstrated previously (23). The cRNA product typically ran as two bands (lower band not shown), and we now believe that the main, upper cRNA band may be a complex formed between either the cRNA product and the template vRNA or between two cRNA molecules. Such duplexes are surprisingly stable in 16% polyacrylamide–7 M urea gels (29). When the RNA template was omitted from the influenza virus transcription reaction, no product was observed (Fig. 3, lane 15).

FIG. 3.

Effect of point mutations in the conserved sequence at the 5′ end of vRNA in the [α-³²P]ATP incorporation assay. Except for the 2′ (G→U) vRNA mutant (lane 3), all vRNA templates were present in excess in the influenza virus transcription reactions. mRNA, cRNA, and the origin are indicated. The signal at the origin is thought to be due to nonspecific binding of radiolabel and/or transcription product from residual endogenous vRNA. Lanes 1 and 16, wild-type (WT) 49-mer; lanes 2 to 14, point mutants; lane 15, no template; lane 17, 2′ (G→C) mutant.

Except for the position 2′ (G→U) mutant, all mutants were templates for cRNA synthesis in the in vitro transcription assay (Fig. 3, lanes 2 to 14). The amount of 2′ (G→U) mutant template vRNA used in the reaction shown in Fig. 3, lane 3, was approximately 10-fold lower than that used for the other mutant RNAs because of the difficulty in synthesizing RNA with uridine at position 2 by use of T7 RNA polymerase. This problem occurred only in the [α-³²P]ATP incorporation assay, which requires template concentrations severalfold higher than those used for the RT-PCR detection of mRNA. The absence of a cRNA signal for the 2′ (G→U) mutant was therefore due to the absence of sufficient template vRNA. Instead, we tested an alternative mutant, 2′ (G→C), which was synthesized more efficiently by T7 RNA polymerase. This alternative mutant was a template for cRNA synthesis in the [α-³²P]ATP incorporation assay (Fig. 3, lane 17). The 2′ (G→C) mutant (717 nt long) was also synthesized and tested in the RT-PCR assay, in which it also failed to produce polyadenylated RNA under conditions in which cRNA was made (data not shown).

To quantify the polyadenylation activities of these mutants relative to that of the wild type, the polyadenylated products were analyzed by PhosphorImager analysis (Fig. 4). For polyadenylated mRNA synthesis, mutants 1′ (A→U) and 10′ (A→U) had mRNA levels equivalent to those of the wild-type construct. Mutants 4′ (A→U), 5′ (G→U), and 6′ (A→U) each made polyadenylated products but with a reduced efficiency, about half that of the wild-type template. In contrast, the polyadenylation activities of mutants 2′ (G→C), 3′ (U→A), 7′ (A→U), 8′ (A→U), 9′ (C→A), 11′ (A→U), 12′ (G→U), and 13′ (G→U) were severely affected. The level of mRNA production by these mutants was <20% the wild-type level.

FIG. 4.

Quantitation of the effect of point mutations (1′ to 13′) in the 5′ end of the 49-mer vRNA-like template on polyadenylation activity assayed by the [α-³²P]ATP incorporation assay. The mRNA products (high-molecular-weight smear; Fig. 3) from different vRNA templates were quantified by PhosphorImager analysis. Polyadenylation activities relative to that of the wild-type (WT) 49-mer (100%) are shown. The average and standard deviation for the mutants were derived from three or more independent assays. The point mutants used in the experiments are indicated. The activities at positions 4′, 5′, and 6′ were each statistically different from that of the wild type (P, <0.01).

DISCUSSION

Previously we demonstrated that polyadenylated products could be synthesized in vitro in ApG-primed influenza virus transcription reactions (23). Results obtained with two vRNA-like point mutations (positions 3′ and 4′) in that investigation (23) support the hypothesis that polymerase binding to 5′ sequences of the template is required for mRNA synthesis (4, 30). Such binding was obligatory for polyadenylation even with templates with wild-type complementary sequences (10 to 15 and 11′ to 16′) and a correctly positioned U₆ track, which were transcribed to make cRNA (23). Moreover, extending the length of the base-paired region was shown to destroy gene expression in an in vivo study even when wild-type conserved sequences were retained (14). Therefore, the duplex region is not responsible for blocking the progression of the polymerase and causing polyadenylation.

