The True Story and Advantages of RNA Phage Capsids as Nanotools

Pili-specific RNA phages, currently the most promising virus-like particle (VLP) carriers, are nonenveloped, spherical viruses with T = 3 icosahedral symmetry and diameters ranging from approximately 28 to 30 nm. RNA phage particles are composed of 178 chemically identical coat protein (CP) molecules, or 89 CP dimers, and one copy of maturation, or A, protein, which replaces a single CP dimer. The phage's monopartite, positive-sense, single-stranded (plus-ssRNA) genome is approximately 4 kb in size and serves as messenger RNA for the synthesis of the capsid-forming CP as well as three other viral proteins: the maturation, replicase, and lysis proteins (fig. 1).

The first Escherichia coli-infecting RNA phages that played an important role in VLP technology were identified in the early 1960s and included the f2 [1], MS2 [2], R17 [3], fr [4], M12 [5], and Qβ [6] phages. The early history and basic properties of RNA phages were reported in detail in two books [7,8] and in one original review [9].

Later, other RNA phages that became more common as VLP carriers were described, including the E. coli phages SP, FI [10] and GA [11], the Caulobacter crescentus phage φCB5 [12], the Pseudomonas aeruginosa phage PP7 [13], the broad host range, P-pili-specific phage PRR1 [14], and the Acinetobacter phage AP205 [15,][16].

According to the most current ICTV taxonomy release [17], RNA phages are members of the family Leviviridae. This family has not been assigned to any higher viral order and comprises two genera: Levivirus, which includes the species BZ13 (first mentioned in Inokuchi et al. [18]) and MS2, and Allolevivirus, which includes the species FI and Qβ. Other common RNA phages were mentioned in earlier ICTV taxonomy releases until 1998.

According to current NCBI Taxonomy browser information [19], the Levivirus genus consists of the BZ13 and MS2 species. The BZ13 species includes the phages GA and JP34, among others, as subspecies, while the MS2 species includes the phages f2, fr, M12, and R17, among others. The Levivirus genus also includes some unclassified members such as the Acinetobacter phage AP205 and the Pseudomonas phage PP7.

Based on the NCBI Taxonomy classification of the Allolevivirus genus, the FI and Qβ species include FI group subspecies (including phages SP, TW19, and TW28, among others) and Qβ group subspecies (including phages MX1 and ST, among others). Some species, including the Caulobacter phage φCB5 and P-pili-specific phages, predominantly the Pseudomonas phage PRR1, remain in the NCBI classification as unclassified Leviviridae members.

Generally, the current classification is based on serological typing [20] and has divided RNA phages into four serogroups, namely, serogroups I to IV [for a detailed discussion, see [21]]. This classification scheme has been confirmed by studies of the template specificity of RNA phage replicases [22], genetic analysis [23,24], and physiochemical data, e.g. by resistance to high hydrostatic pressure [25]. The RNA phages MS2, GA, Qβ, and SP are typically recognized as reference strains for the sero- and genogroups I, II, III, and IV, respectively. Therefore, group I and II members belong to the Levivirus genus, while serogroup III and IV members belong to the Allolevivirus genus. Ecological and wastewater studies have revealed that RNA phages from groups II and III are associated with human waste, whereas group I and IV members are predominantly associated with animal waste [21]. However, this distribution is not absolute and requires further refinement [26,27].

The genomic structures of the Levivirus and Allolevivirus genera members demonstrate some differences (fig. 1). In addition to the standard maturation protein, CP, and replicase subunit, the Allolevivirus genome encodes a C-terminally extended CP known as the minor A1 protein, which appears as a result of ribosomal read-through of a leaky opal termination codon of the CP gene [28] and is essential for the formation of viable Qβ particles in vivo [29,30,31]. The A1 protein is incorporated in 3-10 copies per virion, or in 12 copies in accordance with a recent study [32], it is required for infection, but its precise function is not known [for more references, see [8,29,30,31,32,33]]. A recent electron microscopy visualization of foreign epitopes carried by A1 protein within infectious Qβ particles showed that the A1 protein molecules are occupying corners of the Qβ icosahedron [32]. The lysis protein forms pores in the cellular membrane, leading to activation of autolysins and, eventually, cell lysis [34].

The current NCBI classification comprising two genera as well as four serogroups are adequate in the case of coliphages; however, attempts to assign phages from other hosts to Levivirus, Allolevivirus, or ‘unclassified' members have been rather artificial and lack solid rationale aside from historical considerations. In our opinion, a new classification system based solely on sequence similarities should be employed. For example, Levivirus and Allolevivirus genera could be left only for coliphages, and new genera could be introduced based on sequence similarities among conserved replicase protein sequences.

RNA phages were the first classic models used in early molecular biology and are considered ‘instrumental in the making of molecular biology' [35]. They provided the scientific community with purified RNA and markedly contributed to the decryption of the genetic code, the understanding of RNA translation and replication mechanisms, and the elucidation of virus-host interactions and the self-regulation of biological systems [for details and references, see [7,8,9,35]]. The RNA phage MS2 was the first organism with a fully sequenced genome [36]. The capsids of the RNA phages R17 and f2 were among the first observed virions with resolved icosahedral symmetry [37], after the classical work by Caspar and Klug [38] on the structural analysis of plant viruses. RNA phages have also presented substantial background for studies on phylogeny and genome evolution [33].

Furthermore, RNA phages paved the way for antisense-based gene therapy via the generation of the so-called ‘mRNA-interfering complementary RNA (micRNA) immune system' for the prevention of phage SP proliferation [39,40].

