Structured RNAs in flavivirus genomes
What is structured in the untranslated regions of flavivirus genomes, how exoribonuclease-resistant RNAs stall Xrn1, how these elements are distributed across mosquito-borne, tick-borne and insect-specific lineages, and how they can be designed from scratch.
The genome of a flavivirus is a single positive-strand RNA of about eleven kilobases with one long open reading frame and two untranslated regions. The untranslated regions are not spacers. They carry structured elements that the virus depends on, i.e. elements that resist a host enzyme, that bring the two ends of the genome together, or that bind host proteins, and most of them are conserved as a fold rather than as a sequence. This page collects the work on this site on these elements, from comparative analyses across virus lineages to the design of such a fold from scratch.
How an RNA stops an enzyme
The clearest case is the exoribonuclease-resistant RNA (xrRNA). The host exoribonuclease Xrn1 degrades RNA from the 5′ end. In a flavivirus the enzyme stalls at a defined position in the 3′ untranslated region, and the RNA downstream of that position survives as subgenomic flaviviral RNA (sfRNA), which interferes with the antiviral response of the host.
The resistance is mechanical, not chemical. The element folds into a ring whose closure is held by a pseudoknot, the 5′ end of the element is threaded through that ring, and the construction holds against the enzyme pulling on it. Sequence identity matters only where it maintains this topology, which is why xrRNAs are hard to find by sequence comparison alone and why they were found by looking for the fold.
What is conserved across the lineages
Mosquito-borne flaviviruses such as dengue, Zika, West Nile, Japanese encephalitis and yellow fever virus build their 3′ untranslated regions from a small set of recurring modules: one or more xrRNAs, dumbbell elements and a terminal stem-loop. The number of copies and their arrangement differ between lineages, the vocabulary does not. The book chapter Functional RNA structures in the 3′UTR of mosquito-borne flaviviruses describes that vocabulary for the group.
The comparative analysis Comparative genomics of flavivirus 3′ UTR RNA structures (Viruses 2019) extends the question to tick-borne, insect-specific and no-known-vector flaviviruses. It found xrRNA-like elements distributed far more widely than the lineage-specific view had suggested, and a repeated, cascaded organisation of duplicated elements in insect-specific viruses. Within one virus species the same logic applies: RNA structure conservation and molecular epidemiology of TBEV (Virus Evolution 2022) shows a strongly conserved core domain near the 3′ end of tick-borne encephalitis virus and a variable region upstream of it that is still built from the same modules, combined with a phylogeny of the virus in space and time.
New viruses, same toolkit
Several of the papers here start in the field rather than at the computer, with a virus isolated from mosquitoes or ticks, and then ask whether the new genome carries the known elements.
- Insect-specific flaviviruses from Zambia and their exoribonuclease-resistant RNAs (Viruses 2020). Two viruses from Culex mosquitoes that replicate in mosquito cells but not in mammalian or avian cells. Their 3′ untranslated regions carry xrRNA-like folds, dumbbells and a terminal stem-loop, and the predicted xrRNAs stall Xrn1 in vitro.
- Mpulungu virus and unique xrRNAs in a novel African tick flavivirus (Scientific Reports 2021). A virus from a Rhipicephalus tick in Zambia that groups with Ngoye virus outside the established tick-borne flaviviruses. Two xrRNA elements in its 3′ untranslated region, both confirmed biochemically.
- Xinyang flavivirus and a likely tick-only orthoflavivirus clade (Journal of General Virology 2024). A virus from ticks in Henan Province, China, that joins the same branch. Its dinucleotide composition and a missing furin cleavage site in prM argue for a lineage without a vertebrate host, yet the 3′ untranslated region keeps the usual architecture of pseudoknot-containing stem-loops and a dumbbell.
Taken together these three viruses say something that no single genome could: the structured 3′ untranslated region belongs to the core of the flavivirus genome, not to the part of it that tracks the vertebrate host.
Other elements, other hosts
Not every structure in these genomes is an xrRNA.
- RNA–RNA interaction analysis of Japanese encephalitis virus (Nucleic Acids Research 2023) measures the long-range interaction that cyclizes the genome. The two terminal regions bind with nanomolar affinity in a one-to-one complex, and the affinity drops when the conserved cyclization sequence is disrupted.
- A conserved G-quadruplex in the Zika virus 3′ terminal region (Biochemistry and Cell Biology 2024) shows that a conserved G-rich segment forms a quadruplex in the context of a longer transcript and that the human helicase DDX17 unfolds it.
- Musashi binding elements in Zika virus 3′UTR (Scientific Reports 2019) asks where the binding motif of the host protein Musashi-1 is accessible rather than merely present, using opening energies against shuffled controls.
- 3′ UTR-biased siRNA production in an insect-specific flavivirus (Journal of Virology 2024) finds that more than 95 % of the small interfering RNAs derived from Kamiti River virus map to its long 3′ untranslated region, and that this bias does not disappear when sfRNA production is reduced. The region dominates the antiviral RNA interference response for a reason that sfRNA biogenesis does not explain.
A related element outside the 3′ untranslated region is the internal ribosome entry site of bovine viral diarrhea virus, a pestivirus. Secondary and tertiary structure of the BVDV IRES (RNA Biology 2022) combines SHAPE probing with structure modelling and supports a pseudoknot in its domain III.
From comparison to design
If the function of an xrRNA follows from its topology, the fold should be designable. Rational design of mechanically active RNAs (Nucleic Acids Research 2026) reduces natural mosquito-borne xrRNAs to a symbolic description that keeps the three-way junction, the two pseudoknots and the length constraints between them, generates sequences under those constraints, and filters the candidates by three-dimensional modelling and molecular dynamics. Of the two pseudoknots the second decides the mechanical resistance. The third construct in the series stalls Xrn1 as efficiently as the natural element while most of the recognisable sequence signal is gone. The earlier preprint note records how the project was framed, and the German mini-review Strukturierte RNAs in Viren summarises the idea for a broader audience.
Methods and further reading
The structures on this page were found with comparative genomics and thermodynamic structure prediction. Some were then tested in the laboratory, e.g. the xrRNA candidates of the Zambian isolates and of Mpulungu virus, which were shown to stall Xrn1 in vitro, and the designed elements of the 2026 design paper. Others, e.g. the elements of Xinyang virus and the conserved architectures of the comparative studies, rest on prediction and comparison alone, and the posts say which is which. The methods behind the predictions are described in the tutorials on RNA structure prediction with ViennaRNA, and the terms used here, e.g. 3′ UTR, stem-loop, dumbbell and pseudoknot, are defined in the glossary. All posts on this subject are collected on the topic pages flavivirus and xrRNA. I was guest editor of a Viruses special issue on functional RNAs in virology on the same subject.