Stability and translation are essential for HCV replication

To further disentangle the contribution of miR-122 on the processes of translation and/or stability/replication, a bicistronic luciferase reporter construct with two copies of the HCV 5′UTR was designed (Bi-Luc-SG, Fig. 2a). The first cistron consisted of HCV 5′ UTR (UTR1) harboring the cis-acting elements essential for RNA synthesis fused to a PV IRES allowing miR-122-independent translation of the Nano luciferase (NLuc) gene. Translation of the second cistron comprising a firefly luciferase (FLuc) fused to the HCV NS proteins NS3-5B was initiated by the HCV IRES contained in the second copy of the HCV 5′UTR (UTR2). Since translation of the replicase proteins NS3-5B is a prerequisite for RNA replication, UTR1 and UTR2 thereby allowed studying the relative contribution of miR-122 to genome stability/replication and translation, respectively, on genome replication.

Open in a separate window

Fig. 2

Dissection of stability and replication using a dual-luciferase replicon.a Illustration of the Bi-Luc-SG construct. miR-122 binding to UTR1 acts on RNA stability and replication (as measured by Nano luciferase activity), translation of non-structural proteins is driven by UTR2 (as measured by firefly luciferase activity). b Northern blot analysis of HCV RNA stability over 72h. Stability of different Bi-Luc construct RNAs was analyzed in the presence of miR^mut or miR^WT. Band #1: full-length RNA, #2: truncated product. U:A mutations in SLI of UTR2 (red dots) prevented the formation of truncated RNA. c Illustration of miR^WT (green) and miR^mut (black) binding upon introduction of a matching point mutation in the target (underlined). d, e Translation assay of point mutants cloned into the MBR of UTR1 or 2. The activity of Nano (d) and firefly (e) luciferase was measured in the presence of miR^mut or miR^WT. f Assessment of the full intracellular replication cycle of the Bi-Luc-SG constructs after electroporation with miR^mut or miR^WT. Replication was monitored with the nano luciferase signal. Mean values (±SD), n=3, duplicates. RLU relative light units, ΔGDD replication deficient mutant. For translation, statistical significance was determined for miR^WT against miR^mut, for replication miR^WT and miR^mut conditions were tested against ΔGDD. *P<0.05, **P<0.01, ***P<0.001

Northern blot analysis revealed that two identical copies of the 5′UTR yielded truncated fragments at late stages of the replication cycle, likely due to sequence recombination (Fig. 2b, band #2). Therefore, the stem sequence in SLI of UTR2 was mutated from C:G to U:A, inactivating replication functions of UTR2³. This mutation prevented recombination without significantly affecting the replication efficiency of the bicistronic replicon (Fig. 2b).

To assess the individual contribution of miR-122 to HCV stability/replication and translation, we mutated UTR1 and UTR2 individually in the Bi-Luc-SG construct (Fig. 2a) by introducing a point mutation into the seed sequence of both miR-122-binding sites, complementary to the change in miR^mut (Fig. 2c). Therefore, supplementation of miR^WT or miR^mut would only rescue miR-122-dependent functions in the respective matching UTR1 or 2 variants. Indeed, addition of either miR variant stimulated luciferase activities to varying extent early after transfection, either due to enhanced translation efficiency or RNA stability (Fig. 2d). For further control we deleted UTR1 and replaced UTR2 by an EMCV IRES to exclude crosstalk between UTR1 and UTR2 (Supplementary Fig. ²). These constructs revealed very similar results concerning miR-122 dependency as expected, apart from undetectable stimulation of translation by miR-122 upon deletion of UTR1, suggesting that in short-term experiments, the stabilizing effect of miR-122, preventing degradation by Xrn1, seems to dominate in these constructs, probably due to their excessive length. However, none of the replicons harboring a mutant UTR sequence replicated to significant levels in the presence of miR^WT or miR^mut (Fig. 2e), demonstrating that both functions of miR-122 in stability and translation substantially contributed to efficient HCV replication. We further could rescue replication of a Bi-Luc-SG replicon harboring two mutant UTRs with miR^mut (Supplementary Fig. ^3a), as well as constructs with one mutant UTR by simultaneous addition of miR^WT and miR^mut (Supplementary Fig. ^3b), albeit with very low efficieny. This was likely due to a generally lower efficiency of miR^mut even in the context of a monocistronic replicon (Supplementary Fig. ^3c) and a reduced stimulation by miR^WT upon co-application of miR^mut in case of UTR1+2^WT Bi-Luc-SG (Supplementary Fig. ^3d).

SLII is affected by alternative conformations

After establishing that miR122-driven translation stimulation is important to HCV replication, we investigated the underlying mechanism. Since the MBR itself is dispensable for IRES activity², the effect of miR-122 binding on the target sequence must be relayed from DI (nt1-47, Fig. 1a) to the downstream sequence to positively modulate translation. The current structural paradigm proposes a relaxed state of DI except for the stable SLI stem-loop. In contrast, the complementary region representing the 3′ end of the viral negative strand RNA (3′(-)) is engaged in a conserved stem-loop (SLIIz′, Fig. 3b), which is a cis-acting replication element (CRE) crucial for RNA amplification³. Since complementarity of sequence also favors similar structures (except for U:G rich sequences), a structure similar to SLIIz′ might also form in the positive strand 5′ UTR, which would consequently prevent correct IRES formation. As the MBR is part of this alternative structure, miR-122 binding to this sequence could prevent misfolding of the IRES. Based on this hypothesis in vitro folding experiments of HCV RNA and in silico prediction of the conformation of DI and II were performed. Selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE) reactivity data was used as constraint to assist structure prediction by RNA structure (Fig. 3c). In line with our hypothesis, the algorithm predicted an alternative stem-loop, encompassing nucleotides 21–105 as energetically most favorable option, from here on referred to as SLII^alt (Fig. 3c), which partly resembled SLIIz′ in the negative strand RNA. Importantly, SLII^alt also included parts of SLII, preventing the formation of a functional IRES. Alternative labeling by DMS and NAz produced both a pattern matching the prediction obtained with NAI (Supplementary Fig. ^{4a, b}).

