key: cord-0880104-ammf0hff
authors: Huston, Nicholas C.; Wan, Han; Strine, Madison S.; de Cesaris Araujo Tavares, Rafael; Wilen, Craig; Pyle, Anna Marie
title: Comprehensive in-vivo secondary structure of the SARS-CoV-2 genome reveals novel regulatory motifs and mechanisms
date: 2021-01-01
journal: Mol Cell
DOI: 10.1016/j.molcel.2020.12.041
sha: ddccbc9c209235bd3216b23cea0ad33ffa8aaad6
doc_id: 880104
cord_uid: ammf0hff

SARS-CoV-2 is the positive-sense RNA virus that causes COVID-19 disease. The genome of SARS-CoV-2 is unique among viral RNAs in its vast potential to form RNA structures and yet, as much as 97% of its 30 kilobases have not been structurally explored. Here, we apply a novel long amplicon strategy to determine for the first time the secondary structure of the SARS-CoV-2 RNA genome at single-nucleotide resolution in infected cells. Our in-depth structural analysis reveals networks of well-folded RNA structures throughout Orf1ab, and reveals aspects of SARS-CoV-2 genome architecture that distinguish it from other RNA viruses. Evolutionary analysis shows that several features of the SARS-CoV-2 genomic structure are conserved across beta coronaviruses and we pinpoint regions of well-folded RNA structure that merit downstream functional analysis. The native, secondary structure of SARS-CoV-2 presented here is a roadmap that will facilitate focused studies on the viral life cycle, facilitate primer design, and guide the identification of RNA drug targets against COVID-19.

SARS-CoV-2 is the positive-sense RNA virus that causes COVID-19 disease. The genome of 26 SARS-CoV-2 is unique among viral RNAs in its vast potential to form RNA structures and yet, as much 27 as 97% of its 30 kilobases have not been structurally explored. Here, we apply a novel long amplicon 28 strategy to determine for the first time the secondary structure of the SARS-CoV-2 RNA genome at 29 single-nucleotide resolution in infected cells. Our in-depth structural analysis reveals networks of well- 

Two independent biological replicates of in cell SHAPE-MaP data were generated and analyzed 120 using the ShapeMapper pipeline (Smola et al., 2015b) . Comprehensive datasets were obtained, with 121 median effective read depth > 70,000x and effective reactivity data for 99.7% (29813/29903) of 122 nucleotides in the SARS-CoV-2 genome in both replicate experiments. To check the SHAPE-MaP data 123 quality, we analyzed the relative mutation rates of NAI-treated and DMSO-treated RNA samples, 124 revealing a significant elevation of mutation rates for NAI-treated samples (Fig 1B, p-value < 0.0001).

This confirms that the full-length SARS-CoV-2 RNA was successfully modified in vivo and that these 126 modifications were encoded as cDNA mutations.

To understand the relative SHAPE reactivity agreement within local regions of the genome, we 128 calculated Pearson correlation coefficients between two biological replicates. The Pearson's correlation 129 across the entire span of Orf1ab is 0.628 (Fig S1A) , consistent with those previously reported for 130 reactivities calculated from in vivo modified RNAs of this size (Smola et al., 2016) . Across the sub 131 genomic RNA ORFs, the Pearson 's correlation is poor (Fig S1B) . We believe this reflects the fact that 132 Amplicons 13 -16 will amplify both full-length and sub-genomic RNAs, and the difference in context will 133 result in different secondary structures (Tavares et al., 2020) . For this reason, despite the fact all data 134 have been obtained globally, subsequent discrete structural analysis will focus on shared features of 135 the viral termini and the Orf1ab region. 

The 5' genomic terminus includes seven regions that have been identified and studied in other 146 coronaviruses (Yang and Leibowitz, 2015) . While sequence conservation suggested that these 147 elements might be conserved in SARS-CoV-2, our consensus structure prediction shows this to be the 148 case, and we derived a specific experimentally-determined secondary structure for this section of the 149 genome. The in-vivo SHAPE reactivity data correspond well with the resulting structural model (Fig 1C,   150 inset) and the low overall Shannon entropy values in this region (determined from base pair probability 151 calculation during the SuperFold prediction pipeline (Smola et al., 2015b) ) support a well-determined 152 structure for the 5' genomic terminus (median Nuc(1-400) = 2.7x10 -5 ; global median = 0.022).

