Structural and Dynamic Asymmetry in Icosahedrally Symmetric Virus Capsids

Abstract

A common characteristic of virus capsids is icosahedral symmetry, yet these highly symmetric structures can display asymmetric features within their virions and undergo asymmetric dynamics. The fields of structural and computational biology have entered a new realm in the investigation of virus infection mechanisms, with the ability to observe symmetry-breaking features. This review will cover important studies on icosahedral virus structure and dynamics, covering both symmetric and asymmetric conformational changes. However, the main emphasis of the review will be towards recent studies employing cryo-electron microscopy or molecular dynamics simulations, which can uncover asymmetric aspects of these systems relevant to understanding viral physical-chemical properties and their biological impact.

Graphical abstract

Introduction

The structure of virus capsids has been of intense interest dating back to the invention of the electron microscope in the 1930s. The size and shape diversity of virus structures was a revealing feature, as both rod-like, spherical and tailed viruses were observed in the following decades. As electron microscopy techniques improved, largely through the use of negative staining, the symmetry features of capsids became more apparent. Watson and Crick first proposed that spherical viruses could have icosahedral symmetry [1], but it was Caspar and Klug [2] that resolved the range of sizes and the accommodation of greater than 60 subunits in the capsid shell, through their principle of quasi-equivalence. In a nutshell, the theory states that the protein subunits exist in both pentameric and hexameric arrangements, but the protein-protein interfaces in the pentamers and hexamers are very similar though slightly different (i.e. quasi-equivalent). This allows for a scalable design in which a variable numbers of hexamers can be packed between the 12 vertices, which are each occupied by a pentamer. The classification of these structures is characterized by a triangulation (T) number. The high resolution realization of this arrangement was enabled by X-ray crystallography, which was used to first determine the structure of tomato busy stunt virus (TBSV) at 2.9 Å in 1978 [3] and soon thereafter the structure of southern bean mosaic virus (SBMV) at 2.8 Å [4]. Both TBSV and SBMV have T=3 capsid organization, and these structures allowed for the visualization and detailed molecular understanding of the packaging arrangement within and between pentamers and hexamers for the first time.

The high symmetry (60-fold) of icosahedral capsids makes them tractable macromolecular complexes to pursue structural studies by both X-ray crystallography and cryo-electron microscopy (cryo-EM). This is because the icosahedral symmetry can be exploited through icosahedral averaging, to increase the sampling and enhance the resolution. In this approach each asymmetric unit is treated as an independent unit, which as mentioned provides an approximately 60-fold increase to the data set. However, the resultant structures solved employing icosahedral averaging will, by definition, be perfectly icosahedrally symmetric. Any local structural variation between the different asymmetric units will be lost in the averaging, and any non-symmetric features, such as portals, tails, sub stoichiometric accessory proteins and regions of viral genomes will not be resolvable. While X-ray crystallography was the dominant technique in solving virus structures through the 1980’s and 1990’s, cryo-EM has become the leading technique in solving virus structures in the current millennium. According to the virus structure database (VIPERdb) [5] there are over 800 virus structures solved, with slightly more than half coming from cryo-EM, and of those cryo-EM structures 65% have 5 Å resolution or better and 33% have resolutions exceeding 3.5 Å.

While native structures and structure of virus-like particles have been invaluable in our understanding of icosahedral virus organization, the forefront of structural virology and the focus of this review is to understand the dynamics of virus capsids. Viruses are highly constrained systems, due to limitations on genome size, and therefore the viral proteins and the capsid itself, must be dynamic and multifunctional to carry out the many processes (e.g. assembly [6], maturation [7], genome release, disassembly [8]) required to proliferate a viral infection. This review will focus on two techniques which are capable of revealing asymmetric features and dynamics, which are molecular simulations and cryo-EM. The “resolution revolution” in cryo-EM [9] has advanced the technique to the point where virus structures can be resolved without icosahedral averaging to allow for asymmetric features to be observed [10]. Molecular simulations are also not constrained by symmetry, though it was only in recent years that computational power has been sufficient to conduct atomistic simulations of full icosahedral capsids [11]. Nonetheless much of the research advances in modeling capsids is built upon pioneering work which involved both coarse-grained modeling and icosahedrally constrained approaches. This review will highlight foundational studies that have advanced our understanding of virus particle dynamics, including both older studies involving coarser modeling approaches and lower resolution structures as well as the recent advances which can provide full atomic detail without symmetry constraints.