Open in a separate window

Fig. 3

Structural assessment of the HCV 5′ UTR.a Schematic overview of the HCV 5′ UTR and IRES. The miR-122-binding sites are indicated in green. Domain I and II are indicated above (DI and DII). b Schematic overview of the 3′ UTR of the negative strand RNA, which is complementary to the 5′UTR of the positive strand. Note the divergence in secondary structures, which is due to the different function in replication (3′(-)-strand) and translation (5′(+)-strand). c Representative autoradiography image showing the in vitro SHAPE reactivity of nucleotides 1–120 of a HCV reporter replicon. The SHAPE data were used to predict the secondary structure (minimum free energy, MFE) of DI and II with the RNA structure web tool and were visualized with R2R. The nucleotides of SLII^alt are colored according to their respective normalized reactivity. Yellow and red colors indicate unpaired nucleotides. The SHAPE data are also shown in context of the canonical fold of DI+II. d, e RNA conformation of DI+II (nt 1–120) or DII alone (nt 40–120), monitored by 1D ¹H NMR spectra of imino protons in a temperature gradient. f Top conformations calculated by RNA structure and their respective free energy values (−ΔG[kcal/mol]) for DI+II or DII alone

To gain deeper insights into the impact of DI on the folding of SLII, we subjected fragments of DI+II (nt 1–120) or DII (nt 40–120) to 1D-¹H NMR analysis. While the 1D NMR spectra of the DII imino protons were not significantly influenced by a decrease of experimental temperature (Fig. 3e), we observed line broadening of the imino proton resonances of the DI+II fragment at lower temperatures (Fig. 3d). This could indicate exchange between several conformers in solution at lower temperatures, or aggregates of RNA. To differentiate between these options, we performed size-exclusion chromatography combined with online detection by multi-angle laser light scattering (MALLS) and refractometry. The DI+II fragment was present as a monomer or dimer, both binding a copy of miR-122 not affecting their ratio (Supplementary Fig. ^{5a, b}), whereas DII was exclusively monomeric and did not bind miR-122, as expected (Supplementary Fig. ^5c). However, native gel electrophoresis at 4°C with DI+II displayed only a single band, irrespective of the preincubation temperature (Supplementary Fig. ^5d), further arguing against oligomerization. In addition, in silico folding analysis predicted six structures within the top 10% of the minimal free energy structure for the DI+II fragment, in contrast to only three for the DII fragment (Fig. 3f), suggesting that the presence of DI increased the variety of alternative conformations of the 5′ terminus of the HCV RNA. Some of the energetically most favorable interfered with the formation of SLII.

We reasoned that binding of miR-122 would prevent the formation of alternative RNA structures in the positive strand RNA, and thereby support establishment of the IRES. To experimentally validate this assumption, we tried to assess the modulation of structures in the HCV 5′UTR by miR-122 using SHAPE. In vitro, miR-122 binding by annealing was efficient, indicated by an almost complete loss of SHAPE signal in the MBR, whereas, only moderate changes in reactivity were observed in the rest of the UTR (Supplementary Fig. ^6a). Since in vitro binding of miR-122 might not reflect the situation in cellulo, we performed in cell SHAPE, which failed to detect any signal (Supplementary Fig. ^4c). This result was in line with the previous data, identifying multiple secondary structures throughout the HCV genome, but not in the 5′UTR¹⁸. In addition, not many changes can be expected in the SHAPE signal, since most single stranded regions that are accessible by SHAPE in SLII^alt and SLII are fully or partially overlapping (e.g. nt 53–60, 71–73, 80–86, 93–97, 105–108, Fig. 3b), with only a few exceptions (e.g. 63–64). Moreover, phylogenic analysis suggests that, due the high conservation of the primary sequence across all HCV genotypes, the sequence of the 5′ UTR is compatible with SLII^alt folding. However, due to the few differences in primary sequence, which are also clustered within the bulge region of SLII^alt, we were not able to detect significant covariance to ascertain evolutionary conservation of this structure (Supplementary Fig. ^6b). However, several compatible mutations were detected in the corresponding complementary 3′ negative strand structure SLIIz′, rather arguing for selective pressure to keep this structure intact (Supplementary Fig. ^6c).

In conclusion, our data suggest that DI, being essential for functions in the viral negative strand, interferes with the folding of the functional HCV IRES, thereby impairing genome translation, which is prevented by miR-122 binding. However, we were not able to directly detect the influence of miR-122 on downstream structures with the biochemical measures we implemented so far.

Mutations in DI favoring SLII formation stimulate translation

The previous results suggested a negative influence of DI on the conformation of the IRES, which can be prevented by miR-122 binding. To test this assumption, we designed a panel of DI mutants, all localized outside SLII and the IRES, which should either support or prevent the formation of SLII, according to secondary structure predictions (Fig. 4a, b). SLII formation was favored either by deleting parts of DI (Δ10, Δ15, Δ20), by replacing 5 or 10 positions with A (pA5, pA10) or by introducing mutations engaging DI into compensatory stem-loop structures (C, SL). As control, one mutant was designed, which was not able to fold stably into either SLII^alt or SLII (M25), and another one was predicted to stabilize an SLIIz′-like fold (inhibitory stem-loop, ISL), preventing the formation of SLII. As expected, all mutants were severely impaired in replication or did not replicate at all, most likely due to the destruction of essential parts of SLIIz′ in the negative strand (Supplementary Fig. ^7a, Fig. 3b), therefore we focused our analysis on translation efficiency and stability (Fig. 4c, S​S33b–d).

Open in a separate window

Fig. 4

DI mutants and their influence on translation using the Luc-SG replicon.a Illustration of the HCV 5′ UTR and IRES and location of the inserted mutations (orange). b Sequence alterations of DI mutants compared to WT. The miR-122 seed binding sites are marked by green boxes. Note that all mutants lack wild type miR-122 binding, due to mutations in the seed sequence. c Translation assays in Hep3B cells, using firefly luciferase (FLuc) reporter replicons and a capped Renilla (RLuc) control. miR^mut or miR^WT were co-electroporated, and luciferase activity was measured after 1, 2 and 4h. All constructs were replication deficient ΔGDD mutants, to exclude potential effects of early replication. Mutations are described in the “mut” row, and a pictogram of the predicted minimum free energy structure in the “folding” row. Mutants enhancing SLII formation are depicted in green, SLII^alt stabilizing in red. Mean values (±SD), n=3, duplicates. RLU relative light units. Statistical significance of the difference between WT and DI mutants in the presence of miR^mut is indicated. The reference graph used to calculate is given in each subpanel in light gray. *P<0.05, **P<0.01, ***P<0.001

Indeed, translation efficiency as measured by firefly luciferase activity, was strongly dependent on miR-122 for the WT UTR, whereas none of the mutants was influenced by miR-122, due to the lack of binding sites (Fig. 4b, c; Supplementary Fig. ^7b). Still, all mutants designed to favor the SLII formation translated comparable to WT in the presence of miR-122 (Δ10, pA5, pA10, C, SL) or even substantially more (Δ15, Δ20) at early time points (Fig. 4c). In contrast, translation of the M25 was comparable to WT in the absence of miR-122 and could not be induced by miR-122, due to the absence of binding sites. Interestingly, the ISL mutant exhibited strongly decreased IRES-mediated translation far below the basal WT levels, in line with the predicted secondary structure, favoring SLII^alt formation (Fig. 4c).

