Targeting G-quadruplexes in the rhinovirus genome by Pyridostatin inhibits uncoating and highlights a critical role for sodium ions.

18 The ~2.4 µm long rhinovirus ss(+)RNA genome consists of roughly 7,200 nucleotides. 19 It is tightly folded to fit into the ~22 nm diameter void in the protein capsid. In addition to 20 previously predicted secondary structural elements in the RNA, using the QGRS mapper, we 21 revealed the presence of multiple quadruplex forming G-rich sequences (QGRS) in the RV- 22 A, B, and C clades, with four of them being exquisitely conserved. The biophysical analyses 23 of ribooligonucleotides corresponding to selected QGRS demonstrate G-quadruplex (GQ) 24 formation in each instance and resulted in discovering another example of an 25 unconventional, two-layer zero-nucleotide loop RNA GQ stable at physiological conditions. 26 By exploiting the temperature-dependent viral breathing to allow diffusion of small 27 compounds into the virion, we demonstrate that the


QGRS-mapper identifies potential GQ-forming sequences in all RV genomes. 134
Employing extensive bioinformatics analysis, Lavezzo and colleagues recently 135 predicted GQs in viral genomes of human virus species, including RVs 34 . We independently 136 confirmed and extended their findings for all RV genome sequences available in Genebank 137 by using a local copy of the same program that runs on the quadruplex-forming G-rich 138 sequence (QGRS) mapper web-based server 35 . This software has been originally developed 139 to identify putative QGRS in DNA, where the G-score reflects the stability of a predicted GQ,140 which usually increases with the number of G-tetrades and decreases with the loop size. The 141 respective search motif G ≥3 L 1-7 G ≥3 L 1-7 G ≥3 L 1-7 G ≥3 takes into account that most experimentally 142 identified DNA GQs conformed to short-looped (1 to 7 nucleotides) structures with three and 143 occasionally more G-tetrades. However, RNA GQs are generally more stable than the 144 corresponding DNA GQs 36 ; two-layer RNA GQs are therefore not uncommon 37 , and the 145 overall higher stability of RNA GQs allows the insertion of larger loops compared to DNA 146 GQs 38,39 . Recently, an atypical RNA GQ with no first loop has been described 37 . Therefore, 147 we adjusted the search parameters of the QGRS mapper to allow the identification of 148 putative unconventional long-loop and zero-nucleotide loop G-quartets, including those 149 featuring only two layers. We then plotted the respective QGRS prediction G-scores ≥ 10 150 against their positions in the genomic RNA sequences (Fig. 1a). Strikingly, the vast majority 151 of rhinoviral genomes feature only QGRS predicted to form just two-layered GQs.  and RV-B4 are singled out by the additional presence of one and two three-layer GQs, 153 respectively. 154 Furthermore, most RV-ABC genomes feature at least one and up to seven putative 155 zero-nucleotide loop GQs without marked conservation. The number of potential QGRS 156 varies from 6 to 19 (mean 12) for the RV-As, from 9 to 19 (mean 13) for the RV-Bs, and from 157 10 to 23 (mean 15) for the RV-Cs. The augmented prevalence of QGRS for the RV-C 158 species is likely attributable to their genomes' distinctly higher GC content (43 % for 159 38 % for RV-A, and 38 % for RV-B 40 ). The analysis further revealed four highly conserved 160 QGRS ( Fig. 1a -asterisk), all located in the open reading frame. The slight differences in 161 their position between the genera (RV-A, RV-B, and RV-C) result from clade-specific 162 insertions and deletions. An additional conserved QGRS is uniquely present in the 163 upstream and close to the second highly conserved QGRS motif. By contrast, only 164 moderately conserved QGRS were predicted in the 5´ untranslated region (UTR), which 165 comprises essential regulatory elements such as the 5´ cloverleaf structure and the IRES, 166 required for replication and cap-independent translation 41 . No single QGRS was predicted 167 within the short 3´ UTR, featuring a highly conserved hairpin involved in virus replication 42 . 168 169 NMR analysis demonstrates the folding of synthetic ribooligonucleotides representing 170 selected QGRS of RV-A2 RNA into GQs, which is differentially affected by PDS. 171 For RV-A2, QGRS-mapper predicted 11 QGRS with G-score ≥ 10 (Supplementary 172 Fig. S1). We chose to study the candidates with the lowest and highest G-scores, G11 and 173 G20, the latter also representing one of the four highly conserved GQs (see above). 174 Assuming the almost invariable adoption of parallel GQs by RNA, G11 would give rise to an 175 unusual monomeric two-layer GQ bearing a zero-nucleotide loop 3 in combination with a 176 long loop 1 (Table 1; a schematic illustration of the likely G11 and G20 RNA GQ structure is 177 displayed in Supplementary Fig. S1). To confirm their GQ-forming propensity, we collected 178 1 H NMR spectra using 250 µM of the derived synthetic ribooligonucleotides in 10 mM sodium 179 phosphate (pH 7.4), 100 mM KCl, similarly as described elsewhere 43 . Each ribonucleotide 180 was denatured at 95 °C for 10 min, followed by cooling to 4 °C for 10 min to favour 181 intramolecular annealing into GQ and possibly other secondary structure(s) in the presence 182 of monovalent cations. A high concentration of K + usually stabilises GQs, as shown in other 183 instances 36 . G11 and G20 displayed 1 H signals of the bulk of the imino proton in the region 184 between 10.4 -12.4 ppm, which is the typical NMR signature of GQs  coloured area). The lack of sharp signals can be explained by G-register exchange dynamics 186 44 since certain G runs have more than two Gs, indicating the possible presence of 187 alternative conformations. 188 Peaks in the canonical Watson-Crick base pair region at 1 H NMR shifts larger ~12.4 189 ppm (olive-coloured area) are observable at low (4 °C) but not at high temperatures (34 °C, 190 the optimal temperature for RV-A2 replication), indicating that these secondary structure 191 elements (presumably hairpin(s)) are less stable than the GQs. Within the range assigned to 192 GQs, the second peak (11.9 -12.4 ppm) observed only for G11 (Fig. 1b) exhibited unaltered 193 intensity in an HDX experiment performed for 7 h at 37 °C. The solvent inaccessibility of 194 these imino protons indicated that they were most likely also part of the GQ core 195 ( Supplementary Fig. S2a). In the G20 spectrum, three distinct peaks (one within and two 196 next to the upfield broader of the GQ region) completely disappeared with increasing 197 temperature, indicating that the corresponding imino protons are conceivably not 198 involved in the G-tetrad core formation. The addition of 500 µM pyridostatin (PDS), which 199 specifically binds with high affinity to DNA and RNA GQs but not to other nucleic acid 200 secondary structures except i-motif forming DNA 45 , resulted in substantial changes of the 201 imino 1 H peaks for both synthetic RNAs ( Fig. 1b and Supplementary Fig. S2b), indicative of a 202 specific interaction. 203 The only other 1 H-NMR study we are aware of using PDS and a GQ was performed 204 with a three-layer DNA GQ derived from the src kinase gene (SRC). This showed an upfield-205 shifting of GQ imino proton signals stemming from PDS binding to the top G-quartet in a 206 stacking mode 46 . However, the differential effect observed by us suggested that the binding 207 mode of PDS may vary for G11 vs G20. We hence used molecular docking to assess this 208 possibility further. For lack of detailed structural information about the G11/G20 system, we 209 instead in silico evaluated the interaction of a crystallographic model of a pseudorabies virus 210 RNA GQ molecule composed of two G-quartets (PQS18-1, GGCUCGGCGGCGGA) (PDB 211 6JJH 47 ) and PDS. Using increasing concentrations of PDS in this in silico analysis, we 212 unexpectedly discovered that it could form defined dimers mostly stabilised by π-π 213 interactions. This was then experimentally substantiated by UV-absorption spectroscopy, 214 which clearly demonstrated a concentration-dependent self-association of PDS (K a ~42 µM) 215 in the buffer chosen for NMR spectroscopy ( Supplementary Fig. S3). 