RNA phages, especially the phage MS2, have also markedly contributed to ecological and disinfection studies via their use in the development of numerous physical and chemical methods for genome inactivation, from early attempts [41] to recent systematic studies [42]. The RNA phages MS2, GA, Qβ, FI, SP, and PP7 are still efficiently used as surrogate models for the control of viral contamination in food production and storage, in industrial and clinical applications, and on health care personnel [for two recent examples, see [43,44]], in addition to serving as viral and microbial source tracking materials in wastewater [for references and discussion, see [45]]. It is broadly accepted that RNA phages are fully adequate surrogates for human enteroviruses in studies of virus contamination [46,47]. RNA phages are also often used as internal controls of extraction/amplification efficiency in modern RT-PCR kits for the surveillance of emerging pathogens, including the Ebola virus [48].

RNA phages were the first examples of an ‘operon' mechanism of gene regulation by ‘self' proteins [49]. This mechanism can be described as a full-cycle regulated biological system, where gene regulation is performed by two phage proteins: CP and replicase [50,51]. The CPs of most RNA phages have been shown to repress translation of the replicase gene by binding to an RNA hairpin as an operator at the start site of the replicase gene [7,8,9]. The ability of RNA phage CP to recognize the corresponding operator stem-loop led to the development of an efficient methodology based on the tethering of CP to CP-operator-tagged RNAs. Using this technique, mRNAs that are tagged with the operator sequence are highly specifically recognized by CP, which can be fused to fluorescent or other functional probes. This CP-operator tethering methodology enables imaging of the processing, export, localization, translation, and degradation of operator-tagged mRNA in living cells (see the review by Lampasona and Czaplinski [52] and a recent protocol from Bensidoun et al. [53] as an example). Furthermore, the tethering technique allows affinity purification of the desired RNA-protein complexes [54]. The tethering methodology mostly exploits the CP-operator composition from phage MS2, although PP7-based methodology, including the use of simultaneous MS2 and PP7 two-color labeling [55], has been recently described. A phage R17-based technique was also used in the early development of the tethering approach [56]. It is noteworthy that the tethering technique has recently been applied to further develop the highly productive CRISPR-Cas9 technology [57].

The expression of RNA phage CP genes in E. coli led to the high-level production of correctly self-assembled icosahedral capsids that were morphologically and immunologically indistinguishable from virions in the case of MS2 [58,59] and fr [60] of group I, JP34, an intermediate between groups I and II [61], GA of group II [62], Qβ of group III [63], SP of group IV [64], the unclassified Levivirus members PP7 [65,66] and AP205 [67,68], and the unclassified Leviviridae phages PRR1 [69] and φCB5 [70]. It is noteworthy that not only E. coli cells but also the E. coli-based cell-free translation system can be used for efficient in vitro production of MS2 VLPs [71,72].

Coinfection with two phages led to the production of mixed particles only in the case of the closely related phages MS2 and fr from the serogroup I and not in the case of the more distantly related phages fr and GA; however, the reassembly of recombinant fr and GA CP dimers in vitro allowed the generation of the mixed particles in both spherical and rod-like configurations [73].

Highly efficient production of VLPs was also achieved in the yeast species Saccharomyces cerevisiae and Pichia pastoris for phages MS2 [74], Qβ [75], GA [76], fr, AP205, PP7, and φCB5 [77]. Attempts to prepare RNA phage SP VLPs in S. cerevisiae and P. pastoris were unsuccessful [77].

As mentioned above, the RNA phages R17 and f2 played a unique role in revealing the virion's icosahedral symmetry [37]. The place of RNA phages in the global history of viral architecture was presented in a recent review [78]. Early electron microscopy studies demonstrated clear paracrystalline arrays of virions in E. coli cells infected with the RNA phages f2 [79], µ2 [80], and R17 [81]. Similar paracrystalline structures were later found in E. coli cells expressing the CP gene and producing VLPs; for example, see the electron micrograph of a slice of a cell filled with phage Qβ VLPs in Pumpens and Grens [82].

Fine 3D structures of the most typical RNA phage representatives have been resolved by X-ray crystallography (fig. 2) and were found to be very similar, despite the marked diversity in the primary structures of their CPs (fig. 1). The first 3D structure resolved was for MS2 virions. At the time, this structure showed no similarity to any other known viruses or proteins of any origin. The MS2 virion structure was first determined at a resolution of 3.3 Å [83,84,85] and then refined to a resolution of 2.8 Å [86]. Next, the crystal structure of an MS2 capsid with amino acid (aa) exchanges in the FG loop was resolved [87]. Historically, the first phage MS2 crystals and preliminary X-ray examination data were obtained in Walter Fiers' lab in 1979 [88].

According to the 3D structure of MS2, the 180 CP subunits of the virion are arranged in dimers as initial building blocks and form a lattice with a T = 3 triangulation number (fig. 2). The CP subunit consists of a five-stranded β-sheet facing the inside of the particle and a hairpin and two α-helices on the outside.

The structure of a recombinant capsid of the RNA phage fr was determined by X-ray crystallography at a resolution of 3.5 Å and was shown to be identical to the protein shell of the native virus [89,90]. This was followed by determination of the structure of the GA phage, which showed some structural differences compared to MS2 and fr phage/VLPs, especially in the N- and C-terminal regions [91]. The structures of virions and recombinant capsids of the Qβ phage were resolved at a resolution of 3.5 Å [92,93]. These structures differed from previously determined RNA phage structures by the presence of stabilizing disulfide bonds on each side of the flexible FG loop, which covalently links CP dimers. A comparison with the structure of the related phage MS2 shows that although the fold of the Qβ CP is very similar, the details of the protein-protein interactions are completely different [93].

