Mutational Analyses of Packaging Signals in Influenza Virus PA, PB1, and PB2 Genomic RNA Segments

Abstract

The influenza A virus genome consists of eight negative-sense RNA segments that must each be packaged to produce an infectious virion. We have previously mapped the minimal cis-acting regions necessary for efficient packaging of the PA, PB1, and PB2 segments, which encode the three protein subunits of the viral RNA polymerase. The packaging signals in each of these RNAs lie within two separate regions at the 3′ and 5′ termini, each encompassing the untranslated region and extending up to 80 bases into the adjacent coding sequence. In this study, we introduced scanning mutations across the coding regions in each of these RNA segments in order to finely define the packaging signals. We found that mutations producing the most severe defects were confined to a few discrete 5′ sites in the PA or PB1 coding regions but extended across the entire (80-base) 5′ coding region of PB2. In sequence comparisons among more than 580 influenza A strains from diverse hosts, these highly deleterious mutations were each found to affect one or more conserved bases, though they did not all lie within the most broadly conserved portions of the regions that we interrogated. We have introduced silent and conserved mutations to the critical packaging sites, which did not affect protein function but impaired viral replication at levels roughly similar to those of their defects in RNA packaging. Interestingly, certain mutations showed strong tendencies to revert to wild-type sequences, which implies that these putative packaging signals are critical for the influenza life cycle.

Influenza A virus is a negative single-stranded RNA virus whose genome consists of eight segments. Each viral RNA (vRNA) segment carries the coding sequence for one or two proteins in negative-sense orientation, flanked by 19- to 58-base untranslated regions (UTRs) at the 3′ and 5′ ends (12). The UTRs include highly conserved sequences of 12 and 13 nucleotides (nt) at the extreme 3′ and 5′ termini, respectively, that are partially complementary to each other and are believed to form a bulged duplex “corkscrew” structure that is essential for vRNA replication and transcription (1, 3, 5, 6, 11, 14, 17, 21). During viral assembly, the vRNAs are packaged in the form of viral ribonucleoprotein complexes. Recently, Noda et al. (19) used electron microscopy to examine serial sections through influenza virions and observed a well-organized internal structure that appeared to comprise one central viral ribonucleoprotein surrounded by seven others in an orderly, parallel bundle. How these eight vRNAs are assembled and incorporated into viral particles is still not well understood.

The preferential incorporation of influenza vRNAs (but not of their replicative intermediates, viral mRNAs, or host RNAs) into virions implies a highly selective mechanism of packaging. Indirect evidence suggests, moreover, that the packaging mechanism may discriminate individually among the eight vRNAs. For example, individual vRNAs may be expressed at different concentrations in infected cells but are roughly equimolar within viral particles (24); additionally, a defective interfering vRNA with partial deletions in its coding region was found to compete preferentially for packaging with its wild-type counterpart but not with other segments (4, 20).

Viral genomic packaging generally depends on the recognition of distinctive cis-acting packaging signals within the RNA. Recently, the cis packaging elements for influenza NA, HA, M, NS, PA, PB1, and PB2 segments have been found to reside at both ends of each vRNA, including the UTR, along with up to 80 bases of adjacent coding sequences on any given segment (7, 8, 15, 16, 25). In keeping with their biological function, these regions near the two segment termini show low nucleotide sequence variation among strains of influenza A (9). We have recently shown (15) that these bipartite packaging signals are noninterchangeable among segments, suggesting that those in any given segment form a uniquely interdependent pair that must be presented in concert to the viral packaging machinery. This may indicate that specific intrastrand physical and/or functional interactions between these ends, perhaps including but not limited to the bulged duplex structure formed by the conserved UTRs, may play important roles in packaging (15). Exactly how these cis-acting signals interact with each other and with other viral components to affect packaging is not yet understood.

We and others have developed assays to quantify the relative packaging efficiencies of individual influenza vRNA-like reporters that encode green fluorescent protein (GFP) (15). Using such an assay, we previously mapped the minimal regions that can support efficient packaging of the PA, PB1, and PB2 segments (15), each of which encodes one of the three protein subunits of the vRNA-dependent RNA polymerase. To further characterize the packaging signals, we have now introduced scanning mutations and small deletions across these active regions and examined their effects on packaging using our reporter assay. In general, we find that the 5′ regions are less tolerant of mutation than are the 3′ regions and that the 5′ region of the PB2 vRNA is more sensitive to mutation than those of the PA or PB1 vRNAs. We have extended these findings by introducing conservative or silent point mutations into certain critical residues and then testing their effects on viral growth kinetics. Mutations found to impair packaging also inhibited viral replication and, in certain cases, showed a strong propensity for reversion to the wild-type sequence. These findings help define more precisely the functional packaging signals within each segment and verify their critical role in influenza virus replication.

