Mapping RNA-capsid interactions and RNA secondary structure within authentic virus particles using next-generation sequencing

To characterize RNA-capsid binding sites genome-wide within mature RNA virus particles, we have developed a Next-Generation Sequencing (NGS) platform: Photo-Activatable Ribonucleoside Cross-Linking (PAR-CL). In PAR-CL, 4-thiouracil is incorporated into the encapsidated genomes of authentic virus particles and subsequently UV-crosslinked to adjacent capsid proteins. We demonstrate that PAR-CL can readily and reliably identify capsid binding sites in genomic viral RNA by detecting crosslink-specific uridine to cytidine transitions in NGS data. Using Flock House virus (FHV) as a model system, we identified highly consistent and significant PAR-CL signals across virus RNA genome indicating a clear tropism of the encapsidated RNA genome. Certain interaction sites correlate to previously identified FHV RNA motifs. We additionally performed dimethyl sulfate mutational profiling with sequencing (DMS-MaPseq) to generate a high-resolution profile of single-stranded genomic RNA inside viral particles. Combining PAR-CL and DMS-MaPseq reveals that the predominant RNA-capsid sites favor double-stranded RNA regions. We disrupted secondary structures associated with PAR-CL sites using synonymous mutations, resulting in varied effects to virus replication, propagation, and packaging. Certain mutations showed substantial deficiency in virus replication, suggesting these RNA-capsid sites are multifunctional. These provide further evidence to support that FHV packaging and replication are highly coordinated and inter-dependent events. Importance Icosahedral RNA viruses must package their genetic cargo into the restrictive and tight confines of the protected virions. High resolution structures of RNA viruses have been solved by Cryo-EM and crystallography, but the encapsidated RNA often eluded visualization due to the icosahedral averaging imposed during image reconstruction. Asymmetrical reconstructions of some icosahedral RNA virus particles have revealed that the encapsidated RNAs conform to specific structures, which may be related to programmed assembly pathway or an energy-minima for RNA folding during or after encapsidation. Despite these advances, determining whether encapsidated RNA genomes conform to a single structure and determining what regions of the viral RNA genome interact with the inner surface of the capsid shell remains challenging. Furthermore, it remains to be determined whether there exists a single RNA structure with conserved topology in RNA virus particles or an ensemble of genomic RNA structures. This is important as resolving these features will inform the elusive structures of the asymmetrically encapsidated genomic material and how virus particles are assembled.


Abstract 13
To characterize RNA-capsid binding sites genome-wide within mature RNA virus particles, we 14 have developed a Next-Generation Sequencing (NGS) platform: Photo-Activatable 15 Ribonucleoside Cross-Linking (PAR-CL). In PAR-CL, 4-thiouracil is incorporated into the 16 encapsidated genomes of authentic virus particles and subsequently UV-crosslinked to adjacent 17 capsid proteins. We demonstrate that PAR-CL can readily and reliably identify capsid binding 18 sites in genomic viral RNA by detecting crosslink-specific uridine to cytidine transitions in NGS 19 data. Using Flock House virus (FHV) as a model system, we identified highly consistent and 20 significant PAR-CL signals across virus RNA genome indicating a clear tropism of the 21 encapsidated RNA genome. Certain interaction sites correlate to previously identified FHV RNA 22 motifs. We additionally performed dimethyl sulfate mutational profiling with sequencing (DMS-23 MaPseq) to generate a high-resolution profile of single-stranded genomic RNA inside viral 24 particles. Combining PAR-CL and DMS-MaPseq reveals that the predominant RNA-capsid sites 25 favor double-stranded RNA regions. We disrupted secondary structures associated with PAR-CL 26 sites using synonymous mutations, resulting in varied effects to virus replication, propagation, 27 and packaging. Certain mutations showed substantial deficiency in virus replication, suggesting 28 these RNA-capsid sites are multifunctional. These provide further evidence to support that FHV 29 packaging and replication are highly coordinated and inter-dependent events. 30 31 Importance 32 Icosahedral RNA viruses must package their genetic cargo into the restrictive and tight confines 33 of the protected virions. High resolution structures of RNA viruses have been solved by Cryo-EM 34 and crystallography, but the encapsidated RNA often eluded visualization due to the icosahedral 35 averaging imposed during image reconstruction. Asymmetrical reconstructions of some 36 icosahedral RNA virus particles have revealed that the encapsidated RNAs conform to specific 37 structures, which may be related to programmed assembly pathway or an energy-minima for RNA 38 folding during or after encapsidation. Despite these advances, determining whether encapsidated 39 RNA genomes conform to a single structure and determining what regions of the viral RNA 40 genome interact with the inner surface of the capsid shell remains challenging. Furthermore, it 41 remains to be determined whether there exists a single RNA structure with conserved topology in

Introduction 46
Flock House virus (FHV) is a non-enveloped, single-stranded positive-sense RNA (+ssRNA) virus 47 from the family Nodaviridae. The small bipartite genome comprising RNA 1 (3.1kb) and RNA 2 48 (1.4kb) is packaged into a 34 nm non-enveloped T=3 icosahedral virion. Only two non-structural 49 proteins are produced by FHV: the RNA-dependent RNA polymerase (RdRp) and sub-genomic 50 RNA encoded protein called B2. The B2 protein was discovered as the virus's approach to evade 51 the invertebrate anti-viral RNA silencing machinery (1, 2), which thereafter led to the discovery of 52 similar mechanisms in plant cells (3). FHV is perhaps the best studied alphanodavirus and 53 provides a powerful model system by virtue of its small genome size (4.5kb), genetic tractability 54 and ability to infect Drosophila and mosquito cells in culture and whole flies (reviewed in (4, 5)). 55 More recently, FHV has been adapted into medical field. FHV-related vaccine developments 56 utilized either the viral particle as antibody-display system (6), or the viral RNA as trans-57 encapsidated chimeric viral vaccine platform (7-9). 58

