More Secrets of the Virus World

It's generally conceded that viruses evolve more rapidly than host cells, but the rates vary tremendously depending on the type of virus. Generally, large DNA viruses that infect algae (the phycodnavirus family) are considered to have some of the slowest rates of change, whereas the fastest-to-change viruses tend to be small RNA viruses that infect animal cells (e.g., HIV). In terms of substitutions per nucleotide per cell infection (s/n/c), one recent study found rates of 10⁻⁸ to 10⁻⁶ s/n/c for DNA viruses and 10⁻⁶ to 10⁻⁴ s/n/c for RNA viruses, which means the fastest-mutating viruses change 10,000 times faster than the slowest-mutating viruses.

Given the ultra-rapid rate of change of RNA viruses and their generally impressive level of adaptation to host-cell environments, one might expect a virus like HIV-2 to show a codon usage bias similar to that of the host. And that's approximately true.

HIV-2 codon usage (left), in DNA format (T for U), versus overall human-cell codon usage (right).

The above graph shows codon usage for HIV-2 on the left and codon usage for human cells on the right. (HIV is an RNA virus, but codons are shown here in DNA format, with T in place of U.) R-squared/adjusted comes to 0.2204, so we can't very well say confidently that the codon values are highly correlated. But if you look at the smaller bars (not the "peaky" ones), they tend to taper down on the left, just as on the right.

It might be instructive to go from one of the fastest-changing viruses in the biosphere (HIV) to one of the slowest, and see how its codon usage compares to that of its host. This time, we're looking at the large DNA virus known as PBCV-1 (left) versus its Chlorella host (an alga, right):

Codon usage in Paramecium bursaria Chlorella virus 1 (PBCV-1), left, and Chlorella variabilis strain NC64A, right.

These two data sets are not only not correlated, they appear to be anticorrelated, which is quite unexpected. Bear in mind, PBCV-1 is relatively large, with a genome of 330,601 base pairs encoding hundreds of proteins (and ten tRNAs). Thus the pattern shown here isn't likely to be random noise. Note that PBCV-1 has a genomic G+C content of 40%, versus 61% for the host, which is a pretty sizable separation. It's almost as if PBCV-1 has spent part of its life coexisting with an entirely different host.

Which brings me to the final and most intriguing (I might even say shocking) graphic, which compares codon usage in PBCV-1 virus with codon usage in Chlorella's own host, Paramecium.

Codon usage in PBCV-1 virus (left) and Paramecium (right).

Recall that when it is not free-living on its own, the tiny unicellular Chlorella alga has an endosymbiotic relationship with the comparatively much larger unicellular ciliate protist, Paramecium. That is to say, Chlorella can live inside Paramecium. Chlorella allows Paramecium to thrive in high-sunlight/low-nutrient conditions, whereas Paramecium, in return, gives the non-motile Chlorella free transportation and protection against viruses. (PBCV-1 can infect free-living Chlorella, but does not infect Chlorella living inside Paramecium.) As far as I know, no one has ever reported that PBCV-1 virus can infect Paramecium. Supposedly, it infects only free-living Chlorella And yet, we find that the pattern of codon usage in PBCV-1 is very strongly correlated with the pattern of codon usage in Paramecium. (R-squared/adjusted: 0.527.)

Paramecium filled with Chlorella cells.

This chart is a real shocker from a couple of standpoints. First, as I say, PBCV-1 virus is not known to infect Paramecium. And yet codon usage patterns in the virus are much more closely aligned to Paramecium's patterns than to Chlorella's. Notice that AAA is the No. 1 most-used codon in PBCV-1 as well as Paramecium. Seven of Paramecium's top ten codons are in PBCV-1's top ten.

Secondly, Paramecium doesn't use the standard genetic code! It uses the Ciliate Code (Translation Table 6), in which TAA and TAG encode glutamine instead of serving as stop codons. (TGA is the one and only stop codon in Table 6.) If Paramecium used the standard genetic code, the alignment of the two organisms would be even stronger.

Also interesting is that PBCV-1 and Paramecium are quite far apart in G+C content (the former is 40%, the latter is 28%).