MATERIALS AND METHODS

Cells.

293T (human epithelial kidney) cells were grown in Dulbecco's modified Eagle's medium (with a high glucose concentration) supplemented with 10% heat-inactivated fetal bovine serum. Madin-Darby canine kidney (MDCK) and Madin-Darby bovine kidney (MDBK) cells were maintained in Eagle's minimal essential medium supplemented with 5% fetal bovine serum. After infection with either influenza virus or virus-like particles, MDCK and MDBK cells were grown in L-15 medium containing 15 mM HEPES, pH 7.5, nonessential amino acids, 0.75 g of NaHCO₃ per liter, and 0.125% (wt/vol) of bovine serum albumin.

Plasmids.

The 17-plasmid (18) and the 8-plasmid (10) influenza A virus reverse-genetic systems were obtained from Y. Kawaoka (University of Wisconsin) and G. Hobom (Justus Liebig University, Germany), respectively. PA 66-G-50, PB1 66-G-50, and PB2 0-G-100 are the parental GFP reporter constructs of PA, PB1, and PB2, respectively, and have been described previously (15). Mutations and deletions were created by PCR-based mutagenesis using primers containing the corresponding changes. The primer sequences will be provided upon request. All mutations have been confirmed by sequencing.

Generation of influenza virus-like particles.

293T cells in six-well plates were transfected with the eight plasmids that are required to generate infectious influenza viruses (10), together with 1 ug of GFP reporter construct as described above, using TransIT-LT1 transfectin reagent (Panvera, Madison, WI) according to the manufacturer's protocol and as described before (15).

Determination of relative packaging efficiency.

Aliquots of viral supernatants from 293T cell transfection were used to infect MDBK or MDCK cells in six-well plates at 37°C for 1 h. After replacement of the infection medium with fresh L-15 medium, cells were further incubated at 37°C for 15 h and harvested for immunostaining. Cells were fixed in 4% paraformaldehyde, permeabilized in 0.2% saponin, and incubated first with mouse anti-influenza A NP antibody (Serotec) and then with a secondary R-phycoerythrin-conjugated anti-mouse immunoglobulin G polyclonal antiserum (BD Pharmingen). After extensive washing with phosphate-buffered saline, cells were subjected to flow cytometric analysis to quantify expression of both GFP (green) and NP (red) proteins. The percentage of NP-positive cells that were also GFP positive was used as a measure of the packaging efficiency of each mutant reporter vRNA and was generally expressed as a percentage of the value obtained for the corresponding wild-type reporter.

Quantitative real-time RT-PCR.

As described previously (15), total RNA was extracted from cells by using RNAbee reagent (TEL-TEST, Inc.) according to the manufacturer's instructions. The total RNA was cleared of possible plasmid DNA contamination by incubating samples for 30 min at 37°C with DNase I, which was then inactivated by heating samples to 85°C for 15 min. Reverse transcription (RT) was conducted with strand-specific primers for NP (5′ CAGGATGTGCTCACTGATGC 3′) and GFP (5′ CAGAAGAACGGCATCAAGCG 3′), using the SuperScript III first-strand synthesis system according to the manufacturer's protocol (Invitrogen). The quantitative real-time PCR was carried out with a 20-μl reaction mixture with gene-specific primers for NP (5′ CAGGATGTGCTCACTGATGC 3′ and 5′ TTCTCCGTCCATTCTCACCC 3′) or for GFP (5′ GCTGACCCTGAAGTTCATCT 3′ and 5′ GGACTTGAAGAAGTCGTGCT 3′), using SYBR green DNA dye (Invitrogen). The PCR conditions were 50°C for 2 min, 95°C for 2 min, and 45 cycles of 95°C for 15 s, 55°C for 30 s, and 72°C for 30 s. The plasmids pPolI-NP and pPolI-PA(474-G-239) were used as standards for NP and GFP genes, respectively.

Determination of vRNA synthetic activity.

