An in vitro network of intermolecular interactions between viral RNA segments of an avian H5N2 influenza A virus: comparison with a human H3N2 virus

Cyrille Gavazzi; Catherine Isel; Emilie Fournier; Vincent Moules; Annie Cavalier; Daniel Thomas; Bruno Lina; Roland Marquet

doi:10.1093/nar/gks1181

. 2012 Dec 5;41(2):1241–1254. doi: 10.1093/nar/gks1181

An in vitro network of intermolecular interactions between viral RNA segments of an avian H5N2 influenza A virus: comparison with a human H3N2 virus

Cyrille Gavazzi ¹, Catherine Isel ¹, Emilie Fournier ¹, Vincent Moules ², Annie Cavalier ³, Daniel Thomas ³, Bruno Lina ², Roland Marquet ^1,^*

PMCID: PMC3553942 PMID: 23221636

Abstract

The genome of influenza A viruses (IAV) is split into eight viral RNAs (vRNAs) that are encapsidated as viral ribonucleoproteins. The existence of a segment-specific packaging mechanism is well established, but the molecular basis of this mechanism remains to be deciphered. Selective packaging could be mediated by direct interaction between the vRNA packaging regions, but such interactions have never been demonstrated in virions. Recently, we showed that the eight vRNAs of a human H3N2 IAV form a single interaction network in vitro that involves regions of the vRNAs known to contain packaging signals in the case of H1N1 IAV strains. Here, we show that the eight vRNAs of an avian H5N2 IAV also form a single network of interactions in vitro, but, interestingly, the interactions and the regions of the vRNAs they involve differ from those described for the human H3N2 virus. We identified the vRNA sequences involved in five of these interactions at the nucleotide level, and in two cases, we validated the existence of the interaction using compensatory mutations in the interacting sequences. Electron tomography also revealed significant differences in the interactions taking place between viral ribonucleoproteins in H5N2 and H3N2 virions, despite their canonical ‘7 + 1’ arrangement.

INTRODUCTION

Influenza A viruses (IAVs) are a serious threat to human health (1). They are responsible for the majority of seasonal flu cases and for occasional flu pandemics that can have high case fatality and mortality rates (1). The IAV genome consists of eight single-stranded viral RNA (vRNA) segments of negative polarity ranging form 890 to 2341 nt, each encoding at least one essential protein in the antisense orientation (2). In the viral particles, each vRNA is associated with a trimeric viral polymerase complex and numerous copies of the nucleoprotein (NP), forming a viral ribonucleoprotein (vRNP) (2,3). The polymerase subunits, basic protein (PB2), PB1 and acidic protein (PA), are encoded by the three largest vRNAs, named vRNAs 1, 2 and 3, respectively. The polymerase complex recognizes the 5′ and 3′ termini of the vRNAs that are highly conserved amongst IAV strains and among the eight vRNA species (4–6). The 5′ and 3′ termini, which are 13 and 12 nt long, respectively, are partially complementary and impose a hairpin structure to the vRNPs (7,8).

While influenza B viruses almost exclusively infect humans, the main reservoir of IAVs is constituted by wild aquatic birds, such as ducks, geese and swans, and the IAVs are much more diverse in the waterfowl than in the human population (9). A human IAV pandemic can arise when a flu strain jumps from an animal species to the human species and is transmitted efficiently in the human population (1). An avian influenza virus can be transmitted directly or indirectly to humans, without undergoing major genetic changes: it is thought that such a process gave rise to the ‘Spanish flu’ in 1918. Alternatively, when an animal and a human flu strain co-infect an intermediate host, such as a pig, the vRNAs from the two flu strains can mix to generate new strains (a process named genetic reassortment) that can spread from the intermediate host to humans. If the reassortant viruses contain ‘animal’ hemagglutinin (HA) and neuraminidase (NA) segments that encode virus surface proteins for which humans have limited or no preexisting immunity, it can cause a pandemic in the human population: this happened in 1957 (the ‘Asian flu’), in 1968 (the ‘Hong Kong flu’) and in 2009 (the ‘swine flu’) (1,10).