Subsequently, we assessed the impact of the mutations on RNA stability and structure for a selected set of mutants and chose Δ20 (strongest stimulation), C (stimulation comparable to WT with miR^WT), and ISL (strong decrease in translation efficiency). We assessed short-term RNA decay kinetics by northern blot and found no correlation between stability and translation efficiency (Supplementary Fig. ^7b–d). SHAPE analysis was performed to validate the successful generation of the desired structures by mutations in DI. Due to the limitations of SHAPE analysis discussed above, Δ20 and C did not show substantial differences compared to the WT. However, the ISL mutation, which was predicted to create a perfect mirror image of SLIIz′, indeed substantially changed the SHAPE pattern (Supplementary Fig. ^8a–c).

To exclude inadvertent effects related to the changes in primary sequence of the MBR, we aimed to introduce mutations in SLII affecting SLII^alt and chose residues that are not in contact with the ribosome¹⁹ (SLII^alt mut, C65G+G102C, Supplementary Fig. ^9a). Indeed, SLII^alt destabilization lead to a miR-122 independent ~2-fold increase in early translation (Supplementary Fig. ^{9c, d}). However, since these mutations were predicted to affect SLIIz′ folding (Supplementary Fig. ^9e), they consequently abrogated replication (Supplementary Fig. ^9f). Importantly, introducing compensatory mutations predicted to restore SLII^alt formation (SLII^alt rev, C49G, Supplementary Fig. ^9b) again reduced translation efficiency (Supplementary Fig. ^{9c, d}), but partially restored replication (Supplementary Fig. ^9f).

Conclusively, mutations preventing or favoring alternative folds within DI and II substantially increased or decreased translation efficiency of the HCV IRES, independent from miR-122 binding. Our results demonstrate that the primary sequence of DI indeed negatively influences the activity of the IRES element by disturbing the fold of SLII, due to the essential role of the complementary sequence in RNA replication.

miR-122-independent translation facilitates HCV replication

To analyze how differences in translation efficiency exerted by DI mutations impacted overall replication efficiency, we used the Bi-Luc-SG construct (Fig. 5a). We compared a WT version of UTR1 and UTR2, keeping stability/replication and translation dependent on miR-122 with mutants harboring a UTR2 mutant either stimulating (Δ20, C) or reducing (ISL) translation efficiency independent from miR-122. NLuc activity at early time points after transfection was very similar and enhanced by miR-122 binding for all variants, reflecting increased RNA stability (Fig. 5b). FLuc activity was strongly stimulated by miR-122 binding in case of WT (Fig. 5c), whereas only a slight increase in FLuc acitivity was observed for the mutants due to the stabilizing effect of miR-122 upon binding to UTR1. Mutants Δ20 and C showed FLuc levels comparable to WT in the presence and absence of miR-122, respectively, whereas FLuc levels of ISL were strongly reduced, in line with the previously demonstrated impact of these mutations on the folding of SLII. At later time points, no replication was observed for any construct in the absence of functional miR-122 (Fig. 5d, miR^mut), again demonstrating the important contribution of miR-122 to stability and replication of HCV genome. Yet, the mutants, all incapable of miR-122 binding, showed a divergent behavior: While the ISL mutant, impairing SLII formation, remained replication deficient even in the presence of miR-122, mutations Δ20 and C restored viral replication due to their capability to efficiently support SLII folding (Fig. 5c).

Open in a separate window

Fig. 5

Translation stimulation by DI mutations in the Bi-Luc context.a Schematic representation of the dual-luciferase construct (Bi-Luc-SG) and insertion region of the mutations (black box). The in silico predicted MFE structures of the inserted mutants are given in the table on the right. Mutants enhancing SLII formation are depicted in green. b, c Translation assay of selected mutants cloned into UTR2. The activity of Nano and firefly luciferase was measured in presence of miR^mut or miR^WT. d Assessment of the full intracellular replication cycle of the dual luciferase constructs after electroporation with miR^mut or miR^WT. Replication was monitored via the Nano luciferase signal. Mean values (±SD), n=3, in technical duplicates. RLU relative light units, ΔGDD replication-deficient mutant. *P<0.05, **P<0.01, ***P<0.001

These data again emphasized the contribution of translation stimulation by miR-122 to HCV replication and provided further experimental evidence suggesting that mir-122 binding suppresses unfavorable folding of SLII.

DI interferes with 80S ribosome assembly

In the HCV IRES, SLII drives the assembly of the 48S pre-initiation complex with the 60S ribosome²⁰ (Fig. 6a). Hence, we hypothesized that introduction of mutations like Δ20, as well as miR-binding should enhance 80S monosome complex assembly, if these conditions favor SLII formation. Conversely, the ISL mutant should impair this transition.