The level of vRNA synthesis was quantified by real-time RT-PCR in a five-plasmid expression system. 293T cells were transfected with five plasmids, including four protein-coding plasmids (PB2, PB1, PA, and NP) and one NA vRNA-coding plasmid, each at a 1 μg concentration. An enhanced-GFP (EGFP)-coding plasmid at 0.1 μg is also included as an internal control. The quantitative real-time RT-PCR was conducted essentially as described above, with the differences in the primers used for RT and real-time PCR. We used strand- and sense-specific oligonucleotides in the RT reaction to detect influenza vRNA (5′ AGCGAAAGCAGG 3′ and 5′ AGCAAAAGCAGG 3′), cRNA (5′ AGTAGAAACAAGG 3′), or mRNA [oligo(dT)]. A GFP-specific primer (5′ GGACTTGAAGAAGTCGTGCT 3′) was also included in the RT reaction for vRNA or cRNA analysis. The gene-specific primers for the real-time PCR were 5′ TTTGGGATCCTAATGGATGG 3′ and 5′ GTTGAACGAAACTTCCGCTG 3′ for NA or 5′ GCTGACCCTGAAGTTCATCT 3′ and 5′ GGACTTGAAGAAGTCGTGCT 3′ for GFP. The NA vRNA levels, shown as threshold cycle values, were normalized by GFP RNA level.

To compare polymerase function by use of a luciferase (LUC) assay in the five-plasmid transfection system, cells were transfected with 0.1 μg of the LUC reporter construct (vNA-LUC or cNA-LUC) (23) and 0.25 μg of each of the protein expression vectors encoding PB2, PB1, PA, and NP. A β-galactosidase (β-Gal) expression plasmid was included as an internal control for normalizing transfection efficiency. At the indicated times, cells were harvested for a LUC assay and normalized to β-Gal activity.

RT-PCR and sequencing.

vRNA was extracted using a Qiagen vRNA kit according to the manufacturer's protocol. The RT reaction was conducted using an influenza-specific primer and the SuperScript III first-strand synthesis system (Invitrogen). The target fragments were amplified by PCRs with specific primers, purified from agarose gel, and sequenced. All primer sequences will be provided upon request.

Sequence alignments.

Full-length sequences of PA and PB1 segments from diverse influenza A isolates, comprising all available serotypes from mammalian or avian hosts, were obtained from the NIH NCBI influenza virus resource database (http://www.ncbi.nlm.nih.gov/genomes/FLU/Database/select.cgi?go=1) and were aligned using the MUSCLE (multiple sequence comparison by log expectation) algorithm associated with that resource. Percentage of variation was calculated for each nucleotide position within relevant portions of 5′ sequences.

RESULTS

We have previously shown that all of the cis-acting signals required for efficient packaging of PA, PB1, and PB2 segments map within approximately 100 bases of the 3′ and 5′ termini of each segment, encompassing not only the 3′ and 5′ UTRs but also various lengths of adjacent coding sequences at each end (Fig. 1A). In particular, optimal packaging of PA or PB1 requires 40- and 66-nt coding sequences at the 3′ and 5′ ends of the vRNAs, respectively, whereas the PB2 segment is unique in requiring 80 nt of coding sequence at the 5′ end but none at the 3′ end (Fig. 1A) (15). The specific features within these terminal regions that give rise to their packaging activities remain unknown. To address that question, we carried out higher-resolution mutational mapping of the packaging signals in the PA, PB1, and PB2 segments by introducing a series of clustered point mutations and small deletions (normally within UTRs) that together scan across the regions known to be required for efficient packaging. These changes were introduced into the parental reporter vRNA constructs PA 66-G-50, PB1 66-G-50, and PB2 0-G-100. The reporters were named by indicating the identity of the parental vRNA, followed by the number of bases of viral coding sequence at the 3′ end, the letter G (denoting GFP), and the number of bases of viral coding sequence at the 5′ end. These constructs have previously been shown to be efficiently packaged (i.e., at >76% relative packaging efficiency) (15). Because the distal 12 or 13 bases at the extreme termini are critical for many aspects of viral gene expression, we did not target those residues for mutagenesis but focused instead on the remainder of each UTR as well as on the adjacent coding sequences. The substitution mutations were each designed to replace five to eight consecutive bases with their Watson-Crick complements (e.g., G to C, A to U, and vice versa). Each resulting mutant vector was transiently transfected along with an eight-plasmid reverse-genetic system into 293T cells. During subsequent culture, reporter vRNAs were replicated and packaged into the released viral particles. Viral supernatants were harvested at 48 h posttransfection and then used to infect fresh MDBK cells at a multiplicity of infection (MOI) of 1.6 to ensure that excess helper viruses were present, if necessary, to replicate and transcribe the reporter. The relative packaging efficiencies of individual vRNAs were then calculated as the percentages of NP-positive target cells that were also GFP positive, expressed as percentages of the efficiencies observed for the corresponding wild-type reporters. Since each reporter vRNA is tested in the presence of the respective wild-type vRNA segment, the packaging efficiency that we determined here illustrates the ability of a mutant vRNA to compete with wild-type RNA for virion incorporation.

