Packaging signal of influenza A virus

The fragmented nature of the IAV genome confers significant evolutionary advantages to the virus by allowing for easy exchange of gene segments when two or more influenza viruses infect the same cell. However, it undeniably complicates virion assembly, as a nascent influenza virus particle must incorporate at least one copy of each vRNA in order to become replication-competent and thus fully infectious. Inagaki et al. found that the two virus-like RNAs bearing the same terminal sequences competed for incorporation into virions [18]. Based on multicolor single-molecule fluorescent in situ hybridization (FISH), Chou et al. studied the composition of viral RNAs at single-virus particle resolution [19]. The results demonstrated that most of the virus particles package a full set of gene segments and only one copy of each RNA segment is packaged per virion [18, 19].

The genome of influenza C and D virus (ICV and IDV) consists of only seven vRNAs. However, most of the ICV and IDV packaged eight RNPs arranged in the “7 + 1” pattern [10]. The influenza A viruses (IAV) that fail to package the full genome and are noninfectious also exist [20]. Experiments of Nakatsu et al. demonstrated that some influenza A and B virion packaged vRNPs less than eight [21]. Su et al. found that destroying the ubiquitination in M2 of IAV resulted in the production of defective virion particles that lacked vRNPs [22]. These results suggested that the mechanism of influenza virus genome-packaging is flexible and the vRNP number that incorporated by influenza virions is variable [21, 23].

Two different models have been proposed for packaging the RNPs into newly assembling virus particles: the random-incorporation model and the selective-incorporation model [24‐26]. The random-incorporation model suggests that each viral RNA segment possesses a common structural feature that allows random incorporation of RNA segments into virions. The selective-incorporation model suggests that unique packaging signals present in vRNAs lead to the incorporation of a set of all eight RNA segments into a virion [13]. Modern day instrumentation as well as biochemical assays and viral assays have provided much evidence to support the selective-incorporation mechanism model of the influenza A genome [9, 26, 27].

For example, Yutaka Fujii et al. generated seven and six segment viruses in the background of A/WSN/33 (H1N1) virus and found that the efficiency of infectious virion production is proportional to the number of different vRNA segments, indicating that vRNA segments contribute individually to virion production [28]. Octaviani et al. coinfected MDCK cells with a swine-origin H1N1 and an avian H5N1 virus. Among 59 viral clones the author examined, there were only 33 different genotypes, much less than 254 in theory. And 15% of the viral clones obtained all of their genes from the H5N1 virus [29]. More recently, Cobbin et al. demonstrated that PB1 gene of A/Udorn/307/1972 (Udorn, H3N2) was preferentially co-packaged with NA gene of the same virus when two NA genes existed [30]. All these data suggested that the genome incorporation of IAV is selective (Fig. 1).

The genome vRNAs of IAV share the same organization: a central open reading frame (in the antisense orientation) flanked at both ends by untranslated regions (UTRs) (19–58 nts) [4, 31, 32]. The UTRs of IAV genes consist of two parts: the motif sequences which are highly conserved among viral strains and among the eight segments themselves, located at the 3′ and 5′ termini of every segment; and the segment specific noncoding regions (ssNCRs), the length and sequences of which are specific to each vRNA and between species [26, 33, 34]. The earlier studies identified that the packaging signal consists of a stretch of noncoding regions and the adjacent coding regions at the 3′ and 5′ ends of each vRNA. Whereas, recent studies found that specific sequences within the internal coding regions also play important roles in the genome packaging of influenza A virus (Fig. 3).

Studies demonstrate that untranslated regions of the IAV genome vRNA are more conserved, having a lower evolutionary rate compared to coding regions [52, 53]. Marsh et al. identified the conserved regions in the polymerase gene segments by analyzing approximately 600 vRNA from avian IAV per segment. They found that the conserved nucleotides are located within the terminal packaging signals and the identified regions also extend to other subtypes of viruses, for example, human IAVs [54]. Gog et al. found that most codons showed very little synonymous variation within terminal packaging sequences among identical segments and among different influenza viruses [55]. But not all the codes in termini are conserved. Hutchinson et al. demonstrated that mutations of the highly conserved codons in packaging signals of M gene reduced virus growth and showed defects in virion assembly and genome packaging. However, mutation in codons that are nonconserved had little effect on the characteristics of the virus [44]. Moreover, there are some conserved codons formed clusters in the middle of coding regions that unidentified functions, most notably in the PA gene [55]. Besides, Gavazzi et al. identified that some interactive regions between PB1 and NS genes is not widely conserved among IAVs [38].