Open in a separate window

Fig. 6

Ribosome profiling.a Schematic representation of translation initiation and assembly of the 80S ribosome. The 48S preinitiation complex is constituted of the 40S ribosomal subunit, eIF3 and the (eIF2)–GTP-tRNAP^MetP. Hydrolysis of GTP leads to release of eIF2. The L-shape of SLII is crucial for this process. Hence, the ISL mutant, misfolded WT and GMP-PNP, inhibiting GTPase activity, stall the reaction at this stage (bar-headed line). In contrast, miR-bound WT and Δ20 should activate eIF2-release, since SLII is effectively formed (green arrow). Subsequently, eIF3 is released and the 60S subunit is recruited to form the 80S complex. The following transition from initiation to elongation can be inhibited by CHX (bar-headed line). b Monitoring translation by luciferase assay in HeLa cell extracts, using Luc-SG in presence of miR^mut or miR^WT, DI mutants, GMP-PNP or CHX. c Analysis of HCV RNA distribution in ribosome profiles of HeLa cell extracts. Total lysates incubated with HCV in vitro transcripts were recorded; Relative abundance of HCV RNA from sucrose fractions was analyzed by qRT-PCR. Represented are percentages relative to the total amount in the gradient. The assay was performed with either no inhibitor, GMP-PNP or CHX in the reaction mix. The HCV RNA profile for the control reactions is shown in the top panel. Total RNA was analyzed by denaturing agarose gel electrophoresis (lower panel) to confirm the presence of 18S and 28S ribosomal RNA and distinguish precisely 40/48S and 80S ribosomes. Note that in fraction 4 and 9 the CHX data points are obscured by the GMP data. d Comparison of translation initiation efficiency of Luc-SG in presence of miR^WT. The control with miR^mut is shown as dashed line for comparison. e, f As in (d), comparing the WT reporter replicon and DI mutants. Mean values (±SD), n=3, in technical duplicates. RLU relative light units. Significance was determined compared to the miR^mut control. The reference graph used to calculate is given in light gray. *P<0.05, **P<0.01, ***P<0.001. 40/48S and 80S containing fractions are highlighted by yellow or orange boxes, respectively. Note that results shown in panels (d–f) were obtained in absence of CHX

To address, whether changes in DI had the propensity to modulate SLII function, 80S ribosome assembly was analyzed by polysome profiling. In vitro transcripts of Luc SG replicons (WT, D20, and ISL) were translated in vitro using miR-122-deficient HeLa cell extracts. Replicon translation efficiency was assessed by measuring firefly luciferase activity and showed comparable levels with those obtained in Hep3B cells (Fig. 6b). In vitro translation extracts were analyzed by sucrose density gradient centrifugation. Polysome profiles were continuously recorded and fractions collected based on the elution time (Supplementary Fig. ^10a). Distributions of HCV RNA genomes were quantified in each fraction by qRT-PCR. Distribution of ribosomal RNAs was determined by agarose gel electrophoresis and allowed to distinguish fractions containing RNA associated with the 40S ribosome (18S rRNA) from those containing RNA associated with the 80S ribosome, i.e. engaged in translation. The fractions containing 40/48S and 80S ribosomes were further distinguished by adding 5′-Guanylyl imidodiphosphate (GMP-PNP) or cycloheximide (CHX), respectively (Fig. 6c). GMP-PNP prevents 60S recruitment by inhibition of eIF2 release, whereas CHX stalls the 80S complex via abrogation of the peptidyltransferase reaction²¹. Accordingly, the assembly of 80S ribosomes, indicated by the presence of 28S rRNA in higher density fractions, was impaired by GMP-PNP, but not CHX (Fig. 6c).

As predicted, addition of miR-122 to the WT replicon resulted in less HCV RNA present in the 40/48S fraction, concomitant with a higher amount in the 80S fraction, indicating a more efficient association with the 80S ribosome (Fig. 6d). In line with this observation, a 40S ribosomal subunit marker (rps3), was detected in higher density fractions after addition of miR-122, corroborating the results of the RNA gels (Supplementary Fig. ^10b). The shift in 28S/18S rRNA, 40/48S:80S ratio and rps3 was even more pronounced for the Δ20 construct (Fig. 6e, Supplementary Fig ^10b). In contrast, the ISL mutant was markedly impaired in its ability to initiate translation, since most viral RNA was retained at the 40/48S stage and only minute amounts of HCV RNA were detectable in higher density fractions (Fig. 6f). Expectedly, the distribution of eIF3, an essential translation initiation factor binding to DIII, was not altered in the mutants (Supplementary Fig. ^10b). In contrast, rps3 was detected in the 80S ribosomal fraction (28S/18S rRNA, Supplementary Fig. ^10b). These experiments were confirmed in presence of CHX, to stabilize transiently formed 80S subunits and prevent loss of signal due to polysome formation (Supplementary Fig. ^10c). However, no significant differences were detected, perhaps due to the short translation time used in this experiment.

Overall, these results suggest that mir-122 binding stimulates the formation of SLII thereby favoring the assembly of the 80S ribosome and translation initiation. In addition, translation enhancement after miR-binding is mimicked by the SLII-forming mutant Δ20 but disabled in the ISL mutant.

Generation of Huh7.5^ΔmiR cells by lentiviral transduction

Lentiviruses were produced in HEK293T cells by CaPO₄ transfection with pCMV-ΔR8.31 (HIV gag-pol), pMD2.G (VSV-G), and the pLenti-CRISPR-miR-122 vector in a 3:1:3 ratio, respectively. pCMV-ΔR8.31 and pMD.2G and were kindly provided by Prof. Didier Trono (University of Lausanne, Switzerland). Cell-free supernatants were harvested 48, 56 and 72h post transfection and used for transduction of Huh7.5 cells. Transduced cells were selected by supplementing the culture medium with 2μg/ml puromycin. Clones were produced by seeding single cells in a 96-well plate, using a limiting dilution approach.

Reporter viruses and plasmids

FL HCV luciferase reporter viruses (Luc-FL, PI-Luc-FL) were based on JcR2a³⁸. To achieve better signal at early timepoints, the Renilla luciferase was exchanged for a smaller and brighter Nano luciferase (Promega). pCMV-ΔR8.31 (HIV gag-pol) and pMD2.G (VSV-G) were used as packaging plasmids for lentivirus production. pLenti-CRISPR-miR-122 was produced by cloning a duplex of miR-122-specific sgRNA oligos (gRNA miR-122 sense oligo: CACCGAGTTTCCTTAGCAGAGCTG, gRNA miR-122 antisense oligo: AAACCAGCTCTGCTAAGGAAACTC), designed with E-CRISP (www.e-crisp.de) into the pLenti-CRISPR plasmid (Addgene #52961) via BsmBI (NEB). pFK-I₃₈₉Luc-ubi/NS3-3′/JFH1 (Luc-SG) was described before³⁹. All mutants within the HCV 5′ UTR were generated by PCR-based mutagenesis and inserted with SbfI and XbaI (NEB). pFK-I₃₄₁PILuc-ubi/NS3-3′/JFH1 (PI-Luc-SG) has been used in a previous study⁴⁰. It was generated by replacing the HCV IRES of pFK-I₃₈₉Luc-ubi/NS3-3′/JFH1 with the HCV 5′ UTR-Polio IRES cassette from pFK-I₃₄₁PILuc/NS3-3′/JFH1⁴¹ using SbfI and XbaI (NEB). pRL-CMV (GenBank AF025843) was used for in vitro transcription of Renilla luciferase mRNA, as spike-in template for translation assays. pUC18 vectors were used for transcription of HCV 5′ UTR truncations (1–120nt, 40–120nt) for NMR analysis. The fragments were amplified from pFK-I₃₈₉Luc-ubi/NS3-3′/JFH1 plasmids with primers harboring SbfI and SmaI sites.