FIG. 1.

Mutational analyses of packaging regions in the PA, PB1, and PB2 segments. (A) Minimal sequences required in cis for the efficient packaging of PA, PB1, and PB2 segments, as determined in an earlier study (15). These regions encompass the UTRs at both ends (depicted as boxes, with dark gray shading to demarcate the terminal 12- or 13-base sequences common to all segments) as well as 0 to 80 bases of adjacent coding sequences (shown as a thick gray line). The lengths of various sequences (in nucleotides) are noted above each segment. The internal viral coding sequences found to be unnecessary for packaging are depicted as dashed lines. (B) Packaging efficiencies of mutant forms of the PA, PB1, and PB2 reporters. The structures of these mutants are detailed in Fig. 1C. Packaging efficiencies were determined using an assay described previously (15) and are expressed relative to that of the corresponding wild-type reporter. (C) Functional architecture of packaging regions from the PA, PB1, and PB2 segments. The complete nucleotide sequences of the minimal 5′ and 3′ packaging regions contained in each of the three wild-type reporters are shown, with the coding sequences highlighted in yellow and the interposed GFP cassette in green. Bases targeted by each of the point mutations (m) or internal deletions (d) are indicated by brackets above or below the wild-type sequence, respectively. Point mutations replaced each G with C or A with U or vice versa. The relative packaging efficiency of each mutation, expressed as the mean for 3 independent determinations, is indicated within the bracket. Mutations that reduced packaging efficiency to 2% of the wild-type level or below are indicated in red.

By introducing up to 10 such mutations into each active terminus, we interrogated roughly 75 to 100 individual residues in each of the three segments. Figure 1B summarizes our results by showing the relative packaging activity of each mutant as measured in our single-round transduction assay. Sequences of the characterized regions and the localization of each mutation are shown in Fig. 1C. In our previous deletional mapping analysis (15), PA reporter vRNA containing as few as 15 nt of 3′ coding sequence still retained about 50% relative packaging efficiency, suggesting that these sequences play a less significant role in packaging. We therefore did not create mutations to cover the entire sequences (a few nucleotide gaps in between the mutations). Consistently, all the mutations within the region had no more than a moderate effect, reducing packaging efficiency to 43 to 85% of the wild-type level. Therefore, few, if any, of the 3′ coding bases of PA RNA appear to be specifically required for packaging. In contrast, most mutations within the 5′ PA coding sequence reduced packaging to below 30% and in some cases much lower. For example, the m3, m5, and m15 mutants had packaging efficiencies of 4%, 0.3%, and 2%, respectively. We further confirmed that, in addition to the coding sequences, specific residues in the UTRs of PA vRNA are also important for packaging. Thus, mutation m13 in the short 3′ UTR of PA resulted in a significant decrease in packaging (to 6% of the wild-type level), and mutations in the longer 5′ UTR (e.g., m17, m18, or m20) were even more poorly packaged (0.4 to 2%).

Similarly, none of the 3′-end mutations in the PB1 segment reduced packaging efficiency to below 2% of the wild-type level, whereas five of the nine mutations in the 5′ end did so (Fig. 1B). In our previous deletional mapping, a PB1 reporter vRNA with only 33 nt of coding sequences at its 3′ end retained about 55% relative packaging efficiency, whereas a reporter with 15 nt showed markedly deficient packaging (10%) (15). This reduction in packaging is most likely attributable to sequences around the site of m10, since m10 alone decreased packaging to ∼10%, whereas all the other mutations across the 66-nt 3′ coding sequence retained at least 30% relative packaging. Scanning mutations across the 5′ 40-nt PB1 coding sequences, on the other hand, implicated m3, m12, m13, and m14 sites as critical packaging sequences, in that each caused a profound loss of packaging (to <2% of the wild-type level). In addition, the UTRs also proved important for PB1 packaging, as mutation of the 3′ UTR (m11) or deletion within the 5′ UTR (d2) yielded only 8% or 0.6% relative packaging efficiency, respectively.

