Chapter 1: Production and Purification of Viruses for Structural Studies

Advances in protein production and purification techniques over the past two decades have allowed the structural study of numerous proteins and macromolecular assemblages that would have otherwise been intractable to the necessary approaches (detailed in the following chapters). This chapter focuses on the production and purification of intact viral capsids (particles) with/without genome for structure determination. The production and purification of viral proteins for structure determination by X-ray crystallography and NMR spectroscopy are the subjects of Chapters 7 and 8, respectively. Crystallization is often considered a method of purification and a function of purity, often of a protein or virus capsid, and, as such, sample preparation for structure determination by X-ray crystallography places high demands on sample quality. Screening trials to identify the optimal crystallization conditions also require large quantities of sample compared with the majority of other structure determination approaches discussed in the subsequent chapters of this monograph. Virus samples produced for such analyses also have to be both stable and soluble in their storage buffer since degradation and aggregation are detrimental to the crystallization process. Hence this chapter will focus on methodologies to produce and purify virus capsids (Figure 1) in quantities suitable for structure determination by X-ray crystallography, with the premise that such a sample would also be suitable for structural or biophysical analysis using other methodologies.

Most viruses are considered hazardous material in their wild-type (wt) infectious form (for information on safe handling and containment of infectious microorganisms and hazardous biological materials, see http://www.cdc.gov/biosafety) and are therefore often studied in a recombinant form. Significant effort has been extended into the development of heterologous expression systems to produce recombinant viral proteins which will assemble into viral capsids. The system selected for use is often dependent on the properties of the viral genes and the environmental requirements of the final product. However, the most important factor to consider is the capacity of the host cells to translate the RNA transcript, to ensure proper folding of the gene product and to sustain the protein(s) expressed in an intact and functional state.¹  Protein expression systems contain at least four general components: (1) the genetic elements necessary for transcription/translation and selection; (2) in vector-based systems, a suitable replicon: plasmid, virus genes, etc.; (3) a host strain containing the appropriate genetic traits needed to function with the specific expression signals and selection scheme; and (4) the culturing conditions for the transformed cells or organisms.² 

Since most viruses currently studied are of human or animal origin, mammalian tissue culture is an ideal source to generate viral capsids for structural studies which are generally aimed at functional annotation. In this system, proper folding is achieved and modifications such as complex glycosylation, phosphorylation, acylation, acetylation and γ-carboxylation are obtained. However, yields can be low, depending on gene product(s), ranging from 0.1 to 100 mg L⁻¹ of culture volume. For some of the structural approaches discussed in Section 1 of this monograph, low yields may not be a problem since small amounts of sample are adequate. However, low yields can become problematic in crystallization, especially with a virus that does not have an established crystallization condition. In such a situation, numerous preparation steps may be required to obtain the quantities needed to screen crystallization conditions efficiently. Supplies and reagents can then become expensive, depending on individual cell line requirements. In addition, considerable time and resources can be spent on the construction of a suitable expression system and equally on optimization for suitable yields. In such situations, it is always advisable to seek the expertise of an established molecular biologist before designing new constructs.

Established cell lines and protocols exists for many different tissue systems and, although most of these cell lines are derived from human or mouse tissues, other mammalian cell culture lines are available, such as monkey, raccoon, horse, pig and rabbit. The American Type Culture Collection (ATCC) has over 3400 cell lines from 80 different species, including over 950 cancer cell lines (http://www.atcc.org/). Other cell suppliers include the Health Protection Agency Culture Collections (HPACC; http://www.hpacultures.org.uk/), the German Research Center for Biological Material (DSMZ; http://www.dsmz.de/) and the Riken BioResource Center Cell Bank (Riken; http://www.brc.riken.jp). It is strongly recommended that investigators purchase cell lines from recognized centers such as these listed above to ensure pure, authentic and quality controlled cell lines. The decision to use cells directly from an organism, i.e. primary cells or an immortalized cell line, should be based upon requirements of the virus system and available current protocols. As discussed below, there are three main approaches for virus production in mammalian cell lines: (i) infection of permissive cell lines with wt virus, (ii) transfection of cells with plasmid constructs containing viral genome sequences and (iii) viral vector systems which expression heterologous viral genes.