In the present study, we aimed to study the possible role of residues in the conserved 5′ end of vRNA in polyadenylation. Specifically, we wished to distinguish between two alternative mechanisms by which a 5′-bound polymerase might cause polyadenylation. In one model, a complex composed of the polymerase proteins and the conserved 5′ arm of the vRNA template could act in cis to transcribe the 3′ end of the same vRNA template. Later, because of steric hindrance, the polymerase would be unable to transcribe its binding site, leading to polyadenylation by stuttering on the U_5–7 track (Fig. 5A) (4). Alternatively, the 5′-bound complex could merely act as a block to a second, trans-acting polymerase (i.e., a transcribing polymerase not attached to the 5′ arm of the active template), which would be unable to displace the 5′-bound polymerase, resulting in stuttering on the U_5–7 track (Fig. 5B). If polyadenylation were carried out by a trans-acting polymerase, then the ability of the 5′ end of the vRNA to bind polymerase would alone determine whether a given template would produce polyadenylated transcripts. Conversely, if polyadenylation were to require a cis-acting polymerase, then both the ability of the 5′ end of the vRNA to bind polymerase and the ability of the 5′ end of the vRNA to interact with the 3′ end of the same template would be required.

FIG. 5.

Two possible models for the mechanism by which a 5′-bound polymerase causes polyadenylation of influenza virus mRNA. (A) Transcription is initiated by a 5′-bound, cis-acting polymerase (P). Throughout transcription, the polymerase remains attached to the 5′ end of the template. As a result, the polymerase is unable to transcribe the site to which it is attached. Instead, polyadenylation of mRNA occurs by reiterative copying of the U_5–7 track. (B) Transcription is initiated by a trans-acting polymerase (P₁), which is bound to the 5′ end of a different vRNA template. When transcription is blocked by a 5′-bound polymerase (P₂), the trans-acting polymerase (P₁) starts polyadenylation by reiterative copying of the U_5–7 track.

In both the RT-PCR and the [α-³²P]ATP incorporation assays, polyadenylated products from templates with mutations at positions 1′, 4′, 5′, 6′, and 10′ could be detected (Fig. 2 and 4). In contrast, long (717-nt) vRNA templates with mutations at positions 2′, 3′, 7′ to 9′, and 11′ to 13′ could not make detectable levels of mRNA in the RT-PCR assay (Fig. 2) under conditions in which cRNA was synthesized. In good agreement with the RT-PCR assay, the [α-³²P]ATP incorporation assay showed that the polyadenylation activities of the short (49-mer) vRNA mutants with mutations at positions 2′, 3′, 7′ to 9′, and 11′ to 13′ were severely affected. Less than 20% the wild-type polyadenylation activity was detected in these mutants while cRNA was still efficiently being synthesized (Fig. 4). When the results of the two assays are taken together, residues at positions 2′, 3′, 7′ to 9′, and 11′ to 13′ are more important than residues at positions 1′, 4′ to 6′, and 10′ for the polyadenylation of mRNA.

Previously (4, 5), we demonstrated that the conserved sequence at the 5′ end of the vRNA is involved in transcription initiation and polymerase binding (Table 1). When we compare those results with the present results, except for position 1′, all the residues which are important for polymerase binding (i.e., 2′, 3′, 7′ to 9′, and 11′) are also essential for polyadenylation activity. The correlation between polymerase binding and polyadenylation activity confirms our initial conclusion for the two 5′ mutants at positions 3′ and 4′ (23) that the RNA polymerase has to bind to the 5′ end of the vRNA template for polyadenylation to occur. It also supports our earlier binding study (4) in which the precise residues involved in polymerase binding differed slightly from those identified in an alternative in vitro assay (30). Mutating these positions disrupts the interaction between the polymerase and the template and prevents polyadenylation. Interestingly, mutations at positions 4′ to 6′ had slightly reduced polyadenylation activity in the [α-³²P]ATP incorporation assay (Fig. 4). This result was not obviously apparent in the RT-PCR assay because this assay is unsuitable for the detection of small quantitative differences. Although mutations at positions 4′ to 6′ are clearly compatible with both polymerase binding and polyadenylation activity (Table 1), there are subtle differences in the pattern and intensity of binding of the individual polymerase proteins among these templates in the cross-linking assay (4). Thus, mutations at positions 4′ to 6′ may have a small effect on polyadenylation by an as-yet-unknown mechanism.

TABLE 1.