Individual features that typify coronavirus structures are evident in the secondary structure of 154 the SARS-CoV-2 5'-UTR with good SHAPE reactivity agreement (Fig 1C, inset) . For example, a 155 trifurcated stem is observed at the top of SL5 (Fig 1C) , including UUCGU pentaloop motifs in SL5A and 156 SL5B, and a GNRA tetraloop in SLC, as predicted in other coronaviruses. Previous reports suggest that 157 SL5 may represent a packaging signal for GroupIIB CoVs (Chen and Olsthoorn, 2010) . Similarity 

The 3' genomic terminus includes three well-studied stems, including the bulged-stem loop 164 (BSL), Stem Loop 1 (SLI), and a long bulge stem that includes the hypervariable-region (HVR), the 165 S2M domain, the octanucleotide motif (ONM) subdomains, and a pseudoknot (Yang and Leibowitz, 166 2015). The consensus structure recapitulates the secondary structure of all the three stems with good 167 SHAPE reactivity agreement (Fig 1D, inset) and overall low Shannon entropy (median Nuc(29,472-29,870) = 168 J o u r n a l P r e -p r o o f 0.016). While the BSL is well determined in our structure model, the low reactivity for bulged 169 nucleotides suggests the possibility of protein binding-partners (Fig 1D) .

A pseudoknot structure is proposed to exist between the base of the BSL stem loop and the 171 loop of SL1 in coronaviruses (Yang and Leibowitz, 2015) . While pseudoknot formation is mutually 172 exclusive with the base of the BSL, studies in MHV have suggested that both structures contribute to 173 viral replication and may function as molecular switches in different steps of RNA synthesis (Goebel et 174 al., 2004) . However, our in vivo determined secondary structure is inconsistent with formation of the 175 pseudoknot (Fig 1D) . The low SHAPE reactivities for the nucleotides at the base of the BSL support 176 formation of the extended BSL stem, while high-reactivities of the nucleotides in the loop of SLI indicate 177 that it is highly accessible. Using the SHAPEKnots program (Hajdin et al., 2013) ), we found that a 178 pseudoknot is never predicted in three windows that cover the pseudoknotted region. Taken together, 179 our data strongly support the extended BSL conformation, indicating it is probably the dominant 180 conformation in vivo.

The third stem in the 3' UTR includes three sub-domains. The HVR, poorly conserved across 182 group II coronaviruses (Goebel et al., 2007) , is predicted to be mostly single-stranded in our secondary 183 structure, and the high reactivities across the span of this region lends strong experimental support for 184 an unstructured region (Fig 1D) . The fact that this region is relatively unstructured may explain why it 185 tolerates deletions, rearrangements, and point mutations in MHV (Goebel et al., 2007) .

The S2M region is contained within the apical part of the third stem. We observe that the first 187 three helices of S2M from SARS-CoV-2 exactly match the crystal structure determined for S2M from 188 SARS-CoV (Robertson et al., 2005) . However, our in-vivo secondary structure model deviates 189 significantly at the top of the stem (Fig 1D) . It is possible that the SARS-CoV-2 S2M folds into a unique 190 S2M conformation despite differing by only two bases, both of which are transversions. (Fig 1D, 

Finally, we predict a different structure for the terminal stem in the viral 3'UTR (adjacent to the 196 poly-A tail) than previously reported for other coronaviruses (Zust et al., 2008) . However, structure 197 prediction of the complete stem is not highly accurate, as reactivity information for the downstream 198 stem (nts 29853-29870) is occluded by primer binding and is not constrained by experimental data (Fig   199   1D ). In addition, the complete stem region (nts 29472-nts 29870) is predicted to have high Shannon 

The PRF element previously characterized in SARS-CoV is proposed to contain three parts: an 210 attenuator stem loop (AS), a conserved heptanucleotide "slippery" sequence (HSS), and a H-type 211 pseudoknot (Plant and Dinman, 2008) . We performed SHAPEKnots predictions (Hajdin et al., 2013) 212 over four windows that cover the pseudoknotted region in the SARS-CoV-2 genome. We found that the 213 pseudoknot is successfully predicted in 3 out of 4 windows tested. Moreover, the nucleotides predicted 214 to be involved in the pseudoknotted helix have low SHAPE-reactivity (Fig 2A, red lines) . The frame-

shifting pseudoknot was thereafter included as a hard constraint during secondary structure prediction.

