The myriad roles of RNA structure in the flavivirus life cycle

Figures & data

Figure 1. Flavivirus genome organization.

The viral RNA genome (vRNA) of flaviviruses has a single open-reading frame (ORF) flanked by highly structured 5´ and 3´ UTRs that possess cis-acting RNA structural elements required for viral translation, RNA replication, and assembly. The Zika virus (ZIKV) genome is depicted. A) In the linear (translation-competent) form of the vRNA, the capped viral genomic RNA contains stem-loop A (SLA) and SLB in the 5´ untranslated region (UTR), which lie just upstream of the capsid coding-region hairpin (cHP). The 3´ UTR contains a variable region, with one or more stem-loop structures (SL-I and SL-II), and a dumbbell (DB) region, containing one or more DB sequences, all of which can form local pseudoknot (PK) structures. These structures are followed by the short hairpin (sHP) and the 3´ stem loop (3´SL) at the 3´ terminus of the genome. B) In the circular (replication-competent) form of the viral genome, long-range RNA-RNA interactions between elements in the 5´ and 3´ UTRs (5´-3´CS, 5´-3´DAR, 5´-3´UAR) facilitate cyclization of the vRNA. These long-range interactions partially disrupt the 3´SL structure, and fully disrupt the SLB and sHP structures. UAR, upstream of AUG region; DAR, downstream of AUG region; CS, cyclization sequence.

Table

Box 1. How is RNA structure determined experimentally? Traditionally, RNA structure has been analysed via indirect methods that modify unpaired and/or unbound ribonucleotides. For example, Selective 2´ Hydroxyl Acylation analysed by Primer Extension (SHAPE) reagents preferentially react with the 2´OH group of unconstrained ribonucleotides to form 2´-O-adducts, while dimethyl sulfate (DMS) reacts with specific N atoms in the rings of unpaired adenines and cytidines [6–8]. In both cases, these modifications can be read as terminations of reverse transcription (RT) on the treated RNA. Coupled with high-throughput sequencing, it is now possible to apply these methods on a genome-wide scale. To increase robustness and permit data collection from RNAs with multiple modifications, these approaches utilize mutational profiling (MaP), an approach where the RT conditions are altered such that it introduces a mutation at the modification site on the RNA instead of terminating the reaction [9,10]. As a result, each modification results in a base change that can be detected during sequencing. Advances in sequencing technologies have also resulted in the proliferation of multiple direct methods to assess genome-wide RNA structure, such as psoralen analysis of RNA interactions and structures (PARIS) and cross-linking of matched RNAs and deep sequencing (COMRADES) [11,12]. These more direct methods utilize chemical probes that crosslink base-paired ribonucleotides followed by sequencing of the crosslinked pieces of RNA. However, both the indirect and direct approaches suffer from a shared limitation in that the resulting structural data is an average of all RNA conformations present during treatment. Excitingly, recent advances in the bioinformatic pipelines to analyse this structural data, like DRACO, a tool that deconvolutes different RNA conformations from MaP datasets, may aid in dissecting the heterogeneity of the RNA structurome [13]. Finally, the most direct approaches to assess RNA structure are methods such as cryo-electron microscopy and x-ray crystallography, which have recently been applied in the analysis of the flavivirus SLA structure [14,15]. However, identifying conditions to analyse RNA structures via these methods is not trivial, and at present is largely limited to isolated local structures. Furthermore, additional structure-stabilizing sequences or proteins are often required to sufficiently stabilize the RNA of interest, selectively constraining the RNA in a particular conformation, which may or may not be a biologically relevant conformation [14,15]. As such, combinations of RNA structure-probing techniques (more thoroughly reviewed in [16]), are likely necessary to further our understanding of flavivirus RNA structures.

Table 1. RNA structures in the flavivirus genome and their roles in the viral life cycle.