Effect of point mutations at the 5′ end of vRNA on polymerase binding, transcription initiation, and polyadenylation

Position	Resulta in the following assay:
	Polymerase binding (4)	Transcription initiation (5)	Polyadenylation (this study)
1′  A→U	−	±	+
2′  G→Ub	−	−	−
3′  U→A	−	−	−
4′  A→U	+	+	+
5′  G→U	+	+	+
6′  A→U	+	+	+
7′  A→U	−	−	−
8′  A→U	−	−	−
9′  C→A	−	−	−
10′ A→U	+	+	+
11′ A→U	−	−	−
12′ G→U	+	ND	−
13′ G→U	+	−	−

^a+, the base change was tolerated; −, the base change was not tolerated; ±, partially tolerated; ND, not determined. 

^bAn RNA template with a G→C mutation was studied in the [α-³²P]ATP incorporation assay. 

The position 1′ mutant showed poor polymerase binding in an earlier polymerase binding study (4) but had wild-type polyadenylation activity here (Table 1). However, one should also note that a template with the same mutation showed 48% transcription initiation activity compared to the wild type (5), suggesting that the polymerase could, in fact, tolerate a position 1′ (A→U) mutation and initiate transcription. Evidently, this degree of interaction is sufficient for polyadenylation.

For position 12′ and 13′ mutants, earlier data suggested that these positions are not crucial for polymerase binding (4). The greatly reduced polyadenylation activity of these mutants here, however, clearly shows that polymerase binding is insufficient, by itself, for polyadenylation. We previously showed that base pairing between positions 11′ to 13′ and 10 to 12 of the 5′ and 3′ vRNA termini, respectively, is involved in transcription initiation (4, 5) (Table 1). Therefore, disruption of the RNA duplex region by mutation of position 12′ or 13′ could prevent a 5′-bound polymerase from interacting with and transcribing the 3′ end of the template to which it is bound. As a result, polyadenylation would not occur. Presumably, mutation of nonconserved bases in the duplex (positions 14′ to 16′ and 13 to 15) would also prevent polyadenylation in the same way. Indeed, such mutations interfere with gene expression in in vivo experiments (15). In addition to base pairing, it is possible that specific sequences in the duplex region are required for polyadenylation. For example, in an in vivo assay, although most alternative base pairs were tolerated, in one instance, when all 5 bp of the duplex were exchanged, gene expression was greatly reduced (15). For the position 11′ mutant, we could not distinguish whether the reduction of polyadenylation activity was due to the disruption of the RNA duplex region or the disruption of the interaction between the RNA polymerase and the template. Nevertheless, results obtained with the position 12′ and 13′ mutants suggested that the 5′ and 3′ ends of vRNA must interact with each other to initiate transcription and lead to polyadenylation. These results support the hypothesis that polyadenylation is caused by cis-acting polymerase.

When we compare the roles of individual residues within the conserved 5′-end sequence of the vRNA in polymerase binding (4), transcription initiation (5), and polyadenylation (this study) (Table 1), we have a clearer picture of how the 5′ end of the vRNA may be involved in mRNA synthesis (Fig. 5A). First, the polymerase has to bind to the 5′ end of the vRNA. Second, the 5′ end of the vRNA template, with bound polymerase, has to interact with the 3′ end of the same template by forming an RNA duplex (residues 10 to 15 pairing with residues 11′ to 16′). This interaction between the segment termini has been implicated in cap primer utilization (6) and transcription initiation (4, 5). After the initiation of transcription, the duplex melts as the polymerase proceeds along the template, remaining bound to its 5′ binding site. Throughout chain elongation, the cis-acting polymerase complex remains bound to the 5′-end sequence. As a result, the polymerase is unable to transcribe through its 5′ binding site, which remains part of the complex. Thus, polyadenylation occurs through reiterative copying of the U_5–7 track by a polymerase restrained by being bound to an adjacent 5′-terminal sequence (Fig. 5A). The requirement in polyadenylation for a complex composed of both 5′ and 3′ termini and the RNA polymerase is also consistent with the fact that polymerase-free RNP is unable to adopt a circular conformation (or panhandle or RNA fork) (12) and the fact that this circular conformation is most abundant when mRNA production is at its highest (9).

In conclusion, our results show that, in addition to providing an RNA polymerase binding signal, the 5′ end of the vRNA template also has to interact with the 3′ end of the same template for polyadenylation to occur. These results support the hypothesis that a 5′-bound, cis-acting polymerase complex is required for influenza virus polyadenylation.