Although the majority of viruses currently studied are obtained from recombinant expression systems (see below), direct infection of cell lines with wt virus can be used to generate suitable quantities of sample for structural studies under certain conditions and for well-characterized viral systems. For example, the human rhinovirus 3 (HRV3) virion particles used for determining its structure were purified from virus-infected HeLa cells (immortalized human cancer cells). The atomic structure of HRV3 was initially determined to 3 Å,³  and later refined to 2.15 Å.⁴  It was reported that 10–12 L of HeLa cells (at 6–8×10⁵ cells mL⁻¹) were used to generate the amount of virus necessary to carry out crystallization and structure determination. Echovirus-1, also of the Picornaviridae family, was also successfully produced in HeLa cells for its structure determination to ∼3.55 Å resolution.⁵ 

In the use of plasmid constructs, one or more plasmids usually containing capsid proteins alone and, if needed, replication factors, are used to transfect cells, which results in the assembly of virus-like particles (VLPs). Often, another plasmid is added when a packaged gene is desired, e.g. reporter gene, or if genome is needed to produce stable virions. Recovered virus can either be purified for structural studies or, if infectious, used to infect permissive cells for continual propagation of virions. As an example, molecular clones containing the capsid sequence of canine parvovirus was used for the transfection of Norden Laboratories feline kidney cells (NLFK)⁶  to produce particles for X-ray crystallographic structural studies to 3.2 Å resolution.⁷  For the crystallographic structure determination of the immunosuppressive strain of minute virus of mice (MVMi), infectious virions were harvested from plasmid transfected cell lines and subsequently propagated in a permissive cell line to produce virus for crystallization.⁸ 

The development of heterologous surrogate expression systems for virus capsid production has enabled researchers to overcome the lack of efficient expression in homologous systems for several viruses of interest. As an example, for hepatitis C virus (HCV), a herpes simplex virus-1 (HSV-1)-based amplicon vector system that expresses HCV capsid proteins and the two envelope proteins, E1 and E2, under the HSV-1 IE4 promoter was developed.⁹  This system has several advantages; (i) the ability to infect a wide range of cells, without the limitation of transfection efficiency, including primary cells in a quiescent state, (ii) the simplicity of cloning desired genes into amplicons, (iii) the high capacity of incorporation of exogenous sequences in the vector genome and the transfer of high copy numbers of the exogenous gene and (iv) the potential for using amplicons in vaccine design and development.¹⁰  A mini-review has covered HSV amplicons from genomes to engineering.¹¹  Norovirus is another example of a non-cultivable virus that remained refractory to structural studies due to the lack of a reverse genetics system and a permissive cell line until recent advances. A novel expression strategy, which combined the use of a two baculovirus transactivation system to deliver viral cDNA and an inducible DNA polymerase (pol) II promoter, led to the ability to grow this virus in several cell lines, including HepG2, BHK-21, COS-7 and HEK293T cells.^12,13 

Among the microbial eukaryotic host systems, yeasts can combine the advantages of unicellular organisms (e.g. ease of genetic manipulation and growth) with the capabilities of a protein processing typical of eukaryotic organisms (e.g., protein folding, assembly and posttranslational modifications).¹⁴  The majority of recombinant proteins produced in yeast have been expressed using Saccharomyces cerevisiae. More commonly referred to as baker's or budding yeast, S. cerevisiae was the first eukaryote to have its entire genome sequenced¹⁵  and is still today considered a model organism. A scientific database has been established for S. cerevisiae and is available at http://www.yeastgenome.org/. With its biochemistry, basic genetics and cellular biology already well established, this simple eukaryote has become a major tool in answering questions of fundamental biological importance and is a central player in post-genomics research.

Appealing aspects of the yeast expression system are its rapid cell growth (with a doubling time of ∼90 min), simple growth media, secretion of recombinant proteins to the medium and glycosylation capability. N-linked glycosylation is minimal with high mannose, but O-linked modifications appear similar to mammalian cells. Phosphorylation, acetylation and acylation are also present. Protein yields are comparable with the baculovirus system (see below) at ∼10–200 mg L⁻¹ depending on recombinant gene properties. Issues in large-scale protein production involving S. cerevisiae appear to be hyperglycosylation and retention in the periplasmic space.^16,17  This ultimately leads to a loss of final protein due to retention and degradation. The search for alternative hosts has led to the use of ‘non-conventional’ yeasts in expression protocols. The most established examples include Hansenula polymorpha, Pichia pastoris, Kluyveromyces lactis, Yarrowia lipolytica, Pichia methanolica, Pichia stipitis, Zygosaccharomyces rouxii, Zygosaccharomyces bailaii, Candida boidinii and Schwanniomyces (Debaryomyces) occidentalis.¹⁴  These systems are broken down even further into two categories: methyltrophic, e.g. P. pastoris, and non-methyltrophic, e.g. S. cerevisiae. These categories are based on the fermentation processes involved and generally dictate the promoter that should be used in the experimental design. The choice of yeast host is one of the most important determinants of the success of the entire project, and many reviews debating the subject can be found in the current literature. Generally, the expression of foreign proteins in yeasts consists of (i) cloning of a foreign protein-coding DNA sequence within an expression cassette containing a yeast promoter and transcriptional termination sequences and (ii) transformation and stable maintenance of this DNA in the fusion host.¹⁴  The transformation process is highly dependent on the yeast strain and detailed studies should be conducted in order to achieve high-efficiency transformation.

