Influenza virus segment 5 (+)RNA - secondary structure and new targets for antiviral strategies

Influenza A virus is a threat for humans due to seasonal epidemics and occasional pandemics. This virus can generate new strains that are dangerous through nucleotide/amino acid changes or through segmental recombination of the viral RNA genome. It is important to gain wider knowledge about influenza virus RNA to create new strategies for drugs that will inhibit its spread. Here, we present the experimentally determined secondary structure of the influenza segment 5 (+)RNA. Two RNAs were studied: the full-length segment 5 (+)RNA and a shorter construct containing only the coding region. Chemical mapping data combined with thermodynamic energy minimization were used in secondary structure prediction. Sequence/structure analysis showed that the determined secondary structure of segment 5 (+)RNA is mostly conserved between influenza virus type A strains. Microarray mapping and RNase H cleavage identified accessible sites for oligonucleotides in the revealed secondary structure of segment 5 (+)RNA. Antisense oligonucleotides were designed based on the secondary structure model and tested against influenza virus in cell culture. Inhibition of influenza virus proliferation was noticed, identifying good targets for antisense strategies. Effective target sites fall within two domains, which are conserved in sequence/structure indicating their importance to the virus.

Influenza is a common disease that spans the globe and is dangerous for human health and life. Even seasonal flu is life threatening for young infants, the elderly, and immunocompromised patients. The most dangerous outbreaks are global pandemics that appear sporadically when immunologically novel influenza strains (e.g. from animal populations) gain the ability to infect humans, causing severe illness and complications in healthy populations.
Influenza is a single-stranded negative sense RNA virus with segmented genome (viral RNA -vRNA). The vRNA genome of influenza is divided into eight segments, which are independent transcription-replication units. Each step in the influenza life cycle is dependent on RNA. vRNA is used to generate two types of (+)RNA: cRNA (complementary RNA), and mRNAs (messenger RNA). cRNA is used as a template for replicating vRNAs while mRNA encodes viral proteins. More than 8 viral proteins are encoded through alternative splicing and translation initiation. Recent studies are revealing the roles of vRNA and mRNA secondary structure during different steps in the viral life cycle 1,2 . Knowledge about secondary structure provides insights to better understand virus biology and could be used for designing new therapeutics; however, additional work is needed to experimentally model influenza structure.
Antisense oligonucleotide (ASO) drugs are designed to alter RNA expression and function, thus allowing one to influence pathological processes beyond protein or receptor involvement 3 . Oligonucleotides are a major focus of drug development; numerous clinical trials have been performed and many oligonucleotide therapeutics are in phase III development. Several oligonucleotide therapeutics are FDA approved 4,5 . Oligonucleotide strategies offer advantages in comparison to other existing approaches against influenza. In fast evolving viruses, such as influenza, targeting surface proteins is challenging, as each season novel proteins appear in circulating strains. In contrast, the RNA of influenza, due to combination of evolutionary constrains, possess features that remain constant between strains. One such feature is the need of influenza to conserve RNA secondary structure. Therefore, viral RNA secondary structure determination is a key component in the design of effective inhibitory oligonucleotides.

Results
Chemical mapping of segment 5 (+)RNA. CMCT (modifies N3 of U and N1 of G when unpaired), DMS (methylates N1 of A and N3 of C when unpaired), and kethoxal (modifies G when unpaired) 22,23 and SHAPE mapping (modifies all flexible, e.g. single stranded, bases) 24 were applied to probe the folding in (+)RNA5 and (+)RNA5-ORF. Before mapping, appropriate folding conditions were found to yield a single RNA conformation, as assessed by non-denaturing agarose gel. Both (+)RNA5 and (+)RNA5-ORF adopt a single conformation in buffer containing 100 mM KCl (or 300 mM) and 5 mM MgCl 2 . Mapping was carried out at 23 °C and 37 °C to assess the impact of temperature on folding. This range of temperatures was used because there are influenza strains that can effectively replicate in low temperature (e.g. cold adapted viruses) and others in high temperature (e.g. avian strains) 25,26 ; it is thus valuable to assess how RNA folding may be affected by different temperatures. To read out mapping results, reverse transcription was used with fluorescent labeled primers (Supplementary Table S2) followed by capillary electrophoresis. The analyzed RNAs undergo selective modifications, showing evidence of these being structured RNA molecules. There are relatively short regions that show reactivity at 23 °C (Figs 1, 2, Figures S2-S4, Supplementary Data 1). There are some small differences in reactivity at 37 °C, but they do not significantly alter the general reactivity pattern (Supplementary Data 1). Circular folding of segment 5 (+)RNA. To predict the secondary structure of segment 5 (+)RNA based on the experimental probing data, the results of chemical mapping were used to constrain predictions in the RNAstructure 5.5 program. SHAPE data were loaded as pseudoenergy constraints (the energy contribution of SHAPE reactive nt were penalized) and DMS, CMCT, and kethoxal modifications were included as chemical mapping constraints (reactive nt are forbidden to be in Watson-Crick base pairs flanked by Watson-Crick base pairs). Additional constraints were added to constrain pairing in previously-modeled conserved structural motifs in regions: 1051-1171 nt (1077-1197 (+)RNA5-ORF) pairs 1081/1191, 1083/1190 and 1085/1189 ((+)RNA5 numbering) were added as base pairing constraints 8 .
