More barrels from the viral tree of life

There is now unmistakable evidence for a revolution in our comprehension of the viral world. Despite their importance for disease and the insights obtained by studying their fascinating intricacy, viral relationships have long been unclear. The article by Khayat et al. (1) in a recent issue of PNAS marked a milestone in progress. Viral origins have been masked by their diversity and complexity, the way they were studied, and their small size. Viruses evolved cunning solutions to the problems of packing information into genomes of limited size and to entering and exploiting their host. Until fairly recently, each family was studied separately by its own group of scientific admirers. Structurally, viruses could only be imaged by low-resolution electron microscopy (EM), with most too large to tackle with crystallography. Advances in imaging techniques, changes in the way that viruses are studied, and increasing structural evidence all are contributing to major change. There is now excellent evidence to support the idea that viruses fall into lineages identifiable by features in common with ancient precursors (2). The best-supported lineage has “double-barrel trimer” coat proteins and icosahedral virions. Its members infect Gram-positive and Gram-negative bacteria, as well as animals, and include very large viruses infecting insects, algae, and amoebae. Structures are at hand for coat proteins of viruses infecting animals, bacteria, and algae, but that for an archaeal virus was lacking. This crucial missing link is found in the thermophilic archaeal Sulfolobus turreted icosahedral virus (STIV) (1). We now have solid structural evidence that at least one lineage originated billions of years ago, before life separated into its three domains: Bacteria, Archaea, and Eukarya.

To appreciate this achievement, one must recognize the dramatic progress in describing viral architecture since the first crystal structure of a complete viral particle, or virion, appeared (3). This plant virus had an eight-stranded barrel, with a “viral” jellyroll fold, for its coat protein. Because nearly all other viruses, including simple animal viruses, had this fold, a possible relationship was appreciated immediately (4). Larger viruses, like adenovirus, were much harder to tackle, because they were then beyond the scope of crystallography. Even determining the structure of the coat protein alone, as now done for the STIV major coat protein (MCP) (1), was challenging for molecules like the 967-residue adenovirus hexon (5). Its first low-resolution images revealed that this molecule, although trimeric, has a pseudohexagonal shape. Its shape was used to decipher EM micrographs of negatively stained viral fragments (6) and show how hexons closely pack in the adenovirus virion (Fig. 1).

Fig. 1.

Organization of the coat proteins in PRD1 (and adenovirus) (6, 17). The 240 closely packed pseudohexagonal trimers (P3 or hexon; blue) form arrays of 12 in the facets, which curve to bring the proteins together at the icosahedral edge. An additional distinct location between trimers numbered 1–4 yields facets of any size. The elongated PRD1 tape-measure protein (red) stretches from the vertex to the middle of an edge, where it forms a hook. [Figure courtesy of Drs. Nicola G. A. Abrescia and David I. Stuart, University of Oxford, Oxford.]

An important question in viral architecture is how multiple copies of the coat protein locate correctly for error-free assembly. An answer came from Caspar and Klug's idea of quasi-equivalence (7), which postulated that they lie in similar environments. This scheme worked for simple virions where, for example, 180 proteins form three sets with exact 60-fold symmetry (described by the triangulation number, T = 3). For larger virions like adenovirus (nominally T = 25), it was difficult to understand how accurate assembly could occur. The model gave an answer. The 240 hexons are arranged on the facet like a small crystal (Fig. 1) so that hexons use only four different binding sites. The apparent problem of assembling 1,500 individual proteins accurately into a T = 25 shell (25 × 60 subunits) reduced to placing hexon trimers into four related locations (the penton base protein at the vertex is different). This picture was satisfying, because it suggested that similar principles could explain the puzzle presented by extremely large viruses, such as the insect virus that had been elegantly revealed to be T = 156 by Wrigley in 1969 (8). Adenovirus-like architecture is ideal for large virions because it uses few distinct locations (Fig. 1). With only five sites, hexons can yield facets of infinite size (6). The hexon crystal structure later revealed that each subunit has two vertical barrels (the double barrel), each forming one corner of the hexon base (5). Further details of the virion itself were only revealed after advances in EM provided three-dimensional images.

Current spectacular progress in structural virology (9) would be impossible without cryo-EM as a counterpart to crystallography. Here, flash-frozen virions are protected from radiation damage and imaged without stain, like a computerized axial tomography scan, but using electrons to get views of many particles in different orientations. The first EM reconstruction of the adenovirus virion at 35-Å resolution (10) confirmed its model but showed that the facets were rounded, rather than planar, to bring edge hexons into more intimate contact (Fig. 1). Instead of the 27 virions used then, hundreds of thousands are used today to obtain resolution <10 Å (11). Cryo-EM can image very large virions (>1,000 Å in diameter), does not require large quantities or crystals, and even reveals protein secondary structure. Spectacular examples of the many virions imaged are Paramecium bursaria Chlorella algal virus 1 (T = 169) (12) and the largest one known, mimivirus (5,000 Å in diameter, T = 1179) (13). Combining EM with crystallography gives even more power, as shown by the approach to STIV (1). For STIV, EM was particularly valuable as MCP crystallized as a monomer.