Both authentic virions of FHV and the related Pariacoto virus have been reconstructed by cryo-59
EM and X-ray crystallography to reveal highly ordered dodecahedral cages of RNAs (10, 11). The 60 X-ray structure of FHV virion showed electron density at the icosahedral 2-fold axis, which was 61 modelled as an ordered RNA duplex of approximate 20 nucleotides (12). This would account for 62 1800nts (more than one third) of the viral genome, implicating a highly-ordered and specific set 63 of interactions between the viral protein capsid and the encapsidated genome. Interestingly, and 64 recombinantly expressed virus-like particles (VLPs) of FHV also exhibit a similar dodecahedral 65 RNA cage despite packaging predominantly cellular RNAs indicating that viral capsid may either 66 impose structure upon the encapsidated RNA or select for natively structured host RNAs such as 67 ribosomal RNAs (13,14). However, as these structures are obtained with icosahedral averaging, 68 we still do not know what regions or sequences of viral genomic RNA comprise the RNA cage. 69 Furthermore, it remains to be determined whether there exists a single RNA structure with 70 conserved topology in FHV virions, or rather an ensemble of related genomic RNA structures. 71 The FHV encapsidation process also remains largely unknown. One molecule of each RNA 1 and 72 2 is specifically encapsidated into virus particles (15), while subgenomic RNA 3 is excluded (16). 73 Several components of the capsid protein such as the arginine-rich motif and the C-terminal 74 FEGFGF motif have been demonstrated to be essential determinants of packaging specificity of 75 RNA 1, RNA 2,. It was also speculated that FHV packaging process may be in 76 close association with viral replication and/or translational events (20-23). In the virus genome, 77 one stem-loop structure in RNA 2 proximal to 5' end was demonstrated to be required for RNA 2 78 packaging (24). However, it remains unclear whether there are similar packaging sites on RNA 1 79 or 2, and how these sites interact and thus recruit capsid protein to fulfill virus encapsidation. 80 Next-generation sequencing (NGS) in combination with crosslinking techniques provides a high-81 throughput approach to study transcriptome-wide RNA-protein interactions (reviewed in (25) In an analogous fashion to PAR-CLIP, here we applied the same principles to study the interaction 91 of FHV genomic viral RNA in the context of assembled authentic virions. Unlike the complex 92 cellular micro-environment, authentic virions represent a highly simplified enclosure with few well-93 defined components (viral RNA and capsid). Therefore, we are able to screen for specific in virion 94 RNA-capsid interaction events without interference from other cellular components. Furthermore, 95 since viruses can be readily separated from other cellular components, we avoided the need of 96 immunoprecipitation for RNA recovery, and thus largely simplify PAR-CLIP methodology. This 97 method is hence named 'PAR-CL'. 98 Using FHV as a model system, PAR-CL methodology was validated by determining that the 99 increased U to C (U-C) mutation rate was highly specific to crosslink between viral RNA and 100 capsid. We noticed that the intensity of PAR-CL signals was subjected to the dose of 4SU  CL sites showed evidential deficiency in RNA replication, suggesting these sites serve a 111 multifunctional role in both virus packaging and replication. This provides further evidence to 112 support that FHV packaging and replication are highly synchronized and inter-dependent events. 113

Photoactivatable-ribonucleoside-enhanced crosslinking (PAR-CL) in virus particles 115
PAR-CLIP (Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation) 116 is a well-established method for identification of RNA-protein binding sites and can provide 117 nucleotide-resolution through analysis of specific uridine to cytosine transitions that occur at the 118 site of RNA-crosslinking during cDNA synthesis (26,30,31). Here, we simplified the technique 119 by applying a similar approach to purified authentic virions of RNA virus, thereby removing the 120 necessity of immunoprecipitation, and hence deriving the name "PAR-CL". A schematic of the 121 process is illustrated in Figure 1. of each nucleotides found in the mapped reads was enumerated using samtools and the mutation 136 rates at each genomic position was calculated. 137 To validate the PAR-CL methodology, we first sought to determine if there was a substantial 138 increase in U-C mutation rate as a specific consequence of 4SU-capsid crosslink. We performed 139 a series experiments in which wild-type (wt) FHV without 4SU (4SU-) or 4SU-containing FHV 140 (4SU+) were treated with (UV+) or without (UV-) UV irradiation. As illustrated for FHV RNA 1 in 141 Figure 2a, we plot the measured U-to-C (U-C) mutation rate across the genome and calculate 142 the fold change at each U position. A small number of positions, such as nt. 1259 on RNA 1, 143 showed high U-C mutation rates in both conditions. This possibly reflects the selection of a 144 minority variant during virus passaging. Other than this, we did not notice an increased U-C 145 mutation rate for UV-irradiated FHV in the absence of 4SU (4SU-/UV+). This indicated as 146 expected that UV irradiation alone was not sufficient to induce novel U-C mutations. Similarly, we 147 measured the influence of 4SU substitution in FHV genomic RNA without UV exposure 148 (4SU+/UV-) (Figure 2b). We also did not notice an increase in U-C mutation rate. We only 149 observed increased U-C mutations when 4SU and UV irradiation both were present (4SU+/UV+) 150 (Figure 2c). This confirms the elevated U-C mutation rate is indeed a specific result of 4SU-151 induced crosslinking. The FHV RNA 2 data of these experiments is shown in Supplemental 152 Figure S1. 153 Histograms of the U-C mutation rate frequencies at all genomic U positions are shown in Figure  154 2d. This demonstrated that under 4SU+/UV+ condition, more U positions exhibited high U-C 155 mutation rate (≥0.3%) than controls (4SU-/UVand 4SU+/UV-). Interestingly, we noticed reduced 156 U-C mutation rates when wt FHV was exposed to UV (4SU-/UV+), for an unknown reason. We 157 also sought to determine if 4SU incorporation and UV exposure would induce any non-specific 158 (non-U-C) mutations. A histogram of the frequencies of all non-U-C mutations (A,C,G mutations 159 and U-A, U-G mutations) over all genomic positions is shown in Figure 2e. Importantly, 160 4SU+/UV+ FHV did not show any significant change in non-U-C mutations. We therefore 161 conclude that the increased U-C mutation rate is specific to 4SU-induced crosslinking. 162

163
Magnitude of PAR-CL signals was associated with 4SU dose and incubation time. 164 Our PAR-CL method requires no immunoprecipitation to recover and enrich for crosslinked RNAs. 165 However, this permits wild type uridine and/or uncrosslinked 4-thiouridine to persist in the RNA 166 pool which may dilute the PAR-CL signal. To investigate optimum conditions for PAR-CL, we 167 conducted three parallel experiments (Figure 3). S2 cells were infected with FHV and incubated 168 with 4SU at 100µM (4SU16h) or 150µM (4SU1.5X). Viruses were harvested from infected cells The same results were observed when we plotted the frequency of PAR-CL signals for these 177 three experiments, as well as two controls (Figure 3b). This shows that while the mean PAR-CL 178 signals and the distribution under all three conditions (4SU16h, 4SU1.5X, and 4SU40h) and a 179 4SU+/UV-control were all comparable, the magnitude of outliers showed correlation to 180 experimental conditions (4SU40h > 4SU1.5X > 4SU16h). This indicates that only certain 4SU 181 substitution sites are available for crosslink and therefore sensitive to the varied 4SU 182 concentrations/incubation times. Again, for an unknown reason, the 4SU-, UV+ control ( Figure  183 3b) showed a slightly lesser than 1-fold change PAR-CL signal. Importantly, this does not interfere 184 with our interpretation of PAR-CL signals in crosslinking samples. 185 We sampled the top 5% of PAR-CL signals in each conditioned experiment (Figure 3c) and 186 concluded that, the 4SU40h group showed significantly higher PAR-CL signals than the rest.