The most probable, dominant structure of the PRF region, extracted from the full-length in vivo 217 secondary structure, is shown in Fig 2A. In our model, the SHAPE reactivity and Shannon entropy 218 calculation support a well folded AS immediately upstream of the HSS (Fig 2A, 2C) . The AS has been 219 J o u r n a l P r e -p r o o f demonstrated to be important for attenuating frameshifting in SARS-CoV (Cho et al., 2013) , and 220 previous reports suggested that the AS structure is not well conserved between SARS-CoV and SARS- 

Overall, the dominant structure predicted for the H-type pseudoknot in our structural model 225 differs from the one proposed for SARS-CoV. In SARS-CoV-2, SL1 is well folded, as indicated by 226 SHAPE reactivity mapping (Fig 2A) and Shannon entropy (Fig 2C) . However, the region reported to 227 contain the SL2 stem (Rangan et al., 2020 , Plant et al., 2005 is predicted as single-stranded in our 228 consensus structure. Indeed, the dominant structure predicted for the PRF contains a different stem, 229 which we designate SL3, and this includes the downstream pseudoknot arm (Fig 2A) . The single-230 stranded region expected to contain SL2 is not well-determined in our consensus structure, as 231 indicated by Shannon entropy mapping to the region (Fig 2C) .

As SuperFold calculates a partition function, low probability base-pairing interactions can be 233 captured during structure prediction steps. We therefore checked the partition function output for 234 alternative, low probability base-pair interactions captured for the PRF region. We found that the single-235 stranded region (Fig 2A) forms base-pairing interactions with as many as 6 different regions in the 236 SARS-CoV-2 genome (Fig S2A) . Among these possible interactions is a PRF structure containing the 237 three-stemmed pseudoknot conformation identified across coronaviruses, including a helical SL2 ( Fig   238   2B ) (Plant et al., 2005) . The median base-pairing probability calculated for SL2 is 20% (Fig 2D; 239 individual base-pairs indicated with grey dots). In contrast, the SL3 stem is predicted to form with at 240 least 80% base-pairing probability.

The apparent pairing promiscuity and low SHAPE reactivities within the SL2 region suggests 242 that the PRF region has complex conformational dynamics that are not accurately represented by the To that end, we re-calculated the partition function for a 749nt window in the 246 SARS-CoV-2 genome that surrounds the PRF (Fig S2A) . This partition function calculation was then 247 inserted into an ensemble structure modeling framework implemented within RNAstructure (Ding and 

Using this mode of analysis, a single conformational cluster overwhelmingly dominates the PRF 250 conformational ensemble. As implied by our previous analysis (Fig 2A) , this conformational cluster 251 contains the AS, a single-stranded HSS, SL1, the pseudoknotted helix, and SL3. However, the SL2 252 region is base-paired with a region located 470nt upstream (Fig S2B) . The second-best populated 253 cluster contains a nearly identical domain architecture, except that the SL2 region is base-paired with a 254 region 260nt upstream (Fig S2C) . Together, these two clusters represent 99.2% of the PRF 255 conformational ensemble. The least populated cluster is the one that contains the SL2 region imbedded 256 in the canonical three-stemmed pseudoknot conformation, representing 0.8% of the PRF 257 conformational ensemble (Fig S2D) .

Taken together, these data suggest that the PRF of SARS-CoV-2 in infected cells includes a 259 well-folded AS, SL1, and the pseudoknot helix, but that the region containing the putative SL2 is 260 conformationally variable, with the potential to form a diversity of long-range interactions. Therefore, the 261 three-stem pseudoknot conformation that is conventionally used to characterize β-coronavirus PRFs 

Here we report the first in vivo-derived, SHAPE-constrained secondary structural model that 

To discover additional, well-folded RNA structures within the SARS-CoV-2 genome, we 279 calculated the local median Shannon Entropy and correlated these values with experimentally-280 determined SHAPE reactivities (Fig 3A) . Only regions with both median Shannon entropy and SHAPE 281 reactivity signals below the global median for stretches longer than 40nt, and which appear in both 282 replicate data sets, were considered well-determined and stable. In total, we identify 40 such regions in 283 Orf1ab (Fig 3A, shaded) . Hereafter, any structured region that meets these above criteria will be 284 referred to as "well-folded."