Global Morphological Changes

Structural studies dating back to the 1990’s were able to reveal that many virus capsids do not exist in a single conformation, but can undergo structural transitions and these transitions are often correlated with a change in infectivity of the particle, a process known as maturation [12]. While many of these structures were resolved by cryo-EM at lower resolution (> 10 Å), it is instructive to look at some earlier studies that laid a foundation for understanding physical features of these transitions and also motivated advances in experimental and modeling approaches that have led to the current state-of-the-art. The bacteriophage HK97 (T=7) is a model system for understanding large scale capsid reorganization, as the capsid protein fold is observed in many other phages as well as viruses that infect eukaryotes and archaea [13]. This system has been extensively studied to understand virus maturation and related morphological changes to the capsid. HK97 assembles in a precursor shell, termed prohead I, followed by cleavage and digestion of a region of the capsid protein, known as the delta-domain, by a virally encoded protease [14]. This state with the cleaved delta domains is termed prohead II and is considered a metastable particle. Prohead II is relatively spherical and has a diameter of 54 nm, and has a high resolution structure determined by X-ray crystallography (Figure 1A) [15]. The packaging of the viral DNA genome through a portal at the 5-fold vertex triggers expansion of the capsid leading to the fully packaged, infectious and mature state, termed head II (Figure 1B) [16]. Head II has a diameter of 66 nm, has a faceted/polyhedral capsid shape and is stabilized by inter-protein crosslinks throughout the capsid shell forming a “chain-mail” organization of crosslinks [17]. The maturation transitions can also be stimulated in-vitro by lowering the pH followed by neutralization which appears to generate an analogous transition to the DNA stimulated maturation. The dynamics of the transition have been studied experimentally using scattering methods and cryo-EM and several intermediate states have identified and characterized [18–21]. In these approaches the structural resolution of the intermediates has been too low to provide atomic detail, and structural details of the transitions between states can only be inferred.

Figure 1.

Structural dynamics of viruses. A) The prohead II structure of bacteriophage HK97 (PDB ID: 3e8k) [15]. B) The head II structure of of HK97 (PDB ID: 1OHG) [16]. C) The icosahedrally averaged X-ray structure of FHV (PDB ID: 4FTB). D) EM map of heat treated (“puffed”) particle of FHV (EMD-9732) [38]. The red density on the right is believed to be RNA extruded from the capsid.

Computational approaches have the ability to resolve the molecular motions in high temporal and spatial resolution in physics-based models, especially when there are known endpoints of a transition. The first such study to examine the HK97 maturation transition utilized a coarse-grained approached and simplified model (normal mode analysis (NMA) of an elastic network model (ENM)), but interestingly showed the transition between prohead II and head II states could be described by two icosahedrally symmetric normal modes [22]. This profound result was qualitatively different to studies on other viruses (cowpea chlorotic mottle virus (CCMV ,T=3) and nudaurelia capensis virus (NωV, T=4)) which undergo swelling transitions, but retain spherical morphologies. In CCMV and NωV a single mode could characterize the transition because the change was primarily expansion without significant shape distortion. Whereas HK97 undergoes both expansion and faceting (Figure 1A–B) and a second mode is required to describe the morphological change. In all of these systems, both experimentally determined end-point structures were themselves icosahedrally symmetric, which imposes a limitation in the NMA approach. NMA of an icosahedrally symmetric structure will generate some modes which are icosahedrally symmetric and other modes which have lower symmetries, but only the icosahedrally symmetric modes will contribute to symmetry-preserving transitions as modes with different symmetries are orthogonal (perpendicular) to the icosahedrally symmetric transition. In a first study that removed the constraint of icosahedrally symmetric transition pathways, May et al., performed molecular dynamics (MD) simulations of the transition using a different type of coarse-grained model, known as a structure-based model [23] or a Gō-like model. They were able to capture many transitions between the prohead II and head I/II state and observed that non-icosahedral motions were present along the pathway though (again owing to icosahedrally symmetric end states) the dominant motions along the pathways were icosahedrally symmetric [24]. In a subsequent study, May et al. [25], conducted atomistic simulations of the HK97 maturation transition, but under icosahedrally symmetric constraints using the rotational symmetry boundary condition (RCSB) method [26]. In that work, they constructed a low-energy pathway between immature and mature capsid structures, and umbrella sampling simulations were performed to estimate the free energy change associated with the maturation process, which was in reasonably good agreement with differential scanning calorimetry (DSC) measurements [27]. The umbrella sampling simulations were conducted using a constant-pH approach [28]. which allowed for the effects of pH on the free energy landscape to be estimated based upon calculated residue pK_a values. Ultimately, they proposed a model to relate the in vitro (pH driven) and in vivo (DNA driven) pathways, in which packaging of DNA would induce upward shifts of the acidic residues pK_as, driving the transition in the same manner as acidification.