Consistent PAR-CL signals indicates structural tropism of encapsidated viral RNAs 194
We applied the 4SU40h condition in three parallel experiments: three separate S2 cell cultures 195 were incubated with virus and 4SU, and individually purified viruses were exposed to UV and 196 thereafter proceeded to sequencing. Similar to before, PAR-CL experiments were conducted in 197 pairs, with each PAR-CL dataset comprised of one crosslinked sample (4SU+/UV+) and control 198 with the same sample but without UV irradiation (4SU+/UV-). To ensure reliable mutation rate 199 calculation, we selected for U positions with coverage of at least 10,000 reads. This allowed us 200 to detect reliable mutation profiles over U34 -U3034 on RNA 1, and U9 -U1337 on RNA 2. The 201 PAR-CL signals of these three experiments were compared on each U position on viral RNA 202 genomes (Figure 4a, b). We observed good Pearson's correlation coefficient (≥0.6) between 203 these replicates (Supplemental Figure S4). To validate the consistent PAR-CL signals, the 204 signals over every U position were box-plotted over the triplicates (Supplemental Figure S5).

205
This allows us to readily measure the mean signal strength and signal variation over the triplicates. 206 In order to distinguish reliable crosslinking sites and avoid potential false positives, we removed 207 any PAR-CL signals in crosslinked sample by applying a conservative background threshold filter, 208 retaining only the highest 5% of PAR-CL signals in the uncrosslinked control sample. (Figure 4c, 209 d, Supplemental data 3). Among the most consistent PAR-CL sites (passed background 210 threshold in all three replicates), we identified 20 sites in RNA 1 and 8 sites in RNA 2 that showed 211 the highest average PAR-CL signals (Figure 4e, Supplemental data 3). T-test revealed most of 212 these sites have significantly (P<0.05) higher PAR-CL signal than average. As these same sites 213 consistently displayed significant PAR-CL signals over parallel replicates, this indicates a set of 214 consistent RNA-capsid interactions in authentic FHV virions, which further indicates a structural 215 tropism of FHV RNA in association with the topology of virus capsid shell. 216 217

Probing FHV in virion RNA secondary structures with DMS-MaPseq. 218
We sought to understand if there is any sequence motif among the PAR-CL sites. Significant (>2σ) 219 PAR-CL sites (28 sites from RNA 1 and 15 from RNA 2) and their flanking sequences were 220 analyzed with Discriminative Regular Expression Motif Elicitation (DREME, (39)) for possible 221 sequence motif identification (Supplemental Figure S6). However, no common motif was 222 identified. This led us to hypothesize that the mechanism of RNA recognition by FHV capsids may 223 be related to similar RNA structures rather than sequences. To reliably predict the RNA secondary 224 structures of PAR-CL signals, we sought to determine the secondary structure of FHV RNA in 225 authentic virions. 226 Dimethyl sulfate mutational profiling with sequencing (DMS-MaPseq) provides a reliable and high 227 throughput method to probe RNA secondary structures in vivo (40-42). The resulting constraints 228 provided improvement to thermodynamic map and free energy-based secondary structure 229 prediction. We performed DMS-MaPseq using the TGIRT TM -III enzyme but in combination with 230 ClickSeq to generate RNAseq libraries ("TGIRT-ClickSeq") (Figure 5a), demonstrating that 231 TGIRT TM -III enzyme is compatible with ClickSeq. DMS-MaPseq induces RNA modifications to 232 unpaired adenines and cysteines (and guanine to a lesser level (43)) across the viral genome. 233 Therefore, as expected, in comparison to untreated control virus (DMS-) DMS-treated FHV 234 (DMS+) has a higher average mutation rate over genomic A/C positions (Figure 5b). Similarly, 235 we plotted the frequency of mutation rates over A/C or G/U positions and only noticed a significant 236 higher mutation rate frequency over A/C positions (Figure 5c). We analyzed A or C positions with 237 at least 10k read coverage, which corresponds to nt. 14 -3043 on RNA 1 and nt. 11 -1378 on 238 RNA 2. Similar to PAR-CL data, DMS-MaPseq signal represents the mutation rate fold change 239 between DMS-treated virus and untreated control virus, on all genomic A or C positions. Likewise, 240 we removed potential false-positive signals by applying a background noise threshold, retaining 241 only the genomic sites with mutation rate higher than this threshold. The resulting DMS-MaPseq 242 profile of FHV (Figure 5d and e) showed clear signals up to 100-fold change over both RNA 1 243 and 2. The un-refined DMS-MaPseq profiles with background noise, and mutation rate 244 comparison between DMS-treated and untreated viruses are shown in Supplemental Figure S7. 245