To understand architectural organization of the overall "structuredness", or base-pair content 

While analyzing the resulting secondary structural map, we noticed that the SARS-CoV-2 296 genome contains long-stretches of short, locally-folded stem loops (for example - Fig 3B) with few long-297 J o u r n a l P r e -p r o o f distance base-pairing interactions. To determine if this was a quantifiable feature unique to the SARS-

CoV-2 genome, we calculated the distance between base-paired nucleotides for every base-pairing 299 interaction in our SARS-CoV-2 structural model. We compared these base-pairing distances to those 300 we calculated from published full-length structural models for HCV (Mauger et al., 2015) and Dengue 301 virus (Dethoff et al., 2018) , that used the same structure prediction pipeline and constraints. Interestingly, 302 the median base-pairing distance in our SARS-CoV-2 consensus model is 25nt, and is significantly 303 smaller than the median base-pairing distance in the HCV (median=40nt) and Dengue virus 304 (median=33nt) consensus models (Fig 4B) . This suggests SARS-CoV-2 has fewer long-distance base-305 paring interactions compared to Dengue and HCV genome.

We also calculated the median base-pairing distance for the well-folded regions of the SARS-

CoV-2 genome and compared the result to well-folded regions previously identified using the same Low 

Using an "All β-Coronavirus" alignment, we observed a significantly lower dS for double-330 stranded codons when compared to single-stranded codons in our consensus model (Fig 4D) . In 

When analyzing relative dS within individual protein domains, we observed significantly 338 decreased dS for double-stranded codons in Nsp1, Nsp2, Nsp3, Nsp4, Nsp6, Nsp8, Nsp12, Nsp13, 339 and Nsp15 (Fig 4F) . Consistent with this, Nsp1, Nsp6, Nsp8, and Nsp12 have >50% of their 340 nucleotides localized within well-folded regions (Fig 4A, black bars) . Taken together, this suggests that 341 certain protein-coding domains contain regions of RNA secondary structure that are conserved across 342 β-Coronaviruses. For example, Nsp8, which is the most well folded domain in SARS-CoV-2, is likely 343 well-folded in other β-Coronaviruses.

By contrast, the base pairing content of Nsp5, Nsp7, Nsp9, Nsp10, Nsp14, and Nsp16 does not 345 appear to be conserved, as there is no significant difference in dS (Fig 4F) . Consistent with this, Nsp14 346 and Nsp16 were shown to have <15% of their nucleotides in well-folded regions, while Nsp10 does not 347 contain any well-folded nucleotides (Fig 4A) . Not only does this analysis support the observation that CoV-2 life cycle, we next applied our dS analysis to each of the 40 discrete well-folded domains (Fig   355   3A , Table S1 ). Four regions showed significantly decreased dS at double-stranded codons across the 356 β-coronavirus alignment (Fig 5A, 5B) . Among those well-folded domains, region 25 and 34 are found 357 at protein domain boundaries. Region 25 ends exactly at the Nsp8/9 domain boundary, while Region 34 358 spans the Nsp12/13 boundary. Region 23, 34, and 36 (Fig 5C, 5E, 5F ) contain a series of stem-loops 359 with small bulges. Region 25 contains a long-range duplex that closes a clover-leaf like structure with 8 360 stem-loops radiating from a central loop (Fig 5D) . This hub, or multi-helix junction, might represent a 361 promising drug target, as multi-helix junctions often contain binding pockets with high binding affinity 362 and selectivity for small molecules (Warner et al., 2018) .

Within the Sarbecovirus subgenus, we were able to identify five well-folded regions with 364 significantly decreased dS in double-stranded codons (Fig 6A, 6B ). Among these well-folded domains,

Region 24 contains two discrete multi-helix junctions, each with at least three stems radiating from 366 large central loops (Fig 6C) . Region 27 contains a series of six stem-loops (Fig 6D) . Region 15, like 367 Region 24, contains several well-determined long-range duplexes that segment the region into two 368 discrete multi-helix junctions (Fig 6E) . Region 22 contains a series of well-folded loops and it spans the 369 Nsp5/6 boundary (Fig 6F) . Region 30 is a single stem-loop with bulges that divide the stem into distinct 370 duplexes (Fig 6G) 

To look for evolutionary evidence that directly supports conservation of specific base-pairing 