216 The analysis of the best docking poses for the monomeric and dimeric PDS and 217 PQS18-1 revealed two distinct classes of binding modes ( Supplementary Fig. S4): i) end-218 stack (a and c), and ii) groove/loop-binding (b and d). In both instances, the 2-(quinolin-4-219 oxy)ethanamine moiety of PDS interacts with the nucleobases via Van der Waals forces and 220 an additional contribution of electrostatic interactions between the charged amino groups of 221 PDS structure and the phosphate groups of the GQs. These binding modes may be variably 222 exploited for the interaction with G11 and G20, presumably resulting in the idiosyncratic 223 PDS-induced peak shifts. 224 Altogether, these results provided the initial bioinformatics analysis demonstrating the 225 likely formation of two-layered GQs by both ribooligonucleotides under favourable conditions 226 for the folding of such structures.  We next verified the tentatively assigned folding topology of G11 and G20 by circular 235 dichroism (CD) spectroscopy. CD is a gold standard to determine the strand orientations of 236 GQs. These might be parallel, as commonly adopted by RNA and DNA, or anti-parallel and 237 hybrid, which, with a few exceptions 48 , are observed only for DNA. Telomeric-repeat-238 containing RNA (miniTERRA ; Table 1), which adopts a three-layer parallel conformation 49 , 239 served as a positive control. 240 The CD profiles of miniTERRA, G11, and G20 taken at 25 °C in either 100 mM 241 sodium or 100 mM potassium phosphate buffer (pH 7.4) in each instance displayed the 242 signature of an all-parallel topology (ellipticity shows a negative band at 240 nm and a 243 positive band at 265 nm). The monovalent Na + and K + cations differentially stabilise GQs 244 owing to their different ionic radius and hydration free energies 50 . Generally, potassium 245 promotes folding and stabilises (ribo)oligonucleotide GQs to a larger extent than sodium 36, 51 , 246 though some exceptions have been reported 52 . In accordance with the latter, the degree of 247 GQ formation of G20 was noticeably the same irrespective of the presence of Na + or K + as 248 indicated by the practically unchanged CD spectrum (Fig. 1c). By contrast, as found for the 249 vast majority of GQ forming RNAs, the miniTERRA control and the G11 ribooligonucleotide 250 showed a higher proportion of GQ structure in K + compared to Na + . The trough at 210 nm 251 only evident for G11 and G20 is likely attributable to the additional presence of structural 252 elements comprising A-type duplex RNA regions as also revealed in the above NMR 253 analysis at low temperature. 254 Na + and K + differently impact the physicochemical parameters of G11, G20, and 255 miniTERRA ribooligonucleotides. 256 As seen in Figure 2a, the CD melting and annealing spectra are almost 257 superimposable for the G11 and G20 ribooligonucleotides and the miniTERRA control when 258 dissolved in a sodium-containing buffer. The lack of a marked hysteresis indicates 259 thermodynamic equilibrium between denaturation and refolding in the examined temperature 260 range and the used ramp rate and is typical for intramolecular GQs 53 . The situation for 261 miniTERRA was strikingly different in the potassium buffer, where the unfolding/annealing 262 profiles exhibited a marked hysteresis (Fig. 2a); this is typically observed for intermolecular 263 GQ formation, where the shape of the melting and refolding curves depends on the speed of 264 temperature ramping 53 . Another possibility is a slow conformational change of an 265 intramolecular GQ from one topology to another 54 . However, according to CD spectroscopy. 266 miniTERRA in the presence of either cation (Na + , K + ) formed a parallel GQ (Fig 1c)  267 exclusively. Since the multimerisation of miniTERRA could be ruled out by a subsequent 268 electrophoretic analysis (see below), the relatively slow ramp rate of 1 °C/min apparently still 269 did not allow the reversible refolding of this three-layer intramolecular GQ. This is in line with 270 findings by others that multiple layers of tetrades are particularly prone to hysteresis 55 . By 271 contrast, in the presence of K + , the CD melting/cooling profiles of both two-layered GQs 272 exhibited just a rather subtle (G11) to negligible hysteresis (G20). The latter's overlapping 273 heating and cooling curves, as also seen with Na + buffer, reinforces the notion that G20 274 forms a monomeric (see below) intramolecular GQ in presence of each of these cations. 275 Others have shown that K + -containing solutions stabilise the parallel topologies of 276 two-and three-quartet RNA GQ much more than Na + -containing solutions, with the 277 difference in the melting temperatures ∆T m ranging from 15 to more than 30 °C 36, 51 . Such 278 drastic change in T m (determined at the intersection of the second derivative with the x-axis 279 in Fig 2a) was also evident for the three-layer miniTERRA GQ in K + compared to Na + . By 280 contrast, the melting temperatures of G11 and G20 in the K + containing buffer were just ~ 4 281 °C higher than those determined in Na + buffer. 282 The limited hysteresis of G11 melting/cooling might indicate the formation of a dimeric 283 GQ resulting from end-to-end stacking of two intramolecular GQs. This gains additional solid 284 support from native gel electrophoresis, which clearly shows the occurrence of dimers in the 285 presence of potassium ions for G11. Under these conditions, G20 and the miniTERRA 286 control gave rise to just a single band, indicating that they existed primarily as monomers 287 ( Supplementary Fig. S5). Based on these results, the slight hysteresis of the G11 melting 288 can be best interpreted by assuming a fast monomer refolding, followed by slower 289 oligomerisation. 290 The adoption of a GQ conformation by the two RV-A2 derived RNA sequences (G11 291 and G20) and the positive control (miniTERRA) was then confirmed with a Thioflavin T (ThT) 292 light-up assay with all measurements done at room temperature. ThT end-stacks to RNA 293 GQs, which strongly increases its fluorescence 56 . As seen in Fig. 2b and 2c, no significant 294 fluorescence was detected upon incubation with the negative control (C for G substituted 295 miniTERRA) ribooligonucleotide (see Table 1). In contrast, a clear signal was obtained for all 296 three tested GQs, whose magnitude was quite comparable for G11 in Na + (left panel of Fig.  297 2b) and K + buffer and about 1.5 and 2-fold higher in the presence of K + for miniTERRA and 298 G20, respectively (left panel of Fig. 2c). The ThT fluorescence intensity was strongest for 299 miniTERRA and approximately 3 to 4-fold lower for G11 and G20 (but still 30 to 40 times 300 higher than for the negative control), roughly correlating with their respective G-scores. Since 301 the same amounts of oligonucleotides were employed, this indicated that a fraction of G11 302 and G20 adopted a non-GQ conformation as already observed in the 1 H-NMR analysis. The 303 coexistence of A-type RNA (hairpin) structure and a two-layer GQ structure has been 304 recently described by Lightfoot and others 37 . Based on the ThT light-up probe, G11, but not 305 G20 would be almost indifferent to the choice of the two cations, which is opposite to the 306 results of the label-free CD analysis. Most likely, the conformation of the respective GQs 307 subtly differs when bound to Na + (which coordinates with the Gs in the middle of a tetrade) or 308 K + (residing between G-tetrads), which may variably affect the binding affinity for ThT giving 309 e.g. the impression of an apparent higher stability of G20 in K + buffer. However, this does not 310 change our overall conclusion that GQ structures are formed with both cations in this assay. 