This system is extensively used for studying biological processes in higher eukaryotes and also allows replication of eukaryotic viruses. The first eukaryotic virus for which replication and genome encapsidatation was conducted in S. cerevisiae was brome mosaic virus (BMV), a positive strand RNA [(+)RNA] virus that infects plants.^18,19  The BMV VLPs were subsequently purified for structure-to-function studies using cryo-electron microscopy (cryo-EM) studies.¹⁹  Other (+)RNA viruses that have been successfully replicated in S. cerevisiae include the plant viruses tomato bushy stunt virus and carnation Italian ringspot virus and animal viruses Flock House virus (FHV) and Nodamura virus.²⁰  Human papillomavirus-16 (HPV-16) VLPs have also been successfully expressed in the yeast system²¹  in addition to the bovine papillomavirus-1 (BPV-1).^22,23  The yeast virus L-A was isolated and purified from S. cerevisiae and the structure was solved to 3.4 Å resolution.²⁴ 

Originally isolated from the alfalfa looper (Autographa californica) insect, Autographa californica multiple nucleopolyhedrovirus (AcMNPV) is the most widely used and best characterized baculovirus for recombinant gene expression (a recent review on baculovirus molecular biology is available²⁵ ). The rather large genome (∼134 kbp²⁶ ) can stably accommodate an insertion of ∼38kb,²⁷  making expression of large genes possible. This virus is also known to infect several other insect species including Spodoptera frugiperda. The most commonly used insect host cell lines, Sf9 and Sf21AE, are derived from S. frugiperda pupal ovarian tissue²⁸  and the BTI-Tn-5B1-4 line, also known as ‘High 5 cells’, derived from Trichoplusia ni egg cell homogenates.²⁹  The wt nucleopolyhedrovirus (NPV) produces small inclusion bodies composed of a polyhedron protein which allows for the encapsulation of many virions into a crystalline protein matrix. This protein is expressed in the very late phase of gene expression and is controlled by a very strong promoter, the polydron promoter (a review on baculovirus late expression factors is available³⁰ ). The baculovirus expression vector system (BEVS)^31,32  takes advantage of this very strong polyhedron promoter to drive foreign protein expression. It has also been shown that the non-structural p10 protein is expressed at similar levels in the same very late phase of expression. Both proteins have been shown to be non-essential in the production of baculovirus particles,^33,34  making the replacement of their open reading frame (ORF) ideal for use in foreign gene expression.

The coupling of the very strong polyhedron promoter with a foreign gene-coding region results is the production of high levels of recombinant protein (∼5–200 mg L⁻¹) in a relatively short amount of time using the BEVS. Since the baculovirus genome is generally considered too large to insert the foreign gene of choice by direct ligation, transfer vectors are used. There are many different vectors available for gene insertion, which are variants of a basic design (a review appeared recently³⁵ ). These offer single gene, multiple genes and fusion gene expression. Multiple copies of the promoter can also be engineered into BEVS for the expression of multiple recombinant proteins concurrently in infected cells,^36,37  which permits the assembly of structures that are made up of heterologous proteins, such as viruses.