DMS-MaPseq resolved FHV RNA secondary structures reveals that PAR-CL sites favor 247
double stranded structures and are highly clustered. 248 We incorporated the DMS-MaPseq data into free energy based thermodynamic prediction, by 249 introducing a series of "soft" constraints. Only the most significant (>2σ) DMS-MaPseq sites (60 250 sites in RNA 1 and 30 sites in RNA 2) were forced as unpaired constraints in RNAstructure Web 251 Server (44) with "Fold" algorithm (44, 45). Regardless of their DMS-MaPseq signals, the remaining 252 genome positions were left without any constraints, to allow maximum prediction flexibility. We 253 thereby constructed a DMS-MaPseq-imposed secondary structure map of complete FHV RNA 254 genome (snapshots in Figure 6, full-scaled maps of RNA 1 and RNA 2 were also provided in 255 Supplemental data 1 and 2). Despite the low number of introduced constraints, we were able to 256 greatly improve the thermodynamic mapping of FHV RNAs. With the 60 RNA 1 constraints, 37% 257 (1145/3107) of nucleotides underwent refolding compared to the unconstrained model, yielding 258 different paired/unpaired patterns. Similarly, with the 30 RNA 2 constraints, 20% (273/1400) 259 nucleotides underwent refolding. The dot-bracket maps comparing the differences between 260 unconstrained and constrained folding can be found in Supplemental Figure S8. 261 In combination with PAR-CL data, we noticed that the significant PAR-CL sites heavily favored 262 double-stranded base-pairing. In RNA 1, among the 28 most significant (>2σ) PAR-CL sites, 22 263 are located in double-stranded regions, whereas 3 sites were 1 nt. adjacent to double-stranded 264 stems. In the much shorter RNA 2, 8/15 of most significant PAR-CL sites are located in dsRNA 265 stems, whereas 3 are 1 nt. adjacent. In Table 1, we illustrate the detailed structures of 16 PAR-266 CL sites (11 on RNA 1 and 5 on RNA 2) that presented with highest consistency and average 267 PAR-CL signals (Figure 4a-d). 268 We also noticed that the distribution of PAR-CL signals was uneven and highly clustered. 269 Numerous PAR-CL stems showed more than one PAR-CL sites with >1σ significance 270 (Supplemental data 1, 2, some examples were listed in Table 1). We calculated the average 271 shortest distance between adjacent PAR-CL sites. On RNA 1 (3107 nt.), among 721 uridine sites, 272 102 showed >1σ significant PAR-CL signal. The average shortest distance between these PAR-273 CL sites is 7.4 nucleotides, which is substantially shorter than the average shortest distance of 274 102 random uridines (30.47 nucleotides). On RNA 2 (1400 nt. genome with 351 uridines), the 275 average shortest distance among 45 PAR-CL sites (>1σ significance) was 8.8 nt., which is also 276 shorter than the average shortest distance of 45 random uridines (31.11 nt.). 277 Notably, by combining PAR-CL data and DMS-MaPseq-imposed RNA structure, we are able to 278 characterize a stem loop site which is structurally near identical to a previously predicted stem 279 loop (nts. 168-249) on RNA 2 (24) (Supplemental Figure S9). This stem loop site, as well as the 280 flanking sequence (nt. 210-249) has been determined to be essential for RNA 2 encapsidation. 281 We identified three PAR-CL signals within this region, consistent with role of this stem loop site in 282 RNA 2 packaging. 283 284 285

Structurally-disrupted PAR-CL sites impact FHV lifecycle and fitness.
To determine whether the identified PAR-CL sites have a biological function, we selected 11 PAR-287 CL sites from RNA 1 and 5 PAR-CL sites from RNA 2 as our candidate sites (Table 1). Referring 288 to the DMS-MaPseq-corrected FHV RNA structure maps (Figure 6), we introduced synonymous 289 mutations to disrupt the double-stranded RNA regions of the PAR-CL sites (or the nearest stem 290 of certain PAR-CL sites, i.e. U2515 on RNA1, U534, U903, U968, and U1155 on RNA2). The 291 predicted structure of these PAR-CL sites, primers, and replaced nucleotides are listed in Table  292 1. Plasmids containing these point mutations were transfected into S2 cells. Each transfection 293 consisted either a mutated RNA1 and wild-type RNA2, or a mutated RNA2 and wild-type RNA1 294 (Figure 7a). After transfection, induction and incubation, cell viability of each transfected mutant 295 was determined with alamarBlue assay (Figure 7b). Almost all mutant virus transfections showed 296 reduced cytopathic effect compared to transfection with wild-type FHV RNA. Notably RNA 1 297 mutants U159, and U1233 resulted in little to no detrimental effect to S2 cells. 298 Total cellular RNA was extracted from transfected cells and in-column DNase digestion was 299 conducted to remove remaining plasmids. From each transfection, 200ng of purified RNA was 300 used as template for RT-PCR to detect FHV RNA (Figure 7c). We noticed that accumulation of 301 FHV RNA 2 was unaffected by any PAR-CL mutants, while RNA 1 accumulation varied drastically 302 among RNA 1 mutants. Notably, RNA 1 mutant U1233 yielded undetectable levels of RNA 1 and 303 RNA 3, while RNA 2 production was less affected. RNA 1 mutant U159 also produced marginal 304 amount of RNA 1 and RNA 3, and U227 produced substantially less RNA 1 than that of control 305 or RNA 2 mutants. The replication deficiency of these three mutants agreed with our findings of 306 their low virulence (Figure 7b). Interestingly, these three sites are found within or adjacent to 307 previously described FHV RNA regulatory regions (46, 47) (Figure 8). The importance of these 308 three sites in both RNA-capsid interaction shown here and RNA replication regulation indicates 309 that the same motifs in the RNA genome are involved in multiple stages of the viral life-cycle, 310 consistent with the notion that replication and RNA genome packaging are tightly coupled 311 processes (20, 21). 312 To confirm capsid production, we separated cells and supernatant from the transfected cells.

313
Western blots with anti-FHV were used to detect capsid proteins in both cellular components and 314 supernatants (Figure 7d). In the cellular fraction, we readily detected both mature (alpha peptide) 315 and autoproteolytically cleaved capsid protein (beta peptide) in all mutants. However, reduced 316 capsid yields were found in U159 and U1233 mutants, possibly due to the observed RNA 1 317 replication deficiency. In supernatant fractions, the U159 and U1233 mutants resulted in 318 undetectable level of capsid protein, while U227 resulted in detectable but very marginal amount 319 of capsid production. This confirmed that the mutations at these three PAR-CL sites have 320 significant impact on virus production in S2 cells. 321 To expand mutant viruses, we further inoculated naïve S2 cells with equal amount of transfected 322 p0 cell mix. From the inoculated P1 cell culture, we observed different degrees of cytopathic effect 323 (CPE) under microscope (Supplemental Figure S10), which was correlated to earlier findings.