The PB2 segment is unique in that its packaging does not require 3′ coding sequences. The parental PB2 reporter vRNA used in this study contains 80 bases of 5′ coding sequence but no 3′ coding sequence, yet it maintains about 75 to 80% relative packaging efficiency (15). Hence, our analysis focused mainly on the 5′ 80-nt coding sequence that supports efficient PB2 packaging. In contrast to our findings in the PA and PB1 segments, any perturbation of this region (m3, m4, m5, m6, d1, m7, m8, d3, or d4) substantially decreased packaging efficiency (to 0.4 to 2%) (Fig. 1B), suggesting that this entire 80-nt coding sequence may be critical for PB2 packaging. Mutations m1 and m2 within the 3′ UTR also decreased PB2 packaging, to 7% and 3%, respectively.

In Fig. 1C, mutations that reduced packaging by 2 logs or more are highlighted in red. Notably, all of the most deleterious mutations mapped to the 5′ ends of their respective segments, which accords with our earlier finding that the 5′ regions are quantitatively more important than the 3′ regions for packaging (15). To exclude the possibility that the decreased RNA packaging levels that we observed were due to less efficient transcription or replication of the mutant reporters, we quantified the respective vRNA levels in the transfected 293T cells by real-time RT-PCR. All reporter vRNAs were found to be expressed at comparable levels (data not shown). Thus far, we have not identified any features of primary sequence, or any potential secondary structures, that might account for the uniquely deleterious effects of this subset of mutations. Nonetheless, our data identify a distinct subset of sequences within the bipartite packaging regions of PA, PB1, and PB2 that may have particular relevance to the packaging process.

To assess whether the identified packaging signals are evolutionarily conserved, we aligned the full-length sequences of PA and PB1 segments from more than 580 influenza A viruses (representing all serotypes from different hosts) archived in the NCBI influenza virus resource database. Illustrative results for selected subregions of the PA and PB1 vRNA 5′ termini, shown in Fig. 2, indicate the extent of interstrain sequence variation observed at each base position in the alignment. As expected, the terminal sequences of all three segments are in general highly conserved, indicating that our present findings for the WSN strain may also be applicable to other strains. Each of the most deleterious mutations that we tested (i.e., those that reduced packaging by 2 logs or more, indicated in red) targeted at least one highly conserved base, which could hypothetically be responsible for the observed effects on packaging. It should be noted, however, that not all of our most deleterious mutations mapped within the subregions that appeared most tightly conserved, implying that the strict conservation of many bases in the regions that we interrogated reflects selective forces unrelated to packaging.

FIG. 2.

Interstrain nucleotide variation within portions of the PA and PB1 packaging regions. Full-length sequences of PA and PB1 segments from diverse influenza A virus isolates (including all subtypes from various origins) were obtained from the NCBI influenza virus resource and aligned. Portions of the 5′ packaging regions from the PA vRNA (residues 2134 to 2183) and PB1 vRNA (residues 2255 to 2304) of strain WSN are shown, and the percentage of variation observed at each position in the alignment is indicated above, based on 609 PA sequences and 586 PB1 sequences. The locations of mutations tested in this study are indicated below, with those that reduced packaging efficiency to 2% or less indicated in red.

