Integrative Structural Modeling of Viral Replication Complexes in Situ

Introduction

Viral replication complexes (VRCs) are dynamic, multi-protein assemblies that orchestrate genome replication and transcription within infected host cells. Understanding the atomic architecture of these complexes in their native cellular environment is essential for elucidating replication mechanisms and identifying vulnerable interfaces for antiviral intervention [1, 2]. Traditional structural biology methods such as X-ray crystallography and single-particle cryo-electron microscopy (cryo-EM) often require purification of individual components outside the cellular context, which can disrupt labile interactions and omit critical host factors [1]. In situ structural approaches, including cryo-electron tomography (cryo-ET) and targeted in situ cross-linking mass spectrometry (CLMS), now enable direct visualization and modeling of VRCs within intact cells [1, 2]. When combined with deep learning-based protein structure prediction (e.g., AlphaFold2), these experimental restraints can be integrated to generate full-length atomic models of large, multi-protein assemblies [1]. This article reviews the current integrative structural modeling paradigm for VRCs, with emphasis on the combination of cryo-ET and in situ CLMS with computational prediction, and discusses its application to veterinary virology.

In Situ Structural Biology Techniques

Cryo-Electron Tomography

Cryo-ET allows three-dimensional imaging of macromolecular complexes in their native cellular context at nanometer resolution [2]. By tilting a vitrified specimen and recording a series of projection images, a tomographic volume is reconstructed that preserves the spatial organization of VRCs within the host cell [2]. Recent advances in cryo-ET have enabled the visualization of viral nucleocapsid assembly, membrane remodeling, and replication factory architecture [2]. For example, cryo-ET of Ebola virus-infected cells revealed the helical nucleocapsid structure and its interactions with the viral matrix protein, providing insights into assembly and budding [2]. This technique is particularly valuable for studying pleomorphic viruses and those that form large, membrane-associated replication compartments [2].

In Situ Cross-Linking Mass Spectrometry

In situ CLMS captures structural information by introducing chemical cross-linkers into living cells, covalently linking proximal lysine residues within proteins [1]. After enrichment of the target protein, cross-linked peptides are identified by mass spectrometry, yielding distance restraints that report on intra- and intermolecular contacts [1]. This approach can be applied to recalcitrant proteins that are unstable outside the cellular environment [1]. In a landmark study, targeted in situ CLMS was applied to three SARS-CoV-2 proteins (Nsp1, Nsp2, and nucleocapsid N), producing cross-link sets with an average density of one cross-link per 20 residues [1]. These restraints were sufficiently detailed to enable integrative modeling of full-length protein structures [1].

Computational Prediction of Protein Structures

Deep learning methods, particularly AlphaFold2, have revolutionized protein structure prediction by generating accurate atomic models from amino acid sequences [1]. AlphaFold2 uses a neural network trained on the Protein Data Bank to predict inter-residue distances and angles, producing models with near-experimental accuracy for many proteins [1]. In the context of VRCs, AlphaFold2 can predict the structures of individual domains that are then assembled into larger complexes using experimental restraints [1]. For instance, the full-length model of SARS-CoV-2 Nsp2 was built by combining AlphaFold2-predicted domain structures with cross-linking data from in situ CLMS [1]. This hybrid approach overcomes the limitations of both methods: AlphaFold2 provides high-resolution domain models, while cross-links validate and guide the relative orientation of domains [1].

Integrative Modeling Workflow

Integrative structural modeling combines diverse experimental and computational data to produce a single consistent atomic model [1]. The workflow typically involves the following steps:

1. Data acquisition: Cryo-ET provides low-resolution density maps of the VRC in situ [2]; in situ CLMS yields distance restraints between specific residues [1].
2. Domain prediction: AlphaFold2 or other deep learning tools generate atomic models of individual protein domains [1].
3. Spatial assembly: Computational algorithms (e.g., Monte Carlo sampling, molecular dynamics) position the domain models within the cryo-ET density while satisfying cross-linking restraints [1].
4. Validation: The resulting model is assessed for consistency with all input data and refined iteratively [1].