Advances in experimental design such as a wide variety of transfer vectors, simplified recombinant virus isolation and quantification methods, advances in cell culture technology and commercial availability of reagents have led to the increased use of BEVS for recombinant viral capsid protein production. Belyaev and Roy³⁷  were able to construct a multiple gene transfer vector which co-expressed the four major structural proteins of bluetongue virus (BTV) to produce VLPs. These samples permitted structure-to-function correlations for BTV and advanced BTV research in efforts to characterize its assembly properties and in vaccine development.^38,39  VLPs have also been successfully expressed for many other viruses, including Norwalk virus,⁴⁰  HPV and BPV,⁴¹  rabbit hemorrhagic disease virus,⁴²  the adeno-associated viruses,⁴³  avian influenza virus⁴⁴  and MVM⁴⁵  and FHV⁴⁶  which were useful in crystallization studies, producing crystals which diffracted X-rays to beyond ∼3.3 Å. One of the most appealing factors in this system is the presence of post-translational modifications. N- (simple; no sialic acid) and O-linked glycosylation, acylation, acetylation, disulfide bond formation and certain phosphorylation processes are all carried out in a manner similar to that found in mammalian cells. Recently, recombinant baculovirus vectors that contain mammalian expression cassettes for gene delivery and expression in mammalian cells have been developed (detailed information can be found elsewhere^47,48 ). These versatile constructs have been termed ‘BacMams’ to avoid confusion with the original baculovirus that drives gene expression in insect cell lines.

Advances in plant molecular biology and genomics have opened up the possibility of modifying their genomes. A large number of plant viruses studied today are propagated in the host plant. Host plant species can be easily grown, under proper conditions, and readily inoculated with the infecting virus. The disadvantage of this approach appears to be time and resources. Depending on the plant species and desired size for infection, it may take several weeks to obtain optimal conditions. Extensive space and supplies may also be required to generate adequate amounts of infected plants from which virus can be purified.

There is, to date, no general protocol for gene transfer into plants. Each cell type, tissue and plant species requires careful characterization to ensure optimal transfer to attain the highest efficiencies and reproducibility in terms of gene expression.⁴⁹  The genetic information of plants is distributed among three cellular compartments: the nucleus, the mitochondria and the plastids. The plastid is a circular double-stranded DNA (dsDNA) molecule which can account for 10–20% of the total cellular plant DNA content. Development of reliable methods in plastid genome transformation made feasible the targeted manipulation of the endogenous genetic information of plastids and, in addition, the possibility of introducing novel information to be expressed from engineered chloroplast genomes.⁵⁰ 

Inoculation of plant host species through natural transmission routes generally requires an insect vector and is often not feasible for the average researcher. More common practice involves genetic manipulation of plastids from Agrobacterium tumefaciens and microparticle bombardment or biolistics. A. tumefaciens is a rod-shaped Gram-negative ubiquitous soil bacterium that has become a useful tool due to a set of genes (T-DNA) located on the tumor-inducing (Ti) plasmid. These genes are capable of transferring and integrating into a foreign host and have become an ideal vehicle for gene transfer in plant research.^51,52  The use of A. tumefaciens inoculation has successfully been used to express HPV-16 L1 VLPs for use in edible vaccine production⁵³  and the plant geminivirus maize streak virus for structural studies by cryo-EM.⁵⁴  Biolistics involves the high-speed transfer of naked DNA that has been adsorbed on small metal particles, combining biology and ballistics, into plants. This method is very effective especially when the plant strain being utilized is resistant to A. tumefaciens.

Regardless of the method used to introduce exogenous DNA into plants, the gene of interest must be cloned into an expression cassette whose minimal requirements are a promoter and a terminator of transcription functional in the plant system.⁵⁵  Another requirement is the early selection of transformed cells from non-transformed tissue, achieved by the inclusion of a selection marker. A fertile plant can then be grown from these transformed cells. A major disadvantage is that homologous recombination is not efficient in plants and can lead to random insertion of genes and instability. In this case, several independently inoculated plants should be compared for expression levels.

An alternative system to whole plants is the use of plant cell systems. These can be cultivated like mammalian and bacterial cells and also offer eukaryotic post-translational modifications. Virus-based expression systems are also applicable. This is synonymous to using phage viruses in bacteria and provides an alternative to stable genetic transformation in plants. These vectors can be full (DNA-containing) virus vectors or deconstructed vectors, which generally lack several infectious aspects and places the gene of choice between viral DNA elements. Loss of gene insertion can occur with full viral vectors, especially if it is a large insertion, due to systemic movement in the plant cell. The use of deconstructed vectors has produced decent yields of VLP for several mammalian viruses using a tobacco mosaic virus system from Icon Genetics (http://www.icongenetics.com).⁵⁵  Protein production can reach relatively high levels in 3–14 days, depending on the system. For example, Norwalk virus coat protein levels reached ∼20–30 μg g⁻¹ of dry fruit in transgenic tomato plants.