Having found that mutation of the m5 site of the PA vRNA produced a dramatic effect on reporter packaging, we subjected this 5-base sequence to a more detailed mutational analysis (Fig. 3). The m5 mutation coincides with a pair of codons specifying histidine and alanine. (The vRNA sequence depicted, 3′ GUACGU 5′, is in antisense orientation, with the m5 site underlined.) The only possible silent mutation of this dicodon was a G-to-A substitution at the third residue; we found that this mutation (designated m5-1) did not affect packaging in the reporter assay (Fig. 3A). We therefore tested a series of different mutations that resulted in conservative amino acid changes and found that they had various effects on reporter vRNA packaging (Fig. 3A). Two mutants from the latter group (m5-2 and m5-4), with moderate and severe defects in packaging, respectively, were chosen for further characterization. To exclude the possibility that the resultant amino acid changes significantly impaired functions of the PA protein and hence of the viral polymerase (which mediates vRNA replication and transcription), we tested the abilities of these PA mutants to support polymerase enzymatic activities in cells. For that purpose, 293T cells were transiently transfected with a plasmid encoding the NA-based vRNA reporter, along with the wild-type or mutant PA expression vector and three other plasmids encoding the PB1, PB2, and NP proteins. These four proteins (PA, PB1, PB2, and NP) ordinarily suffice to allow synthesis of mRNA, vRNA, and a replicative intermediate (called cRNA) from the reporter template, and those three RNA products can then be quantified individually using quantitative real-time RT-PCR as previously described (15). As shown in Fig. 3B, transfections using the m5-2 or m5-4 mutant form of PA yielded all three RNA products in quantities comparable to those produced by wild-type PA protein. Since this quantitative RT-PCR was conducted at 48 h posttransfection, when the levels of each RNA may have reached saturation, we have also conducted a LUC-based flu RNA synthesis assay to compare the PA polymerase activities at different time points. Using an RNA construct in which the LUC gene is expressed under the promoter of either the viral NA RNA segment (vNA-LUC) (Fig. 3C) or complementary NA RNA (cNA-LUC) (Fig. 3D), the m5-2 and m5-4 PA mutant proteins produced levels of LUC activity comparable to those for the wild-type PA protein at 16, 24, and 32 h posttransfection, while a control reaction lacking PA protein generated the background level of LUC activity. Taken together, these data suggest that the m5-2 and m5-4 mutations do not appreciably perturb the polymerase activity of the PA protein. When we attempted to generate infectious viruses by substituting these mutants for wild-type PA in the 17-plasmid reverse-genetic system, we found that both mutants yielded lower virus titers in the supernatant over a 48-h time course than did wild-type PA (Fig. 3E). Titers for m5-4, moreover, were consistently lower than those for m5-2, indicating a correlation between the severities of the defects in packaging activity and in replicative kinetics.

FIG. 3.

Conservative mutations at the PA m5 site. (A) Various nucleotide changes (boldface and underlined) were introduced in or near the site of the m5 mutation in the PA vRNA segment. The nucleotide sequence, the resultant amino acid sequence of PA protein, and the relative packaging efficiency as determined using the reporter assay are shown for the wild-type and each mutant (m5-1 to m5-5) in vRNA form. (B) RNA synthetic activities of influenza polymerases containing the wild-type (wt), m5-2, or m5-4 forms of the PA protein. 293T cells were transfected transiently with plasmids encoding PB2, PB1, NP, and wild-type or mutant PA proteins; an NA vRNA-encoding plasmid; and an EGFP-encoding plasmid as an internal transfection control. Total RNA was extracted at 48 h posttransfection and subjected to quantitative RT-PCR. The levels of the three alternative NA-specific RNA species (vRNA, cRNA, and mRNA) were quantified individually by real-time RT-PCR and normalized against the internal control EGFP RNA level, and the resulting threshold cycle [C(t)] values are shown here. (C) Influenza vRNA assay based on vNA-LUC reporter RNA. 293T cells were transfected with different PA expression plasmids (wild type, m5-2, or m5-4) or empty vector (-PA), together with PB1, PB2, or NP protein expression vectors and the vNA-LUC reporter construct. A β-Gal expression plasmid was included as an internal control for normalizing transfection efficiency. LUC activity was measured at different time points posttransfection and normalized against β-Gal activity. (D) Influenza vRNA assay based on cNA-LUC reporter RNA. 293T cells were transfected with different PA expression plasmid (wild-type, m5-2, or m5-4) or empty vector, together with PB1, PB2, or NP protein expression vectors and the cNA-LUC reporter construct. A β-Gal expression plasmid was included as an internal control for normalizing transfection efficiency. LUC activity was measured at different time points posttransfection and normalized against β-Gal activity. (E) Kinetics of infectious-particle production. Influenza virus particles containing the wild-type, m5-2, and m5-4 forms of PA were reconstituted using the 17-plasmid system for transfection of 293T cells as described in Materials and Methods. The quantities of released infectious particles at various time points after transfection were quantified by a plaque assay with MDCK cells. The results shown here are the averages for at least two independent experiments for each time point.

We isolated 10 plaques for each of the two mutants after 48 h of growth in culture, amplified the PA vRNA segment by using RT-PCR, and then subjected the products to sequence analysis. Strikingly, in 5 out of 10 plaques from m5-2, the mutated bases had reverted to the wild-type bases, indicating a strong selective pressure favoring the wild-type sequence at these positions. Because the m5-2 mutant functions comparably to wild-type PA protein in supporting vRNA synthesis, this selective pressure most likely reflects the fitness advantages of maintaining the integrity of the packaging signal.