The secondary structure of (+)RNA5 appears circular in folding (Fig. 1 Figure S5, Fig. 1). These data showed that (+)RNA5 is globally stable at elevated temperatures and only four regions appear to be labile.
The possibility of preferential local folding in segment 5 (+)RNA was also considered for the entire (+)RNA5. For that purpose, beside experimental constraints, a maximum paring distance of 600 nt was set in RNAstructure 5.5 during the prediction of secondary structure: here, the greatest allowable distance between base pairs is 600 nt. All other experimental constraints were also incorporated in this calculation. The consequences of this additional Figure 1. Self-folding of (+)RNA5 predicted by RNAstructure 5.5 using experimental data obtained in 23 °C as constraints. Strong DMS, CMCT and kethoxal modifications, as well as SHAPE reactivities converted to pseudo-free energies were used. Additionally medium reactivity DMS, CMCT, kethoxal and results from RNase H cleavage in presence of DNA oligonucleotides (neither used in modeling) are also annotated. Red lines indicate regions of structure that differ at higher temperature (37 °C). The numbering of (+)RNA5 is from its 5′ end. The AUG start codon spans nucleotides 46-48. constraint are the disappearance of helix 244-281/1478-1533 and the appearance of several new motifs on the 5′ and 3′ ends (Supplementary Figure S7). Besides these few novel folds, the remaining motifs are identical as the calculation without the 600 nt distance constraint.

Base pair probabilities.
To assess prediction quality and identify well-defined structural regions we calculated the secondary structure partition function using RNAstructure 5.5 and, from this, determined the base pair probabilities for model pairs 28 . For the partition function calculations, experimental data were included. Results indicate that there are several regions with paired and unpaired nucleotides of more than 80% probability: 1-342, 388-492, 1480-1565 (Fig. 3). Additionally, all single stranded regions are well defined by having low probability of pairing.
Conservation of (+)RNA5 model structure throughout influenza A strains. 15533 unique sequences of (+)RNA5 were acquired from the NCBI Influenza Virus Resource 29 . Aligned sequences were analyzed with respect to the experimentally constrained (+)RNA5 structure model to identify the level of base pair  (Fig. 4). The most common pairs are GC pairs, followed by AU pairs in roughly similar amounts (37.7% and 34.6%, respectively; Supplementary Data 2). After Watson-Crick pairs, GU pairs are the most commonly observed (13.3%) followed by CA pairs (6.9%) and others in much lower abundance.
The constraint of codon evolution in much of this sequence makes the identification of statistically-significant compensatory mutations difficult; however, we can assess the mutual information (MI), to suggest co-evolution of paired nt in the alignment 30 . The only extended regions with higher-than-average MI (Supplementary Data 2) were the noncoding nt that form the cRNA panhandle (nt 6-14/1560-1551) and the short hairpin between the panhandle (17-26/38-30) and NP start codon ( Fig. 1), which are free of the constraint of maintaining codons.

Hybridization to Isoenergetic Microarrays -oligonucleotide accessible regions. (+)RNA5
accessibility to oligonucleotides was studied using isoenergetic microarrays and RNase H cleavage induced by DNA oligonucleotides [31][32][33] . Short penta-and hexanucleotide probes that are 2′-O-methyl-RNA with LNA and 2,6-diaminopurine modifications at selected positions were used to build isoenergetic microarrays. Hybridization of duplex probe/RNA are engineered to have near equal free energies [34][35][36] ; thus, only the secondary structure of the RNA should influence binding of these isoenergetic probes, not the sequence-specific stability of duplex formation [31][32][33] . Hybridization of (+)RNA5 to the isoenergetic microarray was conducted at 23 °C and 37 °C in buffer containing 300 mM KCl, 5 mM MgCl 2 and 50 mM HEPES pH 7.0. Only strong and medium binding probes were considered in the analysis of structure (Tables 1 and S4, Figure S9).