The segmented nature of the IAV genome confers an obvious evolutionary advantage, but it generates a formidable problem for the viral RNA packaging machinery (3). Natural IAV infections are initiated at low multiplicity of infection; therefore, viral particles that do not possess a complete set of vRNAs will fail to replicate. IAVs rely on a segment-specific packaging mechanism to ensure that each viral particle contains one copy of each vRNA (3). Indeed, IAVs normally incorporate one and only one copy of each vRNA (11–14).

Defective interfering RNAs are generated in cell culture and in vivo under high multiplicity of infection conditions (15–24). They are generated during vRNA replication and contain large deletions in the central coding region(s) (15–17,19,20,22–24). Importantly, they specifically compete for packaging into the viral particles with their parental vRNA (18,19,21,23), indicating that vRNAs contain segment-specific packaging signals. Sequencing of the defective interfering RNAs indicated that these sequences usually reside within ∼100–300 nt from each end of the vRNAs. More recently, reverse genetics experiments, mainly performed on the A/WSN/33 (H1N1) strain (WSN), and to a lesser extent on the A/PR/8/34 (H1N1) strain (PR8), revealed the existence of segment-specific packaging signals in each vRNA, usually located within ∼100 nt from the vRNA ends (4,25–32). The packaging sequences appear to be multipartite and discontinuous (3). No precise boundary could be mapped in these experiments: successive deletions from the middle to the border of the coding regions resulted in a gradual decrease of the packaging efficiency. Importantly, some, but not all, conserved codons, which tend to accumulate in the vRNA terminal regions, have been shown to be important for optimal packaging of several vRNAs (33–35).

How the packaging signals ensure selective packaging of a set of eight different vRNAs is still unknown, but the existence of intermolecular interactions between vRNAs is an attractive hypothesis, especially as no viral or human protein specifically recognizing an IAV packaging signal has been identified (3,26). However, no such interactions have been identified in IAV viral particles. We therefore recently developed an in vitro approach, which showed that the eight vRNAs of the A/Moscow/10/99 (H3N2) strain (MO) are involved in a single interaction network (36). The vRNA regions involved in several interactions were identified and were found to overlap with the regions containing the WSN/PR8 packaging signals, even when they could be narrowed down to a few nucleotides, suggesting that the vRNA network detected in vitro is relevant to IAV vRNA packaging (36). Combined with electron tomography, which revealed the internal organization of MO virions, our in vitro data suggested that the eight vRNAs are selected and packaged as an organized supramolecular complex held together by direct base pairing of the packaging signals (36).

Here, we applied the same strategy to an avian IAV: the A/Finch/England/2051/91 (H5N2) strain (EN). We showed that the eight vRNAs of the EN strain are also involved in a single network of RNA interactions in vitro. We grossly defined the regions of the vRNAs involved in each interaction, and in several cases, we identified the interacting regions at the nucleotide level. Surprisingly, neither the interaction network nor the regions of the vRNAs involved in the intermolecular interactions is conserved between the EN and MO strains. Electron tomography also revealed significant differences in the organization of vRNPs in EN and MO virions. Thus, our data suggest that the packaging signals might not be conserved between these viral strains. The precise identification of the nucleotides involved in the vRNA/vRNA interactions in vitro will allow testing their role in IAV replication.

MATERIALS AND METHODS

Plasmids, cloning and site-directed mutagenesis

The EN complementary DNAs were cloned between the BsmbI sites of pUC2000 vectors starting from the corresponding reverse genetic plasmids. These vectors were obtained as previously described and contain a T7 promoter, the cloning cassette of pHW2000 containing two BsmbI sites, and a unique restriction site (Eco47III, Bsh1236I or Ecl136II) between the PstI and EcoRI sites (36). Sixty-seven mutants with deletions ranging from 11 to ∼400 nt have been obtained by polymerase chain reaction using strictly complementary primers bridging the region to be deleted. After 35 polymerase chain reaction cycles, the parental plasmids were digested with DpnI, and the amplification products were used to transform One Shot MAX efficiency DH5α-T1 (Invitrogen) Escherichia coli cells. Fifteen mutants bearing point substitutions were obtained using the same strategy.