The use of Escherichia coli in the laboratory setting is a definite hallmark in biotechnology and almost marks the birth of this field. It is generally the preferred prokaryotic expression system due to (i) rapid (∼30 min) and high-level expression (50–500 mg L⁻¹) as a result of the speed of cell growth to high density, (ii) low complexity and low cost of growth media and (iii) the ability to target proteins to the desired subcellular localization.⁵⁶  However, the system has many disadvantages when used for the production of large eukaryotic proteins, as follows. (i) The cytoplasm is a reducing environment that strongly disfavors the formation of stable disulfide bonds. This can be detrimental to the formation of important assembly interactions, especially in a large macromolecule such as the viral capsid, which will affect stability and proper folding. The creation of certain strains of E. coli with mutations in thioredoxin reductase (encoded by trxB) and glutathione reductase (gor) aid the expression of proteins whose solubility depend upon an oxidative environment.^57–59  (ii) Wt E. coli lacks the ability to phosphorylate tyrosine residues, although strains have been engineered which allow this modification.⁶⁰  (iii) The overproduction of heterologous proteins in E. coli often results in misfolding and segregation into insoluble inclusion bodies. A number of techniques are cited in the literature, which discuss the conversion of inactive protein, expressed in an insoluble fraction, into a soluble and active form.⁶¹  (iv) Of great concern to virologists is the lack of post-translational modification in bacteria, such as glycosylation, acetylation and amidation. This can alter many of the functional properties of viral proteins and also structural features.

The selection of the promoter is also critical in bacterial expression system design. It is generally controlled by a regulatory gene or inducer, that is either inherent in, or supplied to, the host. The most widely used promoters are the lactose (lac)⁶²  and trytophan (trp)⁶³  promoters. Stronger, more tightly regulated promoters, trc and tac, have been created from the lac and trp promoters, but have incomplete repression in the induced state. This is not an issue when the gene product is not toxic to the cell. Another factor to consider is subcellular localization. Recombinant proteins may be directed to one of three compartments: cytoplasm, periplasm or the extracellular medium. Proteins found in the cytoplasm may require extra purification steps from inclusion bodies and refolding. This can hinder proper folding and ultimately cause issues in assembly when dealing with viral proteins. Periplasmic targeting offers advantages in proper folding due to the oxidative environment. However, proteins must change conformation to be shuttled across the cytoplasmic membrane and can be degraded due to incompatibility with the membrane. Since E. coli does not secrete many proteins into the extracellular fluid, there is less proteolytic activity and thus less degradation. Secretion also makes purification easier as less undesired proteins are present. The disadvantage is the low yield due to successful passage across both the inner and outer membranes.

Although the E. coli system offers many advantages in cost and quantity produced, the expression of complete viruses is not common due to the complexity of the interactions often required for the assembly of viruses. However, viruses that will self-assemble into VLPs with a monomeric unit of their capsid protein can be successfully expressed. For example, Chen et al.⁶⁴  successfully expressed small VLPs of HPV-16 from one of two virally encoded capsid proteins, L1, in an E. coli system. The VLPs were successful crystallized and the structure of the HPV16 L1 capsid was determined to ∼3.5 Å resolution. Bacteriophages naturally use E. coli as a host and have been used to produce many wt viruses which have been successfully used for structure determination studies by many different approaches, including X-ray crystallography and cryo-EM. A classic example is ΦX174, for which virions produced in E. coli were used to grow crystals which diffracted X-rays to ∼2.7 Å⁶⁵  and were used for its structure determination.⁶⁶ 

This system utilizes purified components from cell homogenates that are necessary for protein synthesis, allowing the study of many biological processes free from the complex reactions that occur in a living cell. Cell-free synthesis of infectious virus has been useful in studying the mechanism of viral replication and assembly, and also screening anti-viral drugs. The methodology has been used for the production of poliovirus⁶⁷  and encephalomyocarditis virus.⁶⁸  Yields have not yet been optimized, however, for the amounts required for structural studies by X-ray crystallography, but enough virions can be produced for cryo-EM applications for a well-characterized viral system.

Isolation of infectious virions from patient samples, such as blood, feces or urine, is generally used in initial attempts to determine the presence of virion particles. This technique was used for discovering the newly described human bocavirus, which was isolated from nasopharyngeal aspirates,⁶⁹  and capsids were subsequently visualized by negative stain electron microscopy.⁷⁰  This method does not generally produce a large enough amount of sample for structural studies. In addition, safety concerns associated with the handling of wt infectious virus isolated from patient samples can restrict their use to facilities with established containment appropriate for the biosafety level of the virus system of interest. In general, once described and genetically characterized, viruses isolated from tissue samples are expressed in cells using one of the methods described above for further molecular and structural studies.