RNase H cleavage was used to validate several oligonucleotide probe binding sites (Figs 1 and 3, Supplementary Figure S8). 19 DNA oligonucleotides were used to test accessibility of the (+)RNA5 regions identified by microarray mapping. Generally, RNase H cleavages were observed at the expected single stranded regions: after nucleotides 157, 287, 356, 365, 422, 604-605, 620, 642, 760, 1153, and 1154 ( Fig. 5 and Supplementary Table S5). Additional sites in single stranded regions: 58-60, 923, 925, 997-999 were cleaved by RNase H, showing that these DNA oligonucleotides also bind partially in these regions. Cleavages were also observed after 262-264 (internal loop) and 847 nt. Cleavage after 847 was within a helix, but DNA oligomer that induced that cleavage, CTCTCCATT, is complementary to a helix/loop region. Weak cleavages were observed in single stranded regions after 473, 705, 729, 880 and 1181. Nucleotides 1060-1068 are sequestered in a helix interaction, and were not accessible to complementary oligomer AAGGTCCTC. Oligonucleotides CAAATCCTC and GGATCTATT did not cause cleavage (complementary to regions: 396-404, 943-953, respectively), even though (in proposed secondary structure) complementary regions are single stranded; this suggests the possibility of tertiary, long distance interactions, occurring in these probe-inaccessible regions.    35,36 , in parenthesis, the site of binding for which calculation was done.
The probe binding sites: deduced (confirmed by RNase H cleavage) and probable (assigned after analysis of the most possible alternative binding sites in terms of hybridization duplex free energy and target structure) were marked on the (+)RNA5 secondary structure ( Fig. 5 and Table 1, Supplementary Table S5). The binding of modified probes are almost identical in 23 °C and 37 °C. Only one probe, AcCcUg (nt in capital letters are 2′-O-methyl-RNA nucleotides, while lower case letters indicate LNA nucleotides) differs in binding intensity to (+)RNA5: it binds strongly in 23 °C and with medium intensity at 37 °C. Accessible sites: 154, 155, 156, 264, 365, 601, 605, 630, 845, 880, 921, 923 are in single stranded regions (Fig. 5). Binding to the microarray was also observed at regions closing a hairpin loop or internal loop: sites 644, 645, 846, 956. Generally, helical regions of (+)RNA5 were not accessible for probes. Additionally, several predicted hairpin loops, for example 467-477 and 1016-1029, are not accessible to probes. Tertiary structure and local non-canonical interactions could be affecting binding in these hairpins.
Inhibitory effect of oligonucleotides targeting segment 5 (+)RNA in cell culture. The highly conserved (across type A influenza strains) secondary structure of segment 5 (+)RNA presented in this work allow to designate new targets for oligonucleotides. The new target regions, could be alternative to those previously found for the NP (+)RNA 8,17 and confirmation of determined secondary structure elements. Antisense oligonucleotides (ASOs) were designed to be fully complementary to segment 5 (+)RNA of strain A/California/04/2009 (H1N1) ( Table 2). All synthesized and tested ASOs were 2′-O-methyl RNA with LNA nucleotides to stabilize binding to target regions. ASOs were designed to target single stranded regions that we found to strongly bind microarray probes and/or are well-defined as single-stranded (based on the chemical mapping of segment (+)RNA5) (Fig. 5). Two control ASOs were also used: oligonucleotide N 17 (sequence nonspecific to influenza virus RNA) and Neg 5 (complementary to a region of (+)RNA5 that forms a helix: 1262-1272) ( Table 2, Fig. 5). The nucleotide sequence conservation of each site was calculated ( Table 2). The average nucleotide sequence conservation for the entire segment 5 (+)RNA is 87.6% and 89.7% for ORF. Average nucleotide sequence conservation for ASO binding sites is slightly higher than the overall conservation: 90.1% (ranging between 85.1% and 95.2%).
ASOs at a final concentration of 750 nM were tested against influenza virus proliferation in MDCK cells at 24 h post-infection using immunofluorescence assays and quantitative RT-PCR (qRT-PCR). 9 of 21 tested ASOs showed inhibition of influenza virus replication (Fig. 6). ASO 727A inhibited virus propagation more than seven-fold. Oligonucleotides targeting the previously described 121 nt motif in (+)RNA5 (2A, 3A, 2B) inhibited    and two negative control oligonucleotides (Neg 5 and N) were tested for cytotoxicity ( Figure S10). None of them caused differences in cell viability at concentration from 0.1 to 4 uM, suggesting that the inhibitory effect of the ASOs on influenza A/California/04/2009 H1N1 replication is not due to affecting cell viability, but is due to the specific targeting segment 5 (+)RNA.

Discussion
The secondary structure of segment 5 (+)RNA based on experimental structure mapping, sequence/structure comparison and free energy minimization revealed structure motifs that are conserved across diverse influenza A strains. These motifs are thermodynamically stable and have high calculated base pair probability. Additionally, the discovered motifs persist at both low and high temperatures. All these results point to these stable and conserved motifs being functionally significant.