In vitro transcription, RNA purification and electrophoretic mobility shift assay

DNA templates digested with Eco47III, Bsh1236I or Ecl136II were either co-transcribed in vitro by pairs, and RNA complexes were analysed directly, or vRNAs were synthesized individually, purified, and incubated with another vRNA, followed by analysis of the complexes by native agarose gel electrophoresis. Co-transcription experiments were performed essentially as described previously (36). Briefly, linearized plasmids (∼200 ng of each plasmid) were incubated for 3 h at 37°C and treated with ribonuclease-free DNase I before analysis of the RNA complexes. Alternatively, in vitro transcription of individual vRNAs was performed with ∼30 µg of linearized plasmids essentially as described previously (37), except that the concentration of each nucleotide triphosphate (NTP) was 8 mM, and that 0.5 µg of yeast inorganic pyrophosphatase (Roche) was included in the reaction mixture. Following DNase I treatment, vRNAs were extracted with phenol/chloroform, ethanol precipitated and purified on a TSK G2000SW column (Tosoh Bioscience) as described previously (37). The integrity of the vRNAs was routinely checked by denaturing polyacrylamide gel electrophoresis. Pairs of purified vRNAs (2 pmol each) were denaturated for 2 min at 90°C in 8 µl of water and cooled on ice. Two microlitres of 5-fold concentrated buffer (final concentration: 50 mM sodium cacodylate, pH 7.5; 40 mM KCl and 0.1 mM MgCl₂) were added, and the samples were incubated for 30 min at 37°C or 55°C, then analysed on 0.8% agarose gels containing 0.01% (w/v) ethidium bromide after addition of 2 µl of loading buffer [40% (v/v) glycerol, 0.05% (w/v) xylene-cyanol and 0.05% (w/v) bromophenol blue]. Native gel electrophoresis of the RNA complexes was performed at 4°C in a buffer containing 50 mM Tris, 44.5 mM borate and 0.1 mM MgCl₂. The RNA weight fraction (%) of each band in a lane was determined, and the percentage of intermolecular complex was determined by dividing the weight fraction of the corresponding band by the sum of the weight fractions of all bands in the lane.

In oligo-mapping experiments, vRNAs were incubated in the presence of 10 pmol of an oligodeoxyribonucleotide (ODN) complementary to one of the vRNAs. A trace amount (10 000–20 000 cpm) of ³²P-5′-labelled ODN was included to be able to check its binding to the targeted vRNA. To that aim, after imaging under ultraviolet light, gels were fixed for 10 min in 10% (v/v) trichloroacetic acid, dried at room temperature and submitted to autoradiography.

Electron microscopy (EM) and tomography

Electron microscopy (EM) and electron tomography of EN virus budding from Madin-Darby canine kidney (MDCK) cells were performed as described previously (36,38).

RESULTS

Optimization of the conditions for studying interactions between EN vRNAs

We recently studied the interactions taking place between the vRNAs of the IAV MO strain during in vitro transcription (36). Mutational analysis of three of these interactions showed that they involved the terminal regions of the vRNAs open reading frames (36), which are known to contain packaging signals [in the case of the human PR8 and WSN IAVs (3)], suggesting that the interactions detected in vitro are relevant to vRNA packaging. When we applied the same protocol to EN vRNAs, we observed 15 intermolecular interactions, but in only three of them did the complex represent ≥20% of the RNA mass. Besides, we observed that several EN vRNAs, but no MO vRNAs, interacted with a green fluorescent protein (GFP) messenger RNA control under these conditions (data not shown). Finally, we introduced a series of ∼200-nt deletions covering the complete vRNA 8 non-structural (NS), but none of these deletions had a significant effect on its interaction with other EN vRNAs (data not shown). Thus, we concluded that the experimental conditions that allowed for specific interactions between MO vRNAs are not well suited for EN vRNAs.