Nevertheless, strain-specific differences exist in packaging signals. As studies have shown that the codons that appear to be important for vRNA packaging in WSN are not in PR8, suggesting that not all packaging signals are conserved [26, 54]. Fujii et al. showed that except the 16–20 nt, the packaging process of NS gene did not absolutely require specific sequences at positions 16–35 nt [56]. It has been reported that nucleotide length of the coding regions within packaging signals was as important as the nucleotide sequences [39, 43]. Zhao et al. observed that mutant viruses possessing similar lengths of the ssNCR to that of the wild-type virus replicated to a level close to that of the wild-type virus, while the other mutant viruses with either shorter or longer ssNCRs showed greater reductions in virus replication efficiency [47].

Together, these results suggest that the incorporation of vRNA is not critically dependent on the wild type nucleotide sequences, and the secondary structure may play a more important role in the process [57]. A conserved RNA pseudoknot structure, predicted in the 5′ end of the NP packaging signal region, was shown to affect virus replication [58]. Stem-loop and hairpin structures that are extremely conserved were predicted within the 3′ and 5′ ends of M packaging signal regions [59, 60]. The viruses with structure-disrupting mutations exhibited lower replication efficiencies and a significantly reduced median plaque size relative to the wild-type virus; when the structure was restored, the mutant virus could replicated at levels comparable to the wild-type virus [59].

The motif sequences in the 3′ and 5′ ends of each IAV vRNA are partially complementary to each other. They can form a bulged duplex “corkscrew” structure which is essential for vRNA transcription and replication [61‐63]. Therefore, is compatibility required between other parts of the packaging signals at the ends of vRNA? Zhao et al. generated a series of mutant HA plasmids in the context of the WSN virus, that is, one end of the H1-UTR was replaced with the corresponding UTR of other subtype-specific IAV (H2 to H7 and H9), while the other end was unchanged. The results showed that all of these single-end HA UTR substitution viruses were successfully rescued and could replicate efficiently [47]. Liang et al. constructed six hybrid GFP plasmids that carried mismatched 5′ and 3′ packaging signals that were derived from two different vRNAs of the polymerase gene segments (e.g. the 5′ packaging signal from PB2 and the 3′ packaging signal from PB1 or PA). Interestingly, all of these artificial reporter vRNAs were poorly packaged [34]. The different results found by Zhao and Liang may be due to the methods and segments difference.

Replacing the wild-type ssNCRs at both ends of the WSN HA vRNA with the corresponding ssNCRs of the H2, H3, H5, and H9 subtypes resulted in lower replication efficiency of the mutant viruses than the wild-type virus, especially at earlier time points post infection. And the 3′ end ssNCR substitution viruses demonstrated a more drastic reduction [47]. Barman et al. successfully improved the replication of A/Anhui/1/2013(Anhui/1, H7N9) influenza vaccine virus in eggs by more than a twofold higher titer than that of wild-type by constructing a chimeric gene with the coding sequence of the A/Anhui/1/2013 (Anhui/1, H7N9) NA vRNA and the packaging signals from the PR8 NA vRNA [64]. Recently, stem-loop structures were found at the terminal packaging sequences in both M and PB2 vRNA. Disrupting the predicted secondary structures significantly attenuated the infectivity of the mutant virus and increased the production of defective virus particles [60, 65]. Spronken et al. obtained similar results by studying the conserved hairpin structure in the 967–994-nt of the M vRNA using a compensatory mutagenesis approach [59]. These data suggest that packaging signals of IAVs are very important for efficient viral replication and incorporation.

Extensive evidence indicates that protein incompatibility among both polymerase subunits and packaging signals of vRNAs are restricting factors for reassortment between two viruses [50, 66‐68]. Several reports indicated that in natural or experimental competitive situations, the number of reassortant viruses was very low, and the reassortant genotypes were not random [26, 67, 69]. Essere et al. reported that an avian HA segment of H5N2 could not be incorporated alone into the H3N2 genetic background when competing with the HA segment of the H3N2 virus. However, introducing 5′ and 3′ packaging sequences of the H3N2 HA into an otherwise H5N2 HA backbone or introducing five silent mutations into the H3N2 M segment was sufficient to disrupt this limitation [69].

Recent studies found that the HA segment containing matched packaging signals relative to the background of the virus was packaged preferentially, but there was no preference for homologous packaging signals for the NA and NS segments [70]. These results indicate that the NA and NS segments could move between human H3N2 and H1N1 lineages without restrictions of packaging signal mismatches, while the movement of the HA segment would be limited [70]. By swapping the packaging sequences of NS and HA gene and destroying the intrinsic packaging sequences of the two segments, Gao et al. created a chimeric HA and NS segment with packaging signals of the NS and HA gene. The rewired viruses could be rescued successfully and replicate efficiently, but they lost their ability to independently reassort their HA or NS gene [71].