The effect of RNA folding on sequence evolution was most apparent in the UTRs. Elsewhere, however, there were specific pairs that showed greater-than-average MI (Supplementary Data 2). For example, the bp between nt 397-435, which occurs in the long hairpin formed between nt 388 and 444 (Fig. 1), has both higher-than-average MI and a high bp probability. Here, the high MI is possible because nt 397 is the first codon position for arginine, which, unlike most first codon positions, can vary from between two residues (A and C), but maintain the codon; nt 435 is the third codon position for threonine and can vary between any base. Thus, the increased flexibility of sequence evolution here, allows for the identification of pairing constraints, on top of the need to maintain codons. These results suggest that, while important, the coding constraints in segment 5 play a dominant role in sequence evolution. Indeed, a previous study of the suppression of synonymous codon use in segment 5 found only moderate suppression across the body of the sequence, but significant suppression toward the regions abutting the UTRs 7 .
The segment 5 (+)RNA construct comprises both cRNA and mRNA; thus the secondary structure studied in vitro highlights properties of both of these functional RNAs. Like all orthomyxoviruses, influenza A vRNA segments are transported from the cytoplasm into the nucleus, where they are used for replication 37 . Viral mRNAs, after being posttranscriptionally matured, are exported back to the cytoplasm for translation 38 . The processes of posttranscriptional maturation and export are both mediated by numerous RNA-protein associations-all of which may be affected by (pre)-mRNA structure. Once in the cytoplasm, additional posttranscriptional control mechanisms work to modulate gene expression. Many of these may also be influenced by RNA folding. The cRNA remains in the nucleus, where it is used to replicate additional vRNAs. In addition to the viral polymerase, cRNA also interacts with multiple copies of NP to form a ribonucleoprotein complex (cRNP). Cryo-electron microscopy, revealed that cRNP exhibits a circular, filamentous double-helical organization with defined termini, containing the viral RNA-dependent RNA polymerase at one end and a loop structure at the other end 39 . It was also noticed that sufficient flexibility exists within cRNP to form secondary RNA structures and favor suggestion of importance of cRNA secondary structure motifs. Recent work on NP associations in the RNP revealed that specific regions of RNA were dynamically interacting with protein 40 . Protein-free regions of cRNP can adopt structures that play roles and, vice versa, cRNA structure can affect NP binding dynamics: thus influencing replication rate and RNA stability, for example.
Our (+)RNA5 secondary structure model has circular folding that recapitulates the cRNA panhandle structure that is likely recognized by the viral polymerase, which distinguishes cRNA from vRNA. Interestingly, this interaction, as well as a downstream hairpin that occurs just before the NP protein start codon, were some of the best-characterized structures in our model: they had the highest mutual information scores of any base-paired regions (due to the lack of coding constraints) as well as high base pair conservation and partition function probability. The ability to identify these interactions in silico, based on probing data on "naked" RNA, lacking all interacting proteins, suggests that these motifs can form independently of the cRNP and form in advance of polymerase binding; this supports the idea that cRNA panhandle structure plays important function in the specificity of recognition by the viral polymerase. In addition, the novel hairpin identified nearby to the panhandle is also likely to play important roles, possibly in mRNA5, in the regulation of translation: e.g., the hairpin is directly upstream of the NP start codon, which is itself embedded within a long hairpin (Fig. 1).
Most revealed (+)RNA5 motifs remain the same, independent of the consideration of preferential local folding and also independent of the influence of the UTRs. Interestingly, only when local folding is constrained the structural motif at the 3′ end (nt 1478-1528) is formed (Supplementary Figure S7); this motif was predicted by two independent bioinformatics analyses and is highly conserved among influenza A strains 7,18 . It is possible that, preferentially local folding or circular folding motifs exist in vivo at certain times of the virus life cycle, and are promoted by viral or/and cellular factors. Hypothetically, changes of structure from circular to linear or linear to circular could take part in the life cycle of influenza virus; however, it likely only induces relatively small changes in global folding.
Local secondary structures of the UTRs and possible structurally labile regions revealed by temperature changes in (+)RNA5 could be functional, for example in mRNA5 (by influence on translation speed or interactions with translational factors) or in cRNA5 on replication (by different interactions with NP or in recognition and replication by the viral polymerase). Interestingly, a hairpin in region 577-593 is formed in both the UTR-deleted and temperature elevated models. This hairpin, with a 7-nt hairpin loop (CGGUGCA; Fig. 1), is shifted to have a 4-nt hairpin loop (CGGU; Supplementary Figure S5). This hairpin at nt 577-593 was also predicted by others 18 . It was noticed by these authors that a deletion of nt 583-585 occurs in an H1N1 strain isolated from Mongolia in 1991 18 . This deletion could influence the 4-nt hairpin loop (Supplementary Figure S5) by destabilizing it; this deletion will not, however, affect the 7 nt hairpin (Fig. 1), which may be an adaptation of certain strains. The sequence of nt 577-593 in the A/California/04/2009 H1N1 strain (GCCGCAGGUGCUGCGGU) also favors the alternative 7 nt hairpin.
