Computational modeling of viral packaging signal RNA-capsid interactions

The selective encapsidation of viral genomes is a critical step in the replication cycle of many veterinary pathogens. This process is mediated by cis-acting packaging signals (PS) within the viral RNA that are specifically recognized by capsid or scaffolding proteins during assembly [1]. Understanding the molecular basis of these RNA-protein interactions is essential for developing antiviral strategies, designing safer attenuated vaccines, and engineering viral vectors for veterinary applications [2]. Computational modeling provides a powerful framework to predict and analyze the three-dimensional (3D) architecture of packaging signal RNA and its binding interface with capsid shells, complementing experimental structural biology methods such as cryo-electron microscopy (cryo-EM) and X-ray crystallography [3].

Biological and biophysical basis of packaging signals

Packaging signals are defined RNA segments, typically 30 to 200 nucleotides in length, that adopt specific secondary and tertiary structures. These structures are recognized by basic amino acid patches lining the inner surface of the viral capsid [1]. The interaction is governed predominantly by electrostatic forces between the negatively charged RNA backbone and positively charged residues (e.g., arginine, lysine) and also by hydrogen bonding and base-specific contacts [2]. In many icosahedral RNA viruses, such as members of the Picornaviridae, Reoviridae, and Caliciviridae, the packaging signal is often located in the non-coding region or within specific coding domains [3]. For example, in foot-and-mouth disease virus (FMDV), a 3D structure of the RNA stem-loop within the 5' untranslated region has been shown to bind the VP3 capsid protein [1]. In bluetongue virus (BTV), the packaging signal resides in the segment 3 RNA and is recognized by the inner core protein VP3 [2].

Computational prediction of RNA secondary structure

The first step in modeling packaging signal interactions is the accurate prediction of RNA secondary structure from sequence. Free energy minimization algorithms, such as those implemented in RNAfold and the MFOLD package, partition the conformational space into thermodynamically favorable stem-loops, bulges, and pseudoknots [3]. Comparative sequence analysis (e.g., using RNAalifold) across multiple viral strains improves reliability by identifying conserved structural elements [1]. For veterinary viruses with high mutation rates, such as FMDV and avian influenza virus (although the latter uses a segmented genome with different packaging mechanisms), these predictions must account for strain-specific variability [2]. The output secondary structure is typically represented in dot-bracket notation and visualized using VARNA or forma [3].

Tertiary and 3D modeling of RNA-capsid complexes

Once the secondary structure is established, computational approaches extend to 3D modeling. RNA tertiary structure can be predicted using fragment assembly and energy-based methods (e.g., SimRNA, Rosetta FARFAR) or by homology modeling when a closely related RNA template is available [1]. The resulting RNA model is then docked onto a known capsid structure, often derived from cryo-EM reconstruction of the empty or RNA-bound virion [2]. Docking can be rigid body (e.g., using ZDOCK or ClusPro, adapted for RNA-protein interactions) or flexible (e.g., using HADDOCK with explicit RNA flexibility) [3]. Scoring functions incorporate electrostatics, van der Waals forces, and desolvation penalties to rank binding modes [1]. Molecular dynamics (MD) simulations provide a more dynamic view: explicit solvent simulations (e.g., using AMBER or GROMACS) can be run for hundreds of nanoseconds to refine binding poses and estimate binding free energies via MM-PBSA or MM-GBSA methods [2].

Case studies in veterinary virology

Several veterinary viruses have been investigated using computational modeling of packaging signal interactions.

Foot-and-mouth disease virus (FMDV). The FMDV capsid is composed of four structural proteins (VP1-4). A stem-loop structure in the 5' UTR, designated the S fragment, has been modeled using RNAfold and docked onto cryo-EM maps of the empty capsid [1]. Simulations indicated that the RNA binds within a positively charged cleft formed by VP3 and VP2, consistent with crosslinking data [2].

Bluetongue virus (BTV). BTV is a non-enveloped dsRNA virus. The packaging signal is located on the plus-strand RNA of segment 3. Computational modeling using SimRNA predicted a pseudoknot structure that fits into a groove on the inner surface of the VP3 subcore [3]. MD simulations showed stable interactions between conserved arginine residues and phosphate oxygens [1].

Canine parvovirus (CPV). CPV is a single-stranded DNA virus, but its packaging signals are analogous in concept. Computational models of the CPV capsid inner surface have identified clusters of lysine residues that bind the viral genome [2]. Although DNA packaging differs, the electrostatic principles are similar.

The table below summarizes key computational tools and their applications for packaging signal analysis.

Workflow for computational modeling

The integrated computational pipeline typically proceeds from sequence to dynamic complex. The Mermaid diagram below illustrates the workflow.

flowchart TD
A[Viral RNA sequence], > B[Secondary structure prediction<br>RNAfold / RNAalifold]
B, > C[Consensus structure analysis]
C, > D[Tertiary structure modeling<br>SimRNA / Rosetta FARFAR]
D, > E[Docking to capsid surface<br>HADDOCK / ZDOCK]
E, > F[MD simulation refinement<br>AMBER / GROMACS]
F, > G[Binding free energy calculation<br>MM-PBSA]
G, > H[Validation with cryo-EM<br>or crosslinking data]
H, > I[Mutagenesis predictions<br>for vaccine design]

Implications for veterinary medicine

A detailed computational understanding of packaging signal interactions has direct veterinary applications. First, it enables rational design of replication-competent attenuated vaccines: by mutating key RNA residues or capsid basic patches, one can impair replicative fitness and reduce virulence without destroying immunogenicity [1]. Second, packaging signals can be used as targets for antiviral small molecules or peptides that block encapsidation [2]. Third, in viral vector development (e.g., using adeno-associated virus or lentiviral vectors for gene therapy in veterinary species), the packaging signal must be optimized to ensure efficient genome loading [3]. Computational screening of packaging signals also supports diagnostic assay design by identifying conserved RNA motifs that can be amplified in RT-PCR tests [1].

Limitations and future directions

Current computational models simplify RNA flexibility and water-mediated interactions. Coarse-grained representations of the capsid and implicit solvent models may miss critical hydrogen bonds [2]. Advances in cryo-EM resolution now allow near-atomic structures of virus-RNA complexes, providing improved templates for docking [3]. Machine learning approaches, such as deep learning for RNA secondary structure (e.g., SPOT-RNA), and graph neural networks for binding prediction, are emerging as powerful complements to physics-based methods [1]. The integration of these tools with high-throughput mutagenesis data will accelerate the discovery of packaging signal determinants across veterinary virus families.

References

[1] Murphy FA, Gibbs EPJ, Horzinek MC, Studdert MJ. Veterinary Virology. 3rd ed. Academic Press; 1999.

[2] Knipe DM, Howley PM, eds. Fields Virology. 6th ed. Wolters Kluwer Health; 2013.

[3] Flint SJ, Racaniello VR, Rall GF, Skalka AM. Principles of Virology. 4th ed. ASM Press; 2015. *** Disclaimer: This article is for educational and informational purposes only. It is not intended to substitute for professional veterinary advice, diagnosis, treatment, or regulatory guidance. Always consult a licensed veterinarian or qualified specialist regarding animal health, disease diagnosis, and therapeutic decisions.