Localized/Asymmetric Conformational Transitions

Many virus capsid undergo conformational changes during their life cycle which may not involve large scale capsid size and shape changes or uniform (icosahedrally symmetric) changes across all asymmetric units. These changes can be promoted by receptor interactions, driven by changing chemical environments within the cell and utilized to overcome barriers on the pathway to virus replication. The animal virus Flock House virus (FHV, T=3) is a well-studied system [29] which undergoes capsid conformational changes to disrupt host cell membranes (Figure 1C). The FHV capsid has a 44 amino acid peptide (γ), which is observed on the capsid interior but is exposed or released to the capsid exterior under acidic (e.g. endosomal) conditions [30,31]. Interactions between γ and lipid bilayers have been characterized in-vitro and through MD simulation [32–36], though the mechanism of membrane disruption remains unresolved. Most recently a heat-treated FHV particle, which resembles a previously observed infection intermediate, known as an “eluted particle” [37], was characterized by cryo-EM in both icosahedrally averaged and asymmetric reconstructions [38]. The eluted particle icosahedrally averaged structure did not show density for the γ-peptides, while gold-labeled anti-γ antibody binding detected some γ peptides remain capsid associated, but far less than 180 copies. The non-uniform expulsion of γ peptides is consistent with previous studies on authentic eluted particles which expelled 25% of the peptides [37]. Further heating to 75 °C induced more drastic conformational changes, where a “puff” of density was observed above a 2-fold axis, in an asymmetric cryo-EM reconstruction with 26.2 Å resolution (Figure 1D). The exciting inference in this study is this puff is a portion of the RNA viral genome, which is supported by RNase A treatment.

Many other non-enveloped viruses have capsid structural elements which transition from internalized to externalized under specific conditions. Several members of the picornavirus family, including poliovirus (PV), rhinovirus and hepatitis A, all externalize membrane disrupting components of the capsid [39]. In PV (pseudo T=3), receptor binding triggers expansion of the capsid to a state known as 135S particle. In the 135S state the small VP4 protein (69 residues) and the N-terminus of VP1 are translocated from the capsid interior to exterior. A 5.5 Å resolution structure of the 135S particle was recently resolved using cryo-EM with icosahedral averaging, in which the orientation of the N-terminus of VP1 could be observed to protrude through a pore at the quasi 3-fold vertices [40]. In the presence of receptor coated liposomes, attachment between virus and membrane was observed and umbilical-like connections spanning 50 Å were observed by cryo-electron tomography [41]. RNA is transferred into the liposomes and is protected from RNase digestion, leading to a model in which the connectors are formed from protein and RNA, with VP1 and VP4 being the likely protein components. With the numerous other structural studies on PV and other related picornavirus intermediates [42–45], including the post RNA release state (80S) [44,46], there is now a detailed picture emerging of the molecular reorganization required for a capsid to go from the mature state prior to cell engagement through RNA release into the cytoplasm.