A previous bioinformatics analysis showed a possible structured region for coding region 1-160 of (+)RNA5 (region 47-206 using entire (+)RNA5 numbering) 7 . Our experimental data, combined with new analyses, also proposed a secondary structure for this region. Interestingly, this region showed high base-pair conservation. Additionally, there are several compensating mutations for pairs: 5-56, 8-52, 24-38, 88-120, 93-116, 95-114, 64-180, 74-169, 93-116 and 95-114. Beside the (+)RNA panhandle structure, there are two hairpins: one just before the START codon and another that includes the START codon (Fig. 1). The two last nucleotides of the START codon are paired, making the closing base pairs of a long hairpin. Interestingly, for Flaviviruses it was showed that a small hairpin before a START codon influenced interactions with the ribosome, modulating viral mRNA translation rates 41 . The possible involvement of the (+)RNA5 START codon in a hairpin structure (nt 47-104) and preceding small hairpin (nt 16-39) could both potentially regulate translation in influenza.
Microarray mapping results show which regions from the full-length (+)RNA5 are accessible. Binding sites for microarray probes are in agreement with the proposed secondary structure model (Figs 1 and 5); however, not all single stranded regions are oligonucleotide binding sites. It is possible, that some regions could have additional tertiary interactions that inhibit binding (e.g. a 3D structured loop) or alternative intermolecular interactions. Such a pattern of oligonucleotide binding is typical for structured RNA 11,12,31,33 . The accessibility to oligonucleotide binding gives information on regions that are prone to intermolecular interactions, as well as regions that are mostly single stranded or weakly paired. This opens up the possibility for discovering new antiviral targets that can cause misfolding or block active regions of conserved motifs, which could then inhibit virus proliferation.
ASOs were designed on the basis of secondary structure of the entire segment 5 (+)RNA. ASOs (2A, 3A and 2B), that target a 121 nt conserved motif of (+)RNA5 and were previously tested on a modified, single-cycle infectious influenza A virus (sciIAV), inhibited the propagation of a wild influenza virus (Figs 6 and 7). Direct comparison showed that the level of inhibition was smaller for the wild type virus than for the previously studied sciIAV 8 . Better fitness of the wide type virus and/or different experimental conditions could explain this difference. Several new target regions were also revealed in our current study (Figs 6 and 7). The most effective ASO, 727A, targeted a long internal loop (nt 722-732). Another internal loop was bound by the effective ASOs 400A (398-405), 600A (591-608). Additionally, ASOs 615A, 640A targeting hairpin loops (609-621, 628-644, respectively) also decreased propagation of the virus. There was no correlation between the effectiveness of an ASO and its binding site sequence conservation (Table 2 and Figs 6 and 7). All sites were, on average, quite conserved. This suggests the most effective ASOs could have broad applications for targeting various influenza strains.
A region that overlapped (was one nt longer) the binding site of ASO 640A (628-645), was previously targeted using an antisense phosphorothioate-modified oligonucleotide (named NP628) specific to the influenza H5N1 virus (strain A/Tiger/Harbin/01/2002 H5N1) 17 . NP628 showed the best antiviral effect of three other phosphorothioate oligonucleotides tested by these authors. In addition, liposome mediated introduction of NP628 into mice partially protected them from lethal H5N1 virus challenge and resulted in a lower viral titer. The effectiveness of both oligonucleotides, 640A and NP628, are likely due to the accessibility of the (+)RNA secondary structure in this region.
Internal loops or hairpin loops were targeted by ASOs 760A (745-763), 2A (1146-1156) and 3A (1141-1156). Results for ASOs show that two domains are good targets for oligonucleotides -the previously described region formed by nt 1079-1194 8 and the new region spanning nt 554-763. These two domains could take part in RNA-RNA or RNA protein interactions and be important in the influenza virus life cycle.