324
P1 mutant viruses were nuclease treated, PEG precipitated, and purified with PES membrane 325 protein concentrator to remove potentially unassembled capsid subunits. The presence of virus 326 particles was confirmed with SDS-PAGE (Figure 7e), and virus yield was calculated by 327 densitometry. Similar to before, we failed to detect virus production of U159 and U1233 mutants, 328 while U227 mutant resulted a marginal virus production which can only be detected by western 329 blot but not with SDS-PAGE. This result also agreed with our western blot analysis (Figure 7d). 330 We further tested P1 mutant virus relative virulence by infecting cells with mutants at MOI = 1 331 (Supplemental Figure S11). Most mutant viruses still resulted in varied but inferior virulence, in 332 comparison to wild type virus. 333 In this study, we demonstrated that PAR-CL can be used as a reliable and convenient method to 335 screen for capsid-interacting sites on viral RNA genomes. PAR-CL data analysis features low 336 background noise and thus, highly distinguishable signals. Therefore, PAR-CL signals are highly 337 specific and representative of consistent crosslinking events between virus RNA and capsid. We 338 showed that under the optimized experimental condition (4SU40h)

PAR-CL methodology 353
Photoactivatable nucleoside analogs were successfully utilized in the past to enhance 354 crosslinking efficiency and hence, providing approaches to study RNA-RNA and RNA-protein 355 interactions (reviewed in (48)). Thionucleobases such as 4-thiouracil (4SU) and 6-thioguanosine 356 (6SG)) allows for highly effective crosslinking at 330-365 nm excitation spectrum (49), as well as 357 advantages such as minimum nucleoside structure perturbation (48, 50), lower cytotoxicity (26, 358 32, 33, 51), and less photochemistry and/or photodamage (48, 50). Importantly, the 4SU/6SG 359 incorporated RNA can lead to specific base mismatches during reverse transcription (U-C, and 360 G-T)(27-29), which enables high-throughput screening as indications of crosslinking. This is best 361 illustrated with PAR-CLIP (PhotoActivatable-Ribonucleoside-enhanced CrossLinking and 362 ImmunoPrecipitation) technology (26), which allows for pinpointing crosslinking sites at nucleotide 363 resolution. PAR-CLIP has been successfully applied in the past to identify crosslinking sites of 364 Argonaute 2, embryonic lethal abnormal vision (ELAV) protein and pumilio homologue 2 (PUM2), 365 insulin-like growth factor proteins (26, 52 (between 100-400 x 10 6 cells (30)). 374 The unique aspect of our simplified PAR-CL (PhotoActivatable-Ribonucleoside-enhanced 375 CrossLinking) method is that we applied the similar PAR-CLIP principles to an RNA virus (FHV), 376 which can be easily separated from cellular components. Crosslinks within purified virus particles 377 allow us to: (1) eliminate the need for immunoprecipitation to recover crosslinked RNA; (2) look 378 for specific in virion interactions between viral RNA genomes and viral capsid proteins; (3) study 379 a reductionist and highly controlled microenvironment. The greatly simplified PAR-CL 380 methodology, in combination with ClickSeq library construction technology (36), granted the ability 381 to conduct an experiment with as little as 2 µg of purified FHV particles. A single T25 flask of S2 382 cells can generate ample amount of pure 4SU-containing viruses to conduct multiple PAR-CL 383 experiments. 384 In our PAR-CL method, the final pool of purified viral RNA can comprise large number of wild type 385 uridines, or uncrosslinked 4SUs. As a consequence, the signal of any randomly generated, non-386 specific crosslinking event will be largely diluted into background level. Only the consistent 387 crosslinking sites present due to homogeneity in RNA-capsid interactions within a viral population 388 can readily provide distinguishable PAR-CL signals from background. Therefore, in contrast to 389 the canonical PAR-CLIP approach where only cross-linked RNA fragments are sequenced, we 390 are also able to identify regions of the viral genome where there is no reproducible PAR-CL signal, 391 either due to a lack of RNA-capsid interactions or heterogeneous interactions. This is best 392 illustrated in Figure 3b, where the background noise levels are largely unchanged, with or without 393 crosslinking. 394 In both PAR-CLIP and PAR-CL, there are intrinsic limitations of 4SU-induced crosslinking. Firstly, 395 crosslinking is only limited to U positions. Any potential interaction between protein and other 396 nucleotides is undiscoverable. Next, 4SU crosslinking with protein is affected by reactivity of 397 amino acid side chains (27, 29), with aromatic amino acids (phenylalanine, tyrosine, and 398 tryptophan) being predominant targets but also lysine and cysteine (27). Consequentially, not all 399 RNA-protein interactions can be depicted by PAR-CL or PAR-CLIP, and certain interactions may 400 not result in crosslinking. 401

FHV PAR-CL experiments and data analysis 403
Several approaches were used to ensure reliable interpretation of PAR-CL signals on FHV: 1) to 404 ensure reliable interpretation of mutation rate, we limited our analysis to genomic positions with 405 at least 10k coverage. For this reason, our FHV PAR-CL experiments reliably covered U34 -406 U3034 on RNA 1, and U9 -U1337 on RNA 2. However, it is possible that we omitted potential 407 capsid interaction sites out of our analyzed range. 2) We previously noticed that certain point 408 mutations may be selected by virus and could be associated with defective interfering RNA 409 generation (34). In this study, we also noticed substantially increased mutation rates on certain 410 genomic positions (such as U1259 on RNA1, as illustrated in Figure 2a-c). Thus, to eliminate 411 virus intrinsic mutational events, we avoided to use U-C mutation rate as a measurement. Instead, 412 we decided to use fold change of U-C mutation rate, between crosslinked virus and uncrosslinked 413 virus control, as our PAR-CL signals. 3) Because our PAR-CL signal corresponds to the fold 414 change of U-C mutation rates of two datasets, a substantial PAR-CL signal can be a consequence 415 of three scenarios: a high U-C rate in crosslinked virus, a much lower U-C rate in uncrosslinked 416 control, or both. To minimize the possibility of false positives, we introduced a background 417 threshold. Only the PAR-CL signals above this threshold were taken into our further consideration, 418 as they represent mutation rates distinguishable from background fluctuation range (illustrated in 419 Supplemental data 3). Together, we believe these three quality control measurements provided 420 stringent analysis to our PAR-CL data to reveal truly biologically relevant FHV RNA-capsid 421 interaction sites. 422

DMS-MaPseq and FHV secondary structure mappings 424
Several studies have proposed lowest free energy-based FHV local or whole genome secondary 425 structure predictions, with the focus on viral RNA intracellular arrangement and replication 426 regulations (24,46,47,53). In vivo RNA chemical probing methods such as DMS and SHAPE 427 allow for structure-specific chemical modifications to be screened by next generation sequencing 428 techniques (40, 54, 55). Using DMS-MaPseq in authentic FHV virions, we are able to provide 429 experimental validation of the RNA structures inside virus particles. With the same rationale of 430 PAR-CL, we also applied stringent quality control measurements to ensure reliable interpretation 431 of mutational profiles generated by DMS-MaPseq: A/C error rates were only analyzed over 432 positions with at least 10k coverage (A14-A3043 on RNA 1, and C11-A1378 on RNA 2); fold 433 change of A/C mutation rate was regarded as DMS-MaPseq signals instead of actual mutation 434 rate; similar background noise threshold was also applied to prevent potential false positives. 435 Canonically, DMS-MaPseq data is imposed upon thermodynamic prediction by enforcing 436 unpaired constraints on any position with a signal above a given threshold (40)