Asymmetric Structural Features Revealed by Cryo-EM

Technological and methodological advancements in cryo-EM have enabled researchers to move beyond icosahedrally averaged structures and can now determine high resolution structures without imposed symmetry [47,48]. This breakthrough has allowed for the observation of unique features in capsids which do not obey icosahedral symmetry and accompanying insights into infection mechanisms to be made. An example system where asymmetric reconstructions have advanced our understanding is in the RNA bacteriophage MS2 is member of the Levivirus family, which has a T=3 capsid (Figure 2A). MS2 virions contain a single copy of a maturation (“A”) protein, and while it was long known that the A-protein is required for attachment to its receptor, the E. coli F-pilus, and infectivity, the structural basis for this requirement could not be resolved through symmetry averaged structures [49]. Using cryo-electron tomography (ET), an asymmetric structure of MS2 bound to a pilus was determined at 39 Å resolution [50]. Despite the low resolution, this study provided strong evidence that the receptor binding occurs at a 2-fold axis. It was proposed that the A-protein would replace a coat protein dimer in the capsid, thereby breaking the icosahedral symmetry. In a study examining the emergence of disinfectant-resistant MS2, cryo-EM was used to reconstruct the asymmetric structure of wild-type and mutant MS2 virions at 10.5 and 9.5 Å resolution, respectively. [51] This study confirmed the location and role of the A-protein to bind its host, as the most significant structural changes between wild-type and disinfectant-resistance viruses were in the A-protein. The position of the A-protein was were further validated in a subsequent cryo-EM asymmetric reconstruction at 8.7 Å resolution (Figure 2B) [52]. In this higher resolution map, the A-protein could be clearly identified to be replacing a coat-protein dimer and protruding away from the capsid, poised for receptor binding. This study also revealed the RNA genome to be well ordered, with 44 stem loops (SL) interacting with coat protein dimers. These SL-dimer interactions were believed to be packaging signals, which are specific interactions that promote genome packaging and capsid assembly [53]. A higher resolution asymmetric reconstruction at 3.9 Å, was able to trace 80% of the genome, providing the most detailed image of full virion to date [54]. Similar features have been observed in the related Levivirus Qβ in a 3.7 Å resolution asymmetric reconstruction [55].

Figure 2.

Asymmetric features in icosahedral virus capsids. A) The icosahedrally averaged X-ray structure of MS2. (PDB ID:2MS2) [49]. B) The EM map of the MS2 internal RNA density (EMD-3404) [52]. The protruding blue density on the right is the A-protein. C) The FCV capsid (PDB ID: 6GSH) [58] with capsid proteins colored red, blue and grey. On the right side of the capsid, the portal is observed (PDB ID:6GSI) with receptor proteins colored yellow and the portal forming VP2 proteins in green. D) Top down view of the FCV portal complex (PDB ID:6GSI) [58].

Recent advances in data processing algorithms, known as masked classification or subparticle classification [47], have enabled the observation of unique portal structures in viruses such as herpes simplex virus 1 [56,57] and Feline calicivirus (FCV) [58]. FCV is an RNA virus with a T=3 capsid (Figure 2C). FCV contains a minor capsid protein (VP2) at low copy number, which is believed to be encapsidated inside the mature virus. When FCV is incubated with its cellular receptor, the particles undergo a conformational change resulting in a pore at the 3-fold symmetry axis constructed by 12 VP2 subunits surrounding the pore (Figure 2D). This unique structure is the first observation of a portal complex in a small RNA virus, and from the structure a model is suggested in which the VP2 portal would penetrate endosomal membranes and form a channel for genome release in the cytoplasm.

Dynamics of Icosahedral Viruses from MD Simulations

MD simulations are a powerful tool for studying biomolecular systems. The method relies on a potential function (force field) to describe the interactions within the system, from which forces are derived that are used to propagate the system forward in time using Newtonian dynamics. In these classical models, the system description can be fully-atomistic or can be of lower resolution (coarse-grained (CG)), which can drastically reduce the computational costs, though typically with some loss of accuracy. Historically, CG models were the reasonable and feasible choice to study the dynamics of assembled icosahedral virus capsids. In particular, structure-based models have been successful in quantitatively predicting mechanical properties of viruses and found useful in interpreting atomic force microscopy (AFM) experiments [59–62]. However, the rapid increase in computational power in recent years has enabled researchers to model complete virus capsids or virus-like particles at fully-atomistic resolution, and compute dynamics on long enough timescales (hundreds of ns – microsecond) to generate results that can help inform and interpret experimental efforts.