To extend this analysis still further, we then tested the effects of mutating putative packaging signals in the context of the authentic PA, PB1, or PB2 segment, focusing only on sites where synonymous mutations could be used. As indicated in Table 1, we designed one additional mutant form of each of these vRNAs, in each case altering 5 to 8 bases to produce silent mutations in two to four consecutive codons. These silent mutations (designated PA-m15s, PB1-m12s, and PB2-m6s) roughly corresponded in location to the highly deleterious PA-m15, PB1-m12-m13, and PB2-m6 mutations described above and so targeted sites that our initial reporter-based studies had strongly implicated in packaging. When introduced into their respective reporters, each of these silent mutations reduced packaging to 2% or less of the wild-type efficiency (Table 1). We next introduced these same mutations into vectors encoding the authentic, full-length vRNAs and then used the 17-plasmid transfection system to create viruses carrying each mutant vRNA in place of its wild-type counterpart. When viral titers were determined for the supernatants at 48 h posttransfection, each of the mutants was found to impair virus production, yielding titers 4- to 67-fold lower than the wild-type level (Table 1). We then plaque purified each mutant, confirmed that the expected mutations were present, and used these to infect MDCK cells at an MOI of 0.001. As shown in Fig. 4, all three mutant viruses grew more slowly than the wild type, with PB2-m6s being the most impaired: at 32 h postinfection, both PA-m15s and PB1-m12s produced l-log-lower viral titers, and PB2-m6s produced a 1.5-log-lower titer, than the wild-type level. Taken together, these data demonstrate that the cis-acting sequences that we have identified do indeed affect viral growth, most likely by being part of the packaging signals that determine the incorporation of their respective genomic RNA segments into virion particles.

TABLE 1.

Silent mutations in the coding sequences of PA, PB1, and PB2 segments

Virus	Wild-type sequence	Protein sequence	Mutant sequencea	Relative packaging efficiency (%)b	Virus titer in 293T supernatant (no. of PFU/ml)c
Wild type				>75	2 × 106
PA-m15s	GCATTGAGA	671ALR673	GCTCTCCGT	2	5 × 105
PB1-m12s	GAGCTCAGACGG	752ELRR755	GAATTGCGTAGA	0.4	3 × 104
PB2-m6s	TCTAGC	741SS742	AGCTCG	0.4	2.4 × 105

^aChanged nucleotides are underlined.

^bRelative packaging efficiency was determined as described in Materials and Methods.

^cWild-type and mutant viruses were reconstituted in 293T cells by using the 17-plasmid system, and the infectious particles in the supernatants collected at 48 h posttransfection were quantified by a plaque assay.

FIG. 4.

Growth curve analysis of viruses harboring translationally silent mutations in vRNA packaging signals. Influenza viruses were reconstituted using the 17-plasmid system, either with exclusively wild-type (WT) vRNAs or with one of the three silent mutants shown in Table 1 substituted. MDCK cells were infected with these viruses at an MOI of 0.001. Supernatants were collected at the indicated times following infection, and virus titers were determined by plaque assays. The results, expressed in numbers of PFU per ml, are the averages for triplicate assays, with standard deviations shown. hpi, hours postinfection; •, WT; 218, PB2-m6s; ▴, PA-m15s, □, PB1-m12s.

DISCUSSION

Packaging signals of influenza vRNAs have a bipartite organization, residing within the UTR and adjacent coding sequences (7, 8, 15, 16, 25) at both ends of each vRNA. The vRNA termini are believed to associate physically with each other during certain phases of the viral life cycle, presumably through base pairing of the distal 12 or 13 complementary bases in each UTR (2, 5, 6, 13, 22), and this interaction is clearly required for key steps in vRNA synthesis that are carried out by the viral polymerase. Whether such interterminal contacts are also important for packaging is not yet known, though the observation that the 3′ and 5′ packaging regions of a given vRNA function are functionally interdependent (15) may indicate that this is the case. Previous mutational dissections of the UTRs have localized features that are essential for vRNA transcription and replication, and some molecular structural information has been obtained for their distal termini in duplex form, but the physical and functional topologies of the remainder of each packaging region have not been analyzed in detail, and it is not yet clear how they are recognized by the viral packaging machinery.

