There is a twist on transduction, using viruses to transfer genes between cells. It’s a kink that was first discovered in 1974, but it still is not well understood. Barry Marrs of Saint Louis University was analyzing strains of a bacterium called Rhodobacter, and found that certain strains of this organism were capable of transferring genes to other strains with surprising efficiency. In some ways, Marrs’ discovery seemed very similar to transduction: something involving DNA in protein envelopes was involved, and random pieces of the cells’ genome were carried in these envelopes.
Despite the superficial similarity to transduction, it was apparent that something unique was going on. Transduction is a rare event—for every transducing virus produced, a hundred thousand killer viruses are made. However, Marrs and others observed that the gene transfer was never associated with any lethality; it was as if the viruses were only carrying cellular DNA, never viral DNA. Other differences were noticed as well. The “gene transfer agents,” or GTAs as Marrs called them, were far smaller than any known virus. Although they could be observed by electron microscopy, and had the same shape as a regular virus, they were relatively tiny; here is a transmission electron micrograph of purified GTAs, showing their resemblance to tailed viruses.
A normal virus could hold a few dozen genes, but the GTAs could only carry four or five at a time. This small size makes it clear that whatever GTAs may be, they are not viruses. Remember that a virus particle is a protein envelope carrying the genes for making more virus particles—so the DNA encoding the virus must fit within a virus particle. We know the genes required for making GTAs, and how long a piece of DNA that requires—and it is impossible for the DNA encoding GTA production to actually fit in a GTA particle! It's as if the instructions for making a box were far, far larger than the box itself.
So what are GTAs? This is still an open question. We know from experiments and the analysis of genomes that there are many different types of GTAs, and they are found in Bacteria and Archaea. They always “look” like viruses. The genes encoding GTAs are organized in a pattern reminiscent of some viruses, and some of the individual genes for GTAs are homologous to viral genes--but there's always the problem that there's no way to fit all the genes for a GTA into a GTA.
However, while the execution of the construction program for many viruses has been dissected in excruciating detail, we know next to nothing about how GTAs are actually produced. We don’t even know such basic facts as whether, as with viruses, GTA production results in the destruction of the “donor” cell. The presence of a gene similar to viral “lysis” genes in many GTA gene clusters is suggestive—but it is not a smoking gun, and there's no picture like this
of GTAs escaping a cell. All we really know is that GTA genes are under cellular control, and are typically only expressed when the cells are crowded—conditions which make it more likely for a GTA to find a susceptible target.
The evolutionary history of GTAs is also wrapped in mystery. Their homology to viruses is suggestive of some sort of shared ancestry, but the nature of the connection is obscure. One can make a "family tree" of the proteins in viruses, and you'll see that a handful of GTA proteins fit very nicely in that tree. You can also make a family tree of the proteins in GTAs, and a handful of viral proteins fit right in. So, the only thing we can say is clear is that the viruses and GTAs have some shared family history--but who is descended from whom?
Are GTAs viruses that have been domesticated? Perhaps. The widespread distribution of viruses throughout the “tree of life,” and the relative rarity of GTAs has been used as an argument for this case. In this view, a twig on the family tree of viruses mutated to become GTAs, and they have been preserved for their role in facilitating genetic exchange. On the other hand, there are certain viruses that have genomes containing genes commonly found in viruses as well as genes that are usually used in making GTAs. The earliest reports of GTAs argue that gene transfer is as necessary to Bacteria and Archaea as sex is to eukaryotes. In this view, GTAs predate viruses; viruses are GTAs that "went rogue."
There is another aspect to GTA evolution that is puzzling. Although it is not known for certain, the similarities to viruses suggest that GTA production is fatal. The GTA genes thus serve no benefit to the individual cell—in fact, they are absolutely detrimental. Traditional views of natural selection imply that such genes should not be preserved—to this view, it is every bacterium for itself and damn the rest. The notion that a single bacterial cell should commit suicide and share its genes with its neighbors for their benefit is anathema. However, GTAs may be evidence of “kin selection,” where a gene benefits the species at the expense of the individual. The persistence of GTA genes in many different microbial lineages suggests that this may indeed be the case.
Such a view is heresy--I constantly remind my students that genes work for the good of the individual, never for the good of the species. However, GTAs may force me to put an asterisk by that statement and rethink evolution, or (more likely) redefine "individual" to include "a group of genetically identical yet independently living cells."