pFK-I₃₄₁PINLuc/I₃₈₉FLuc/NS3-3′/JFH1 (Bi-Luc-SG) was generated by insertion of a fusion PCR product of the I₃₄₁PI cassette from pFK-I₃₄₁PILuc/NS3-3′/JFH1 and the Nano Luciferase gene from pNL1.1.TK[Nluc/TK] Vector (Promega) into pFK-I₃₈₉Luc-ubi/NS3-3′/JFH1 before the HCV 5′ UTR. The HCV 5′ UTRs of the first and second cistrons were named UTR1 and UTR2, respectively. All mutants were generated by PCR-based mutagenesis and inserted with SbfI and AflII for UTR1, or AflII and XbaI (NEB) for UTR2. pFK-JcN2a (Luc-FL) was generated from pFK-JcR2a by amplifying the Nano luciferase gene from pNL1.1.TK[Nluc/TK] Vector (Promega) and adding the sequence of the 2A peptide. The NLUC-2A fragment was inserted into pFK-JcR2a with AscI and BsiWI (NEB).

pFK-JcPIN2A (PI-Luc-FL) was generated by replacing the I₃₈₉NLuc HCV IRES fragment from Luc-FL with the I₃₄₁-PI-NLuc cassette from pFK-I₃₄₁PINLuc/I₃₈₉FLuc/NS3-3′/JFH1 with SbfI and BsiWI (NEB).

Oligonucleotides

miR mimics were purchased at Eurofins MWG. Sequences of hsa-miR-122-5p (guide strand) and hsa-miR-122-3p (passenger strand) were acquired from miRbase (www.mirbase.org). However, an additional 3 U was added, since this sequence is the most abundant variant in hepatocytes⁹ (hsa-miR-122-5p WT: UGGAGUGUGACAAUGGUGUUUGU, hsa-miR-122-3p WT: AAACGCCAUUAUCACACUAAAUA). For the inactive mutant miR-122^mut, A4 was changed to U, and the complementary exchange was performed on the passenger strand (hsa-miR-122-5p mut: UGGAGUGUGACAAUGGUGUUUGU, hsa-miR-122-3p mut: AAACGCCAUUAUCACACUAAAUA).

Stem-loop qRT-PCR for miR-quantification

A previously described qPCR detection method for mature miRNA was adapted, using SYBR green instead of TaqMan probes for detection⁴². Total RNA was extracted from cells with TRIZOL (Thermo Fisher Scientific), according to the manufacturer’s instructions. cDNA for miR-122-5p and miR-23b-5p was generated in a single reaction with specific RT primers (hsa-miR-122-5p RT primer: GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCAAACA, hsa-miR-23b-5p RT primer: GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGGTAAT). qPCR primers (50nM, Stem-loop universal reverse qPCR primer: ATCCAGTGCAGGGTCCGAGG, hsa-miR-122-5p forward qPCR primer: GCGGCGGTGGAGTGTGACAATG, hsa-miR-23b-5p forward qPCR primer: GCGGCGGATCACATTGCCAGGG) and dNTPs (0.5mM) were preheated to 65°C and flash-cooled on ice. 1× FirstStrand buffer (Thermo Fisher Scientific), 10mM DTT, 1U/µl RNAsin (Promega), and 2.5U/µl SuperScript III (Thermo Fisher Scientific), and 0.5µg of total RNA were added and samples were transferred into a TPersonal PCR cycler (Biometra) and cDNA synthesis was performed with a pulsed program: 30min 16°C, cycle 60× (30°C 30s, 42°C 30s, 50°C 1s), 85°C 5min and then cooled to 4°C. cDNA was diluted 1:20 and 3µl were used for qPCR. Analysis was performed with the 2× iTaq Universal SYBR green supermix (Bio-Rad) on a CFX96 Touch real-time PCR detection system (Bio-Rad) with the following program: 95°C for 3min, 95°C for 10s, and 60°C for 30s. miR-23b was used as an internal reference, based on previous findings⁴³. The 2^−CT values of RT negative controls were subtracted for each sample. Relative gene expression was determined using the threshold cycle (2^ΔΔCT) method⁴⁴.

In vitro transcription

Sub-genomic reporter replicons: RNA for electroporation was generated from 5µg of pFK-plasmid (linearized with MluI (NEB)) in a 100µl reaction mix containing 20µl of 5× rabbit reticulocyte lysate buffer (400mM Hepes (pH 7.5), 60mM MgCl₂, 10mM spermidin, 200mM DTT), 12.5µl 25mM NTP-solution, 2.5µl RNasin (40U/µl), 0.1U pyrophosphatase (Sigma Aldrich), and 6µl of T7 polymerase. The reactions were performed over night at 37°C. Subsequently, the input DNA was removed using RQ1 RNase-free Dnase (Promega). The RNA was purified by acid phenol:chloroform extraction and isopropanol precipitation. The pellet was washed with 70% ethanol and resuspended in RNAse free water.

Capped Renilla mRNA for translation experiments: For capped Renilla transcripts, the pRL-CMV vector (Promega) was linearized with BamHI (NEB). The setup of the transcription reaction was as stated above, except 12.5mM m7G analog (NEB) were added to the reaction, and the GTP concentration in the rNTP stock was reduced to 12.5mM.

Truncations of the 5′ UTR for NMR investigations

5′ UTR RNA constructs (1–120 nt (DI+II), 40–120 nt (DII)) for NMR studies were transcribed from pUC18 vectors, linearized using SmaI (NEB). 100µg of linearized DNA plasmids were used for 1ml run-off transcription reaction. The reaction mix containing 4mM NTP, 16mM MgCl₂, 34µg of BSA, 3.6U of pyrophosphatase from Becker Yeast, 125µg of T7 RNA polymerase, 5mM DTT, 1mM spermidine, 0.01% Triton X-100, 80mg of PEG 8000, and 40mM Tris at pH 8, was incubated at 37°C for 3h in a Hybrigene HB-3D (Techne) at 20rpm. Then, 50 units of RNAse-free DNAse (Euromedex) were added and incubated 30min at 37°C to stop the reaction. Reaction volumes were 1, 4, or 8ml and RNA constructs were purified using Q-sepharose Hiload column (GE Healthcare) on DuoFlow (Biorad) in denaturing condition. Briefly, the reaction mixture was diluted 10-fold in 20mM Tris buffer at pH 8, containing 10mM EDTA and 8M urea, and loaded on the column previously equilibrated with the same buffer. Then RNA was purified with a gradient from 0 to 1M NaCl in the same buffer. Fractions containing RNA (identified using absorbance at 260nm) were controlled using denaturing polyacrylamide (15%) gel electrophoresis. The fractions with the target transcript were pooled, dialyzed three times against 200 volumes of H₂O, and lyophilized.