In this study, we used scanning mutagenesis to further dissect the packaging regions of the PA, PB1, and PB2 segments. Not surprisingly, our results confirm that the UTR regions are essential for packaging of all three segments. The involvement of the UTRs in many other aspects of viral replication, however, limited their suitability for mutational dissection in our present analysis, which therefore focused mainly on the nearby portions of the coding regions that have been implicated in packaging. We interrogated roughly 60% of the bases within these active coding regions and identified a distinct subset that may contribute specifically to packaging function. These highly deleterious mutations invariably mapped at the 5′ ends of their respective vRNAs, implying that the 3′ coding sequences play a less significant quantitative role in our single-round packaging assay. Within the 5′ coding regions of the PA and PB1 vRNAs, moreover, only about half of the tested coding-region mutations (two of six in PA and four of eight in PB1) produced such marked defects, whereas significant expanses of the remaining coding sequence proved comparatively tolerant of mutation.

This heterogenous pattern of sensitivity may indicate the presence of a few discrete, functional packaging loci within the PA and PB1 coding regions. Such an interpretation must be viewed with caution, however, as it is not yet clear whether any given mutation in our series directly impinges upon a cis-active locus or instead acts indirectly, perhaps by perturbing features of RNA secondary structure that affect other loci nearby. We sought additional evidence of the importance of the mutated bases by aligning the relevant portions of genomes from over 580 influenza viruses from diverse sources and hosts, which revealed that all of the highly deleterious mutations in our series had altered one or more bases that were stringently conserved (Fig. 2). Interestingly, however, such mutations did not all cluster within the most obviously conserved portions of the genome that we evaluated and, conversely, mutations at some highly conserved positions (e.g., the m2 mutations in PA and PB1) caused only modest reductions in packaging. This may highlight the challenges of deducing function from such broad (and admittedly cursory) phylogenetic comparisons, and it offers future opportunities to explore the factors that constrain sequence variation within the packaging regions. In preliminary studies, we have used an RNA structure prediction algorithm (M-Fold) to search for potential conformational features within the terminal regions of all three vRNAs (individually and as intrastrand pairs) that might account for the results of our mutational study but have thus far found none (data not shown). Potential effects on the polymerase subunit proteins encoded by these three vRNAs also cannot be formally excluded in most cases, though we were able to confirm certain key findings using silent mutations (Table 1) and found no evidence that other mutations impaired RNA-synthetic activities of the polymerase (Fig. 3). It remains to be determined whether some of our mutations disrupt hypothetical RNA-protein or interstrand RNA-RNA contracts that might be critical for packaging.

The PB2 segment is unique in requiring no 3′ coding sequences in order to be packaged efficiently; it instead utilizes an unusually long (114-base) 5′-terminal region that extends as much as 80 bases into the 5′ coding sequence. All eight of the 5′ mutations that we tested, which together scanned across this entire 80-base coding region, severely diminished packaging efficiency. This geographically extensive sensitivity to mutation contrasts starkly with the heterogeneous patterns found in the two other vRNAs studied here (Fig. 1C). Moreover, silent mutation in this region of the authentic PB2 vRNA had a somewhat greater effect than that observed with a seemingly comparable silent mutation in the PA or PB1 segment (Fig. 4). Significantly, Muramoto et al. (16) have reported that impaired packaging of the PB2 vRNA leads to defects in packaging other segments, and they have speculated that PB2 may serve as the central element in the presumed multisegment bundle structure visible in the interiors of budding influenza A virus particles by electron microscopy (19). Packaging signals in the PB2 vRNA thus have a distinctive topological organization and other properties that merit further investigation, as they may reflect a unique role for this segment in influenza virus packaging or assembly.

Our results also serve to verify the importance of these putative packaging signals in the context of authentic vRNAs. The initial scanning mutagenesis in this study (Fig. 1) was performed using truncated reporter vRNAs and single-round transduction assays, mirroring approaches used successfully by several laboratories to map influenza virus packaging signals. In subsequent studies, however, we have now introduced comparable mutations into intact, full-length vRNAs, incorporated these as substrates for viral assembly, and monitored the resulting effects on influenza virus replication kinetics in a multicycle infectivity assay. It should be noted that, in our version of the reporter-based assay, each mutant vRNA competes for packaging against its wild-type counterpart, which is expressed at a roughly equivalent concentration in the transfected cell. Because no wild-type competitor is present in our multicycle replication system, however, the magnitudes of the packaging defects produced by any given mutation in these two assays are not directly comparable in absolute terms. Nevertheless, we observed a strong correlation between results for the two assays that clearly supports the biological validity of our findings. These results confirm the applicability of our findings to live influenza A virus and offer proof of the concept that disrupting the packaging signals can decrease influenza virus replication, an approach that might prove applicable to the development of novel antiviral drugs or of attenuated influenza vaccines.