The first atomistic simulation of a complete virus particle was performed in 2006 by Schulten and coworkers on the T=1 satellite tobacco mosaic virus (STMV) and simulated for approximately 10 ns. The simulations were run in an explicit water environment, creating a system size of ~1 million atoms, which was unprecedented at the time [63]. Despite the short timescales, the simulation results clearly demonstrated that capsid stability was greatly enhanced by the presence of viral RNA. In the absence of RNA, the capsid distorted from icosahedral symmetry and collapsed, suggesting the viral genome plays a key role for assembly and stability of STMV virions. In 2009 the southern bean mosaic virus (SBMV, T=3) capsid was simulated and found to be relatively stable in the absence of viral genome during a 20 ns simulation [64]. In this work, the mechanical properties of the capsid were evaluated by a simulated atomic-force microscopy (AFM) nanoindentation approach. While spatial heterogeneity in the elastic constants across the capsid surface were observed, the force loading rates were several orders of magnitude faster than experimental loading rates. In 2012, van der Spoel and coworkers simulated another T=1 satellite virus, satellite tobacco necrosis virus (STNV) capsid over a timescale of 1 μs, which is more than an order of magnitude longer that previous full capsid simulation studies [65]. They studied the effect of removal of bound calcium ions from the capsid and found the capsid swells 2.5% upon Ca⁺² removal, which was consistent with experimental estimates. In a different study by van der Spoel and coworkers, half-capsids simulations of STMV and STNV, both with and without the crystallographically resolved RNA segments, were conducted for 200 ns [66]. Simulation results revealed a clear correlation between the binding site of the resolved (ordered) genome and the distribution of chloride ions, suggesting that it is possible to pinpoint the location of unresolved (disordered) viral genome segments.

In 2014 the poliovirus capsid was simulated for 200 ns, which was a considerably larger system (~6.5 million atoms) than previous full capsid simulation studies. They found the PV capsid to be semi-permeable, with rapid exchange of water between capsid interior and exterior but did not observe ion exchange [67]. In 2017 Nerukh and coworkers performed all-atom MD simulation on one of the smallest viruses, T=1 porcine circovirus type 2 (PCV2) and examined the role of (unbound) solvent ions on the capsid structure on the 10 ns timescale [68]. They found that neutralizing ions on the capsid interior to be critical for maintaining capsid symmetry and structure. In 2018 the T=3 Triatoma virus (TrV) capsid was studied through a multiscale simulation approach including quantum mechanical (QM), classical and CG simulations [69]. Atomistic MD simulation confirmed the existence of an ion channel along the 5-fold axis, whereas QM simulations revealed unidirectional proton conduction from the capsid interior to the exterior. These results lead to a hypothesis that an alkaline environment would induce proton leakage leading to rising pH inside the capsid which could alter electrostatic balance and could destabilization of the capsid structure. Recently, Hadden and co-workers performed MD on the empty T=4 hepatitis B virus capsid (HBV) [70]. This study was one of the most extensive to date, with the large system (~6 million atoms) being simulated for over 1 us. This enabled the observation of large-scale, asymmetric, collective motions, (Figure 3) and provided insights into the resolution limits that may be achievable in cryo-EM

Figure 3.

Dynamics of HBV capsid. A) Local dynamics from root-mean-squared fluctuations (RMSF) of atomic positions, stars indication 5-fold symmetry axis. B). Measurement of opposing pentameric displacements from capsid center reveal dynamic asymmetric distortions in the capsid shape. Figure is modified from [70].

To date, most all-atom MD simulation studies of complete virus capsids have been broadly employed to study the chemical–physical properties of the capsid, for example elastic properties, the effects of solvent (ions and water) on the capsid and the role of metal ions in capsid stability. However, the potential to further reveal functional dynamics of viruses is sure to be enabled by longer timescale computer simulations and information from high-resolution cryo-EM on intermediate states of these particles. The use of all-atom capsid simulations in computer aided drug design for both discovery and mechanistic insights has been initiated [71–73] and will continue to develop into the future.

Summary

Advancements in experimental structural techniques and technology as well as advancements in MD simulation algorithms and computational power have brought the field to brink of being able to resolve dynamics of viral infections. Gaining structural insights into intermediate states of viral structures during infection will provide sufficient data to construct and evaluate models describing the molecular acrobatics required for these transitions. In the near future we expect further insights into the nature of virus dynamics and how these processes are regulated will emerge. Furthermore, one can be optimistic that translational advancements in combating viral disease will progress, as revealing asymmetric features will provide new drug targets which were previously not accessible in structure-based drug design efforts.

Highlights.

• Symmetry breaking events in icosahedral viruses are important for infection

• Advances in cryo-EM technology has enabled asymmetric virus structure determination

• Current MD simulations can compute full capsid dynamics on ns – μs timescales

• Further integration of modeling and experimental data will advance drug discovery