Flock house virus PAR-CL sites and biological indications 447
It has been observed previously that the RNAs of FHV, as well as other Nodaviruses, form a 448 highly ordered dodecahedral cage inside virus particles (10, 56). However, it was not clear 449 whether the dodecahedral RNA cage had a fixed topology. From our PAR-CL data (Figure 4a-d), 450 we can clearly identify highly consistent RNA-capsid interactions over certain genomic positions 451 among multiple replicate experiments. This provides evidence that there is well-defined tropism 452 between FHV RNA cage and capsid shell, at least at a these sites identified here. Among the 453 most consistent and distinguished PAR-CL sites (Figure 4c, d), we noticed that they exhibited a 454 highly clustered pattern. The clustering effect is more pronounced, when taking RNA secondary 455 structures into consideration (Figure 6 and Supplemental data 1, 2). 456 The multiple RNA-capsid interaction sites spanning the whole FHV genome suggest the 457 possibility that FHV encapsidation may require multiple packaging signals to assembly the entire 458 virus genome. Flock House virus genome packaging process may be similar to the two-staged 459 packaging mechanism of MS2 bacteriophage (57). Numerous synergetic high-affinity RNA-capsid 460 interaction sites are required to recruit capsid subunits. These widely-distributed interaction sites 461 facilitate capsid-capsid interactions, which reciprocally mediate RNA folding and tertiary 462 compression of RNA genome. Subsequently, this can be followed by continuous recruitment of 463 capsids on folded RNA to finalize encapsidation process. 464 Several PAR-CL sites also aligned with, or in close adjacent to, known RNA motifs (Figure 8). 465 On RNA 1 (Figure 8a), we could not align any candidate PAR-CL signal to subgenomic RNA 3, 466 which suggests the possibility that the exclusion of RNA 3 during packaging is due to lack of 467 strong RNA-capsid interactions. Interestingly, two most significant PAR-CL sites on RNA 1, U159 468 and U1233, aligned with previously discovered replication regulatory elements: a 5' cis element 469 (nts. 68-205) that is essential for RNA 1 replication and mitochondria-targeting(46), and short 470 distal subgenomic control cis-element (nts. 1229-1239) which mediates subgenomic RNA 3 471 replication (47). Furthermore, U2515 and U2576 were located in the subgenomic promoter region 472 (47, 58) which are also adjacent to a RNA 1 internal cis-acting replication element (intRE, nts 473 2322-2501) (47). Similarly, on RNA 2 (Figure 8b), we noticed PAR-CL site U534 is adjacent to a 474 RNA 2 cis-acting regulatory site (59), and U1155 which is within a site required for specific 475 packaging of both RNAs (18). A previously predicted stem loop site (nts 168-249) on RNA 2 476 serves as a RNA 2 packaging signal (24). This is also the only established FHV RNA packaging 477 signal to date. Our DMS-MaPseq map did predict near identical stem loop structure as previous 478 proposed and we noticed three significant (>1σ) PAR-CL sites were clustered in this critical region 479 (Supplemental Figure S9). Since these RNA-capsid interaction sites are correlated to RNA 480 cellular replication/mitochondrial targeting sites, we suggest they might be multi-functional in virus 481 life cycle, and there can be a strong synergy between protein A-mitochondria localization (9, 60, 482 61), RNA replication, and virus assembly. 483 It was speculated previously that FHV packaging and replication are coordinated events. When 484 FHV and brome mosaic virus (BMV) were co-expressed in plant cells, assembled virions only 485 packed their own respective viral RNAs (20). Intracellular protein-protein interactions between 486 FHV replicase (protein A) and capsid were detected (21). It has also been shown that FHV can 487 ensure genome assembly specificity only when capsids were translated from replicating viral 488 RNAs (23). It was hence suggested that FHV encapsidation may be coupled with the RNA 489 replication. Our FHV PAR-CL experiments directly implicated only one aspect of FHV biology: the 490 RNA sites interacting with capsid proteins. However, upon further analysis and mutational assays, 491 a number of PAR-CL sites clearly indicated their significance in FHV replication and regulation: 492 U159 and U227 mutants showed severe deficiency of RNA 1 replication and virion production, 493 while U1233 entirely abolished viral replication. This provides further evidence that FHV 494 replication and packaging are not sequentially separated events, but rather a synchronized, highly 495 inter-dependent processes. Furthermore, these RNA-capsid interactions are not only important in 496 post-replicational/translational RNA packaging, but may also be essential for multiple aspects of 497 virus early stage activities in host cells. 498

D. melanogaster (S2) cells were regularly maintained and passaged with Schneider's Drosophila 501
Media (Gibco) containing 10% fetal bovine serum, 1 × Antibiotic-Antimycotic (Gibco), 1 × MEM 502 non-essential amino acids solution, and 1 mM sodium pyruvate. 503 As described previously (34), wt Flock House virus (FHV) was generated by transfecting S2 cells 504 with pMT plasmid vectors (Invitrogen) containing respective genomes (NC_004146 for RNA 1, 505 and NC_004144 for RNA 2). Copper sulfate was used to induce the promoter 24 h post 506 transfection, while viruses were allowed to accumulate until 3 days post induction to yield passage 507 0 (p0) virus/cell mixture. The p0 transfected cells and viruses were then used to inoculate naïve 508 S2 cells in a T75 flask for 3 days to yield passage 1 viruses, which were purified and used as FHV 509 inoculum in this study, unless otherwise mentioned. 510 All virus transfections, infections, and passages with S2 cell culture were maintained in 27°C 511 incubator, unless otherwise mentioned. 512 To purify FHV, 1% Triton X-100 was added to the cell culture containing p1 viruses. Cell culture 513 underwent one freeze-thaw cycle, and cell debris was removed with 3000 × g centrifugation. FHV 514 in the supernatant was crudely purified with 4% polyethylene glycol (PEG) 8000 and centrifuged 515 (6000 × g) to remove debris (8). This was followed by DNase I and RNase A overnight digestion, 516 to remove any co-precipitated cellular DNA or RNA. Unless otherwise mentioned, viruses were 517 further purified with a 10-40% sucrose gradient, and ultracentrifuge at 40,000 RPM for 1.5 h. 518 Viruses were then concentrated with 100K MWCO polyethersulfone (PES) membrane protein 519 concentrator (Pierce) and washed three times with 10mM Tris pH 7.4. 520