The RNA samples for NMR studies were resuspended in water, heated at 95°C for 5min, and were rapidly cooled by the addition of ice-cold 2× NMR buffer. NMR buffer composition was 50mM Hepes at pH 6.4, 100mM NaCl, 5mM MgCl₂ and 5% D₂O. The final RNA concentration for each NMR sample was 0.9mM. All NMR experiments were recorded using a 950MHz Bruker Avance III HD spectrometer equipped with a 5mM cryogenically cooled pulsed-field-gradient triple-resonance probe.

SEC–MALLS–RI

Size exclusion chromatography (SEC) combined with online detection by MALLS and refractometry (refractive index—RI) is a method allowing the measurement of the absolute molecular mass of each component of a sample in solution, regardless of its dimensions and shape⁴⁵. SEC was performed with a Superdex 200 column (GE Healthcare) equilibrated with 50mM Hepes buffer (pH6.4) containing 100mM NaCl and 5mM MgCl₂. Fifty microliters of a DI+II and DII solution at 0.77 and 0.8mM, respectively, was injected with or without 1.2 equivalents of miR-122. Separations were performed at room temperature with a flow rate of 0.5ml/min. Online MALLS detection was performed with a DAWN-HELEOS II detector (Wyatt Technology Corp.) using a laser emitting at 690nm, and RNA concentration was measured online by measuring the differential RI using an Optilab T-rEX detector (Wyatt Technology Corp.) and an averaged RI increment, dn/dc, of 0.180ml/g. Weight-averaged molar masses (Mw) were calculated using the ASTRA software (Wyatt Technology Corp.).

Native gel analysis of RNA fragments

15 pmol of DI+II RNA fragment were denatured at 95°C in 15µl of H₂O and flash cooled on ice. 8µl of 3× SHAPE folding buffer (333mM Hepes pH7.5, 333mM NaCl, 20mM MgCl₂) were added and reactions were incubated for 10min at the desired temperature. Samples were put on ice to terminate the folding reaction and split into two tubes. One was supplemented with 3µl of native loading dye (60mM KCl, 10mM Tris pH 7.6, 50% w/v glycerol, 0.01% w/v xylene cyanol, 0.01% w/v bromophenol blue), the other with 12µl LDII (95% formamide, 0.025% SDS, 15mM EDTA, 0.025% w/v xylene cyanol, 0.025% w/v bromophenol blue) as denaturing control.

10% polyacrylamide gels were pre-run for 10min at running conditions. Native TBM (44mM Tris–borate, pH 8.3, 1mM MgCl₂) gels were run for 1h at 100V at 4°C, denaturing urea TBE (89mM Tris, 89mM boric acid, 2mM EDTA, 8M urea) gels 1h, 100V at room temperature. Gels were stained with a 1:10,000 dilution of diamond nucleic acid dye (Promega) in the respective running buffer for 10min at room temperature and visualized on a UV transilluminator (INTAS).

Electroporation of cells

3*10⁶ Hep3B or Huh7.5 cells were resuspended in a total volume of 200µl in “Cytomix” (120mM KCl, 0.15mM CaCl₂, 10mM K₂HPO₄/KH₂PO₄ (pH 7.6), 25mM Hepes (Gibco), 2mM EGTA, 5mM MgCl₂, 5mM glutathione, 2mM ATP) and electroporated (0.166kV, 950µF, 0.2cm cuvettes) with 5µg of reporter replicon RNA and 50pmol wild type or mutant miR-122 mimic. For translation assays, 5µg of a capped Renilla transcript were added as internal control for transfection efficiency. Electroporated cells were seeded onto six-well plates and incubated at 37°C for the desired time.

Luciferase reporter assay

Activity of reporter replicons was measured after cell lysis in 350µl luciferase lysis buffer (1% Triton X-100, 25mM glycylglycin pH 7.8, 15mM MgSO₄, 4mM EGTA, 10% glycerol) per well in a six-well plate. Plates were stored at −20°C until measurement of luciferase activity.

Short-term translation assays

Measurements of firefly and Renilla luciferase activity were performed separately in a Lumat LB 9507 single tube reader (Berthold).

For firefly luciferase, 360μl luciferase assay buffer (15mM K₃PO₄ pH 7.8, 25mM glycylglycine pH 7.8, 15mM MgSO₄, 4mM EGTA) with freshly added 1mM DTT, 2mM ATP was mixed with 100µl of lysate in a luminometer tube. After injection of 200µl of 1mM d-Luciferin (PJK) in 5mM glycylglycine the sample was measured for 20s. Renilla luciferase was measured by adding 100μl luciferase assay buffer supplemented with 1.43μM coelenterazine (PJK) to 20µl of lysate, and an acquisition time of 10s. Non-transfected cells were used to determine the background signal. Firefly luciferase from IRES-dependent translation, was normalized to the respective Renilla value to account for transfection efficiency and cell numbers. Note that capped in vitro transcripts encoding Renilla luciferase were not included in experiments primarily focusing on Northern blot analysis and in experiments using the Bi-Luc-SG replicon.

Dual-luciferase assay

For the dual luciferase replicons, the commercial Nano-Glo Dual-Luciferase Reporter Assay System (Promega) was used. In brief, 40µl of cell lysate were transferred from the six-well plate to a white walled 96-well plate (Falcon) in duplicates. Next, 40µl of ONE-Glo EX Reagent (Promega) were added, mixed and the reaction was incubated for 10min at room temperature before measurement of the firefly luciferase signal. Then, firefly activity was quenched for 10min using 40µl of NanoDLR Stop& Glo buffer (Promega) before determining the Nano luciferase signal. Non-transfected cells were used to determine the background signal. Measurements were performed using a Mitras LB940 plate reader (Berthold).