The crystal structure of the P. aeruginosa phage PP7 was resolved at a resolution of 3.7 Å [94,95]. As in the case of the MS2 and Qβ CPs, the RNA recognition site on the PP7 CP was determined [96]. Recently, a detailed biophysical study of PP7 virions, including the effects of pH and salt concentration on the charge transition from net-positive to net-negative, was undertaken [97]. The crystal structure of another Pseudomonas phage, PRR1, was resolved at a resolution of 3.5 Å and exhibited a binding site for a calcium ion close to the quasi-3-fold axis [69].

The crystal structure of a very distant RNA phage, the Caulobacter phage φCB5, was resolved at 3.6 Å, and the structure of a φCB5 VLP was resolved at 2.9 Å [70]. The structures appeared to be nearly identical, with some differences in the average density of RNA. Unlike in other phages, φCB5 capsids are significantly stabilized by calcium ions, similarly to some plant viruses [98]. Disassembly of these capsids occurs when the calcium ions are chelated with EDTA and/or there is a reduction in the surrounding salt concentration. Another unique feature of φCB5 is the involvement of RNA bases in the stabilization of its interdimer contacts [70].

Recently, the crystal structure of a phage AP205 dimer was solved at a resolution of 1.7 Å and then fitted into a 6.6-Å resolution cryo-EM map [99]. The structure of the AP205 CP dimer can be regarded as a circular permutant relative to the structures of MS2 and other family members (fig. 3). This feature is made possible by the fact that the N-terminus of one monomer in the dimer is in close proximity to the C-terminus of the other monomer. Compared to MS2 and other phages with known structures, the AP205 CP is missing one beta strand in its N-terminus, but it has an extra beta strand in its C-terminus. However, when considered from a 3D perspective, the position of the beta strand is essentially the same in both cases. As a consequence, AP205 has N- and C-termini in the same locations as those occupied by surface-exposed AB loops in other phage capsids. It is notable that the N- and C-termini in other phages are not well exposed on the surface and are clustered around the quasi-3-fold axes. This explains the previous observations that, in contrast to other phages, the AP205 CP can tolerate long additions at its N- and C-termini without compromising capsid assembly [67].

In addition to the structures of self-assembled RNA phage capsids, the crystal structures of unassembled mutant CP dimers of MS2 [100] and GA [62] have also been resolved. These structures showed only minor differences in comparison to their self-assembled counterparts. Surprisingly, the MS2 CP sustained a genetic fusion, resulting in a duplicated CP that folded normally and functioned as a translational repressor due to its physical proximity to the N- and C-termini of the CP [101].

Furthermore, the crystal structure of an icosahedral MS2 capsid that was assembled from covalently joined dimers, or so-called single-chain dimers, was resolved at 4.7 Å [102]. The structure resembled the wild-type (wt) virion except for the intersubunit linker regions, but a fraction of the capsids was unstable in phosphate buffer because of assembly defects [102]. Moreover, the organization of MS2 single-chain dimers into crystals may have resulted in an arrangement of subunits that corresponds to T = 3 octahedral particles [103]. In this case, the arrangement of dimers is somewhat similar to that in normal T = 3 icosahedral particles, except that four FG loops interact near the 4-fold axis of symmetry on an octahedron rather than five FG loops interacting near the 5-fold axis of symmetry on an icosahedron. However, when MS2 CP dimers are not crystallized in the F cubic crystal form, they are assembled into T = 3 icosahedral capsids that are indistinguishable from the wt particles [103].

X-ray crystallography led to a breakthrough in understanding the protein-RNA interactions that occur during translational repression and genome encapsidation. This breakthrough was particularly apparent after the first crystal structure of a complex of recombinant MS2 capsids with the 19-nucleotide RNA operator was resolved at 2.7 Å [104,105,106]. The residues responsible for the protein-RNA interactions were localized by analysis of aa exchanges at positions 45 and 59 [107,108,109]. Furthermore, numerous other mutations responsible for altering the specificity [110] and efficiency [111] of translational operator complexes were identified. The crystal structures of MS2 VLPs complexed with RNA aptamers, which differ in secondary structure from wt RNA [112,113,114] or involve the presence of 2′-deoxy-2-aminopurine at the critical -10 position [115], have also been resolved.

Structures of CPs in complex with operator RNA fragments have also been solved for PRR1 [116], PP7 [117], and Qβ [118]. The CP dimers formed by the RNA phages studied to date all recognize a stem-loop sequence around the replicase start codon. This CP-RNA interaction serves as a mechanism for translational repression and self-genome recognition during virion assembly. Although the overall binding mode of the stem-loop to CP is similar in all the studied cases, the details are surprisingly different among different viruses. A number of nucleotides form sequence-specific and sequence-unspecific interactions with CPs. For CP recognition, the most important nucleotides are two adenosines, one located in the loop region and another in a stem bulge. In MS2, PP7 and Qβ, these adenosines form quite different interactions with the CP dimer (fig. 4). It should be noted that we have failed to identify analogous CP-RNA interactions in the more distantly related phages AP205 and φCB5 [K. Tars, unpubl. observations], suggesting that mechanisms of genome recognition and translational repression may differ significantly among distant Leviviridae family members.