311 We then examined whether the tetrade-associated cation type (K + or Na + ) impacts the 312 interaction of the respective GQs with PDS by conducting a fluorescent indicator 313 displacement (FiD) assay, similar to the one described in refs. 57,58 . FID allows evaluating the 314 relative affinity and selectivity of compounds binding to a GQ 59 . Expectedly, ThT was 315 displaced by PDS in each instance (right panels in Fig. 2b/c). However, while the dose-316 response curve for miniTERRA was comparable in Na + and K + containing buffers, the 317 reduction in fluorescence determined for G11 and G20 was about 2-fold more efficient in the 318 presence of Na + as indicated by the respective IC 50 values (Table 2). Notably, with the lower 319 G-scoring G11, a sharp decrease of the ThT fluorescent signal was already evident at low 320 PDS concentrations. Since ThT was shown not to markedly alter the stability of GQs 60 , this 321 might indicate a higher affinity of PDS for the noncanonical GQ-forming G11 in comparison 322 to miniTERRA and G20 as reflected by the correspondingly lowest IC 50 values ( Taken together, employing several orthogonal assays, we could demonstrate the 331 intrinsic ability of two selected RV-A2-derived QGRS to form RNA GQs. However, within the 332 context of the viral RNA, these and the other predicted candidate sequences may fail to fold 333 into such scaffolds if embedded into or overlapping with alternative secondary structures of 334 higher stability or more rapid formation (and trapped in a metastable state). Furthermore, the 335 G-quartet fold of some untested putative QGRS may be too unstable at the temperature of 336 virus propagation. PDS and other so-called GQ-stabilising compounds can selectively 337 enhance the mechanical and thermal stability of DNA and RNA GQs over the level achieved 338 by K + alone 61, 62 . Thereby, they can also force alternative secondary structures to transform 339 into the compound-stabilised quadruplex conformation 63 . The latter was evident for G20 and, 340 less pronounced, for G11 by the loss of Watson-Crick peaks in the 1 H-NMR spectrum on 341 incubation with PDS ( Supplementary Fig. S2b). 342 It is presently believed that the uncoating of RV-A2 and several other enteroviruses 343 requires a structural switch of the genomic RNA and the transient unfolding of secondary 344 structures for transit as a single-strand through one of the narrow pores formed in the A-345 particles 5, 7, 64, 65, 66 . Therefore, we reasoned that the exposure of the encapsidated viral RNA 346 to PDS by stabilising preexisting GQs and promoting the transition of unstructured and/or 347 alternatively folded QGRS into such unconventional secondary structures might interfere with 348 its in vitro and in vivo uncoating. RVs, in contrast to other enteroviruses such as poliovirus, 349 are readily permeable for monovalent cations such as Cs + already in the cold and also for 350 small organic compounds such as dansylaziridine and ribogreen when incubated at 351 breathing conditions 67, 68, 69 . Capsid breathing describes a transitory expansion of the protein 352 shell with temporary exposure of normally internal amino acid sequences through reversibly 353 formed small holes, commencing at room temperature to around physiological temperatures, 354 dependent on the RV serotype 70 . We thus attempted to deliver PDS to the viral RNA 355 genome within the native virion by exploiting this phenomenon. All incubations were done in 356 phosphate-buffered saline (PBS) to prevent PDS aggregation into long fibrils, as we have 357 recently observed for Tris-but not phosphate-based buffers 71 . 358 First, we confirmed the dependence of PDS delivery to the inside of the capsid as a 359 function of temperature. Purified RV-A2 was incubated with PDS under conditions of strongly 360 diminished capsid breathing (at 4 °C) and a capsid breathing-promoting temperature (at 34 361 °C). Unbound PDS was removed, and a Particle Stability Thermal Release Assay (PaSTRy) 362 72, 73 was performed to determine a possible impact of PDS on temperature-dependent 363 uncoating, a commonly used model for in vitro uncoating of picornaviruses 74 . Fig. 3a  However, the temperatures T max at the peak of the SYTO 82 signal (indicative of the 370 complete conversion of all native virions into permanently porous A-particles) and the peak of 371 the first derivative (at T 50 , corresponding to 50 % RNA accessibility) 73 remained practically 372 unaltered. 373 We interpret this finding as that PDS, by interacting with (potential) QGRS in the 374 genome of RV-A2, directly or indirectly affected RNA contacts with the capsid, thereby 375 resulting in enhanced mobility of the protein shell of the native virus. This allowed 376 appreciable uptake of the SYTO 82 dye into the viron for binding the viral RNA already at a 377 lower temperature compared to the control. However, the PDS-induced effect did not 378 critically impact the rate of the temperature-dependent conversion of native to A-particle as 379 inferred from the unchanged T 50 for this sigmoidal conversion 73 . The heat-triggered RNA 380 release starting at T max (resulting in the subsequent drop of the SYTO 82 signal) was also not 381 affected. As thermal unzipping of secondary structures is required for the exit of the RNA 382 from the capsid in this in vitro uncoating model 64 , the bound PDS evidently did not increase 383 the melting temperature of the viral GQs above one of the most stable non-GQ secondary 384 structures (presumably mostly hairpin loops 75 ) inside the capsid. 385 We then assessed the effect of PDS on the in vivo uncoating of RV. HeLa cells were 386 infected with RV-A2 pre-incubated with PDS and without PDS (control), both at 34 °C for 4 h. 387 This treatment time sufficed for the entry of appreciable amounts of RiboGreen into the 388 capsid of RV-A2 69 . Unbound PDS was removed by centrifugation in an Amicon ultrafilter 389 unit, followed by repeated washing with PBS at 25 °C. The viral samples were then 390 transferred to HeLa cells grown in 10 cm diameter dishes and incubated for 30 min to allow 391 for viral uptake and uncoating in the absence of inhibitory compounds 65 . Supernatant and 392 cells were collected by scraping, and internalised viral material was recovered by five times 393 freezing and thawing. Cell debris was removed, and aliquots of the supernatants were 394 subjected to immunoprecipitation with the subviral (A-and B-particle) specific monoclonal 395 antibody 2G2 76, 77 . From equal aliquots of the precipitated material, protein and RNA were 396 recovered and quantified in Western blots (Fig. 3b) and by RT-qPCR, respectively (Fig. 3c). 