Purification is an essential process for generating virus capsids for structural characterization and is an integral requirement for successful crystallization for structure determination by X-ray crystallography. Prior to the use of the capsids for structural or other biophysical study, the sample should be checked by SDS-PAGE developed by Coomassie blue or silver staining to ascertain purity, by Western blot against an antibody to verify that the capsid viral proteins are present and by negative stain electron microscopy for integrity. The steps involved in a purification protocol are contingent on the nature of the virus under study and the medium from which it is being purified. For example, enveloped viruses may require extra factors, such as detergents for solubilization, which may not be necessary for non-enveloped viruses, or mild solvent conditions may have to be employed when purifying unstable complexes. Samples that are not secreted from cells or are in inclusion bodies will require more steps than those that have been secreted into the cell media. Other factors that must also be taken into account after expression includes solubility, i.e. whether refolding is required, and capsid stability, i.e. whether pelleting will damage the integrity of the assembled capsid. The two most common approaches used for virus capsid purification are ultracentrifugation and chromatography, which are described below.

Ultracentrifugation is the usual technique of choice for the purification of virus capsids, particularly because this approach can utilize their defined size and shape to aid separation from other cellular material. The rate of sedimentation depends upon the size, density and morphology of the capsid, in addition to the nature of the density medium and the force that is applied during centrifugation. Cushions, using sucrose, dextran or Ficoll (GE Healthcare), can be incorporated in virus purification after the initial centrifugation to allow for the collection of morphologically intact capsids, without causing mechanical stress, for further purification. Density gradients, i.e. a variation in density over an area, are often used after the cushion step and are a cornerstone in virus purification. The two types of density gradients routinely used are rate-zonal and isopycnic centrifugation.

In rate-zonal approaches, the sample is layered over a gradient that allows for the separation of particles into bands or zones, based on the particle sedimentation rate. Step gradients generally result in better separation, but linear gradients can also be used. Gradient makers can be used to create linear gradients or steps poured can be allowed to sit vertically overnight (4 °C) to aid diffusion between the boundaries. The rate at which separation occurs is dependent on the particle size, shape, density, force applied and the profile of the gradient medium. In this method, capsids continue to migrate into the gradient, hence an idea of the virus capsid sedimentation velocity is required to ensure that the sample is not pelleted. Sucrose is generally used for virus purification by sedimentation velocity, but established protocols are also available for Ficoll, iodixanol (OptiPrep; Axis-Shield) and dextran detergents may be added, especially if the virus has a tendency to associate with membranes, to separate cellular debris from virus particles. To achieve the best separation, gradients should not be overloaded with sample and while the amount is dependent on the virus being studied, it is suggested that for large swing-out rotors, which hold volumes up to 30 mL, loaded samples should not exceed 5 mg, whereas for smaller capacity rotors, which may hold ∼5 mL, 1 mg or less should be loaded.⁷¹ 

In contrast to rate-zonal gradients, isopycnic gradients separate capsids based on their calculated densities. Samples can be layered on the gradient or mixed directly with the gradient medium since the gradient will equilibrate upon centrifugation. Particles migrate to their density in the gradient and do not migrate further. Cesium chloride (CsCl) is the medium of choice for this application, mainly due to its ability to form dense solutions of up to ∼1.91 g cm⁻³ that extend beyond the range of density values for most non-enveloped and enveloped viruses. Suggestions for the concentrations of CsCl to be used range from 1.32 g cm⁻³ (32% w/v) for virus containing 5% RNA to ∼1.7 g cm⁻³ (55% w/v) for DNA-containing virus.⁷²  A milder method for labile or non-enveloped viruses uses a ‘positive density/negative viscosity’ approach. This involves the layering of potassium tartrate such that it generates increasing density from top to bottom, or the use of glycerol, which yields decreasing viscosity from top to bottom. CsCl can still be used for these viruses, but potassium tartrate or glycerol provides a gentler medium.⁷² 

Chromatographic techniques, which utilize separation on a column, may be implemented as intermediate or final steps in virus capsid purification. There are several properties of viruses which can be exploited to aid in their purification using chromatographic methods, such as size, charge, hydrophobicity and ligand specificity. For excellent guidance in selecting the proper media and more in-depth methods on column chromatography and protein purification, the reader is directed to GE Healthcare Handbooks (www.gelifesciences.com).