Electron cryomicroscopy studies have shown that, in addition to the operator, many other RNA sequences in the MS2 genome are able to bind to the CP dimer [119]. Such studies have allowed 3D visualization of icosahedrally averaged genomic MS2 RNA at a resolution of 9 Å [120]. Recently, direct evidence for packaging signal-mediated assembly of the MS2 phage was presented based on cross-linking studies of peptides and oligonucleotides at the interfaces between the capsid proteins and the genomic RNA of this phage [121,122]. Remarkably, the same CP-RNA and maturation protein-RNA interfaces were identified in every viral particle.

Particles of RNA phages contain a single copy of a maturation protein, also known as the A protein, which binds to genomic RNA and is absolutely necessary for the infectivity of phages by attachment to bacterial pili. Therefore, the A protein must be exposed at both the inner and outer surfaces of the CP shell. Because the capsids of ssRNA phages contain holes at the 5-fold and 3-fold axes of symmetry that are large enough for the diffusion of RNA fragments, it was long believed that the A protein binds in proximity to these axes. However, it was recently shown in cryo-EM studies that the A protein actually replaces a single CP dimer [123,124]; as such, the virion actually contains 178 CP monomers.

Comparative investigations of MS2 VLPs and virions containing a copy of the A protein by SANS (small-angle neutron scattering) revealed some differences, particularly the presence of ‘thin' (preinfection) and ‘thick' (postinfection) capsids, which are not seen by crystallography [125]. The role of the A protein during virus assembly may involve the accumulation of tension that is later used to eject the genomic RNA and the A protein into a host cell.

Species-specific virus-neutralizing antibody responses to RNA phages were studied in the late 1960s [126,127,128,129]. The MS2 phage was characterized as a T cell-dependent antigen that may also function as a T cell-independent antigen when used in high doses [130], with high efficacy in inducing T cell response [131]. However, no B cell epitopes have been mapped for virus-neutralizing antibodies on RNA phage capsids. General observations have shown the conformational nature of the B cell epitopes [our unpubl. data]. It was found that only a synthetic MS2 CP peptide spanning from positions 89 to 108 is capable of inhibiting phage neutralization by specific antiserum and of inducing the production of virus-neutralizing antibodies in rabbits [132] and guinea pigs [133]. Chemical conjugation of an N-acetylmuramyl-L-alanyl-D-isoglutamine (MDP) adjuvant to the 89-108 peptide also markedly increased the induction of virus-neutralizing antibodies in rabbits [134]. Although proper CP epitopes have not been mapped, the epitope-recognizing complementarity-determining region (CDR) peptides from a virus-neutralizing monoclonal antibody specific for phages fr, MS2, and GA were described and documented for their capability to neutralize the abovementioned phages [135]. Recently, CDRs from thermostable single-domain anti-MS2 antibodies obtained by the immunization of llamas were sequenced [136].

Recombinant capsids of fr [137,138] and MS2 [139] were the first VLP carriers proposed for the presentation of foreign immunological epitopes on their surfaces by genetic fusion. To identify appropriate CP regions, oligonucleotide linkers encoding short aa sequences and containing convenient restriction sites were inserted into different regions of the fr CP [140]. Remarkably, this work was based on computer predictions of the spatial structure of the fr CP and was conducted before the first crystal structure of an RNA phage, namely that of the homologous MS2 capsid, had been resolved [83,84,85,86]. Recombinant fr CPs containing 2- to 12-aa-long additions to their N- or C-termini or insertions at position 50 in the RNA-binding region were capable of self-assembly, but this was not the case at positions 97-111 in the αA-helix [140]. The majority of other fr CP mutants demonstrated reduced self-assembly capabilities and formed either CP dimers (aa exchanges at residues 2, 10, 63, or 129) or both dimer and capsid structures (residue 2 or 69) [141]. The FG loop of the fr CP was also recognized as a potential target for insertions/replacements and was initially modified by a 4-aa-long deletion [142]. The deletion variant retained the ability to form capsids, although they displayed significantly reduced thermal stability. Furthermore, the 3D structures of the mutant capsids revealed that the modified loops were disordered near the 5-fold axis of symmetry and were too short to interact with each other [142]. Because of the high importance of the FG loop in capsid stability, further development of chimeras based on the implementation of FG loops was not pursued.

Vectors were constructed for the insertion of foreign sequences into codon position 2 of the fr CP-coding sequence in all possible reading frames [137]. When the preS1 epitope DPAFR from hepatitis B virus (HBV) was inserted as a marker at positions 2, 10, and 129, it appeared in all cases on the particle surface [137]. Attempts to introduce the 40-aa-long V3 loop from the human immunodeficiency virus 1 (HIV-1) gp120 protein into the N-terminus, at positions 10, 12, and 15, or into the FG loop led to unassembled products [K. Tars, unpubl. data].

Attempts to use the N-terminal β-hairpin exposed at the surface of the capsid, namely aa residues 15/16, allowed for the production of MS2 capsids bearing a number of different peptide sequences up to 24 aa residues in length [139]. Foreign epitopes exposed on these chimeras were found to be immunogenic in mice [139]. Mutational mapping revealed residues that are responsible for inter- and intramolecular contacts and therefore for the thermal stability of MS2 capsids [143].