Structure		Location in Genome	Interactor(s)	Role(s) in Flavivirus Life Cycle
SLA		5´ UTR	NS5	Translational shut-off Negative-strand RNA synthesis Capping RO biogenesis? Positive-strand RNA synthesis?
SLB		5´ UTR	NS5	Translational shut-off?
cHP		Capsid coding region	Ribosomes	Start site selection
SL-I		3´ UTR	XRN-1	Generation of sfRNAs (immune evasion)
SL-II
DB-I
DB-II (or DB-like)
3´ UTR	SL-II (ZIKV)	3´ UTR	NS2A	Packaging (specificity)
	DB-I to 3´SL (DENV)
3´SL		3´ UTR	NS5	Negative-strand RNA synthesis Translational shut-off?
Cyclization elements	UAR	5´ and 3´ UTRs	Partner cyclization element	Translational shut-off Genome cyclization Negative-strand RNA synthesis
	DAR
	CS
Long-range RNA-RNA interactions (>0.5 kb)		Multiple throughout genome	Capsid	Packaging (genome compaction) RO biogenesis?
SLA´		3´ terminus of negative-strand	NS5?	Positive-strand RNA synthesis?

Table

Box 2. Can flaviviral RNAs be translated via both cap-dependent and cap-independent mechanisms? In addition to canonical cap-dependent translation, several flaviviral RNAs have been demonstrated to be capable of cap-independent translation in both mammalian and mosquito cells [39–41]. However, translation of uncapped flaviviral RNAs is much less efficient, with approximately 100-fold lower translation than their capped flaviviral RNA counterparts, and the precise mechanism for this noncanonical translation is still unclear [39]. There remains debate whether the 5´ UTR of flaviviruses has internal ribosome entry site (IRES) activity, with the 5´ UTRs of both DENV and ZIKV passing the classical IRES test of being capable of driving translation of an internal reporter in a bicistronic reporter construct [42,43]. However, inclusion of the 5´ UTR alone resulted in significant, albeit weak, cap-independent translation in these studies [42,43]. Moreover, cap-independent activity was negligible in monocistronic reporter RNAs when the canonical m^7GpppG cap was replaced with an ApppA cap structure that cannot interact with the eIF4E cap-binding protein and recruit the ribosome [42]. In general, the ability of uncapped reporter RNAs to be translated appears to be dependent upon both the 5´ and 3´ UTRs of the flaviviral genomic RNA, with the 3´ UTR acting as some kind of translational enhancer [39–42]. Nevertheless, it remains unclear which circumstances would necessitate a requirement for such cap-independent translation. One possible explanation could be that cap-independent translation may provide flaviviruses with a counter defence strategy to mitigate the downregulation of global cap-dependent translation in cells resultant from infection-induced ER stress and/or activation of cellular antiviral responses [40,41,43,44]. Conversely, cap-independent translation could allow uncapped genomes that are packaged into virions to establish a RO upon infection, at which point progeny positive-strand genomic RNAs would be capped via the viral machinery [39–41,43]. In either case, the inefficiency of this approach implies that the function of this secondary noncanonical translation mechanism is more of a complement to primary cap-dependent translation, rather than a direct substitute.

Figure 2. The ‘L/V’ conformations of DENV and ZIKV SLA.

A–C) 2D, 3D and NS5-bound conformations of the DENV ‘L’-shaped SLA structure. The base stem, SSL, and TSL of SLA are coloured in dark green, chartreuse, and pale green, respectively. The AG motif of the TSL is indicated in yellow. The structures of SLA and SLA-NS5 of DENV were generated using PDBs:7LYF and 8GZ9, respectively. In C), the DENV NS5 RdRp and MTase domains are coloured in magenta and pink, respectively; with the Mg²⁺ ion coordinating addition of the viral cap to the 5´ terminus indicated in red, while the SAM donor is indicated in light purple. The 5´ terminus of the DENV genome (nt 1-3) is depicted in dark blue.D–F) 2D, 3D and modelled NS5-bound conformations of the ZIKV ‘V’-shaped SLA structure. The base stem, SSL, and TL of SLA are coloured in dark blue, teal, and pale blue, respectively. The AG motif of the TSL is indicated in yellow. The structure of ZIKV SLA was generated using PDB:7LYG and the latter SLA-NS5 model was generated by aligning a ZIKV NS5 monomer from PDB:5M2X and the ZIKV SLA structure (PDB:7LYG; nt 4-69) with their counterparts in the DENV SLA-NS5 structure (PDB:8GZP). Notably, as the ZIKV SLA structure was determined via appending the BSL to a tRNA, we have less confidence in the modeled interactions between SLA and the MT domain than those between SLA and the RdRp domain. In F), the ZIKV NS5 RdRp and MTase domains are coloured in magenta and pink, respectively. G) Structural alignment of SLA structures from DENV (PDB:7LYF) and ZIKV (PDB:7LYG).