Perhaps at some point in its past, PBCV-1 had a wider host range, one that included Paramecium. It's possible that even today, it has hosts other than Chlorella that have yet to be observed experimentally. Certainly, the pattern of codon usage is consistent with such an idea.

Virus genes don't always come from the host

Usually, when a virus contains a certain kind of gene, and the host contains the same kind of gene, it's assumed the virus got its copy from the host. This is not a terribly safe assumption, however. In some cases it's demonstrably wrong.

For a striking example of how wrong this assumption can be, you need look no further than the tiny Chlorella alga that can be found living symbiotically inside the fresh-water ciliate Paramecium. When it's not living inside Paramecium, Chlorella is subject to infection by PBCV-1 (the Paramecium bursaria Chlorella virus).

Tiny green Chlorella cells can be seen here growing
inside Paramecium. The full Paramecium cell is
shown in the inset at lower left. (Photo by Charles Krebs.)

Both Chlorella and PBCV-1 have a gene for an enzyme called thymidylate synthase, which is the enzyme that produces thymidine monophosphate (dTMP, or just TMP), a precursor molecule for making DNA. Ordinarily, one would assume that the virus picked up the gene for this enzyme from its host at some point in the past. But there's a problem.

The only thymidylate synthase gene in Chlorella's genome codes for a protein with 508 amino acids. The PBCV-1 virus version of this gene codes for a much shorter protein with only 216 amino acids. It turns out there's a perfectly good explanation for the size difference. Like other small algae (such as Micromonas, Ostreococcus, and Bathycoccus) and certain protozoans as well, Chlorella has evolved a bifunctional enzyme. In Chlorella, the same enzyme acts as both a thymidylate synthase and as a dihydrofolate reductase. In most higher organisms, two different enzymes carry out these functions. Organisms that have the dual-function enzyme are presumed to have developed this capability through a gene fusion event sometime in the (most likely distant) past.

It turns out the PBCV-1 virus synthase not only isn't bifunctional, it carries out its thymidylate reaction by an entirely different mechanism than that used in the host enzyme. The host enzyme employs folate (but no flavins) as a cofactor, whereas PBCV-1 is strictly dependent on flavin adenine dinucleotide (FAD), as verified experimentally by Graziani et al. in 2006. We now know that many bacteria use the FAD version of this enzyme (often called ThyX, as disintguished from ThyA, the folate-only enzyme). And the FAD users all have relatively small thymidylate synthases, of about 200 to 300 amino acids.

The above scenario isn't exclusive to Chlorella and PBCV-1. It turns out, certain other small algae (Micromonas, Ostreococcus, and Bathycoccus; all happen to be salt-water algaae) have a bifunctional thymidylate kinase, yet they are subject to infection by viruses that use the much smaller, mono-functional flavin-binding enzyme.

In all these cases, the virus uses an entirely different style of enzyme than the host to carry out TMP production. There is essentially zero chance that the virus derived its enzyme from the host (or vice versa), because the reaction mechanisms of ThyA and ThyX are radically different. (For more detail on this, see the excellent review article at http://www.ncbi.nlm.nih.gov/books/NBK6401/.) These aren't orthologues; these aren't paralogues; these are entirely different enzymes.

So where did the virus get its thymidylate synthase from, if not the host?

If you take the protein sequence for the PBCV-1 thymidylate synthase and run a BLAST search at UniProt.org, the best non-viral hits (in the range of 58% identities, 77% similarities, E-value 10^-69) are for the thymidylate synthases of Prochlorococcus marinus and other cyanobacteria, with cyanophages also scoring high. This makes a great deal of sense, because the photosynthetic Prochlorococcus and its relatives are thought to be some of the most ancient bacteria on earth (possibly going back 3.8 billion years). They're thought to be the ancestors of chloroplasts. At one point, they were almost certainly the predominant life form in the oceans. Since phycodnaviruses (of which PBCV-1 is a member) are thought to be quite ancient, it's entirely possible they got their thymidylate synthase from cyanobacteria. That's certainly what the protein-sequence evidence suggests.

I'll go with the evidence.