397 Comparing these results, it becomes clear that regardless of whether the pre-incubation was 398 carried out in the presence or absence of PDS, the same amount of viral proteins were 399 detected. This excludes any influence of PDS on cell attachment, e.g. via forming virus-400 trapping filaments as observed in Tris-based buffers (see above and ref. 71  The observed effect of PDS in the PaSTRY assay and on the in vivo uncoating of RV-417 A2 was most likely due to the trapping/stabilisation of GQs located in the encapsidated RNA 418 rather than an unspecific binding to other components of the virion. However, in one report, 419 PDS was found to act as a weak inhibitor of the C5 convertase, a component of the 420 complement system 79 . Therefore, to further furnish our hypothesis, we repeated the in vivo 421 analysis at capsid breathing conditions with another frequently used GQ-binding compound, 422 Phen-DC3 80 . RV-A2 was incubated as described above with PDS at 20 µM and 200 µM and 423 in parallel with PhenDC3 at 1 µM and 5 µM, respectively. A mock-treated virus was used as 424 a control. All samples were subjected to repeated centrifugal ultrafiltration to remove the 425 excess of the compounds maximally. HeLa cells were then challenged with these samples 426 and incubated for 9 h to allow for a one-cycle infection. RV-A2 positive cells were determined 427 by fluorescence-activated cell sorting (FACS), using the intracellularly produced VP2 as a 428 readout. Fig. 3e demonstrates a concentration-dependent decrease in the number of cells 429 containing replicating RV-A2 upon pretreatment with PDS. Notably, a significant reduction 430 was also observed with 5 µM of Phen-DC3. This result further strengthened our hypothesis 431 that PDS triggers a structural change in the encapsidated RNA as postulated from the 432 PaSTRY analysis, preventing its orderly egress from the capsid under in vivo conditions. 433

PDS affects the conformation of the free RV genome and reduces viral infectivity in 434
the presence of Na + but not in the presence of K + . 435 We next embarked on an ultrastructural analysis of gently extracted rhinoviral RNA 436 for directly visualising the proposed PDS-induced structural change. The protein shell of RV-437 A2 was proteolytically removed with proteinase K. To additionally examine a possible 438 differential impact of the prevalent extracellular (Na + ) and intracellular (K + ) monovalent 439 cation, the digestion was performed with the purified virus in sodium-as well as in indicated that distinctly more GQs were now available for ThT binding, likely resulting from a 453 conformational transition of kinetically trapped, metastable alternative (e.g. hairpin) structures 454 to the more stable quadruplex structure triggered by the elevated temperature. A similar 455 result was recently obtained in a study examining the kinetic vs thermodynamic control of a 456 sequence within an mRNA able to switch from a hairpin to a GQ, using N-methyl 457 mesoporphyrin (NMM) as light-up indicator 81 . The recorded ThT signal was consistently 458 higher for the Na + compared to the K + -containing samples, despite the greater GQ stabilising 459 propensity of the latter cation, amounting to a ~40 % difference at 30 °C, which diminished to 460 just 9 % for the 60 °C heated samples. It suggests that the viral RNA molecule with intact 3D 461 structure is possibly more compact in the presence of the charge-neutralising K + , leading to 462 reduced accessibility of the already existing GQs for the ThT probe. Binding of ThT to GQs 463 forming during refolding of the heated viral RNA was expectedly much less affected by the 464 monovalent cation type when the probe was already present before full compaction of the 465 nucleic acid molecule. 466 We next incubated the ex virion RNA with 20 µM PDS, again in a Na + or K + -containing 467 phosphate buffer, and submitted it to rotary shadowing. The platinum replicas were then 468 observed by transmission electron microscopy (TEM) (Fig. 4b) The drastic shape change only observed in Na + containing buffer likely results from 487 the rescuing of unstable GQs and/or the shifting of long-lived metastable structures with 488 alternative G-quartet forming ability to the more stable GQ conformation by PDS. This 489 presumably disturbs short and long-range interactions determining the global structure of the 490 RNA genome 85, 86 , similarly as described for certain small molecules on binding to tRNA and 491 riboswitches 87, 88 . Possibly, K + binding pockets in the viral RNA as e.g. found in ribosomal 492 RNA 89 further constrain and condense its native tertiary structure compared to Na + , limiting 493 the access of the compound to these regions to account for the observed different effect of 494 PDS. To substantiate this hypothesis, we assessed the stability of the protein-free (ex virion) 495 RNA by carrying out a differential scanning fluorimetry (DSF) similarly as described by 496 Silvers, Keller 90 . Inspection of the DSF traces obtained in Na + and K + buffer indicated 497 similarly low accessibility of the viral RNA for SYTO 82 at 25 °C up to about 40 °C, attesting 498 that the RNA molecule stays compact and extensively folded within this temperature range 499 ( Supplementary Fig. S8). Here, SYTO 82 intercalates mostly into solvent-accessible double-500 stranded regions located at the periphery, which seems less affected by choice of these 501 cations compared to the accessibility of GQs for ThT under similar conditions (Fig. 4). The 502 subsequent relatively rapid fluorescence increase from the lower baseline (at 44.5 °C and 51 503 °C for Na + and K + , respectively) to the upper base line (= maximal response; at 54.3 °C and 504 57.1 °C for Na + and K + , respectively) relates to the disruption of (mostly) the RNA tertiary 505 structure, now allowing SYTO 82 binding also to internally localised stem regions. The rise 506 was ~2-fold (Na + ) and ~1.2-fold (K + ), indicating a more extensive unfolding in the sodium-507 containing buffer. The higher stability of the tertiary contacts of the viral genome in the 508 presence of K + is illustrated by the around 4 °C higher T m1 . The following progressive drop in 509 emission intensity is due to melting of the more stable secondary structures with the release 510 of SYTO 82, resulting in somewhat closer spaced higher melting temperatures T m2 (60.5 °C 511 for Na + and 63 °C for K + ) for this transition. 512 The above in vitro data derived with the ex virion RNA altogether implies that its 513 distinctly more compact tertiary structure imposed by K + hinders the PDS molecules from 514 arriving at internal regions able to transform into GQs on the binding of this compound. If 515 correct, this should materialise in a distinctly smaller number of PDS molecules associated 516 with the RV-A2 genome in potassium compared to sodium buffer. As assessing this figure is 517 technically less demanding with encapsidated RNA, we quantified the amount of PDS that 518 remained associated with the virions after incubation (20 μM) at 4 °C or at 34 °C for 4 h in 519 Na + or K + containing buffers and subsequent extensive washing to remove any externally 520 deposited traces of the compound. The result further corroborated our hypothesis that K + , while commonly cooperating 528 with PDS in GQ stabilisation 91 , unexpectedly protects the rhinoviral RNA from binding of this 529 compound not only ex virion but also inside authentic viral particles (in virion). The magnitude 530 of the difference of virion-incorporated PDS was quite impressive, as the encased 531 polynucleotide of picornaviruses is already very compact 92, 93 . This might have conceivably 532 mitigated the tertiary structure changes in the presence of Na + vs K + , believed to govern the 533 extent of pyridostatin binding to viral QGRS when free in solution. A monovalent cation-534 dependent breathing activity as an additional source for that difference was ruled out by a 535 nanoDSF analysis of purified virions diluted in 100 mM sodium or potassium phosphate 536 buffer ( Supplementary Fig. S9). 