“This work would have been impossible using the 'intelligent design' approach to evolution!”

We have been somewhat surprised ourselves by the interest shown in viral evolution. Perhaps the main reason is that our findings promise to bring some order to the viral universe. The key idea is that viruses can be grouped into lineages by comparing the structures of their key protein components. In the past, viruses have been seen as complicated things placed apart from other organisms by the view that they are not living. Their genomes are usually sufficiently different from each other that sequence comparison, even at the protein level, reveals few similarities. Historically too, viruses have often been grouped and studied in terms of disease. These factors make viruses somehow "strange" and contribute to the difficulty of placing them in an evolutionary context. The usefulness of the idea that there is likely to be a limited number of viral lineages in our own work is shown by the fact that several citations come from our other papers. While this suggests that some caution is always warranted in interpreting citation data, we find from personal experience that many colleagues are finding the ideas expressed in our paper both useful and stimulating. In our work, we combined structural and genetic data for the coat proteins of dsDNA viruses to relate many different viral families. Other researchers can apply the same approach to further viral proteins and lineages.

The paper shows that many viruses with quite different hosts are related by their use of a common "double-barrel" structural motif in their major coat protein. The investigation had its origins in our discovery that the bacteriophage PRD1 coat protein is similar to that of adenovirus. In our paper, we used a bioinformatics approach using molecular modeling to show that these are just two members of a far more extensive viral lineage. What is striking is that the lineage embraces viruses infecting both Gram-positive and Gram-negative bacteria, as well as animals, and includes very large viruses infecting insects, algae and amoebae. In our studies, we have found that the concept of "viral self" is very powerful. This separates those aspects of a virus that reflect its origins from the many host-related features that arise in parasitic organisms. Taken together, our findings lead to the important general conclusion that apparently unrelated viruses can be linked by comparing the molecular structures of their fundamental components. A very useful specific conclusion is that it now seems highly likely that all large icosahedral viruses with triangular-looking coat proteins are constructed in a similar manner to adenovirus. This proposal is powerful as it has predicted new members of the "double-barrel" lineage that were later confirmed by crystal structures for their coat proteins.

We find that viruses can be classified according to the molecular structures of certain key features. This approach, using morphology, corresponds to the scheme introduced by Carl Linnaeus (1707-1778) for plants and subsequently applied to all other living organisms. Such relationships in viruses have been obscured by their rapid genetic changes, their customary designation as non-living, and their evolution of new features to optimize interactions with their host. For these reasons, and because viruses are so small, their morphology must be compared at the atomic level. Classification may seem quite dull, but it is fundamentally important for understanding and treating infectious diseases. If a newly discovered organism can be classified, it can be combated with drugs successful with its cousins. For example, information about bacteriophage PRD1 may help in controlling dangerous viruses now known to lie in the same lineage, such as African Swine Fever Virus. It has been difficult to develop antiviral agents, but a good start has been made with HIV. If more viruses can be grouped into lineages, then antiviral agents developed and tested for one member can be tried on the entire group. A very important implication is that it may be possible to develop broad-spectrum antivirals.

The roots of this project lie in the crystal structure of hexon, the adenovirus major coat protein, which was determined in the laboratory of Dr. Roger M. Burnett. The first low-resolution images in the late 1970s revealed that the large trimeric molecule is pseudo-hexagonal in its base. Its shape was used to interpret electron micrographs of viral fragments and work out how the coat proteins are organized in adenovirus. It was shown that this architecture was ideal for building icosahedral particles with very large facets, while using only the limited number of hexon-hexon interactions required for error-free assembly. We began our very fruitful collaboration with Drs. Jaana K.H. Bamford and Dennis H. Bamford at the first FASEB Viral Assembly meeting in 1990. This was particularly appropriate, as Dr. Burnett and Dr. Jonathan King (MIT) had established the meeting to bring together workers on human and bacterial viruses. Dr. Stacy D. Benson subsequently determined the crystal structure of the bacteriophage PRD1 coat protein (P3) as a graduate student. The great surprise was that PRD1-P3 has two "viral" barrels or jellyrolls—a "double-barrel" fold—like adenovirus hexon. Although the coats of several small viruses were known to contain this fold, the relationship uncovered by the common double-barrel was a revelation. PRD1 and adenovirus are quite complicated and apparently unrelated dsDNA viruses infecting quite different hosts. In addition, by the time of this discovery, links between different organisms were being more frequently exposed by the rapidly increasing mass of genetic data.

The next stage came with a short article that summarized our developing ideas. Our collaborator, Dr. David I. Stuart (Oxford), thought to test if the PRD1-P3 crystal structure matched the shape of the coat protein in the reconstruction from electron microscopy for a very large eukaryotic virus, Paramecium bursaria Chlorella virus 1 (PBCV-1). The good fit provided experimental evidence for the idea that PBCV-1 was a member of the adeno-PRD1 lineage. The subsequent crystal structure by Dr. Michael G. Rossmann’s group showed that the two proteins were indeed very similar.

We then decided to compare all the known sequences for coat proteins of icosahedral viruses with large facets. This was Dr. Benson’s last project with Dr. Burnett before setting up his own laboratory at Oklahoma State. He used model building to test whether the sequences, in the absence of detectable similarity, were consistent with the structure of that for PRD1. The results were very convincing. We are grateful to the editors of Molecular Cell for encouraging us to present our results as a Hypothesis article. Their interest, and answering the hard questions posed by reviewers and our colleagues, have all been invaluable in refining the ideas presented in our paper.

In many ways, viruses are the last frontier of infectious disease in terms of our ability to identify, classify, and eradicate microorganisms. The possibility of understanding them better is exciting as it should help in developing ways to combat viral infections as easily as antibiotics are now routinely used for bacterial disease. This work would have been impossible using the "intelligent design" approach to evolution!