The addition of the Flag octapeptide to the N-terminus or its insertion into the N-terminal β-hairpin of the MS2 CP prevented self-assembly and proper folding, respectively. However, genetic fusion of the Flag to a duplicated CP-encoding sequence resulted in the synthesis of a protein considerably more tolerant to the structural perturbations and mostly corrected the defects accompanying Flag peptide insertion [144]. The putative protective epitope T1, spanning 24 aa in length and derived from the immunodominant liver stage antigen-1 of the malaria parasite Plasmodium falciparum, was inserted at the tip of the N-terminal β-hairpin (between positions 15 and 16) of the MS2 CP [145].

fr VLPs have shown unusually high capacities as vectors, particularly when the addition of long segments of hamster polyomavirus VP1 to the N-terminus of the fr CP did not prevent VLP self-assembly [146]. When discovered, these findings markedly enhanced interest in using RNA phage coats as potent vaccine candidates.

However, the relatively low tolerance of fr- and MS2-based VLP vectors to long foreign insertions represents a clear limitation of these models. For example, the fr CP failed to form VLPs following the insertion of long HBV preS1 sequences [147]. To overcome this difficulty, Qβ protein A1, namely the C-terminal extension of the Qβ CP within the A1 protein, has been proposed as a site for foreign insertions [148,149]. Potentially, the 195-aa extension of the Qβ CP could be considered an ideal target for insertions. This region was found to contain elements that typically protrude as spike-like structures on the particle surface. Furthermore, the self-assembly capabilities of capsids with mutually exchanged extensions of the Qβ and SP CPs were confirmed experimentally [64]. A following mathematical prediction showed possible colinearity between the Qβ CP extension and the surface-located HBV preS sequence [148]. It is noteworthy that the crystal structure of the read-through domain from the Qβ A1 protein was recently determined at a resolution of 1.8 Å and revealed a fold that is unique among all proteins found in the protein data bank [150].

Qβ CP proteins with additions at their C-termini failed to form particles in most cases. However, such additions were incorporated into particles in the presence of a wt ‘helper' CP to form mixed, or mosaic, particles. Mosaic VLPs differ therefore from chimeric VLPs which are built from identical recombinant CP molecules, without any helper molecules. Such mosaic particles were constructed by either enhancing the level of UGA suppression in the presence of overexpressed suppressor tRNA [151] or by exchanging the UGA stop codon to a GGA sense codon and expressing the extended and helper forms of the Qβ CP from two separate genes. These genes were located either on the same plasmid or on two separate plasmids with different antibiotic resistance genes. Potential insertion sites were mapped by insertion of the 5-aa preS1 DPAFR model epitope and the 39-aa-long HIV-1 gp120 V3 loop [149]. After enhancing UGA suppression, mosaic particles were detected, but the proportion of A1 to helper CP in these particles dropped from 48 to 14% as the length of the A1 extension increased [152]. A model insertion of the preS1 epitope DPAFR located on the particle surface produced specific antigenicity and immunogenicity in mice [152]. The antibody response to the preS1 epitope was higher for self-assembled Qβ-preS1 VLPs than for a nonassembled Qβ-preS1 variant [153]. When the Qβ CP was modified to carry long HBV preS insertions (full-length preS, preS1, or preS2 alone) instead of the A1 extension, mosaic particles formed that had surface-exposed preS, but regular VLPs did not form without the presence of the Qβ CP as a helper [I. Cielens and R. Renhofa, unpubl. data].

On Qβ CP itself, the residues responsible for RNA recognition have been mapped [154,155]. This was helpful for the development of packaging and gene transfer technologies based on the Qβ phage model. The ability of the RNA phage CP to package RNA in vivo [156] demonstrated the potential of RNA phages as gene delivery vectors (see Nonvaccine Applications).

Surprisingly, icosahedral Qβ VLPs have been converted into rods after modification of the FG loop structure [157]. As mentioned above, the appearance of alternate VLP forms of RNA phages was further confirmed by the presence of rod-like structures in the case of the coassembly of phage fr and GA CPs [73].

The MS2 phage was found to be capable of accommodating short (pentapeptide) sequences added to the N-terminus of its CP within viable virions, although not all insertions were genetically stable [158]. Following well-known peptide display systems based on filamentous phages, a novel peptide display platform on MS2 VLPs was developed and successfully tested, including the identification of conformational epitopes [159,160].

As an experimental example of fusion technology, figure 5 compares RNA phage virions, recombinant VLPs, and chimeric VLPs carrying long foreign insertions by electron microscopy. It is remarkable that the surfaces of some insertion-containing VLPs differ considerably from the surfaces of unmodified VLPs by the appearance of distinct knobs that are presumably formed by the inserted sequences. Besides VLP applications, the phage fr CP was successfully used in its nonassembled variant for the fine mapping of epitopes, primarily of HBV-derived proteins[161,162,163,164,165,166].

Chemical coupling of foreign oligopeptides to the surface of VLPs was developed as an alternative method to the genetic fusion of epitope-encoding sequences. Chemical coupling was initially applied to RNA phage Qβ VLPs in 2002 [167] by using an approach initially developed for another broadly used recombinant VLP, HBc (HBV core antigen) [167,168]. Model oligopeptides containing a free cysteine residue at the N-terminus were coupled to an exposed lysine residue on the Qβ VLP surface using the hetero-bifunctional cross-linker maleimidobenzoic acid sulfosuccinimidyl ester. The modified VLPs showed efficient induction of oligopeptide-specific antibodies in mice [167]. This chemical coupling approach initiated the development of a panel of experimental therapeutic vaccines (see early reviews by Bachmann and Dyer [169] and Dyer et al. [170], and a recent review by Bachmann and Jennings [171]). By following the lysine-cysteine oligopeptide coupling methodology, an experimental West Nile virus vaccine was constructed on an AP205 VLP platform [68].