Oligonucleotide 760A partially targets the same region as published antisense oligonucleotide NP749 targeting strain A/Tiger/Harbin/01/2002 17 . NP749 reduced H5N1 virus titer and NP mRNA level. Both results show that targeting this region of different influenza strains is effective and also that this region is functionally important. We could not compare directly the antiviral effect because different conditions and method of analysis used. It is, however, notable that conserved regions for different strains of H5N1 influenza are good targets for antisense oligonucleotides and their effectiveness is correlated to the revealed RNA secondary structure described here. This model will facilitate future research into the functional analysis of influenza RNA as well as its therapeutic targeting. Oligonucleotides synthesis and labeling. Primers for PCR, qRT-PCR and reverse transcription, DNA oligonucleotides for RNase H cleavage, ASOs (Table 2, Supplementary Tables S1-S3) were synthesized by the phosphoramidite approach on MerMade synthesizer. Primers were deprotected and purified according to published protocols 42,43 . Concentrations of all oligonucleotides were measured using a Spectrophotometer UV (NanoDrop2000 Thermo Scientific). Primers for reverse transcription were synthesized with aminolinker on 5′ end and after deprotection and desalting were labeled from the 5′ end with fluorophores: 5-FAM, 5-ROX, 6-TAMRA, and 6-JOE (dyes from ANASPEC). For labeling the reaction mixture containing 300 µg primer (11 µl in H 2 O) and 220 µg one of diluted in DMSO fluorophores (14 µl) and 75 µl 0.1 M sodium tetraborate, pH 8.5 was incubated overnight in room temperature in shaker oscillating at low speed. Labelled primers were precipitated with ethanol and purified by electrophoresis on 12% denaturing PAGE. RNA synthesis. DNA template for in vitro transcriptions of (+)RNA5 and (+)RNA5-ORF were obtained by PCR from vector pPol1 using primers For_c5, Rev_c5 and FM5, RM5, respectively (Supplementary Table S1). DNA was purified using PCR/DNA Clean-up purification Kit (Eurx). In vitro transcription reaction was performed using an Ampliscribe T7 Flash kit (Epicenter) according to manufacture protocol. Product was purified using RNeasy MiniElute Cleanup Kit (Qiagen). Integrity and purity of samples were checked on agarose gel.  Figure S1). For chemical mapping, the above 100 mM KCl buffer was used and for microarrays mapping and RNase H cleavage the higher salt buffer with 300 mM KCl was used.
Chemical mapping. Prior to chemical mapping RNA was folded as was described above. Chemical mapping was carried out with 6 mM DMS (dimethyl sulfate), 16.5 mM CMCT (1-cyclohexyl-(2-morpholinoethyl) carbodiimide metho-p-toluene sulfonate), 6.7 mM kethoxal or 1 mM NMIA (N-methylisatoic anhydride) for 15, 20, 30 or 40 min, respectively, at selected temperature (23 °C or 37 °C). Control reaction were performed in the same conditions but without mapping reagents. Modified nucleotides were read out by primer extension using a stechiometry of 2 pmol primer/1.5 pmol RNA. Primer extension was performed at 55 °C with reverse transcriptase SuperScript III (Invitrogen) using manufacturer's protocol. Primers labeled by 6-JOE were used for detection of modification by DMS, CMCT, NMIA or kethoxal. Primers labeled by 5-FAM were used for detection of reverse transcription products that serve as control of quality of RNA. For each primers labeled with 5-ROX and 6-TAMRA there were prepared ddNTP ladders: (most often ddGTP and ddATP). cDNA fragments were separated by capillary electrophoresis (DNA Research Center, Poznan).
Processing of chemical mapping data. Data were analyzed with the ShapeFinder program 44 .
Quantitative NMIA reactivities for individual datasets were normalized to a scale in which 0 indicates an unreactive site and the average intensity at highly reactive sites is set to 1.0. The normalization factor for each datasets was determined by first excluding the most reactive 2% of peak intensities and then calculating the average for the next 8% of peak intensities. All reactivities were then divided by this average. In this scale, reactivities ≥0.700 are considered as strong, 0.500-0.700 as medium and <0.500 as weak, but all calculated reactivities were used for prediction of secondary structure. Nucleotides with no data were indicated as −999. Normalized SHAPE reactivities from each primer extension reaction were processed independently. DMS, CMCT, kethoxal modifications analysis was equivalent to that described above, but only strong and medium modification were used in prediction. For all SHAPE and DMS, CMCT, kethoxal mapping analysis at least three datasets were obtained from each primer and the average of results was used in the RNAstructure 5.5 45,46 for prediction of secondary structure.
Prediction of secondary structure using chemical mapping results. For determination of RNA secondary structure RNAstructure5.5 program was used with incorporated experimental data. RNAstructure 45 uses thermodynamic parameters 46,47 and free energy minimization for prediction of base pairing 48 . SHAPE reactivities were converted to pseudo free-energy change terms to restrain predictions. For this purpose the text file with normalized SHAPE reactivity (as described above) was input to RNAstructure 5.5 using "Read SHAPE reactivitypseudo free energy" mode with slope 1.6 and intercept −0.6 kcal/mol 49 . DMS, CMCT, and kethoxal mapping data were also introduced at the same time, using "chemical modification" mode to apply only strong DMS, CMCT and kethoxal modifications as constraints 50,51 . Conservation analysis of model structure. All available influenza A strains were downloaded from the NCBI influenza virus resource 29 , only non-redundant, full-length sequences were taken. Genome alignment of influenza sequences was performed using MAFFT 52  RNase H assays. For RNase H assays RNA was folded as described above. Then, 3 pmol of DNA oligonucleotide in buffer with DTT (final concentration 1 mM) and 5 units of RNase H were added to final volume 20 μl. A control reaction was prepared equivalently but without DNA oligomer. Reactions were incubated for 30 min at 37 °C and then RNase H was inactivated by incubation for 10 min at 65 °C. Samples were precipitated with ethanol and cleavage sites were identified with primer extension using stoichiometry 2 pmol primer/1.5 pmol RNA. Primers labeled by JOE were used for detection of RNA cleavage. Primers labeled by FAM were used for control reactions. For each primers labeled with ROX and TAMRA there were prepared ddNTP ladders: most often ddGTP and ddATP.