Total RNA extraction and Northern blot

For each time point, 6*10⁶ Hep3B cells were electroporated with 10µg of replicon RNA. Cells from one electroporation were seeded on a 10cm dish, and incubated at 37°C for 0, 30, 60, and 120min. Cells were scraped in the medium and put on ice immediately. 1ml of the 6ml suspension was transferred to an Eppendorf tube for luciferase assay, to control for translation activity at each time point. The cells were pelleted washed with PBS and taken up in 200µl of Luciferase lysis buffer. The luciferase assay was performed as described above. The remaining sample was spun down and washed with ice-cold PBS with 0.1mM aurintricarboxylic acid. Total RNA was extracted with acid guanidinium thiocyanate–phenol–chloroform and were denatured by treatment with 5.9% glyoxal in 50% DMSO and 10mM NaPO₄ buffer, pH 7.0. Total RNA was separated by glyoxal agarose gel electrophoresis. Then, the RNA was vacuum blotted onto a nitrocellulose membrane and UV crosslinked twice in a Stratalinker 1800 (Stratagene) with the “Auto Crosslink” program. Blotting efficiency and RNA integrity was checked by methylene blue staining of the membrane. After washing the membrane with demineralized water, the blot was cut below the 28S rRNA band and blocked with salmon sperm DNA at 68°C in hybridization buffer. The upper slice was hybridized with a ³²P-body-labeled RNA probe, complementary to nucleotides 5979–6699 of the HCV genome and the lower with a β-actin antisense riboprobe to correct for total RNA amounts loaded in each lane of the gel. Hybridization was performed over night at 68°C. The blot was washed twice with wash buffer 1 (0.3M NaCl and 30mM trisodium citrate, 0.1% SDS) and three times with buffer 2 (0.03M NaCl and 3mM trisodium citrate, 0.1% SDS) at 68°C. Specific bands were quantified after exposure of a phosphor imaging screen using a Molecular Imager FX scanner (Bio-Rad). Uncropped versions of the blots can be found in Supplementary Figs. ¹²–¹⁴.

In silico RNA structure calculation

For in silico folding of the 5′ terminal truncations of HCV, the online versions of RNA structure⁴⁶, mfold⁴⁷ were used without constraints. RNA structure was also used for integration of SHAPE data. Phylogeny analysis and visualization of RNA secondary structures was performed with Infernal⁴⁸ and R2R⁴⁹. The sequences used are shown in Supplementary Data ¹.

Selective 2′-hydroxyl acylation analyzed by primer extension

SHAPE is a method to address RNA structures in solution by chemical modification of flexible/unpaired nucleotides. For readout, an RT reaction with a labeled primer is performed, and the polymerase will stop at each labeled residue. The resulting cDNA fragments are separated on a denaturing gel, and each band stands for single stranded RNA. Generation of 2-methylnicotinic acid imidazolide (NAI) and SHAPE assays were performed as describe before⁵⁰, with minor modifications. Briefly, 1mmol 1,1′-carbonyldiimidazole (Sigma Aldrich) in 0.5ml DMSO was added dropwise to an equal molar amount 2-methylnicotinic acid (Sigma Aldrich) in 0.5ml DMSO, and the mixture was kept at room temperature for 1h to generate NAI. In vitro transcripts or RNA isolated from cells were denatured for 2min at 95°C, flash-cooled on ice and refolded in presence or absence of miR-122-5p at 37°C for 10min in SHAPE buffer (100mM HEPES pH 7.5, 100mM NaCl, 6.6mM MgCl₂). 100mM NAI or DMSO were added and incubated for another 10min. 1µl β-mercaptoethanol was added to quench the reaction. 0.2M NaOAc and 20µg glycogen were added, and RNA was extracted with phenol:chloroform. RNA was precipitated with three volumes of absolute ethanol and washed twice with 70% ethanol. The purified RNA was dissolved in 12µl RNAse-free water. For reverse transcription, 4µl were denatured for 2min at 95°C with 1pmol of a ³²P-labeled primer (HCV-130: AGACCACTATGGCTCTCCCG, HCV-65: CTAGGCGCTTTCTGCGTG) and 0.5M Betaine (Sigma Aldrich). After 2min incubation on ice, 1× FirstStrand buffer (Thermo Fisher Scientific), 10mM DTT, and 0.5mM NTPs were added and the reaction was prewarmed to 55°C for 1min. 20U SSIV reverse transcriptase (Thermo Fisher Scientific) were added and the reaction incubated for 10min at 55°C. Next, 1µl of 4M NaOH were added to hydrolyze the RNA and incubated at 95°C for 5min. The samples were cooled and one volume LDII was added before the cDNA fragments were separated by denaturing urea PAGE on a 10% gel (30W, 180min). The bands were quantified by phosphor imaging using a Molecular Imager FX scanner (Bio-Rad). SHAPE data were analyzed with SAFA⁵¹. The data were normalized to the mean of the top 10% reactivities. Intermediate reactivity (0.2–0.8) is indicated in yellow, highly reactive nucleotides (>0.8) are colored in red. For differential analysis of miR-bound versus free RNA the normalized values were subtracted and all differences of at least 0.3 were considered as significant. Less reactive nucleotides are depicted in green, more reactive in red.

For SHAPE, dimethyl sulfate (DMS) and nicotinoyl azide (NAz) probing of the full-length HCV genome, RNA was transcribed, purified, and folded as previously described¹⁸. Briefly, RNA was in vitro transcribed by runoff transcription with T7 RNA polymerase followed by treatment with RQ1 DNase (Promega). The RNA was then purified using RNeasy columns (QIAGEN) according to manufacturer’s protocol and eluted in ME buffer (8mM MOPS pH 6.5, 0.1mM EDTA). The RNA was then adjusted to a final concentration of 250ng/μL in folding buffer (50mM K-HEPES (pH 7.4), 0.1mM EDTA, 150mM KCl, and 5mM MgCl₂) and heated to 65°C followed by cooling to 37°C over a 45-min time course.

For DMS probing, DMS was added for a final concentration of 0.4%, or an equivalent volume of EtOH was added for unmodified control samples. Samples were incubated at room temperature for 10min, and modification was quenched by addition of 2-mercaptoethanol at 10%. In case of NAz, the sample was exposed to UV light (310nm) to activate the reagent.