537 We then assessed whether the different PDS uptake by RV-A2 as found by MS 538 resulted in consequences on its infectivity by determining the TCID 50 of virus exposed to the 539 compound diluted in the respective monovalent cation-containing phosphate buffer. To avoid 540 any unspecific loss of infectivity due to thermal inactivation, we reduced the incubation 541 temperature to 25 °C and compensated for the resulting diminished breathing by extending 542 the treatment time to 20 h. As shown in Fig. 4d, the virus sample treated with PDS in the 543 presence of Na + exhibited a titer reduction by 2 logs compared to control conditions (the 544 same buffer without PDS). By contrast, PDS treatment in K + buffer did not affect the virus 545 titer as expected from the small amount of the compound detected by MS in the virion of 546 similarly treated virus ( Table 3). Note that the slight reduction of infectivity on changing the 547 internal monovalent cation environment from Na + to K + in the absence of the PDS was not 548 significant. 549

Finally, we explored at what step of the infection cycle PDS might act on RV-A2 in 550
vivo (i.e. without pre-incubation) by a time-of-addition experiment. The virus was bound to 551 the cells for 30 min at 4 °C, PDS was added, and the cells shifted to 34 °C (T0). The same 552 experiment was conducted in parallel, except that PDS was added at T180 and T300, 553 respectively. This roughly corresponds to RV entry and uncoating (T0), RNA synthesis 554 (T180), and assembly (T300) 94, 95, 96 . The cells were maintained for 9 h pi (one full cycle of 555 infection), and viral synthesis was measured by fluorescence-activated cell sorting (FACS). 556 As can be seen in Fig. 4e, the infection rate and viral synthesis was only significantly 557 impacted upon the addition of PDS at T0. 558

DISCUSSION 559
GQs have recently become a focal point as promising antiviral targets 18 . Despite this 560 surge in interest, the possible importance of this unconventional secondary structure remains 561 unexplored in picornaviruses. In this report, we fill this gap by focusing on RVs of the genus 562 Enteroviruses within the Picornaviridae family, which are responsible for more than 50 % of 563 common cold cases 4 . Using QGRS mapper 35 , we identified putative intramolecular QGRS 564 for all completely sequenced rhinoviruses of clade A, B, and C. Altogether, RVs comprised 565 between 6 and up to 23 such QGRS motifs, which, with the exceptions of RV-A41 and RV-566 B4, would give rise to only two-layer GQs. In most genomes, a variable fraction of the 567 predicted GQs would furthermore lack one G-quartet connecting loop. Most of the putative 568 QGRS are not or just weakly conserved, making it unlikely that they play an important role in 569 the virus life cycle. Nonetheless, four of these putative QGRS, all located in the ORF and 570 predicted to fold into conventional two-layer GQs, are highly conserved across all A, B and C 571 types (Fig. 1a); they were also identified in an independent analysis of all human virus 572 families by Lavzzo et al. 34 using a proprietory software. While suggesting a functional 573 relevance for these QGRS, they might instead encode a critical sequence in the 574 corresponding polyprotein as an alternative reason for their strict maintenance, which 575 remains to be investigated. Atypical RNA GQs such as those with a bulged out nucleotide or 576 a vacancy will escape our analysis, but there are currently only a few reports on their 577 existence and possible role 48 . 578 The bioinformatics analysis indicated that targeting of GQs might be a promising new 579 approach for combating RV infections, with the strongly conserved GQs conceivably 580 presenting a high genetic barrier to drug resistance. For this purpose, we chose RV-A2 as 581 one of the best-characterised representative 97 . The 11 putative QGRS within its genomic 582 sequence would all give rise to just two-layer G-quartets ( Supplementary Fig. S1). Using 583 various orthogonal biophysical assays, we demonstrated that the synthetic 584 ribooligonucleotides G11 and G20, representing the QGRS with the lowest and highest G-585 score, respectively, formed all-parallel GQs in the presence of mM concentrations of K + as 586 well as of Na + . This is also the most common conformation of naturally occurring RNA GQs 587 98 . However, in contrast to many other examples 36 , their thermal stability was only 588 moderately (by ~4 to 5 °C) enhanced by K + vs Na + . Also, G11 and G20 differed only little in 589 their stability with respect to the same coordinating alkaline cation, despite the considerably 590 higher G-score of G20. At lower temperatures (0 to 25 °C), a fraction of each 591 ribooligonucleotide folds into an A-type RNA structure (presumably hairpins). These 592 alternative conformers were not further explored as they were unstable at 34 °C, the optimal 593 temperature for RV-A2 replication 99 . This analysis is, to our knowledge, now the second 594 showing that an unconventional, two-layer zero-nucleotide loop RNA GQ as represented by 595 G11 is stable at physiological conditions. Interestingly, while no first loop was present in the 596 example of the previous report 37 , G11 lacks the third loop. 597 The 1 H-NMR analysis evidenced that both, G11 and G20, bound the GQ stabiliser 598 pyridostatin (PDS; 91, 100 ). In all experiments with this compound, we specifically avoided Tris-599 buffer as a solvent since we recently found it to promote aggregation of PDS into variably-600 sized fibres 71 . Using phosphate buffer instead, we noted the formation of PDS dimers, which 601 were the dominant species at concentrations ≥ 200 μM. Molecular docking indicated that this 602 association was largely driven by a π-π stacking interaction between two PDS molecules. An 603 in silico modelling with the structurally well-characterised pseudorabies virus RNA-derived 604 two-quartet GQ molecule PQS18-1 showed that both monomeric and dimeric PDS bind to 605 the exposed G-tetrades on the top or bottom of a GQ and share a second binding mode 606 involving the bases in the groove and a loop. (Supplementary Fig. S4). This is entirely in line 607 with a 19 F-NMR study at a similar ligand (PDS) to RNA ratio 101 as used by us, which besides 608 π-π end-stacking, indicated additional (uncharacterised) binding site(s) for PDS on the 609 employed RNA GQ. Our analysis corroborates and extends previous reports on modelling 610 the GQ-PDS interaction, which proposes either a π-π end-stacking mode 102, 103 or an 611 exclusive binding of PDS to the loop/groove interface 104 , being furthermore restricted to 612 docking of just the monomeric compound. While our findings for PDS are novel, they are not 613 entirely unexpected, as stacking-mediated dimerisation and further aggregation of flat 614 polycyclic aromatic nucleic-acid binding dyes are already known for long (e.g. Bradley and 615 Wolf 105 ). Further, at least one other GQ-ligand (DMSB, a cyanine dye) was shown to bind as 616 a dimeric associate to the terminal G-tetrade and the groove of a DNA GQ 106 . 617 (FHV), a small icosahedral (+)ssRNA virus, by an aziridine derivative, which also led to 631 enhanced capsid mobility, suggested to originate from disrupted capsid protein-RNA 632 interactions in the native particle 108 . By analogy, we concluded that PDS similarly triggered a 633 rearrangement of the encapsidated rhinoviral RNA affecting contacts with the inner surface 634 of the shell 5 , including those mediated by the recently identified enterovirus packaging 635 signals 109 , leading to enhanced capsid breathing. 