Table

Box 3. Is multimerization of flavivirus genomes important for flavivirus infection? Interestingly, approaches to study flavivirus genome cyclization cannot actually discriminate between cis- and trans-interactions; and, as such, it has been hypothesized that these elements may also be involved in genome dimerization or multimerization [65]. Thus, rather than genome cyclization (in cis), this hypothesis proposes that the effects of the cyclization elements could be due to formation of an antiparallel homoduplex dimer between two genomes or head-to-tail concatemers of multiple flavivirus genomes (in trans) [65]. In line with this hypothesis, others have proposed that in the SLA “L/V” conformations of DENV and ZIKV, dimerization may occur via a kissing loop interaction of the SSLs, which was observed at high RNA concentrations in vitro [14]. Disruption of the proposed complementary region, which comprises most of the SSL, also impairs viral replication in DENV, but is rescued when swapped with the corresponding region of ZIKV SLA [14]. However, it is not clear from these experiments whether this effect is due to genome dimerization or simply the importance of these nucleotides in interactions with NS5 [27]. Moreover, it remains unclear if this kissing loop interaction can occur in the tick-borne flaviviruses, which have a much longer SSL [58]. Regardless, it seems clear that one or a few viral genomes is sufficient to establish an infection, suggesting that genome dimer/multimerization is unlikely to be an absolute requirement at least at the early stages of the viral life cycle. Nonetheless, it is possible that genome multimerization may occur in vivo at later stages in the infection cycle, when high local concentrations of viral genomes are produced [65]. Given that sfRNAs, which contain the 3´ cyclization elements, are also produced during infection, to what extent these can interact with full-length genomic RNAs and modulate the infection process is also an open question [25]. Thus, while genome cyclization (in cis) is important for the establishment of an infection, more research will be needed to establish whether genome dimer/multimerization and/or interactions between genomes and sfRNAs (in trans) have a role in modulating the course of flavivirus infection.

Figure 3. Models for the initiation of negative-strand RNA synthesis.

A) In model 1, NS5 bound to SLA initiates negative-strand RNA synthesis directly on the adjacent 3´ terminus of the positive-strand. SLA is released during the elongation phase of RNA synthesis. B) In model 2, the NS5 bound to SLA initiates negative-strand RNA synthesis directly on the adjacent 3´ terminus, but remains bound to SLA during elongation, only releasing SLA when it is itself unwound for completion of the dsRNA replicative intermediate. C) In model 3, the NS5 bound to SLA is transferred to the 3´SL due to the higher affinity of NS5 for the 3´SL following genome cyclization-induced unwinding of SLB. NS5 then initiates negative-strand RNA synthesis directly on the 3´ terminus of the positive-strand. D) In model 4, the NS5 bound to SLA interacts with a second NS5 bound to the 3´SL, allowing the latter NS5 to initiate negative-strand RNA synthesis off the 3´ terminus of the positive-strand. The 5´ terminus of the genome remains associated with the SLA-bound NS5, only releasing SLA when it is itself unwound for completion of the dsRNA replicative intermediate.

Figure 4. Models for the initiation of positive-strand RNA synthesis.