We thus individually transcribed EN vRNAs in larger amounts and purified them, to be able to co-incubate them by pairs under varying conditions. Based on the work on retroviruses, which showed that different conditions have to be used to analyse in vitro dimerization of the genomic RNA of different viruses (39,40), we incubated all possible pairs of EN vRNAs at varying ionic strengths and temperatures. Only a few weak interactions were detected at 37°C (data not shown), but these interactions and additional ones were more clearly detected at 55°C. Surprisingly, they formed also more efficiently at low ionic strength. We thus analysed the interactions between EN vRNAs in a buffer containing 50 mM sodium cacodylate, pH 7.5; 40 mM KCl and 0.1 MgCl₂.

EN vRNAs form a single interaction network

When incubated under these conditions, most individual EN vRNAs migrated as a single band corresponding to monomers (Figure 1 and Supplementary Figure S1). A significant band attributed to homodimeric species was observed with vRNA 8 (coding for NS1 and NS2/nuclear export protein (NEP)), and fainter dimeric bands were observed with vRNAs 4 (coding for HA), 5 (coding for NP) and 7 (coding for M1 and M2). Importantly, when pairs of EN vRNAs were co-incubated, 13 additional bands corresponding to intermolecular complexes were detected (Figure 1A and Supplementary Figure S1). Not all vRNA/vRNA interactions were equally strong: four complexes accounted for 7–17% of the RNA mass in the lane, five accounted for 21–49% and four accounted for 65–82% (Figure 1B). Each vRNA was involved in several interactions: three vRNAs interacted with two partners, one interacted with three partners, three interacted with four partners and one (vRNA 2) interacted with five other vRNAs (Figure 1). Remarkably, the network of interactions linking the EN vRNAs is different from the one linking the MO vRNAs (36). These two networks only share four interactions in common: those between vRNAs 2 and 6 (coding for NA), 4 and 7, 4 and 8 and 5 and 8.

Identification of the regions of the EN vRNAs involved in intermolecular interactions in vitro

The regions of the vRNAs containing segment-specific packaging signals have been mainly studied for WSN and PR8, two human H1N1 strains of IAV (25,26,28–35,41). These regions tend to be located towards the ends of the vRNAs: most packaging signals are located within ∼100 nt from both ends of the vRNA coding regions. In the case of MO, we recently showed that regions of the vRNAs involved in vRNA/vRNA interactions in vitro correspond to regions identified as packaging signals in WSN or/and PR8 (36). We therefore introduced systematic deletions in all EN vRNAs, except vRNA 6, to identify the regions involved in most vRNA/vRNA interactions detected in vitro. Analysis of the weakest interactions appeared technically more difficult, and we therefore did not attempt to identify the regions of vRNA 6 interacting with vRNAs 1 and 2 (Figure 1B).

The vRNA 4/vRNA 7 interaction

We first introduced deletions close to the ends of vRNA 7, without affecting the promoter sequences, and tested their effect on the interaction with vRNA 4. Surprisingly, none of these deletions had a significant impact on the interaction with vRNA 4 (Figure 2A, lanes 10 and 15). We thus also constructed a series of deletion mutants covering the central part of vRNA 7. Remarkably, one of them, extending from nucleotides 514–724 completely abolished the interaction with vRNA 4 (Figure 2A, lane 13). In addition, we observed that all but the 514–724 deletions favoured homodimerization of vRNA 7 (Figure 2A, lanes 1 to 8). Similarly, we constructed a set of eight deletion mutants covering the entire vRNA 4 sequence, except the terminal promoters. Again, deletions close to the ends of the vRNA had no effect on the interaction between vRNAs 4 and 7 (Figure 2B, lanes 12 and 19). The most pronounced effect on this interaction was observed for the deletion covering nucleotides 613–857 of vRNA 4 (Figure 2B, lane 15). However, unlike deletion 514–724 in vRNA 7, deletion 613–857 did not totally abolish the vRNA 4/vRNA 7 interaction. This might be an indication that more than one region of vRNA 4 are involved in this interaction (see later). Accordingly, several other deletions also weakened the interaction between vRNAs 4 and 7, albeit less dramatically (Figure 2B, lanes 13–14 and 16–18).