PAR-CL and ClickSeq 522
S2 cells were maintained in T75 flask until 70% -90% confluency. Cells were infected with purified 523 Flock House virus (p1) at MOI = 1 (34, 62, 63). As an initial dose, 4-thiouridine (Sigma-Aldrich) 524 was supplemented to the cell culture to 100µM as 1 × concentration with virus. An optional "boost" 525 dose of 4-thiouridine can also be supplied 16 h post infection (Figure 3a). Cells and viruses were 526 harvested at 16 or 40 h post infection (Figure 3a). Viruses were purified with methods described 527 above. 528 The nuclease-treated and purified 4SU-containing viruses were placed uncovered over ice and 529 irradiated with 0.15 J/cm 2 (26, 30) of 365nm UV light (3UV-38, UVP). After crosslink, viruses were 530 digested with 8U of proteinase K (NEB) at 37 °C for 30 min. Crosslinked RNAs were extracted 531 and purified with RNA Clean & Concentrator (Zymo Research) to yield RNA template for 532 4SU+/UV+ sequencing library sample. 533 Unless otherwise mentioned, the same 4SU-containing virus without any UV irradiation were 534 prepared in the same way to give RNA template for 4SU+/UV-control library. 535 Both the crosslinked and uncrosslinked viral RNA were used to construct the ClickSeq Illumina 536 libraries per standard ClickSeq method, which is detailed previously (34-36). 250ng of RNA template was used in reverse transcription reaction with 1:35 Azido-NTPs:dNTPs ratio and 538 SuperScript III reverse transcriptase (Invitrogen). 539 Equal molar of each indexed library was pooled and run on a HiSeq 1500 platform (Illumina), with 540 single read rapid run flowcell for 1x150 reads and 7 nucleotides for the index. 541 542

DMS-MaPseq and ClickSeq 543
Dimethyl sulfate (DMS) RNA methylation method was described previously (40,42). In this study, 544 nuclease-treated and purified FHV was supplemented with DMS to 5% final concentration. After 545 5 min incubation at 30°C, reaction was quenched on ice for 5 min with 2 volumes of 10mM Tris 546 pH 7.4 and 30% 2-Mercaptoethanol (BME). RNA extraction was conducted with Quick-RNA Viral 547 Kits (Zymo Research) with additional BME in the extraction buffer. The DMS-control sample 548 comprises the same virus stock with identical treatments as above, but without DMS supplement. 549 Methylated FHV RNA and respective controls were proceeded with similar ClickSeq library 550 construction method. One exception is the use of a high-fidelity and processive thermostable 551 group II reverse transcriptase enzyme (TGIRT-III, InGex) during reverse transcription. 100U of 552 TGIRT-III was mixed with 250ng of RNA template, 0.5mM of AzNTPs/dNTPs mixture 553 (AzNTPs:dNTPs = 1:35), and the following reaction conditions: 5 mM Dithiothreitol (Invitrogen), 554 10 U RNaseOUT (Invitrogen), 50 mM Tri-HCl pH 8.0, 75 mM KCl, and 3 mM MgCl2. The reaction 555 mix was incubated in room temperature for 10 min, followed by 57°C incubation for 1.5 hrs, and 556 75°C termination for 15 min. The terminated reaction was then digested with RNase H to remove 557 RNA template. The purified cDNA was proceeded with click reaction with Illumina adapters and 558 final PCR amplification with indexes. 559 Library pooling and Illumina sequencing platform are the same as above. 560 561

Bioinformatics and data analysis 562
The Illumina sequencing data of both PAR-CL and DMS-MaPseq were subjected to the following 563 bioinformatic pipelines: first the Illumina sequencing adapter sequence "AGATCGGAAGAGC" 564 was trimmed with cutadapt (37) (command line parameters: -b AGATCGGAAGAGC -m 40); then, 565 we used FASTX toolkit (http://hannonlab.cshl.edu/fastx_toolkit/index.html) to remove the 566 remaining random nucleotides from the Illumina adapter sequence and random base-pairing as 567 a result of azide-alkyne cycloaddition from cDNA fragments (command line parameters: 568 fastx_trimmer -Q33 -f 7); a further quality filter was applied to remove any reads that contained 569 more than 4% nucleotides with a PHRED score <20 (command line parameters: 570 fastq_quality_filter -Q33 -q 20 -p 96). The remaining reads were aligned to FHV genomes 571 (NC_004146, and NC_004144). Data generated from PAR-CL experiments were aligned for end-572 to-end matches with Bowtie (v1.0.1) (38) (command line parameters: -v 2 --best). Data generated 573 from DMS-MaPseq experiment were aligned with Bowtie2 (64) to allow longer and gapped 574 alignments (command line parameters: --local). Using SAMtools (65), the aligned reads were 575 binarily converted, merged, indexed, sorted, and mathematically noted. 576 For both PAR-CL and DMS-MaPseq, we excluded any nucleotide location with less than 10k 577 coverage to ensure reliable mutation rate calculation. For PAR-CL, we calculated the mutation 578 frequencies of each of the four nucleotides, as well as the U-C mutation rate at each genomic U 579 position. For DMS-MaPseq, similar analysis was conducted but we focused on the overall 580 mutation rates of A and C genomic positions. Between test group and respective control 581 (4SU+/UV+ and 4SU+/UV-for PAR-CL, DMS+ and DMS-for DMS-MaPseq), we compared the 582 mutation rate at the same genomic position, to yield the fold change map, as presentations of 583 PAR-CL or DMS-MaPseq signals. 584 A background filter was applied to both PAR-CL triplicates (Figure 4a-d) and DMS-MaPseq data 585 (Figure 5d, e), to ensure reliable data analysis and avoid potential false positives. For PAR-CL, 586 the background threshold is determined by bottom 95% of U-C mutation rate in the uncrosslinked 587 control group (4SU+,UV-). In the correspondent crosslinked group (4SU+,UV+), we removed any 588 datapoint with U-C mutation rate below this threshold, as it is indistinguishable from background 589 fluctuation. An example of applying background threshold for PAR-CL data can be found in 590 Supplemental data 3. For DMS-MaPseq, similar background threshold was determined as the 591 bottom 95% of A/C mutation rate, in the DMS-untreated control group (DMS-). Only the datapoints 592 passed the background threshold were used to compile the fold change maps of mutation rate 593 changes. 594 The raw sequencing data for both PAR-CL and DMS-MaPseq experiments are available in the 595 NCBI sequence read archive (SRA) with accession number: PRJNA554838. 596 597 RNA secondary structure prediction 598 RNA secondary structure prediction was conducted with RNAstructure (44) with 310.15 K 599 temperature and maximum loop size = 30. "Fold" (44, 45) and "Partition" (66) were used to 600 prediction the structure of RNA and calculate the base pairing probability, respectively. The most 601 significant DMS-MaPseq signal sites were applied as unpaired constraints in structure prediction.