In size-exclusion chromatography (SEC) or gel filtration, samples are run over a solid-phase column composed of beads with pores of a defined size, thus samples are separated based solely on size. There are two distinct approaches that can be used, (i) group separation, where high or low molecular weight species can be distinguished, or (ii) high-resolution fractionation, in which isolation of individual species occurs based on molecular weight. The first step towards a successful separation is the selection of the appropriate media. The Sephadex G series (GE Healthcare) are useful for group separation applications, whereas for high-resolution fractionation Sephacryl, Superose and Superdex (GE Healthcare) are used. The high-resolution matrices are applicable for the separation of samples in the 1–8000 kDa molecular weight range, which can separate from peptides to large proteins or large complexes. The choice of medium to use for a particular application is dictated by the range of sizes to be separated and a predetermined selectivity curve, which is available from the supplier.

Ion-exchange column chromatography (IEX) separates samples based on differences in the net surface charge and is capable of distinguishing molecules that have minor differences in their charge properties. The functional groups that are bound to the matrix determine the charge of the IEX media. The columns can be either cation exchangers, binding to net-positive surfaces, or anion exchangers, binding to net-negative surfaces. Separation by this technique thus relies on the condition of the sample under certain pH and ionic strength conditions.

Hydrophobic interaction chromatography is an excellent step which can be incorporated between other methods to allow for buffer exchange and concentration of the final product. This method uses the inherent hydrophobic properties of a molecule for binding and its subsequent reversal for elution. Samples can also be separated based on their differing degrees of hydrophobicity. Reversed-phase chromatography is essentially based on the same properties, but in this approach the surface medium is more hydrophobic. This leads to stronger binding interactions, which then requires more stringent elution techniques, for example, organic solvents.

Lastly, ligand specificity or affinity purification involves the use of columns with bound ligands that interact specifically with the capsid of interest, based on biological function or chemical composition. This can result in high purity due to its specificity. There are many ligands available for different applications, including antibodies, enzymes and cell-surface receptor molecules. For virus purification, this can be very useful for the isolation of properly folded capsid components and properly assembled capsids due to the high selectivity involved in the interactions with the column medium.

The adeno-associated virus (AAV) is a small, non-enveloped, single-stranded DNA virus (ssDNA) that belongs to the family Parvoviridae. Several different serotypes are under development as viral vectors for gene delivery applications due to their simplicity, non-pathogenicity and ability to package and deliver non-genomic DNA to non-host cells and tissues. In an effort to improve the efficacy of gene delivery by these promising AAV vectors, there is a need to understand their basic biology, particularly in terms of the capsid structure and its role in dictating the functions of the virus during the infectious process. These include the interactions required for (i) cellular receptor recognition and entry, (ii) trafficking to the nucleus for genome replication, (iii) genome packaging following assembly and (iv) with host cell antibodies, which can lead to neutralization. Hence methods have been developed for large-scale capsid production and purification to facilitate these studies.

The AAVs package a 4.7 kb ssDNA viral genome with two open reading frames (ORFs), rep and cap. The rep ORF codes for four overlapping proteins required for replication and DNA packaging. The cap ORF encodes three capsid viral proteins (VPs) from two alternately spliced mRNAs. One of these mRNAs contains the entire cap ORF and encodes VP1. The other mRNA encodes for VP2, from an alternative start codon (ACG), and VP3, from a conventional downstream ATG. AAV capsids are assembled as a T=1 icosahedral particle (∼260 Å in diameter) from a total of 60 copies of VP1, VP2 and VP3, in a predicted ratio of 1:1:8/10.⁷³  VP3 is a 61 kDa protein that constitutes 90% of the capsid's protein content. The less abundant capsid proteins, VP1 (87 kDa) and VP2 (73 kDa), share the same C-terminal amino acid (aa) sequence with VP3 but have additional N-terminal sequences. AAV capsids can be assembled from heterologous systems from expressed VPs in the absence of genome. An example is given below for the production of VLPs assembled from VP1, VP2 and VP3 of AAV serotype 1 without packaged genome using the BEVS and their purification using ultracentrifugation and chromatographic approaches for structural studies.