Another chemical coupling approach was developed based on the rational design of modified MS2 VLPs that displayed a reactive thiol on the VLP surface as a result of a T15C substitution in the MS2 CP [172,173]. Cysteines are among the most useful functional groups found in proteins as they can bind a variety of metals and react with a large collection of organic reagents, and are therefore obvious targets for protein modification [172]. Two cysteine residues that are present in the wt MS2 CP are internally located and therefore relatively unreactive. Thiolated MS2 VLPs were chemically modified with fluorescein-5′-maleimide to create the first fluorescent nanoparticles [173].

Additional functionalization was achieved in a cell-free protein synthesis platform by the production of MS2 and Qβ VLPs with surface-exposed methionine analogues (azidohomoalanine and homopropargylglycine) containing azide and alkyne side chains [174]. Such VLPs can be used for one-step, direct conjugation schemes to display multiple ligands of interest. Using such technology, proteins including an antibody fragment and granulocyte-macrophage colony-stimulating factor, as well as nucleic acids and poly(ethylene glycol) chains, were displayed on the VLP surface using Cu(I) catalyzed click chemistry [174]. Surface functionalization methodology has not only been applied to the VLP surface but also to the interiors of MS2 VLPs by modification of tyrosine residues via a recently developed hetero-Diels-Alder bioconjugation reaction [175].

The preparation of histidine-tagged MS2 VLPs by the introduction of a His₆ linker between CP codons 15 and 16 to simplify the purification of VLP-covered RNAs is another example of VLP surface functionalization [176]. Moreover, a set of advanced His₆-tagged Qβ VLPs was generated, and their ability to complex metal-derivatized compounds was confirmed [177,178]. In parallel, VLP vector capacity for chemical coupling was broadened by the introduction of azide- or alkyne-containing unnatural amino acids, which was achieved by expression of the Qβ CP gene in a methionine auxotrophic strain of E. coli [179].

A highly specific approach for the further development of VLP vectors was achieved by the asymmetrization of Qβ VLPs after the introduction of a single copy of the maturation protein A2, which allowed the production of VLPs with a single unique modification [180].

Recently, a novel plug-and-display system was established for modular RNA phage VLP functionalization via further decoration of VLPs with peptides of interest [181,182]. The decoration of VLPs is based on the so-called bacterial superglue approach, namely on the ability of a peptide (SpyTag) and a protein (SpyCatcher) to form spontaneous covalent isopeptide bonds between lysine and aspartic acid residues under physiological conditions [183]. The SpyTag and SpyCatcher are split units of the Streptococcus pyogenes fibronectin-binding protein FbaB and can form a highly stable amide bond by an irreversible reaction that occurs within minutes [184]. The SpyTag or SpyCatcher sequences were genetically fused to the N- and/or C-terminus of the AP205 CP. After mixing modified AP205 VLPs with the correspondingly linked peptides, the quantitative covalent coupling of the peptides to the VLPs was observed [181,182].

Efforts to improve the stability of natural RNA phage VLPs started with the development of a methodology that allowed the screening of bacteria for the synthesis of mutant MS2 CPs with altered assembly properties [185] and the selection of D11N variant CPs that formed virions more stable than the wt CP [186]. The introduction of interdisulfide bonds into the 5-fold axis of symmetry to cross-link MS2 VLPs improved their thermal stability to the level of that seen in Qβ VLPs, which possess natural intersubunit disulfide bonds [187]. In contrast, cross-linking at the 3-fold axis of symmetry resulted in variant CPs that were unable to self-assemble [187]. The development of an E. coli-based cell-free protein synthesis system opened a direct avenue for studying the role of disulfide bond formation in the stability of mutant MS2 VLPs in comparison to Qβ and HBc VLPs [188]. Through the construction of a set of Qβ CP mutants, it was found that disulfide linkages are the most important stabilizing elements in VLPs and that interdimer interactions are less important than intradimer interactions for Qβ VLP assembly [189].

Elucidation of the thermal stability of foreign epitope-carrying VLPs, such as the chimeric MS2 VLPs formed by single-chain dimers, in comparison to natural disulfide cross-linked PP7 VLPs [190] is of great importance for the further development of RNA phage-based nanotechnology. The genetic fusions of two copies of the MS2 [159], PP7 [191], and GA or Qβ [I. Cielens and A. Strods, unpubl. data] CPs resulted in a self-assembly-competent single-chain dimer that not only increased thermodynamic stability but also considerably improved tolerance to foreign insertions in the AB-loop. The resultant correctly assembled VLPs encapsidated mostly their ‘own' CP-encoding mRNAs.

Vaccines represent the most advanced field of the RNA phage VLP applications due to the excellent and well-established scaffold properties and structural tolerance to the decoration by foreign immunogenic sequences. Such decoration can be performed both genetically and chemically, and the VLP scaffold may provide foreign epitopes with a strong T cell response. Moreover, RNA phage VLPs can serve as nanocontainers that can encapsulate specific adjuvants, such as immunostimulatory oligodeoxynucleotides or CpGs, as TLR9 ligands [192]. RNA phage VLPs can also be packaged with single-stranded or double-stranded RNA fragments as TLR7 and TLR3 ligands, respectively [193]. Moreover, recombinant RNA phage VLPs contain encapsulated bacterial RNA, which may act as an adjuvant. Reviews on vaccine applications include articles discussing RNA phage VLPs among other VLP candidates [82,169,170,171,194,195,196,197,198,199,200,201,202] and articles focused either on RNA phage VLPs in general [203] or on the use of specific RNA phage species, such as MS2 VLPs, as vaccines [204].