A) In model 1, NS5 bound to SLA initiates positive-strand RNA synthesis directly on the adjacent 3´ terminus of the negative-strand within the dsRNA replicative intermediate. B) In model 2, NS5 is recruited directly to SLA´ to initiate positive-strand RNA synthesis directly on the 3´ terminus of the negative strand. C) In model 3, the NS5 bound to SLA is transferred to SLA´ and initiates positive-strand RNA synthesis directly on the 3´ terminus of the negative-strand within the dsRNA replicative intermediate. D) In model 4, the NS5 bound to SLA interacts with a second NS5 bound to SLA´, allowing the latter NS5 to initiate positive-strand RNA synthesis off the 3´ terminus of the negative-strand within the dsRNA replicative intermediate. It is unclear whether the NS5 originally bound to SLA would dissociate from or remain bound to the elongating NS5 carrying out positive-strand RNA synthesis. In all cases, the positive-strand is displaced from the dsRNA replicative intermediate during synthesis of the new positive-strand.

Table

Box 4. How do cellular RBPs participate in genome compaction? To accomplish the ambitious task of compacting the viral genomic RNA into the assembling virion, it seems likely that cellular RBPs must lend a helping hand. For example, the DEAD box (DDX) helicase, DDX56, relocalizes from the nucleus to the site of WNV assembly during infection, and its helicase activity is important for efficient virion production [93,94]. In general, DDX proteins are known to regulate RNA condensation, as they remodel RNA-RNA and protein-RNA interactions through their helicase activity [95]. As such, it is tempting to speculate that DDX56 promotes genome compaction during WNV assembly via remodelling RNA-RNA and protein-RNA interactions with the viral genomic RNA. Interestingly, several DDX family proteins are also implicated in flaviviral RNA replication, suggesting that genome compaction for RO biogenesis may also require DDX helicase activity [96]. Additionally, cellular RBPs may also promote genome compaction by providing additional protein-RNA interactions to condense the viral RNA. For example, the YBX-1 protein has been shown to interact with the 3´SL, which inhibits viral translation and has a detrimental impact on DENV infection; but this may be mitigated by the generation of sfRNAs, which are known to sequester cellular RBPs [97,98]. On the other hand, YBX-1 is also known to be required for proper virion assembly, through interactions with the viral genomic RNA as well as the DENV capsid and envelope proteins [99,100]. However, it is unclear if the interaction with the viral RNA during virion assembly is solely through the established interaction with the 3´SL, or if YBX-1 interactions provide multivalency for genome compaction by binding to additional sites across the DENV genome [98]. Nonetheless, these examples suggest that cellular RBPs also interact with flaviviral RNAs and help modulate genome compaction during both RO biogenesis and virion assembly.

Siegfried NA, Busan S, Rice GM, Nelson JA, Weeks KM. RNA motif discovery by SHAPE and mutational profiling (SHAPE-MaP). Nat Methods. 2014;11(9):959–965. doi: 10.1038/nmeth.3029 Epub 20140713. PubMed PMID: 25028896; PubMed Central PMCID: PMC4259394.

 PubMed Web of Science ®Google Scholar

Zubradt M, Gupta P, Persad S, et al. DMS-MaPseq for genome-wide or targeted RNA structure probing in vivo. Nat Methods. 2017;14(1):75–82. doi: 10.1038/nmeth.4057 Epub 20161107. PubMed PMID: 27819661; PubMed Central PMCID: PMC5508988.

 PubMed Web of Science ®Google Scholar

Lu Z, Zhang QC, Lee B, et al. RNA duplex map in living cells reveals higher-order transcriptome structure. Cell. 2016;165(5):1267–1279. doi: 10.1016/j.cell.2016.04.028 Epub 20160512. PubMed PMID: 27180905; PubMed Central PMCID: PMC5029792.

 PubMed Web of Science ®Google Scholar

Ziv O, Gabryelska MM, Lun ATL, et al. COMRADES determines in vivo RNA structures and interactions. Nat Methods. 2018;15(10):785–788. doi: 10.1038/s41592-018-0121-0 Epub 20180910. PubMed PMID: 30202058; PubMed Central PMCID: PMC6168409.

 PubMed Web of Science ®Google Scholar

Morandi E, Manfredonia I, Simon LM, et al. Genome-scale deconvolution of RNA structure ensembles. Nat Methods. 2021;18(3):249–252. doi: 10.1038/s41592-021-01075-w Epub 20210222. PubMed PMID: 33619392.