The Mermaid diagram below illustrates this integrative workflow.

flowchart TD
    A[In situ cryo-ET of infected cells], > B[Tomographic reconstruction of VRC]
    C[In situ CLMS of target proteins], > D[Cross-link distance restraints]
    E[AlphaFold2 domain prediction], > F[Atomic models of individual domains]
    B, > G[Integrative modeling: domain docking into cryo-ET density with CLMS restraints]
    D, > G
    F, > G
    G, > H[Full-length atomic model of VRC]
    H, > I[Visualization of subunit interfaces and 3D architecture]

Case Studies

SARS-CoV-2 Nsp2 and Nucleocapsid Protein

Using targeted in situ CLMS, Slavin et al. obtained extensive cross-link sets for SARS-CoV-2 Nsp2 and N protein [1]. For Nsp2, the cross-links revealed a complex topology with long-range interactions, suggesting a role in zinc regulation within the replication-transcription complex [1]. Integrative modeling with AlphaFold2 domain predictions produced a single consistent all-atom model of full-length Nsp2, which identified three putative metal-binding sites [1]. For the N protein, intra- and interdomain cross-links enabled the construction of an N dimer model that can accommodate three single RNA strands simultaneously, both stereochemically and electrostatically [1]. These models demonstrate how in situ CLMS data can be integrated with computational predictions to resolve the architecture of recalcitrant viral proteins [1].

Ebola Virus Nucleocapsid Assembly

Watanabe et al. employed cryo-ET to visualize intracellular Ebola virus nucleocapsid assembly in situ [2]. The tomograms revealed the helical organization of the nucleocapsid and its interactions with the viral matrix protein, providing a structural framework for understanding assembly and budding [2]. Although this study did not use integrative modeling with AlphaFold, the cryo-ET density maps can serve as templates for docking predicted structures of individual nucleocapsid proteins, following the workflow described above [2]. The combination of cryo-ET with computational prediction holds promise for generating atomic models of filovirus replication complexes in their native cellular context [2].

Applications to Veterinary Virology

The integrative structural modeling approach described here is directly transferable to veterinary viruses of economic and zoonotic importance. For example, the replication complexes of porcine reproductive and respiratory syndrome virus (PRRSV), African swine fever virus (ASFV), and avian influenza virus are large, membrane-associated assemblies that have been challenging to characterize by conventional methods [1, 2]. In situ cryo-ET can visualize these complexes within infected livestock cells, while targeted in situ CLMS can provide residue-specific distance restraints for viral proteins that are difficult to purify [1, 2]. AlphaFold2 predictions of viral polymerase subunits, helicases, and accessory proteins can then be integrated with these experimental data to generate atomic models of the entire VRC [1]. Such models can reveal subunit interfaces that may be targeted by antiviral compounds or vaccines [1]. For a broader discussion of viral replication factory formation, see the article on Structural bioinformatics of viral replication factory formation and organelle remodeling. For details on AlphaFold-based modeling of viral glycoproteins, refer to AlphaFold-Based Structural Modeling of Viral Glycoproteins.

Challenges and Future Directions

Despite its promise, integrative structural modeling of VRCs in situ faces several challenges. Cryo-ET resolution is often limited to the nanometer scale, making it difficult to assign secondary structure elements without additional restraints [2]. In situ CLMS requires efficient enrichment of the target protein and may miss transient or low-abundance interactions [1]. Computational prediction of large multi-protein complexes remains computationally expensive, and the accuracy of AlphaFold2 for viral proteins with limited sequence homology can be variable [1]. Future developments in cryo-ET detectors and phase plates will improve resolution, while advances in cross-linking chemistry and mass spectrometry sensitivity will increase the density of distance restraints [1, 2]. Integration with other in situ techniques, such as correlative light and electron microscopy and single-molecule fluorescence, will provide dynamic information to complement static models [2].

Conclusion

Integrative structural modeling that combines cryo-ET, in situ CLMS, and deep learning-based protein structure prediction offers a powerful framework for elucidating the atomic architecture of viral replication complexes in their native cellular environment. The successful application of this approach to SARS-CoV-2 and Ebola virus proteins demonstrates its feasibility and potential for veterinary virology. By resolving subunit interfaces and 3D organization, these models can inform the design of antiviral strategies targeting conserved features of VRCs across diverse viral families.

References

[1] Slavin M, Zamel J, Zohar K, et al. Targeted in situ cross-linking mass spectrometry and integrative modeling reveal the architectures of three proteins from SARS-CoV-2. Proceedings of the National Academy of Sciences of the United States of America. 2021. URL: https://www.semanticscholar.org/paper/378eaaa7967106395db1e452038ad5282cf07f7b

[2] Watanabe R, Zyla D, Parekh D, et al. Intracellular Ebola virus nucleocapsid assembly revealed by in situ cryo-electron tomography. Cell. 2024. URL: https://pubmed.ncbi.nlm.nih.gov/39293445/ *** 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.