All of these researches suggested that IAV packaging signals play a critical role in mediating the reassortment between different viruses, and the importance of packaging signals in the process may be segment dependent. Beyond IAV packaging signals, some amino acids may also be required for packaging of specific RNA segments. Moreira et al. demonstrated that the head and body domain of NP impair genome packaging of influenza viruses; thus, the author referred to this set of amino acids as the “NP packaging code” [72]. Bolte et al. showed that packaging sequence mutation of one vRNA won’t impact the genome packaging of an H7N7 virus, whereas combining the mutation with specific nucleoprotein amino acid will impair the process greatly [73].

A series of studies demonstrated that not all vRNAs are equally important in their role in vRNA incorporation into influenza A virions [35, 49, 54]. A recent study showed that PB2, PA, NP and M play a more crucial role than other vRNAs in the packaging procedure of the PR8 virus [26, 35, 54, 74]. Moreover, reductions in the packaging of the PB2 vRNA reduced incorporation efficiency of PB1, PA, NP, M and NS vRNA in WSN virus significantly, but had little impact on the packaging level of HA or NA vRNA [54]. Nevertheless, another experiment showed that when lacking the HA segment, the growth of the virus was reduced and the packaging of other segments was impaired in WSN virus, particularly the PA, NP, NA, M and NS vRNAs [49].

For all IAV segments, deletions or mutations in their packaging regions affect vRNA incorporation efficiency, not only for the segment in which they reside, but also for other segments. For the three polymerase segments, this impact was most prominent for the PB2 gene [35]. A key region of the open reading frame (1659–1671-nt) in both the WSN and PR8 virus HA segments was identified [49]. Mutations in the region resulted in significant reductions in the packaging of the HA and other vRNAs. The most affected segments were PB1 vRNA in the WSN virus and NA and M vRNAs in the PR8 virus [49]. These results indicate that some cooperation exists between vRNAs and virus-specificity exists [54].

Actually, the favored hypothesis of the highly selective genome packaging mechanism relies on RNA-RNA or protein-RNA interactions (Fig. 1) [17, 34, 35]. As early as 2011, Fournier et al. found that the eight vRNPs of a human H3N2 IAV are interconnected in a “transition zone” at the budding tip of virions [17]. Further studies showed that the vRNAs formed a single network of intermolecular interactions in vitro and the known packaging signals are identified as the strongest interactions [17, 75]. Similarly, for the H1N1 WSN virus, most of the inter-segment interactions were mediated by their 3′ and 5′ terminal packaging signals [76]. Another study demonstrated that the interaction network also existed in an avian H5N2 IAV. However, both the interaction network and the involved sequences of each vRNA varied from those of the human H3N2 IAV mentioned above [77]. By studying interactions between PB1 and NS segments of the avian H5N2 IAV, they further proved that the interaction identified in vitro takes place in infected cells and was crucial for viral replication, and genome packaging and reassortment. Noticeable was that the sequences identified as interaction not located in the packaging signals [38].

It’s not difficult to find that the identified sequences required for the genome packaging are varied from different research groups, and some relevant conclusions seem conflict. Moreover, other recently emerging ideas that are contrary to conventional wisdom are also notable. For example, experiment from Dadonaite et al. revealed that the RNA-RNA interactions are extensive, redundant and complex, rather than a finite set of discrete interactions between packaging signals [57]. Bolte et al. suggested that the packaging of IAV vRNA is governed by the redundant and plastic network of RNA-RNA and potentially RNA-nucleoprotein interactions [73]. Contrary to the classical model that nucleoprotein binds vRNA as a uniform pattern of “beads on a string” [78‐80], the nucleoprotein binds vRNA non-uniformly and without apparent sequence specificity. For example, some sites are enriched and some sites are poor in nucleoprotein association [6, 7, 80]. The sequences with low-NP-binding, including both internal sequences and the traditional packaging signals of the segment, are necessary for virus propagation [6, 7]. Different from the RNA-RNA molecular recognition mechanism in genome packaging of IAV, Venev et al. proposed another hypothesis. They generated model 8-segment virions by using monte carlo simulations. The results demonstrated that selective packaging of the IAV genome can be achieved by self-repulsion of identical segments rather than by molecular recognition [81].