For functional targets, we focused on two well-folded ORF regions, 15 and 22, each of which 388 has strong evolutionary support (Fig 6E, 6F) . LNAs targeted to these regions were designed for 389 maximal structure disruption, hybridizing to the top of the stem loop as well as duplex RNA flanking the 390 loop (Fig 7A, 7B ; red lines). Importantly, we also designed a negative control that targets high Shannon 391 entropy regions immediately downstream of each well-folded region, but still within the ORF (Fig 7A, 

7B; blue lines). We do not expect hybridization of negative control LNAs to have an effect on viral 393 growth unless overall translation is disrupted. We included a scrambled LNA that should not bind to the 394 SARS-CoV-2 genome as a global negative control.

As shown in Fig 7D, the LNA targeting the covarying stem in region 15 results in a 40% 396 decrease in GFP+ cells when compared to the region 15 control and a 35% decrease when compared 397 to the scrambled LNA control. The region 15 control LNA has no effect on viral growth relative to the 398 scrambled LNA control. A similar trend is observed for region 22 (Fig 7E) . The LNA targeting the stem 

Our structural modeling of the PRF suggests it contains a conformationally flexible SL2. In order 403 to evaluate the functional importance of SL2, we tested whether an LNA targeted against the SL2 404 region resulted in a measurable defect in viral growth (Fig 7C; red line) . In addition, we designed an 405 LNA targeted against the PRF pseudoknot (SL1) (Fig 7C; blue line) as disruption of the SARS-CoV 406 PRF has been demonstrated to reduce viral growth (Plant et al., 2013 , Plant et al., 2005 . This LNA 407 results in an 18% reduction in GFP+ cells relative to the scrambled LNA control (Fig 7F) . Interestingly, 

The core of the SARS-CoV PRF, which shares an almost identical sequence with SARS-CoV-2, is 463 predicted to form a three-stem pseudoknot comprised of SLI, SL2, and a pseudoknotted helix, with an 

The P1 stock was used to inoculate Vero-E6 (ATCC) cells for three days. Supernatant was harvest and 663 clarified by centrifuging at 450g for 5min. Clarified supernatant was filtered through a 0.45-micron filter, 664 aliquoted, and stored at -80°C.

Virus titre was determined by plaque assay. VeroE6 cells were seeded at 7.5 x 10 5 cells/well in 666 6-well plates.The following day, media were removed and replaced with 100µL of 10-fold serially diluted 667 viral stock. Plates were incubated at 37°C for 1 hour with gentle rocking. Following the incubation, each 

MarathonRT purification was performed as described in . For each amplicon, 706 500ng of total cellular RNA was mixed with 1µL of the corresponding 1µM RT primer. Gene-specific 707 primers used for RT are listed in Table S2 . Primers were annealed at 65°C for 5min then cooled to 708 room temperature, followed by addition of 8µL of 2.5x MarathonRT SHAPE-Map Buffer (125mM 1M

Tris-HCl pH 7.5, 500mM KCl, 12.5mM DTT, 1.25mM dNTPs, 2.5mM Mn 2+ ), 4µL of 100% glycerol, and constraints. The two pseudoknots tested were the programmed ribosomal frameshifting element that 741 exists at the Orf1a/b boundary, and a pseudoknot in the 3'UTR that was identified in the MHV and B-

CoV genomes (Goebel et al., 2004) . We analyzed all 500nt windows separated by a 100nt slide that 743 contained each of the putative pseudoknots to determine if the pseudoknot was successfully predicted. 

To perform ensemble structure modeling, we followed step 6, 7 and 8 from the Rsample 762 program (Spasic et al., 2018) . To elaborate, first we used the Partition program (implemented in RNA 763 structure v6.1, Mathews (Mathews, 2004) ) to generate the partition saved file (PFS) for the region 764 described. Replicate 1 SHAPE reactivity was used as a soft constraint (using the same slope and 765 intercept as we used in the Superfold prediction) and the pseudoknotted base pairs were forced single 766 strand. The PFS file was used to sample 1000 probable structures in proportion to their Boltzmann 767 weights using the stochastic program (implemented in RNA structure v6.1) (Ding and Lawrence, 2003) .

This sample was then clustered using the hierarchical divisive method (Ding et al., 2005) and was asked 769 to output 10 clusters with a representative conformation. A cluster is defined as a subset of structures 770 with similar base pairs. The PFS file was visualized using IGV v2.8.2(Busan and Weeks, 2017).