Table 1 contains a detailed list of the vaccine candidates that have been constructed from RNA phage VLPs using genetic fusion methodology. First, the success of an experimental human papilloma virus (HPV) vaccine based on PP7 single-chain-dimer VLPs [191,205,206,207,208,209] developed in preclinical studies [210] must be mentioned. A similar HPV vaccine candidate was constructed on MS2 single-chain-dimer VLPs and tested in preclinical studies [210,211]. Both PP7 and MS2 VLP-based vaccines were immunogenic, but the MS2-L2 VLPs induced a broader HPV-neutralizing antibody response. This is likely because of the structural context of L2 display on the VLPs, since L2 was displayed on the AB-loop of the PP7 CP, but at the N-terminus of the MS2 CP [191]. A review on HPV vaccine candidates, including RNA phage VLP-based vaccines, was recently published [212]. A malaria vaccine based on MS2 VLPs has also been reported as very promising [213,214]. As to the selection of optimal VLP carriers for such purposes, AP205 VLPs have demonstrated a high capacity and tolerance to foreign insertions [67]. Moreover, the AP205 VLPs represent good candidates for the construction of mosaic VLPs [215].

The chemical coupling approach was validated by an impressive line of experimental therapeutic vaccines [169,171]. The idea of therapeutic vaccines is based on the assumption that VLP carriers can present surface-displayed self-antigens and to augment their ability to overcome the natural tolerance of the immune system toward self-proteins and to induce high levels of specific autoantibodies [216]. Initially, this approach was planned as a method to replace host-specific monoclonal antibodies in the treatment of acute and chronic diseases, starting with noninfectious diseases [169]. Therefore, the transition from passive administration of monoclonal antibodies to active vaccination against self-antigens was a logical step in drug development, focusing on affordable medicines and broader patient acceptance and regulatory compliance. The induction of autoantibodies might be beneficial under certain physiological conditions in order to remove unwanted excess of a particular self-antigen, such as angiotensin in the case of hypertension. Table 2 presents a detailed list of predominantly therapeutic vaccines and demonstrates the validity of this assumption.

We included the well-known Qβ VLP-based allergy vaccine CYT003-QbG10 that contains an encapsulated CpG sequence, so-called QbG10, in table 2, although it does not carry any attached epitope [217,218,219,220,221]. However, this potentially successful vaccine initially originated from epitope-coupling methodology. Moreover, we included RNA phage VLPs carrying model epitope p33, which is derived from the lymphocytic choriomeningitis virus glycoprotein [222,223,224,225,226,227,228,229], and epitope D2, which is derived from Salmonella [230]. Although these are not actual vaccine candidates, the model epitopes have played a central role in the elucidation of the fine immunological mechanisms that govern responses to chimeric VLPs.

The display of small antigens, such as nicotine [231,232,233,234] or carbohydrate moieties [235,236], on RNA phage VLP surfaces is a novel approach that enabled the generation of strong immunological responses against nonpeptide antigens and paved the way for the development of experimental vaccines against nicotine addiction and cancer, respectively. We recommend a recent review on novel vaccines constructed on RNA phage VLPs [237] as well as special reviews on vaccines against allergies [238,239,240], Alzheimer disease [241], hypertension [242,243,244], influenza [245,246], malaria [247], and nicotine addiction [248,249,250,251,252].

Attention is currently focused on the use of virus-based nanoparticles as potential scaffolds for novel biomaterials and as subjects for nanoscale engineering applications involving exposure to various chemical compounds [253,254,255,256].

Table 3 summarizes RNA phage VLP-based experimental approaches that could be classified as VLP packaging and targeting methodologies. Historically, the idea of encapsulation/targeting by RNA phage VLPs appeared in the early 1990s, only a few years after the vaccine/epitope display approaches described above. A first attempt at RNA phage nanocontainer-targeted drug delivery involved, first of all, encapsulation of a deglycosylated ricin A chain coupled to an RNA operator stem-loop and decoration of the MS2 VLPs by transferrin [257,258,259]. Such structures, called ‘synthetic virions' by the authors, demonstrated high toxicity to leukemia cells carrying transferrin receptor. A detailed review of these and similar pioneering experiments was published in 2002 [260]. Another detailed review devoted to anticancer perspectives of nanoparticles filled with siRNA was recently published [261].

The next step in the development of VLP packaging/targeting technology included the chemical coupling of putative cargo and targeting molecules to the outer and inner surfaces of VLPs, respectively [175,262,263,264,265,266,267]. It has been demonstrated that MS2 VLPs can be conjugated to peptides recognizing human hepatocellular carcinoma cells and can be loaded with vastly different types of cargo, including low molecular weight chemotherapeutic drugs, siRNA cocktails, protein toxins and nanoparticles, resulting in the selective killing of target cells [268]. The packaging of Qβ VLPs with immunostimulatory CpG sequences led not only to the development of potential allergy vaccines (see above) but also strongly contributed to the general understanding of the mechanics of oligodeoxynucleotide-induced stimulation [217,218,219,220,221,222,223,224,225,226,227,228,229,269,270].