636 The proposed conformational change induced by PDS was directly verified by 637 ultrastructural analysis of rhinoviral RNA gently freed from the surrounding capsid by 638 proteinase K digestion. Instead of PBS, the incubation was done in a phosphate buffer 639 containing exclusively Na + or K + as the major extra-and intracellular monovalent cation to 640 assess their potential role. Remarkably, the GQ-specific light-up probe ThT revealed that in 641 both conditions, only a few GQs were already established in the ex virion RNA. Most putative 642 QGRS were apparently sequestered into alternative, long-lived (kinetically trapped), 643 metastable secondary structures, which transformed into thermodynamically more stable 644 GQs on heating and slow cooling of the viral RNA, resulting in the observed massively 645 increased ThT fluorescence emission. Presumably, during positive-strand RNA synthesis in 646 the infected cells, sequences of the nascent (+) strand comprising these QGRS fold much 647 faster into alternative, metastable conformations such as RNA hairpins than into the more 648 stable GQ (tens of microseconds vs hundreds of milliseconds 110, 111 ) as soon as they emerge 649 from the active centre of the viral RNA replicase. If separated from each other by an 650 appreciable activation energy barrier, as indicated by the ThT analysis, these metastable 651 structures, following their rapid encapsidation (which is tightly coupled with replication 112 ), 652 will persist in the progeny virions. The kinetically favoured formation of a metastable hairpin-653 like structure instead of an alternative, more stable, GQ during co-transcriptional folding of a 654 nascent mRNA has been recently reported 81 . 655 Rotary shadowing and AFM demonstrated a profound PDS-induced shape change of 656 the ex virion RNA, in full agreement with our prediction. However, to our surprise, this 657 occurred only in the presence of Na + but not the typically more strongly GQ-stabilising K + . 658 Then, the DSF analysis provided a plausible explanation for this initially puzzling 659 phenomenon, showing that potassium ions considerably strengthened the tertiary structure 660 of the ex virion RNA compared to sodium ions. We speculate that this is due to specific 661 chelation sites for K + in the RV-A2 genomic RNA as described in certain ribozymes 113, 114 662 and ribosomal RNA 89 , which reinforce their tertiary structure. The viral RNA will likewise 663 further compact and rigidify when the postulated pockets are occupied by K + , which we 664 believe renders internally located, potential GQ-forming sequences inaccessible for 665 pyridostatin. In support of this, Favre and coworkers have previously shown that tight folding 666 of various RNA species substantially restricted the intercalation of ethidium bromide 115 . 667 Conversely, these GQ-forming sequences must remain accessible for PDS in the less 668 tightly packed ex virion RNA exposed to Na + (at the same concentration as K + ) to explain the 669 observed drastic structural reorganisation, which we attribute to the transition of QGRS 670 sequestered in the alternative, metastable conformations into GQs promoted by PDS. Apart 671 from its GQ-stabilising effect, this compound likely accelerates the process by its direct 672 participation in the GQ folding, thereby lowering its activation energy, as shown in an optical 673 tweezer study 61 . An analogous refolding of stem-loop structures into GQs induced by the G-674 quadruplex ligand PDP (a PDS derivative) was recently proposed for an RNA of hepatitis C 675 virus 116 , and a profound effect on RNA long-range folding due to extensive differences in 676 secondary structures acquired with GQ formation was predicted in a bioinformatics analysis 677 91 . The unexpected "protective" effect of K + was also clearly evident when RV-A2 was 678 incubated with PDS at breathing conditions, which did not result in a change in infectivity, 679 while the virus titer dropped by two logs on its replacement by Na + in the buffer. Accordingly, 680 mass spectrometry revealed a marked accumulation of PDS inside the capsid only in the 681 presence of sodium and not potassium. It must be emphasised that the mere exchange of 682 Na + for K + in the capsid did not significantly impact the infectivity despite the different level of 683 compactness of the genomic RNA believed to regulate the compound access. 684 How may the binding of PDS to the QGRS of the viral genome under permissive 685 monovalent cation conditions affect its release from the capsid? In the currently favoured 686 model of enterovirus endosomal uncoating, the viral RNA must transiently unfold in order to 687 pass with its 3´ end first through one of the in total 30 small pores (~1.5 nm diameter 7 ) 688 opening permanently at each of the two-fold icosahedral symmetry axis of the subviral A-689 particle. The emerging RNA reaches the cytosol without contact with the endosomal contents 690 by passing through a connecting channel in the endosomal lipid bilayer formed by 5-6 copies 691 of expelled VP4 with a lumen diameter of between approximately 4.6 nm and 12 nm. Based 692 on preliminary data, host factor(s), in coordination with released VP4 acting as chaperone, 693 may pull the genome from the capsid once its unstructured poly(A) tract at the 3´ end 694 appears outside of the endosome 5,7,64,65,66,117,118 . The forced transfer through the narrow 695 capsid opening will result in the reversible unzipping of secondary and tertiary structure 696 elements akin to the electrophoretically driven transport of structured RNA through the α-697 hemolysin nanopore 119 . The average number of PDS molecules (10) incorporated into the 698 capsid of RV-A2 in the presence of Na + is comparable to the number of putative QGRS (11). 699 Assuming that they become folded into GQs as described for the ex virion RNA, it is tempting 700 to speculate that the stabilising effect of PDS prevents their unwinding by the above 701 mechanism. With an effective size of ~2.4 nm, the GQ-compound complex would sterically 702 block the ejection of the genomic RNA through the 2-fold related pore. This scenario closely 703 resembles the one suggested for compound-stabilised GQs located in the ORF of mRNAs on 704 encountering a translating ribosome, believed to plug its entry site featuring a diameter of 705 ~1.5 nm 120 , thereby obstructing further entry of the mRNA. However, though attractive, we 706 consider this rather unlikely, as our PaSTRY experiment with RV-A2 showed that 707 incorporation of PDS did not increase the temperature T max , where RNA release starts 708 ( Supplementary Fig. 6). This implies that PDS-binding by rhinoviral GQs did not raise their 709 stability above the one attributable to the most stable non-GQ secondary structure(s) formed 710 in the encapsidated rhinoviral RNA. We consequently favour an alternative mode of action 711 based on the likely structural reorganisation of the encapsidated genome triggered by PDS, 712 as indicated by PaSTRY and directly visualised with the ex virion RNA. This might 713 conceivably compromise the formation of the well-ordered RNA layer beneath the protein 714 shell of the A particle proposed to guide its ordered egress 5 . In addition or alternatively, it 715 could dislodge the viral RNA´s 3´ end found to exit first 65 from a position believed to reside in 716 the