The next chapter in the story concerns bacteriophage PRD1, first of interest for its internal membrane. Although similar to adenovirus by EM, it was a great surprise when its coat protein (P3) showed a double-barrel-like hexon (14). PRD1 and adenovirus are complicated, and apparently unrelated, viruses infecting quite different hosts, but their structural similarities pointed strongly to a common ancestor. Could other icosahedral viruses be similar? A good fit of the PRD1 P3 trimer to the Paramecium bursaria Chlorella algal virus 1 EM reconstruction was suggestive (2), and the structure of its Vp54 coat protein gave confirmation (15). A bioinformatics approach, testing sequences for compatibility with PRD1 P3, then identified other icosahedral dsDNA viruses with double-barrel trimer coat proteins (14). Predictions that STIV is a member of this lineage (14, 16) are now verified (1).

If the STIV coat protein resembles other double barrels, what else is new? Crystallographers love new folds but, once again, we find that there is much to be learned from each example of a homologous protein. For STIV MCP, the big surprise is that the stable unit is one double barrel (it crystallizes as a monomer). The high degree of internal packing, with 50% fewer cavities than PRD1 P3, explains its stability in an acidic hot spring (pH 2–4, 72–92°C). Its upper loops are shorter than in PRD1 P3 and Paramecium bursaria Chlorella algal virus 1 Vp54, and prolines reduce flexibility. The trimerization loop that stabilizes PRD1 P3 by tripling the buried surface area is missing. That MCP forms a trimer is only known from the EM reconstruction of the virion (16). Why is it not a trimer? Two reasons come to mind. The first is the problem of sealing a virion edge. PRD1 shows how facets are not planar but curve to maximize their mutual interactions (Fig. 1). The facets in STIV do not interlock like PRD1 but are skewed triangles, as in the large insect and algal viruses. Perhaps a less rigid trimer makes the facet more flexible and increases overall coat protein contacts. The deviations and slightly better fit obtained when MCP monomers rather than trimers were fitted to the virion (1) suggests that this is the case. Second, folding a monomer alone at high temperature is simpler (P3 and hexon both need chaperonins, missing here, for trimerization). But what compensates for the missing loop? Surprisingly, the MCP trimer in the virion is not hollow, as would be expected. Perhaps a small cementing protein fills the void.

A major question for assembly is what stops the growth of facets to determine virion size? The crystal structure of the entire PRD1 virion revealed an extraordinary elongated protein (Fig. 1) and suggested the answer (17). If this “tape measure” is common to the lineage, it elegantly solves the problem. Does STIV have this protein, and is there a pentamer of single barrels to seal the capsid at the vertex like PRD1 (17) and adenovirus (18)? Alas, the resolution of its EM reconstruction is still too low to provide answers.

Do viral lineages have practical consequences? An important one is classification, vital for understanding and treating infectious disease. PRD1 contains a unique vertex responsible for packaging DNA (19). If its relatives are similar, drugs inhibiting PRD1 may work on viruses like African swine fever virus (14). Tape-measure proteins could be targeted by disrupting their loops (Fig. 1). Growing evidence supports other lineages, such as that linking herpesvirus to the tailed bacteriophage (2, 9). As common features in a lineage are identified, antiviral agents developed and tested on one member can be tried on all. Broad-spectrum antivirals now seem possible, as effective with viral infections as antibiotics are with bacterial disease.

Finally, confirmation that the PRD1-adenovirus lineage infects all domains has implications for early life. It is striking that many members have membranes and that the related poxvirus (14) forms immature particles with trimers in an external honeycomb scaffold (20). Perhaps early viruses assembled like PRD1 by attaching proteins to a membrane. With evolution, some, like adenovirus, found ways to bypass this two-dimensional organization and lost the membrane. Others, like poxviruses, kept the membrane but found it advantageous to be freed from the constraints of a rigid coat and so discarded it while maturing. The lineage may embrace many more viruses than is currently apparent. One very intriguing putative member is mimivirus (14), whose properties and genome suggest an ancient form between a virus and a bacterium (13). Together, all of these findings cast some light on viral origins. Perhaps viruses are descendants of ancient self-reproducing systems that helped early cells to develop by shuttling genes between the protected environments in divisible lipid vesicles. The evidence points to a primordial role for viruses (and barrels), rather than to a later appearance. These are big conclusions from a small virus, but the work by Khayat et al. (1) illustrates that a picture may be worth 1,000 words, but some images lead to even more speculation.