The use of imaging agents in combination with RNA phage VLPs has contributed to the high-resolution and noninvasive visualization of these particles, as well as to the potential treatment of diseases [254]. The RNA phage VLP-based applications that have been developed as imaging technologies are compiled in table 4. The first studies on the generation of nanoparticles for magnetic resonance imaging applications and the first comparisons of interior versus exterior cargo strategies appeared in the mid 2000s [271,272]. Following these reports, MS2 VLPs were loaded with positron emission tomography radiolabels [273]. To date, MS2 phages and VLPs have played a leading role in bioimaging studies.

The ‘armored' nucleic acids are useful as noninfectious, easily available reagents for the quality control in the diagnosis of pathogenic viruses like HCV (hepatitis C virus). As an example of ‘armored RNA' technology, MS2 VLPs were made to perform in vivo encapsidation of a desired RNA, which was accomplished by including the MS2 operator sequence on the RNA molecule to enable its packaging. The armored RNA was therefore protected from RNase digestion [274]. Recently, the creation of armored dsDNA using HBV and HPV particles was accomplished on the basis of previous work using MS2 VLPs [275]. Armored RNAs have multiple uses, such as for the detection of HCV, severe acute respiratory syndrome coronavirus, and influenza virus RNA [276] or for the creation of unique RNA molecules harboring both tRNA and mRNA functions [277]. Perspectives on the production and practical use of MS2 VLPs for routine diagnostics including food quality control were discussed in a recent review [278].

The next milestones for the development of the RNA phage VLP field are presented in figure 6. The major tendency provides the combination of both (i) decoration of the VLP scaffolds with molecules of interest and (ii) packaging of foreign material into VLP nanocontainers.

Vaccineswill remain a classic area for the application of RNA phage VLPs [279]. For example, clear progress in vaccine formulation and storage was achieved through a MS2 VLP-based papillomavirus vaccine [280]. RNA phage CP properties can be used in combination with other possible vaccine carriers, such as retroviral platforms, with a special goal to improve RNA packaging and delivery [281].

MS2 VLPs [160] and Qβ phage [32] offer favorable alternatives to filamentous bacteriophages for the display of immunologically active peptides. Recently, a tumor-associated antigen was identified using MS2 peptide-display technology [282], confirming the importance of such a methodology for future studies.

RNA phage VLPs are regarded as potential components of future nanomachines, which is a general term for a machine ranging in size from 1 nm to 1 μm. Perspectives on the revolutionary nanoscale engineering of RNA phage VLPs as natural prefabricated scaffolds to contain molecules in precisely defined arrays were described in a recent review [253]. The future of VLP-derived materials [for chemical strategies and details of VLP bioconjugation technologies, see [283]] is quite impressive because the number of methods that can be used to change both the interior and exterior surfaces of capsids by the incorporation of organic and inorganic compounds is unlimited. Further development of VLPs with defined antigenic and immunogenic properties, as well as VLPs with improved packaging and targeting capabilities, will create novel viral nanotechnology applications and is expected to produce nanomachines with rationally designed characteristics. For example, an interesting new approach is exemplified by the use of DNA as a scaffold to arrange MS2 VLPs into one-dimensional arrays with precise nanoscale positioning [284]. Such bioinspired plasmonic nanostructures may provide a flexible design base for manipulating photonic excitation and photoemission [285].

The clear differences that exist in RNA phage VLP structures, stabilities, and reconstruction capacities have provoked extensive studies aimed at identifying novel members of the Leviviridae family. Out of such efforts, 19 new genomes of icosahedral RNA phages [286], genomes from the phages M [287], C-1 and HgalI [288], DL52 and DL54 [289], from a so-called JS-like group [290] and of the phages EC and MB [291], have been described.

A recent survey of metagenomic databases revealed 158 partial single-stranded RNA phage genome sequences belonging to about 120 distinct phylotypes [292]. Sixty-six of the genomes contained a putative open reading frame predicted to be the CP [292]. Novel RNA phage sequences were present in samples collected from a range of ecological niches worldwide, including invertebrates and extreme microbial sediment, demonstrating that they are more widely distributed than previously recognized [292]. These genomes are expected to undergo studies of CP gene expression as well as elucidation of their 3D capsid structures, and will potentially serve in new applications.

3D structural investigations have revealed unexpected capabilities of RNA phage capsids, such as their ability to transform into T = 1 [77] or rod-like [157] structures. Moreover, such studies have contributed worthwhile information towards understanding protein folding and virus assembly [293]. Additional important directions for future studies involve the elucidation of VLP surface properties in the context of the presence of internal RNA, as reported previously [294], and/or the external display of foreign sequences of different origins, such as oligonucleotides, peptides, sugars, and metal ions. Such studies would be helpful for the prediction of VLP characteristics in the context of aqueous and nonaqueous media, as well as under different biological conditions, which should facilitate the use of these particles as vaccines and/or gene therapy tools.

We thank Dr. Maija Bundule, Dr. Indulis Cielens, Dr. Dzidra Dreilina, Dr. Juris Jansons, Dr. Andris Kazaks, Dr. Janis Klovins, Dr. Janis Rumnieks, Dr. Dace Skrastina, Dr. Irina Sominskaya, Dr. Arnis Strods, Dr. Alexander Tsimanis, Dr. Inta Vasiljeva, Dr. Tatjana Voronkova, and Ms. Inara Akopjana (Riga) for their contributions and sharing of unpublished data. We are grateful to Prof. Dr. Rüdiger Schmitt (Regensburg) for the first N-terminal aa sequence of the AP205 CP and to Prof. Dr. Wolfram H. Gerlich (Giessen) for his critical reading of the manuscript and helpful comments.