vicinity of one of the pores opening at the two-fold axis to allow its facile ejection, perhaps 717 directed by electrostatic focusing 121 . Its PDS-driven relocation would result in a high entropic 718 penalty for finding such holes via thermal fluctuation (e.g. Polson and McLure 122 ), critically 719 diminishing the successful vectorial traversal of the viral RNA through the capsid. This 720 problem is presently little appreciated, though one report already highlighted its role in a 721 coarse-grained model of RV-A2 uncoating 123 , and the results with PDS now provide first 722 experimental cues that it might matter. Worth mentioning, the icosahedral (+) ssRNA phages 723 MS2 has remarkably solved this problem by the strong binding of a hairpin at the 3´ end of 724 the viral genome to a single copy of a maturation protein, which is directly incorporated into 725 the capsid, replacing a coat protein dimer at one of the icosahedral two-fold axes, being 726 pulled out alongside the RNA by attachment to and subsequent retraction of a bacterial F-727 pilus 124 . Our reasoning does not contradict the fact that the heat-triggered uncoating of RV-728 A2 is unaffected by PDS, as the substantially increased thermal motion will restore the 729 RNA´s chances of finding a suitable exit pore. We note that several other RNA-binding 730 compounds incorporated into the encapsidated genomic RNA of rhino-and other 731 enteroviruses, such as RiboGreen, SYTO 82, neutral red, proflavine, and acridine orange did 732 not markedly affect their infectivity (the latter three when examined in the dark as they render 733 viruses photosensitive) 66,69,125,126  to a close apposition of viral genome to positively charged regions of the capsid followed by 744 its cracking due to an increased pressure exerted on the capsid from the inside. This is 745 thought to lead to the expulsion of pentamer(s) followed by genome release. While PDS 746 might also interfere with such a process, we consider this mode of uncoating less important 747 for RV-A2, as cryo-EM analysis of low pH treated RV-A2 did not reveal the presence of 748 significant amounts of open particles 5 . 749 Time-of-drug-addition showed that PDS had little consequence for RV-A2 protein 750 production when added after the uncoating stage. All subsequent events required for viral 751 reproduction occur in a high K + and low Na + cytosolic environment, limiting the access of the 752 viral RNA for the compound as found for the ex virion RNA. However, Lu and coworkers 129 753 have recently shown that the enteroviral RNA within infected host cells is represented by an 754 ensemble of 3D structures, which substantially differed depend on the stage of engagement 755 (translation, replication) or when prevented from being packaged. It is, therefore, reasonable 756 to assume that even at the high intracellular K + level, in certain more open conformations of 757 these ensembles, the QGRS will become accessible for the compound, likely leading to 758 stabilisation of the respective GQs. A complete block of PDS binding by intracellular K + is 759 also unlikely based on recent life cell imaging of RNA GQ in the absence or presence of a 760 variant of PDS 130 . Previous experiments with GQs featuring ≥ four G-tetrades showed that 761 they could be unwound by the DHX36 helicase even when bound by GQ-stabilising ligands 762 such as PDS and PhenDC3, with a rate dependent on the thermal stability of the GQ-763 compound complex 131 . We thus believe that the battery of intracellular host cell-derived 764 helicases 132 together with the virus-encoded helicase 2C 133 efficiently disrupt any 765 intracellularly formed, intrinsically weak two-layer rhinoviral GQ even when bound by PDS. 766 In summary, we have shown that targeting QGRS by GQ-stabilising compounds 767 specifically inhibits the uncoating of a common cold virus and provides a mechanistic 768 explanation based on biophysical and ultrastructural analysis. Strikingly, PDS-uptake into the 769 virus occurs in the presence of physiological concentrations of Na + but not K + due to their 770 differential impact on viral RNA compaction rather than GQ formation. Finally, apart from its 771 potential as a new anti-rhinoviral compound, PDS did not interfere with the binding of RV-A2 772 to the host cell and likely preserved the immunogenic epitopes, making these low-infectious

Cells and virus 812
HeLa Ohio cells were originally obtained from ATCC and maintained in DMEM, 813 supplemented with 10 % FBS, 1 % penicillin and streptomycin. Cells were kept in a 814 humidified 5 % CO 2 -containing atmosphere at 37 ˚C. In infection assays, the serum 815 concentration was reduced to 2 % FBS and cells incubated at 34 ˚C, the optimal growth 816 temperature of RV-A2 For virus infection, we used RV-A2, initially acquired from ATCC and 817 propagated and purified following the protocol detailed in 99 . 818

ThT assay for detection of GQ 819
Ribooligonucleotides were diluted in 100 mM potassium or sodium phosphate buffer 820 (pH 7.4) to a final concentration of 5 µM, incubated for 10 min at 90 °C followed by 10 min at 821 4 °C, and mixed with ThT (final concentration 5 µM). Samples were excited at 440 nm, and 822 emission was measured at 490 nm using a PerkinElmer VICTOR Nivo Multimode Plate 823

Reader. 824
For examination of viral RNA, purified RV-A2 (~2 µg) was suspended in 100 mM 825 sodium phosphate buffer or 100 mM potassium phosphate buffer, both at pH 7.4, and the 826 protein shell was digested with 5 µg of proteinase K at 4 °C overnight. The following day, the 827 ex virion RNA samples were ultrafiltered using 100 K Merck Amicon Ultra Filter units 828 according to the manufacturer's protocol, followed by 4 X 400 µl washes with the respective 829 buffers. Samples were mixed with ThT (final concentration 5 µM), and the volume was 830 adjusted to 100 µl. The ThT fluorescence signal was acquired as described above. The 831 sample labelled at 30 °C was maintained at room temperature all the time. The 60 °C 832 labelled sample was incubated at this higher temperature for 10 min and cooled to room 833 temperature on the bench for 30 min, followed by the acquisition of the fluorescence signal. 834

Fluorescent indicator Displacement assay (FiD) 835
ThT-containing ribooligonucleotide samples prepared as above (100 µl Purified RV-A2 (~2 µg) was suspended in 100 mM sodium phosphate buffer or 100 859 mM potassium phosphate buffer, both at pH 7.4, and the protein shell was digested with 5 µg 860 of proteinase K at 4 °C overnight. On the following day, the ex virion RNA samples were 861 ultrafiltered using 100 K Merck Amicon Ultra Filter units according to the manufactures' 862 protocol, followed by 4 X 400 µl washes with the respective buffers. SYTO 82 was added to a 863 final concentration of 5 µM, and the volumes were adjusted to 70 µl with the respective 864 buffers. Three 20 µl aliquots from each of these samples were dispensed into the wells of a 865 thin-walled PCR plate, and the temperature was ramped from 25 -95°C at 1.5 °C / min, and 866 SYTO 82 light-up fluorescence was recorded. Three independent measurements were 867 performed for each condition, and the fluorescence signal means for each condition were 868 displayed. 869

Immunocytochemistry and flow cytometry 875