Hybridization to Isoenergetic Microarrays. Isoenergetic microarrays were prepared and microarrays mapping was performed similarly to published procedures 31,54 . Microarray probes were complementary to (+) RNA5 step-by-step giving that whole studied RNA was scanned (Supplementary Data 3). Probes were 2′-O-methyl oligonucleotide pentamers and hexamers with incorporated LNA nucleotides and 2,6-diaminopurine riboside (LNA or 2′-O-methylated nucleotide) 32,34 . Probes were spotted in duplicate. Spotting buffer, monomer U, and pentamer UUUUU (2′-O-methyl-RNA), which should show no binding to RNA, were also printed on the microarray as internal negative controls. Silanized slides were coated with 2% agarose activated by NaIO4 and microarrays were printed in the European Center of Bioinformatics and Genomics in Poznan, Poland with NanoPrint microarray printer (Arrayit). Printed microarrays were incubated for 12 h at 37 °C in 50% humidity. The remaining aldehyde groups on microarrays were reduced with 35 mM NaBH 4 solution in phosphate-buffered saline solution and ethanol (3:1 v/v). Then slides were next washed (3x) in water at room temperature, next in 1% SDS solution at 55 °C (1x), and finally in water at room temperature (3x) and dried at room temperature.
About 200,000 cpm of 5′ radioactive labeled RNA (ca. 10 nM) was folded as described above before hybridization to isoenergetic microarrays. Hybridizations were carried out in folding buffer for 18 h at 23 or 37 °C in 100% humidity. Microarrays were washed for 3 min at the same temperature and buffer composition as hybridization occurred and then dried by centrifugation. Hybridization was visualized by exposure to a phosphorimager screen and quantitative analysis was performed with ArrayGaugeV2.1. Binding was considered strong, medium and weak, when the integrated spot intensity was ≥1/3, ≥1/9, or ≥1/27 of the strongest integrated spot intensity, respectively. The average of at least three repeats is presented. Alternative binding sites were predicted using RNA/ RNA thermodynamics.

Antiviral test of ASOs in MDCK cells.
One day before transfection, 2.5 × 10 6 cells were seeded in a 100-mm dish. Before each transfection, Lipofectamine 2000 (Thermo Fisher Scientific) was diluted with the Opti-MEM I Reduced Serum Medium (Opti-MEM) and incubated 10 min at RT. Then 15 µl of this solution was mixed gently with 15 µl ASO in OptiMEM and incubated 30 min at room temperature. To each of prepared ASO solution, 300 µl of 7.5 × 10 4 MDCK cells in media was added. Next, the mixture was divided for 100 µl and added to three wells in a 96-well plate. Final concentration of ASO was 750 nM. Cells were incubated at 37 °C in air with 5% CO 2 and after 12 h medium was changed for fresh one followed by incubation of another 6 h. Transfection efficiency was checked on flow cytometer and optimized using fluorescently labeled (TAMRA) oligonucleotide. Several concentrations were tested: 4 µM, 2 µM, 1 µM, 0.75 µM, 0.5 µM, 0.25 µM and 0.1 µM. The efficiency of 750 nM was the highest and similar as for 1-4 µM concentration and therefore was chosen for further experiments. In each experiment, transfection was checked using labeled NEG5 (ROX) under fluorescence microscope.
After 18 h posttransfection with ASO, 0.01 MOI (multiplicity of infection) of A/California/04/2009 (H1N1) was used to infect MDCK cells. Infection medium consist of PBS supplemented with 0.3% bovine albumin (BA) and 1% PS (penicillin-100 U/ml, streptomycin-100 mg/ml). After 1 h incubation at RT on a rocking platform, the infection medium was changed for the postinfection medium containing DMEM supplemented with 0.3% BA, 1% PSG, and 1 mg/ml tosyl-sulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin (Sigma). Infected MDCK cells were next incubated at 33 °C in air with 5% CO 2 for 24 h. Next, cell culture supernatants (CCS) were collected. Viral titers were analyzed with the immunofocus assay and vRNA amount is measured using qRT-PCR.