Two data signatures were used to identify well-folded regions: The first is the SHAPE reactivity 

Local median SHAPE reactivity and Shannon Entropy were calculated in 55nt sliding windows.

The global median SHAPE reactivity or Shannon Entropy were subtracted from calculated values to aid 780 in data visualization. Regions with local SHAPE and Shannon Entropy signals 1) below the global 781 median 2) for stretches longer than 40 nucleotides 3) that appear in both replicate data sets were 782 considered well-folded. Disruptions, or regions where local SHAPE or Shannon Entropy rose above the 783 global median, are not considered to disqualify well-folded regions if they extended for less than 40 784 nucleotides. Arc plots generated from each replicate consensus structure predication were compared 785 for regions that meet sorting criteria described above in order to ensure agreement between secondary 786 structure models generated from each replicate SHAPE-MaP dataset. 

We also generated an "All β-coronavirus Alignment" using the sarbecovirus sequences Table S4 ).

Vero-E6 were grown in DMEM+10% FBS+1% PBS and incubated at 37°C/5% CO 2 .

Approximately 7.5x10 5 Vero-E6 cells were plated per well in a 6-well plate prior to transfection. LNAs 

Graphs and statistical analysis were made using GraphPad Prism 8 and Microsoft Excel v16.

The results are expressed as Tukey plots with median and interquartile range indicated with bars.

J o u r n a l P r e -p r o o f Specific values are reported in the Results. All statistically significant differences were calculated using 853 the unpaired t-test assuming both populations have equal variance unless otherwise stated.

Significance of comparisons is indicated in figures and supplemental data as *, p < 0.05; **, p < 0.01; 855 ***, p < 0.001; ****, p < 0.0001. 

& 861 MOSS, W. N. 2020. An in silico map of the SARS-CoV-2 RNA Structurome

Strong epistatic selection on the RNA secondary structure of HIV

Programmed ribosomal frameshifting in decoding the SARS-CoV genome

Visualization of RNA structure models within the Integrative Genomics

Accurate detection of chemical modifications in RNA by mutational 872 profiling (MaP) with ShapeMapper 2

Genomic variance of the 2019-nCoV coronavirus

Dynamics of translation by single ribosomes through mRNA secondary structures

Group-specific structural features of the 5'-proximal sequences 879 of coronavirus genomic RNAs

Regulation of programmed 881 ribosomal frameshifting by co-translational refolding RNA hairpins

Ribosome pausing, a dangerous necessity for co-translational events

SARS and MERS: recent 885 insights into emerging coronaviruses

Pervasive tertiary structure in the dengue virus RNA genome

A Balancing Act: MDA5 in Antiviral 891 Immunity and Autoinflammation

RNA secondary structure prediction by centroids in a 893 Boltzmann weighted ensemble

A statistical sampling algorithm for RNA secondary structure 895 prediction

An interactive web-based dashboard to track COVID-19 in real 897 time

Small molecules that target group II introns are potent antifungal agents

Role of RNA structures in genome terminal sequences of the 902 hepatitis C virus for replication and assembly

Characterization of the RNA 904 components of a putative molecular switch in the 3' untranslated region of the murine 905 coronavirus genome

A hypervariable 907 region within the 3' cis-acting element of the murine coronavirus genome is nonessential for 908 RNA synthesis but affects pathogenesis

Sequencing and Structure Probing of Long

RNAs Using MarathonRT: A Next-Generation Reverse Transcriptase

Accurate SHAPE-directed RNA secondary structure modeling, including pseudoknots

Targeting the SARS-CoV-2 RNA 917 Genome with Small Molecule Binders and Ribonuclease Targeting Chimera

Evidence for ligandable sites in 920 structured RNA throughout the Protein Data Bank

Structural and functional conservation of the programmed -1 ribosomal 923 frameshift signal of SARS coronavirus 2 (SARS-CoV-2)

The Architecture of SARS-CoV-2

Touchdown PCR for increased specificity and sensitivity in PCR 927 amplification

R-CHIE: a web server and R package for 929 visualizing RNA secondary structures

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor

& 934 ROUSKIN, S. 2020b. Structure of the full SARS-CoV-2 RNA genome in infected cells