Modified RNA was purified by phenol:chloroform extraction, LiCl precipitation, and resuspended in ME buffer at 1µg/µl. Primers were ³²P-end labeled by T4-PNK (NEB), and reverse transcription was performed with 1µg RNA by SuperScript III (Thermo Fisher Scientific) according to manufacturer’s protocol and the addition of 10U Superasin (Invitrogen). Following reverse transcription, RNA was degraded with the addition of NaOH and incubation at 95°C, and cDNA products were resolved by denaturing UREA PAGE on an 8% gel.

Translation assay in HeLa cell lysates

2.5µg of reporter replicon RNA was added to 10µl HeLa lysate (1-Step Human Coupled IVT Kit, Thermo Fisher Scientific) translation mix in a total volume of 15µl, following the manufacturer’s instructions. When needed, miR-122 duplex RNA was incubated at 30°C in the mix, 10min prior to RNA addition. The translation reaction was performed for 30min at 30°C. Then, 235µl of ice-cold lysis buffer, containing 200µg/µl CHX was added to stop the reaction. 10µl of the reaction were used as input sample for RNA extraction and western blot, 5µl for luciferase assay. Tubes were kept on ice until fractionation. 5mM GMP-PNP (Sigma Aldrich) or 500µg/ml CHX (Sigma Aldrich) were added prior to the incubation for experimental determination of the 40/48S and 80S fraction, respectively.

Fractionation on sucrose gradients

Lysates were loaded onto a linear gradient of 17.5–30% in 15mM Tris–HCl pH 7.4, 15mM MgCl₂, 300mM NaCl and subjected to ultracentrifugation at 165,000×g at 4°C using a SW60 rotor (Beckman) for 2.5h. Gradients were then eluted from the top using a Teledyne ISCO gradient elution system. Polysome profiles were obtained by measuring the absorbance at 254nm. In parallel, fractions were collected every 15s (equivalent to ~400µl). Before RNA extraction from the fractions, 20µg of glycogen, 350µl urea buffer (10mM Tris–HCl pH 7.5, 350mM NaCl, 10mM EDTA, 1% SDS, 7M urea) were added. Moreover, 1ng of in vitro transcribed Renilla luciferase RNA was supplemented as internal control. Following phenol:chloroform extraction and ethanol precipitation, the pellets were resuspended in 40µl of RNase-free water. Samples were stored at −80°C until further processing.

Detection of HCV RNA in fractions

HCV genomic RNA extracted from the fractions was analyzed by one-step qRT-PCR using qScript XLT one-step RT-qPCR ToughMix (Quanta Bio-sciences) according to the manufacturer’s instructions. Briefly, 15µl of reaction mixture contained 7.5µl of 2× RT mix, 1µM of JFH1-specific and Renilla-specific primers (JFH1 S146 forward qPCR primer: TCTGCGGAACCGGTGAGTA, JFH1 A219 reverse qPCR primer: GGGCATAGAGTGGGTTTATCCA, Renilla S767 forward qPCR primer: AATCGGACCCAGGATTCT, Renilla A917 reverse qPCR primer: ACTCGCTCAACAACGATTT), 0.27µM of the corresponding probes (JFH1 TaqMan probe: [FAM]AAAGGACCCAGTCTTCCCGGCAATT[TAMRA], Renilla S859 TaqMan probe: [HEX]TCGCAAGAAGATGCACCTGATGA[TAMRA]), 3µl of template RNA, and RNase-free water. To generate a standard curve, serial dilutions of an RNA standard (10⁷–10⁹ copies of HCV RNA and Renilla in vitro transcript) was prepared for each plate. Reactions were performed using the following program: 50°C for 10min, 95°C for 1min, and 40 cycles of amplification: 95°C for 10s and 60°C for 1min. The amount of HCV RNA was normalized to the Renilla control and the input sample.

Immunoblotting

10µl of each fraction were substituted with 2µl of 6× Lämmli buffer with 5% β-mercaptoethanol and denatured for 10min at 95°C. Samples were separated on a 10% SDS polyacrylamide gel and wet-blotted on a PVDF membrane for 1h at 400mA. Blots were blocked with low-fat milk in PBS-T (0.1% Tween) for 30min at room temperature. Primary antibody (goat polyclonal eIF3η antibody (Santa Cruz, N-20)), rabbit polyclonal RPS3 antibody (Abcam, ab140688) was applied in a 1:1000 dilution in milk PBS-T for 1h at room temperature. Blots were washed three times with PBS-T and species specific HRP-coupled secondary antibodies were added in a 1:5000 (rabbit, Sigma Aldrich, A0545) or 1:10,000 (goat, Invitrogen, A15999) dilution. After another three washes, the blot was incubated with Clarity Western ECL Substrate (Promega) for 1–2min and analyzed with an INTAS imager. Uncropped versions of the blots can be found in Supplementary Fig. ¹⁴.

Denaturing RNA gel electrophoresis

18S and 28S ribosomal RNA was detected for each fraction using a denaturing TAE gel approach⁵². The samples were heated to 65°C for 5min in loading buffer LDII (95% formamide, 0.025% SDS, 15mM EDTA, xylene cyanole 0.025% w/v, bromophenol blue 0.025% w/v) and flash cooled on ice. The denatured samples were applied to a 1% TAE agarose gel, containing 10µg/L ethidium bromide. The gel was run for 30min at 100V in 1× TAE buffer and bands were visualized on a UV transilluminator (INTAS). Uncropped versions of the gels can be found in Supplementary Fig. ¹².

Quantification and statistical analysis

Where appropriate, a paired, two-tailed Student’s t-test was performed using Prism 6 software (GraphPad). All data are presented as mean (±SD), the sample size is specified in the corresponding figure legend. P values of <0.05 were considered statistically significant. In graphs, statistical significance is indicated as follows: *P<0.05; **P<0.01; ***P<0.001.

We are grateful to Rahel Klein and Ulrike Herian for excellent technical assistance. We thank Jérome Boisbouvier and Bernhard Brutscher (IBS, Grenoble, France) for aiding in interpretation of NMR data. The study was funded by a grant from the Baden-Württemberg Stiftung to V.L. (BWST_NCRNA_024). This work used the High Field NMR, Biophysical and Cell-Free facilities at the Grenoble Instruct-ERIC Center (ISBG; UMS 3518 CNRS-CEA-UJF-EMBL) with support from FRISBI (ANR-10-INSB-05-02) and GRAL (ANR-10-LABX-49-01) within the Grenoble Partnership for Structural Biology (PSB). C.M. benefited from an ANRS fellowship. A.R. was supported by the Deutsche Forschungsgemeinschaft (SFB 1129 TP13).