 PubMed Web of Science ®Google Scholar

Lee E, Bujalowski PJ, Teramoto T, et al. Structures of flavivirus RNA promoters suggest two binding modes with NS5 polymerase. Nat Commun. 2021;12(1):2530. doi: 10.1038/s41467-021-22846-1 Epub 20210505. PubMed PMID: 33953197; PubMed Central PMCID: PMC8100141.

 PubMed Web of Science ®Google Scholar

Osawa T, Aoki M, Ehara H, Sekine SI. Structures of dengue virus RNA replicase complexes. Mol Cell. 2023;83(15):2781–91 e4. doi: 10.1016/j.molcel.2023.06.023 Epub 20230720. PubMed PMID: 37478848.

 PubMed Web of Science ®Google Scholar

Spitale RC, Incarnato D. Probing the dynamic RNA structurome and its functions. Nat Rev Genet. 2023;24(3):178–196. doi: 10.1038/s41576-022-00546-w Epub 20221108. PubMed PMID: 36348050; PubMed Central PMCID: PMC9644009.

 PubMed Web of Science ®Google Scholar

Berzal-Herranz A, Berzal-Herranz B, Ramos-Lorente SE, Romero-Lopez C. The genomic 3’ UTR of flaviviruses is a translation initiation enhancer. Int J Mol Sci. 2022;23(15):8604. doi: 10.3390/ijms23158604 Epub 20220803. PubMed PMID: 35955738; PubMed Central PMCID: PMC9369090.

 PubMed Web of Science ®Google Scholar

Fernandez-Garcia L, Angulo J, Ramos H, Barrera A, Pino K, Vera-Otarola J, et al. The internal ribosome entry site of the Dengue virus mRNA is active when cap-dependent translation initiation is inhibited. J Virol. 2021;95(5):10–128. doi: 10.1128/JVI.01998-20 Epub 20201209. PubMed PMID: 33298544; PubMed Central PMCID: PMC8092825.

 Web of Science ®Google Scholar

Song Y, Mugavero J, Stauft CB, Wimmer E. Dengue and Zika Virus 5’ untranslated regions harbor internal ribosomal entry site functions. MBio. 2019;10(2):10–128. doi: 10.1128/mBio.00459-19 Epub 20190409. PubMed PMID: 30967466; PubMed Central PMCID: PMC6456755.

 Web of Science ®Google Scholar

Edgil D, Polacek C, Harris E. Dengue virus utilizes a novel strategy for translation initiation when cap-dependent translation is inhibited. J Virol. 2006;80(6):2976–2986. doi: 10.1128/JVI.80.6.2976-2986.2006 PubMed PMID: 16501107; PubMed Central PMCID: PMC1395423.

 PubMed Web of Science ®Google Scholar

Wang T, Merits A, Wu Y, et al. cis-acting sequences and secondary structures in untranslated regions of duck tembusu virus RNA are important for cap-independent translation and viral proliferation. J Virol. 2020;94(16):10–128. doi: 10.1128/JVI.00906-20 Epub 20200730. PubMed PMID: 32522848; PubMed Central PMCID: PMC7394898.

 Web of Science ®Google Scholar

Roth H, Magg V, Uch F, Mutz P, Klein P, Haneke K, et al. Flavivirus infection uncouples translation suppression from cellular stress responses. MBio. 2017;8(1):10–128. doi: 10.1128/mBio.02150-16 Epub 20170110. PubMed PMID: 28074025; PubMed Central PMCID: PMC5225315.

 Web of Science ®Google Scholar

Lott WB, Doran MR. Do RNA viruses require genome cyclisation for replication? Trends Biochem Sci. 2013;38(7):350–355. doi: 10.1016/j.tibs.2013.04.005 Epub 20130612. PubMed PMID: 23768999.

 PubMed Web of Science ®Google Scholar

Filomatori CV, Iglesias NG, Villordo SM, et al. RNA sequences and structures required for the recruitment and activity of the dengue virus polymerase. J Biol Chem. 2011;286(9):6929–6939. doi: 10.1074/jbc.M110.162289 Epub 2010 Dec 25. PubMed PMID: 21183683; PubMed Central PMCID: PMC3044948.