Bridging the gap between 936 in vitro and in vivo RNA folding

Integrative Analysis of Zika Virus Genome RNA Structure 939

OligoWalk: an online siRNA design tool utilizing hybridization 941 thermodynamics

Biological activity and biotechnological aspects of locked nucleic acids

Structural and functional conservation of cis-acting RNA elements in 947 coronavirus 5'-terminal genome regions

Coronaviruses : methods and protocols

Genome-wide mapping of therapeutically-relevant SARS-CoV-2 RNA structures. bioRxiv

Using an RNA secondary structure partition function to determine confidence 954 in base pairs predicted by free energy minimization

Functionally conserved architecture of hepatitis C virus RNA genomes

Probing RNA structure in vivo

FUBAR: a fast, unconstrained bayesian approximation for inferring 962 selection

Functions and regulation of RNA editing by ADAR deaminases

966 Conformational inhibition of the hepatitis C virus internal ribosome entry site RNA

The Coding Region of the HCV 969

Genome Contains a Network of Regulatory RNA Structures

The role of programmed-1 ribosomal frameshifting in coronavirus 971 propagation

A three-stemmed mRNA pseudoknot in the SARS coronavirus frameshift signal

Altering SARS coronavirus 976 frameshift efficiency affects genomic and subgenomic RNA production

RNA genome conservation and secondary structure in 978 SARS-CoV-2 and SARS-related viruses: a first look

MACSE v2: Toolkit for 980 the Alignment of Coding Sequences Accounting for Frameshifts and Stop Codons

In-line probing analysis of riboswitches

RNAstructure: software for RNA secondary structure prediction 985 and analysis

A statistical test for conserved RNA structure shows lack of 987 evidence for structure in lncRNAs

The 989 structure of a rigorously conserved RNA element within the SARS virus genome

Genome-wide probing of 991 RNA structure reveals active unfolding of mRNA structures in vivo

RNA motif discovery by 993 SHAPE and mutational profiling (SHAPE-MaP)

Structural constraints on RNA virus evolution

In vivo analysis of influenza A mRNA secondary structures 997 identifies critical regulatory motifs

Detection of RNA-Protein Interactions in Living 999 Cells with SHAPE

SHAPE reveals transcript-wide interactions, complex 1002 structural domains, and protein interactions across the Xist lncRNA in living cells

Selective 2'-hydroxyl 1005 acylation analyzed by primer extension and mutational profiling (SHAPE-MaP) for direct, 1006 versatile and accurate RNA structure analysis

Nitazoxanide and JIB-04 have broad-1011 spectrum antiviral activity and inhibit SARS-CoV-2 replication in cell culture

Modeling RNA secondary 1014 structure folding ensembles using SHAPE mapping data

The global and 1016 local distribution of RNA structure throughout the SARS-CoV-2 genome

Phylogenetic Analysis with Improved 1018 Parameters Reveals Conservation in lncRNA Structures

Inhibition of HCV translation by 1020 disrupting the structure and interactions of the viral CRE and 3' X-tail

Thermodynamic and 1023 phylogenetic prediction of RNA secondary structures in the coding region of hepatitis C virus

RNA self-1026 assembly contributes to stress granule formation and defining the stress granule transcriptome

Understanding the transcriptome 1029 through RNA structure

Receptor Recognition by the Novel 1031 Coronavirus from Wuhan: an Analysis Based on Decade-Long Structural Studies of SARS 1032 Coronavirus

Principles for targeting RNA with drug-like small 1034 molecules

Jalview 1036 Version 2--a multiple sequence alignment editor and analysis workbench

Genome-wide CRISPR 1043 Screens Reveal Host Factors Critical for SARS-CoV-2 Infection

An Infectious cDNA Clone of SARS-CoV-2

The structure and functions of coronavirus genomic 3' and 5' ends

Structural basis for inhibition of the RNA-dependent 1053 RNA polymerase from SARS-CoV-2 by remdesivir

A cis-acting replication element in the 1055 sequence encoding the NS5B RNA-dependent RNA polymerase is required for hepatitis C virus 1056 RNA replication

A Novel Coronavirus from Patients with Pneumonia in China

MaPseq for genome-wide or targeted RNA structure probing in vivo

Genetic interactions between an 1064 essential 3' cis-acting RNA pseudoknot, replicase gene products, and the extreme 3' end of the 1065 mouse coronavirus genome