Immunofocus assay. Serial dilution of cell culture supernatants were used to infect MDCK cells cultured on 96-well plates. Infection was conducted for 1 h at RT. After incubation, the infection medium was changed for the postinfection medium and infected cells were maintained at 33 °C in air with 5% CO 2 for 8-10 h. Next, 4% paraformaldehyde and 0.5% Triton X-100 in PBS were used to fixed and then permeabilize cells. Cells were incubated with a blocking buffer containing 3% bovine serum albumin in PBS for 1 h. After blocking, the solution was replaced with mouse anti-influenza A monoclonal primary antibody targeting NP protein (MAB8257 Merck Merc Millipore) diluted in the blocking buffer (1 µg/ml) and incubated for 1 h at 37 °C. FITC-conjugated secondary rabbit anti-mouse IgG antibody (AP160F Merck Millipore) diluted in blocking buffer (1:150) was used to detect primary antibody by incubation 1 h at 37 °C. Fluorescent-forming units (FFU/ml) were counted under a fluorescent microscope. The mean titer from triplicates and the standard deviation were calculated with Microsoft Excel software.
Extraction of total RNA from cells. For real-time PCR cells were seeded on 24-well plate. Transfection and infection were conducted as describe above (proportional to 24-well plates format). Following the removal of supernatant, infected cells were washed with PBS and total RNA was extracted from cells using Chomczynski and Sacchi method 57 with TRI Reagent (Sigma-Aldrich). RNA quality was checked on agarose gel. Viral RNA reference standards for qRT-PCR. Viral RNA of segment 7 (vRNA7) was used as a reference standard for the quantitation procedure 58 . To obtain vRNA7 standard first total RNA was extracted from infected MDCK cells. Reverse transcription was carried out using Superscript III reverse transcriptase (Invitrogen) and primer RT (Supplementary Table S3) which is complementary to the 22 nucleotides of the 3′ end of vRNA7. Obtained cDNA was used as a template for PCR reaction with primers carrying T7 RNA polymerase promoter sequence. PCR products were in vitro transcribed using the AmpliScribe T7-Flash Kit (Epicentre Biotechnologies), following manufacturer's protocol. The RNA transcript was purified using the RNeasy MinElute Spin Columns (Qiagen). Quality of RNA sample was verified by electrophoresis on agarose gel and concentration was determined by spectrophotometry. 10-fold serial dilutions of RNA were prepared and reverse transcription with RT primer was carried out. cDNA was then amplified in qPCR reaction to obtain standard curve 59 . qRT-PCR. For cDNA preparation, a separate reverse transcription step was performed using a SuperScript III (Invitrogen) according to manufacture protocol. 500 ng of total cellular RNA serve as temple for reaction and 2 uM of RT primer (Supplementary Table S3) was used. The qPCR step was performed using 5x HOT FIREPol ® Probe qPCR Mix: (Cytogen) on a C1000 Touch TM Thermal Cycler (BioRad) using software version v.3.1.1517.0823 in 96-well format. qPCR reaction components were set-up in triplicate according to the manufacturer's instruction. 1ul cDNA from the RT step and 10 pmol of QF, QR and 10 pmol Taqman probe Q (Supplementary Table S3) were added to reaction Mix to the total volume of 10 ul. Primers were specific to a highly conserved region of the vRNA7. Standard cycling conditions were 95 °C for 15 min, followed by 49 cycles of 95 °C for 20 s, 60 °C for 60 s. A negative water sample was included in each run. The quantity of viral RNA molecules was determined using the vRNA7 standard curve, processed in parallel.
Cytotoxicity assays. MDCK cells were transfected with ASOs in concentration from 1 uM to 4 uM and seeded on 96-well plate as for experiments of antiviral tests described above. After transfection cells were incubated at 37 °C in air with 5% CO 2 and after 12 h medium was changed for fresh one followed by incubation of another 6 h. MDCK cells treated with LPF-2000 were also evaluated. The cytotoxicity of ASOs were evaluated using 1% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (tetrazolium). MDCK cells were incubated for 2 h with tetrazolium at 37 °C in air with 5% CO 2 and after 2 h, solution was changed for DMSO (100 ul/well) and incubated for 15 min. in dark. The rate of formazan formation was determined by measuring the absorbance (570-650 nm) on xMark TM Microplate Spectrophotometer (BioRad). Cells enzymes are capable of reducing the tetrazolium to its insoluble formazan. The 570-650 nm reading value is directly proportional to the number of living cells. Cytotoxicity assays were normalized to viability of cells treated only with Lipofectamine 2000. The mean and standard deviation were calculated with Microsoft Excel software from three independent experiments each containing three technical repeats.