 PubMed Web of Science ®Google Scholar

Upstone L, Colley R, Harris M, Goonawardane N. Functional characterization of 5’ untranslated region (UTR) secondary RNA structures in the replication of tick-borne encephalitis virus in mammalian cells. PLOS Negl Trop Dis. 2023;17(1):e0011098. doi: 10.1371/journal.pntd.0011098 Epub 20230123. PubMed PMID: 36689554; PubMed Central PMCID: PMC9894543.

 PubMed Web of Science ®Google Scholar

Filomatori CV, Carballeda JM, Villordo SM, et al. Dengue virus genomic variation associated with mosquito adaptation defines the pattern of viral non-coding RNAs and fitness in human cells. PLOS Pathog. 2017;13(3):e1006265. doi: 10.1371/journal.ppat.1006265 Epub 2017 Mar 07. PubMed PMID: 28264033; PubMed Central PMCID: PMC5354447.

 PubMed Web of Science ®Google Scholar

Reid CR, Hobman TC. The nucleolar helicase DDX56 redistributes to West Nile virus assembly sites. Virology. 2017;500:169–177. doi: 10.1016/j.virol.2016.10.025 Epub 20161104. PubMed PMID: 27821284.

 PubMed Web of Science ®Google Scholar

Xu Z, Hobman TC. The helicase activity of DDX56 is required for its role in assembly of infectious West Nile virus particles. Virology. 2012;433(1):226–235. doi: 10.1016/j.virol.2012.08.011 Epub 20120825. PubMed PMID: 22925334; PubMed Central PMCID: PMC7119007.

 PubMed Web of Science ®Google Scholar

Overwijn D, Hondele M. DEAD-box ATPases as regulators of biomolecular condensates and membrane-less organelles. Trends Biochem Sci. 2023;48(3):244–258. doi: 10.1016/j.tibs.2022.10.001 Epub 20221104. PubMed PMID: 36344372.

 PubMed Web of Science ®Google Scholar

Krishnan MN, Ng A, Sukumaran B, et al. RNA interference screen for human genes associated with West Nile virus infection. Nature. 2008;455(7210):242–245. doi: 10.1038/nature07207 PubMed PMID: 18690214; PubMed Central PMCID: PMC3136529.

 PubMed Web of Science ®Google Scholar

Michalski D, Ontiveros JG, Russo J, et al. Zika virus noncoding sfRnas sequester multiple host-derived RNA-binding proteins and modulate mRNA decay and splicing during infection. J Biol Chem. 2019;294(44):16282–16296. doi: 10.1074/jbc.RA119.009129 Epub 20190913. PubMed PMID: 31519749; PubMed Central PMCID: PMC6827284.

 PubMed Web of Science ®Google Scholar

Paranjape SM, Harris E. Y box-binding protein-1 binds to the dengue virus 3’-untranslated region and mediates antiviral effects. J Biol Chem. 2007;282(42):30497–30508. doi: 10.1074/jbc.M705755200 Epub 20070828. PubMed PMID: 17726010.

 PubMed Web of Science ®Google Scholar

Bonenfant G, Williams N, Netzband R, et al. Zika virus subverts stress granules to promote and restrict viral gene expression. J Virol. 2019;93(12):10–128. doi: 10.1128/JVI.00520-19 Epub 20190529. PubMed PMID: 30944179; PubMed Central PMCID: PMC6613768.

 Web of Science ®Google Scholar

Diosa-Toro M, Kennedy DR, Chuo V, Popov VL, Pompon J, Garcia-Blanco MA, Coyne CB. Y-Box binding protein 1 interacts with dengue virus nucleocapsid and mediates viral assembly. MBio. 2022;13(1):e0019622. doi: 10.1128/mbio.00196-22 Epub 20220222. PubMed PMID: 35189699; PubMed Central PMCID: PMC8903895.

 PubMed Web of Science ®Google Scholar

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.