Figure 2. — Region 514–724 of vRNA 7 and region 613–857 of vRNA 4 are required for optimal interaction between these vRNAs. Deletions were introduced in vRNA 7 (A) or vRNA 4 (B), and formation of the intermolecular complex was monitored by agarose gel electrophoresis.

The vRNA 1/vRNA 2 interaction

We next introduced sequential deletions in vRNAs 1 and 2 to delineate the regions involved in the intermolecular interaction between these two vRNAs. None of the vRNA 1 and vRNA 2 deletion mutants formed homodimers, and they all migrated as expected based on their size (Figure 3 and data not shown). One deletion in vRNA 2, encompassing nucleotides 1939–2198 had a dramatic effect on the vRNA 1/vRNA 2 interaction (Figure 3A, lane 12). Deletion of nucleotides 785–1170 in vRNA 1 had a similar effect (Figure 3B, lane 6). These two deletions had major effects but did not completely abolish the interaction, suggesting minor contributions from other regions of both vRNAs.

Figure 3. — Region 1939–2198 of vRNA 2 and region 785–1170 of vRNA 1 are required for optimal interaction between these vRNAs. Deletions were introduced in vRNA 2 (A) or vRNA 1 (B), and formation of the intermolecular complex was monitored by agarose gel electrophoresis.

The vRNA 3/vRNA 8 interaction

We next studied the interaction between vRNAs 3 and 8 (Figure 4). As already mentioned, vRNA 8 significantly dimerized, and dimerization was also observed with all vRNA 8 mutants, except those in which nucleotides 235–456 or 435–656 were deleted (Figure 4A, lanes 1–7). Several deletions in this RNA had an impact on the interaction with vRNA 3: deletion of nucleotides 435–656 totally abolished the intermolecular interaction, whereas deletion of nucleotides 235–456 had a drastic effect, although some complex was still detected (Figure 4A, lanes 13 and 14). One possibility is that region 435–456 of vRNA 8, which was deleted in both mutants, is involved in the interaction with vRNA 3, together with downstream sequences. In addition, deletion of nucleotides 635–834 also had a noticeable, although more limited, effect on the vRNA 3/vRNA 8 interaction (Figure 4A, lane 15). Similarly, the analysis of vRNA 3 showed that deletion of nucleotides 125–514 totally abolished the intermolecular interaction (Figure 4B, lane 5), whereas deletions in the 515–2230 region had a limited but significant negative impact on this interaction (Figure 4B, lanes 6–10).

General features of the EN vRNA/vRNA interactions

The strategy detailed previously for interactions between vRNAs 4 and 7, 1 and 2 and 3 and 8 was applied to all interactions detected in vitro, and allowed to gain information on the regions of the vRNAs involved in most of them (Figure 5, Supplementary Figures S2 and S3 and data not shown). In many cases, several (usually contiguous) deletions affect a given interaction, but one deletion has usually a more pronounced effect than the other ones, suggesting that it constitutes the main interacting domain, and only this region has been reported in Figure 5. The deletions that have less pronounced effects may either correspond to secondary interaction sites or affect the interaction indirectly by affecting the vRNA structure. However, in the case of the vRNA 3/vRNA 5 interaction, two non-overlapping adjacent deletions in vRNA 5 completely abolished the interaction with vRNA 3 (Figure 5), suggesting that the sequence of vRNA 5 interacting with vRNA 3 is located close to the junction between these two deletions (i.e. nt 1419). In contrast with our recent report on MO vRNAs (36) and with the location of the PR8 and WSN packaging regions (3), the regions involved EN vRNA/vRNA interactions are usually located >100 nt from the ends of the coding regions. The only exception could be the region of vRNA 5 interacting with vRNA 3, which remains to be delineated more precisely (see earlier in text and Figure 5). Deletion of nucleotides 635–834 of vRNA 8 abolished interaction with vRNA 5 (Figure 5 and Supplementary Figure S3), but deletion of nucleotides 765–877 did not (Supplementary Figure S3, lane 5), indicating that the region of vRNA 8 interacting with vRNA 5 is located >100 nt away from the 5′ end of the coding region of this vRNA. Remarkably, several regions of interaction are located close to the middle of the vRNAs (Figure 5). Deletion analysis of vRNA 1 did not allow to identify a region of this vRNA required for the interaction with vRNA 6 because none of the deletion significantly impaired this interaction. Deleting nucleotides 125–384 in vRNA 2 decreased the interaction with vRNA 6, but the effect was modest. As these two interactions appeared difficult to analyse, we did no attempt to identify the regions of vRNA 6 they involve.