Cells grown in a 6-well plate until 90 % confluent were infected with 0.1 µg RV-A2 876 either untreated (corresponding to an MOI of 1) or pretreated with PDS (at 20 or 200 µM final 877 concentration) diluted in PBS and incubated for 4 h at 34 °C. The unbound PDS was 878 removed by centrifugal ultrafiltration. The same experiment was carried out with Phen-DC3 879 at 1 µM and 5 µM. At 9 h pi the infection medium was aspirated, cells were washed once with 880 PBS and detached with 0.1 % trypsin in 0.05 % EDTA. The trypsin was inactivated with 10 % 881 FBS in DMEM. Cells were harvested by low-speed centrifugation at 300 g for 3 min at 4 °C. 882 The pellet was resuspended in 500 µl ice-cold PBS, followed by the addition of 500 µl 4 % 883 formaldehyde in PBS and incubation for 10 min at 4 °C. This and all subsequent steps were 884 done with gentle rocking. Cells were subsequently washed 3 times with 1 ml of ice-cold 885 PBST (PBS plus 0.1 % Tween-20, pH 7.4) at 4 °C, resuspended in PBS + 0.1 % Triton X-886 100, and incubated for 10 min at 4 °C. The cells were then incubated in blocking buffer (1 % 887 BSA, 0.1 % Tween-20 PBS (pH 7.4)) for 30 min at 4 °C, followed by incubation with 10 µg/ml 888 8F5, a monoclonal antibody specific for VP2 of RV-A2 137 in blocking buffer for 1 h at 4 °C. 889 Cells were again washed 3 times with PBST and incubated for 1 h with goat anti-mouse 890 AlexaFluor 488 antibody diluted (1 : 1,000) in blocking buffer at 4 °C. Samples were then 891 incubated with Hoechst dye solution in PBS (1 : 2,000) for 10 min for staining nuclei, followed 892 by 3 times washing with PBST. They were finally resuspended in PBS and analysed with a 893 BD Bioscience FACSAria III flow cytometer; more than 10 4 events were acquired for each 894 sample. Forward scattering (FSC) vs VP2 (FITC-A) plots were generated by Tree Star 895 FlowJo X v10.0.7 software. 896

Immunoprecipitation 897
Cells grown in a 10 cm culture plate until ~80 % confluent were infected with ~1 µg 898 RV-A2 pretreated ± 200 µM PDS (as for flow cytometry, see above) at 34 °C; 30 min pi the 899 medium was removed and replaced with 1 ml PBS. The cells were gently detached with a 900 cell scraper (Corning) and subjected to 3 freeze/thaw cycles. Cell debris was removed by 901 low-speed centrifugation, and the supernatant was divided into 2 aliquots. From one aliquot, 902 viral uncoating intermediates (i.e. subviral A-and B-particles) were immunoprecipitated using 903 MAb 2G2 77 bound to protein G magnetic beads (Dynabeads-Protein G; Life Technologies). 904 The second aliquot was taken as a negative control by omitting MAb 2G2 but otherwise 905 processing it identically. As a positive control, ∼1 µg of heated RV-A2 (10 min at 56 °C 906 resulting in an almost ~100 % conversion into subviral B-particles) was processed identically. 907 After extensive washing in PBS, the immunoprecipitates were resuspended in 100 μl PBS. 908 Two µl of 5X protein sample buffer were added to 18 µl of each sample and the mixture 909 heated to 95 °C for 10 min. The proteins were separated by SDS-PAGE (10 %) followed by 910 transfer to an Immobilon-P membrane. The Western blot was done essentially as described 911 in 138 . In brief, the viral protein (VP2) was detected with mouse monoclonal antibody (8F5), 912 anti-mouse-horseradish peroxidase and SuperSignal West Pico PLUS chemiluminescent 913 substrate (Thermo Fisher). The signal was quantified using a ChemiDoc Gel Imaging System 914 (Bio-Rad). To investigate the proportion of A-and B-particles, the viral RNA was also 915 quantified in 50 µl of the remaining respective resuspended immunoprecipitates. As internal 916 control and for normalisation, RNA obtained from 100 µl Aichi virus (AiV; a member of the 917 kobuvirus species of the Picornaviridae family), corresponding to 2 X 10 7 TCID 50 , was added 918 to each sample. RNA was recovered by TRIzol (1 ml; Invitrogen) extraction and precipitation 919 (together with GlycoBlue from Invitrogen) following the manufacturer's protocol. First strand 920 cDNA synthesis was carried out with the NEBNext reagent kit (New England Biolabs, UK) 921 using random primers and the samples quantified by qPCR using primers specific for RV-A2 922 or AiV, using a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). 923

Sucrose gradient sedimentation 924
Cells were infected with RV-A2 pretreated ± 200 µM PDS for 30 min at 34 °C as 925 described above (immunoprecipitation method), the medium was removed, and 1 ml PBS 926 was added, followed by gently dislodging the cells with a cell scraper (Corning). The 927 resuspended cells were subjected to 3 freeze/thaw cycles. Cell debris was removed by low-928 speed centrifugation. Five hundred µl of the resulting supernatants containing the (sub)viral 929 particles were deposited onto preformed 10-40 % (w/v) sucrose density gradients made in 930 virus buffer (5 ml of 50 mM NaCl, 20 mM Tris-HCl, pH 7.4) and centrifuged at 4 °C for 30 min 931 in an SW55 Ti rotor (Beckman) at 286,794 g. Aliquots (250 μl) were collected from top to 932 bottom and frozen at -80 °C until further use. Twenty μl of each fraction were deposited onto 933 a methanol-activated Immobilon-P membrane (Millipore) placed in a 96-well Bio-Dot 934 Microfiltration Apparatus (Bio-Rad) for the dot blot analysis. Viral protein VP2 was detected 935 using the VP2-specific MAb 8F5 and IRDye 680RD Goat anti-mouse IgG secondary antibody 936 essentially as described for the development of the Western blot further above. The 937 fluorescent signal acquisition was performed in an Odyssey Infrared Imager (LI-COR). 938

Time-of-drug addition experiment 939
HeLa cells grown in 6-well tissue culture plates to roughly 80 % confluency were 940 challenged with RV-A2 at MOI = 10 for 30 min in infection medium at 4 °C with steady 941 rocking allowing the virus to attach to its receptor while preventing its internalisation for 942 achieving a synchronised infection. Then, the inoculum was removed, cells were washed 943 three times with PBS, a fresh infection medium was added, and incubation continued at 34 944 °C to trigger virus internalisation (T = 0 min pi). Immediately (T0) or after 60, 180 or 300 945 minutes (T60, T180 and T300 respectively), PDS was added to a final concentration of 20 946 µM. At 9 h pi, cells were processed for flow cytometry and immunocytochemistry as 947 described above. Using Tree Star FlowJo X v10.0.7 software, a FITC-A histogram showing 948 the mean fluorescence intensity (MFI) from more than 10 4 events corresponding to de novo 949 produced VP2 was generated upon gating based on FSC and SSC properties. The typical sample volume was 500 µl. Water suppression for 1 H NMR spectra was 956 performed using a double WATERGATE echo with extra water flip-back pulse to avoid 957 saturation of exchangeable hydrogens due to hydrogen exchange 139 . Typically ca 1000 958 scans each were required to obtain a 1 H NMR spectrum with sufficient signal to noise ratio. 959 For the hydrogen-deuterium exchange (DXH) assay, 0.3 mM G11 RNA was prepared in 10 960 mM sodium phosphate (pH 7.4), 100 mM KCl and diluted 1:3 in D 2 O to 0.1 mM G11 RNA 961 (final concentration) immediately before acquisition The signal attenuation of the imino 962 hydrogens due to hydrogen-deuterium exchange was observed in a series of 1 H NMR 963 spectra throughout 9 h. 964

Circular Dichroism (CD) and melting profile 965
Ribooligonucleotides were used at 20 µM in 100 mM potassium phosphate (pH 7.4) 966 or 100 mM sodium phosphate (pH 7.4), as indicated in the figure. Unfolding and refolding 967 experiments were performed with temperature ramping from 25 -90 °C or 90 -25 °C, 968 respectively, at 1 °C/min using a Chirascan plus spectropolarimeter equipped with a Peltier 969 temperature control system from Applied Photobiophysics. The buffer's CD spectrum was 970 recorded identically and subtracted from the spectrum obtained for the RNA-containing 971 solution. Data were zero-corrected at 400 nm. 972