A recombinant baculovirus encoding the AAV1 ORF was constructed using the Bac-to-Bac system (Gibco BRL). The AAV2 capsid ORF in pFBDVPm11⁴³  was replaced by the respective ORF encoding AAV1 capsid proteins derived from pAAV2/1.⁷⁴  Similar mutations were introduced into 50 non-coding and coding sequences to permit the expression of the AAV1 capsid proteins in the insect-cell background,⁴³  and the resulting construct expressed all three AAV capsid proteins, VP1, VP2 and VP3 (Figure 2). DH10Bac-competent cells containing the baculovirus genome were transformed with pFastBac transfer plasmids containing the AAV component insert. Bacmid DNA purified from recombination-positive white colonies was transfected into Sf9 cells using TransIT Insecta reagent (Mirus). Three days post-transfection, media containing baculovirus (pooled viral stock) were harvested and a plaque assay was conducted to prepare independent plaque isolates. Several individual plaques were propagated to passage one (P1) to assay for the expression of the AAV1 capsid genes and a selected clone was propagated to P2 and subsequently amplified to P3 for large-scale virus production.

A titered P3 recombinant baculovirus stock (generated as described above) was used to infect Sf9 cells grown in Erlenmeyer flasks at 300 K using Sf-900 II SFM media (Gibco/Invitrogen Corporation) at a multiplicity of infection (MOI) of 5.0 plaque-forming units (PFU) per cell. The cells were harvested at ∼72 h post-infection (pi), spun down in a Beckman JA-20 rotor at 1090g and resuspended in lysis buffer (50 mm Tris–HCl, pH 8.0, 100 mm NaCl, 10 mm MgCl₂, 0.2% Triton X-100) at a final concentration of ∼1×10⁷ cells mL⁻¹.

The VLPs were released from infected cells by three rapid freeze–thaw cycles in lysis buffer, with the addition of Benzonase (Merck, Darmstadt, Germany) after the second cycle. The sample was clarified by centrifugation at 12 100g for 15 min at 277 K and any resulting pellet was discarded. The cell lysate was pelleted through a 20% w/v sucrose cushion (in 25 mm Tris–HCl, pH 8.0, 100 mm NaCl, 0.3% Triton X-100; buffer A) by ultracentrifugation at 165 000g for 3 h at 277 K in a Beckman Type 70 Ti rotor. This process involved the layering of the sucrose solution at the bottom of a tube containing the clarified cell lysate. The supernatant was quickly discarded after the spin and the pellet was resuspended in ∼1 mL of buffer A, with 1 mm EDTA added, overnight at 277 K. The sample was further subjected to multiple low-speed spins at 10 000g to remove insoluble material. The clarified sample was loaded on to a sucrose-step gradient (5–40% w/v), prepared by layering ∼1.5 mL of a sucrose percentage solution into a Beckman ultraclear tube, beginning with the 40% and ending with the 5% fraction at the top. The gradient was spun at 210 000g for 3 h at 277 K in a Beckman SW 41 Ti rotor. A visible blue VLP band (illuminated by a light source) in the 20% sucrose layer was extracted (with a syringe needle) and dialyzed into 50 mm Tris–HCl, pH 8.0, 15 mm NaCl at 277 K to remove sucrose.

AAV1 VLP was also further purified, following a sucrose gradient, using IEX. As described by Zolotukhin et al.⁷⁵  for recombinant AAV1 vectors, a 5 mL HiTrap Q column (Pharmacia) was equilibrated at 5 mL min⁻¹ with five column volumes of buffer B (20 mM Tris–HCl, pH 8.5, 15 mm NaCl), then 25 mL of buffer C (20 mm Tris–HCl, pH 8.5, 500 mm NaCl), followed by 25 mL of buffer B using a peristaltic pump. The AAV VLP-containing fractions were then diluted 1:1 with buffer B and applied to the column at a flow rate of 3 mL min⁻¹. After the sample had been loaded, the column was washed with 10 column volumes of buffer B and the sample was eluted with buffer C on a Pharmacia ATKA FPLC system, and 0.5–1 mL fractions were collected.

The purity and integrity of the viral capsids were monitored using SDS–PAGE and negative-stain electron microscopy (Figure 2), respectively. The sample was generally buffer-exchanged at 5000g using Amicon Ultra filters (Amicon Ultra-15, 10 kDa molecular weight cutoff; Millipore) into 100 mm HEPES–NaOH, pH 7.3, 50 mm MgCl₂, 0.03% NaN₃ and 25% glycerol and concentrated to a final concentration of ∼10 mg mL⁻¹ for X-ray crystallographic studies or into a different buffer and another desired concentration as appropriate for the study to be undertaken.

The production and purification of virus capsids in quantities suitable for structural characterization provides a means for functional annotation of numerous virus systems. A number of these viruses can be produced using different expression systems and the majority of the viral capsids produced are amenable to purification by two well-developed methods, ultracentrifugation and column chromatography.

This work was supported in part by NSF grant MCB-0718948 and NIH R21 AI81072341.