602
No other constraints applied to the rest genomic sites, regardless of the DMS-MaPseq signals, to 603 ensure the flexibility of algorithm. The predicted structure file was then re-organized and certain 604 nucleotides were highlighted for graphical purposes with StructureEditor (v. Polymerase, NEB), with disrupted RNA structure at each selected PAR-CL sites. These 614 overlapped fragments were then cloned into competent cells with standard In-Fusion HD Cloning 615 (TaKaRa) techniques. The plasmids containing mutated FHV RNA 1 or RNA 2 sequences were 616 then sanger-sequenced and mutation sites were confirmed. 617 To generate mutated viruses, the plasmids containing PAR-CL site mutations were used to 618 transfect S2 cells with above-stated methods. Each mutant transfection consisted of equal 619 amount of one mutated RNA genome with disrupted PAR-CL site, and wt genome of the other 620 RNA (Figure 7a). These p0 mutant viruses were allowed to propagate in cell culture until 3 days 621 post induction. Similar to before, p1 mutant viruses were generated by inoculating naïve S2 cells 622 with p0 cell culture/virus mix. 623 624

Relative virulence of mutant viruses 625
The virulence of p0 PAR-CL mutant viruses was measured via transfecting S2 cells with plasmids 626 containing

SDS-PAGE and western blot 655
After collecting p0 transfections, the cell/virus/supernatant mix was centrifuged at 1000 × g for 10 656 min. Supernatant fraction was removed and collected thereafter. The cell pellet was washed once 657 with 1 × PBS and centrifuged as before. The washed cellular fraction was then resuspended in 1 658 × PBS and 1 × cOmplete (Roche). 150 µL of supernatant (of each sample) was supplemented 659 with 1 × cOmplete and then reduced with vacuum centrifuge prior to SDS-PAGE. 660 All SDS-PAGE assays were conducted with Bolt 4-12% Bis-Tris Plus Gels (Invitrogen).  4-thiouridine (4SU) were supplemented to S2 cells, during Flock House virus (FHV) infection. After incubation, purified viruses are irradiated with 365nm UV to induce crosslink. After proteinase K digestion, crosslinked RNAs are purified and subjected to ClickSeq with Azido-NTPs. In viral genome, the crosslinked sites are characterized with elevated U to C mutation rates.  . U-C mutation rate elevation is specific to 4SU crosslink. Using FHV RNA 1 as an example, several control experiments were conducted to ensure specificity of PAR-CL signals. (a) effects of UV exposure to wt FHV (4SU-) were compared. We did not observe significant U-C mutation rate elevation. (b) effects of 4SU incorporation were compared to FHV without UV exposure (UV-). We did not observe significant U-C mutation rate elevation. (c) only when 4SUcontaining FHV was irradiated with UV, we observed a significant increase in U-C mutation rates. (a-c) orange line represents the average fold change in each experiment. (d) the distribution of U-C mutation frequency: only when both 4SU and UV both presented, did we notice a shift towards higher U-C mutation rates.
(e) crosslink did not impact the rates of other mutations (A,G,C-mutations, or U-A/U-G mutations).  Figure 3. PAR-CL signal intensities correlated with 4SU dose and incubation time. Using FHV RNA 1 as an example, (a) three experimental conditions were tested for impact to PAR-CL signals. We observed that the intensities of PAR-CL signals (i.e. fold change of U-C mutation rate) were related to the concentration of 4SU in cell culture and the time of incubation. (b) with or without crosslink, the average PAR-CL signals among all 4SU-containing FHVs is similar. However, the outliers of crosslinking groups showed a significant higher PAR-CL signals than that of control (4SU+,UV-), and the magnitude of outliers correlated with 4SU concentration and incubation time. (c) We sampled top 5% PAR-CL signals from three experiment groups and determined that optimal PAR-CL signal was achieved under 4SU40h condition, which was applied to all further PAR-CL experiments.  Triplicated PAR-CL signals of FHV RNA 1 and 2 were box-plotted. We removed any PAR-CL signal failed to pass the background threshold. A number of sites on both RNA 1 and RNA 2 showed consistently significant PAR-CL signals, indicating reliable crosslinking sites between RNA and protein. These consistent PAR-CL sites also suggest a strong tropism of FHV RNA cage inside virion. X-axis is not continuous. (e) Among these consistent PAR-CL sites, most of them showed significantly higher PAR-CL signals than the average. *P<0.05; **P<0.01; NS=not significant. .   Snapshots of RNA 1 (left) and RNA 2 (right) are shown. Full scaled maps can be found in Supplementary data 1 and 2. PAR-CL signal sites of different significance were color annotated. The introduced DMS-MaPseq constraints were highlighted by lower case "a" or "c" in red color.  Coomassie-stained α-tubulin as loading control for cellular assay. (e) p1 viruses were purified and filtered with 100 K molecular weight filter, mutant virus production was verified with SDS-PAGE gel. Heat shock protein 70 (hsp70) shown as a loading control. (a) On RNA 1, four highly conserved defective interfering RNA regions (DI regions 1-4)(34) were shown in pink; four RNA 1 replication regulatory elements were shown in orange (5'Cis (46), 5'intRE(47), 3'intRE(47), and (67)); two RNA 3 regulatory elements and a putative RNA 3 subgenomic promoter region were shown in green (RNA3cis(dis) (47), RNA3cis(prox)(47)); the candidate PAR-CL sites (listed in Figure 4e) were shown as red bars on bottom. (b) On RNA 2, three conserved defection interfering RNA regions were shown in pink (DI regions 1-3 (34)); a mid-genome cis-acting replicational element(59) and a 3' cis-acting regulatory element(67) were shown in brown; a RNA 2 packaging signal (24) and a capsid site essential for RNA 1&2 specificity (18) are shown in green.