Figure 5. — Summary of the vRNA regions involved in intermolecular vRNA interactions. vRNAs have different colours and are drawn to scale. Regions defined by deletion analysis are presented by rectangles that have the same colour as the interacting vRNA. Note that when only one of several overlapping deletions affects the interaction between two vRNAs, the region required for the interaction is shorter than this deletion, as the overlapping sequences are not required. Deletions that have limited effects are indicated by question marks. Sequences identified by ODN mapping or/and site-directed mutagenesis are indicated by asterisks.

Precise identification of the sequences of EN vRNAs involved in vRNA/vRNA interactions

We next attempted to precisely identify the nucleotides involved in several vRNA/vRNA interactions. When deletion experiments allowed us to identify the region of one vRNA involved in an intermolecular interaction, we used a series of antisense ODNs spanning this region to refine the interaction zone. Sequence analysis then usually allowed to propose a limited number of possible interactions between the zone identified by ODN scanning on one vRNA and the region identified by deletion on the other vRNA. These possible interactions were then tested with ODNs complementary to the second vRNA. When many interactions were possible, a systematic ODN scan of the longer vRNA was performed.

The vRNA 4/vRNA 8 interaction

A first example of this strategy is shown in Figure 6 for the interaction between vRNA 4 and 8. Deletion analysis indicated that region 257–434 of vRNA 8 interacts with vRNA 4 (Supplementary Figure S2B, lanes 6–8). Two adjacent ODNs complementary to nucleotides 346–375 and 376–405 almost totally abolished the vRNA 4/vRNA 8 interaction (Figure 6A, lanes 7–8). Knowing, that in vRNA 4, the region of interaction is located between nucleotides 368 and 612 (Supplementary Figure S2A), the likeliest interaction is between nucleotides 364–374 of vRNA8 and nucleotides 449–459 of vRNA 4 (Figure 6B): it involves six G-C and four A-U base-pairs, as well as a non-canonical G-A base pair. The existence of this interaction is inferred from the inhibitory effect of ODNs complementary to nucleotides 442–466 of vRNA 4 and to nucleotides 357–381 of vRNA 8 (Figure 6C). Note that the ODN complementary to nucleotides 346–375 of vRNA 8 also directly targets this interaction, whereas the ODN complementary to nucleotides 376–405 hybridizes immediately adjacent to that region and probably has an indirect effect (Figure 6A and C).

The vRNA 5/vRNA 8 interaction

Among a series of nine ODNs targeting region 1165–1419 of vRNA 5 identified by deletion analysis (Supplementary Figure S3B), two adjacent ODNs complementary to nucleotides 1193–1222 and 1223–1252, respectively, dramatically inhibited the interaction between vRNAs 5 and 8 (Figure 7A, lanes 5 and 6). Interestingly, nucleotides 1216–1226 of vRNA 5 are strictly complementary to nucleotides 660–670 of vRNA 8 (Figure 7B), which are located in region 657–764 identified by deletion analysis of this vRNA (Supplementary Figure S3A). As predicted, an anti-vRNA 8 ODN targeting nucleotides 654–676 also abolished the interaction between vRNAs 5 and 8, supporting the existence of the proposed base pairing between these vRNAs (Figure 7C).

Figure 7. — Interaction between vRNAs 5 and 8 involves nucleotides 1193–1252 of vRNA 5 and nucleotides 654–676 of vRNA 8. (A) ODNs were used to map region 1163–1432 of vRNA 5. (B) Proposed interaction between vRNAs 5 and 8. (C) ODNs complementary to vRNAs 5 and 8 were used to test the proposed interaction.