Genetic Drift of Human Immunodeficiency Virus Type 1?

In a recent analysis of nef gene sequences from a human immunodeficiency virus type 1 (HIV-1)-infected individual, Plikat et al. (5) conclude that genetic drift is the most important process of evolution in vivo. While we do not dispute the fact that stochastic processes may be important in shaping the genetic structure of HIV-1, we strongly reject the claim by Plikat et al. that their data can rule out immune-driven positive selection.

The analysis presented by Plikat et al. is based on a comparison of the numbers of synonymous (d_S) and nonsynonymous substitutions (d_N) per site that have accumulated over time. Under positive selection it is predicted that d_N should be greater than d_S because advantageous changes are fixed faster than they are produced by mutation (d_N/d_S > 1.0, where d_S is always assumed to reflect neutral change), which cannot happen under the neutral theory of molecular evolution (2). Plikat et al. are correct in stating that previous analyses of d_N and d_S are potentially biased in that the phylogenetic relationships of sequences are not taken into account: the currently used pairwise methods of analyzing these distances may replicate the same comparisons. Plikat et al. therefore develop a phylogenetic method based on split decomposition, which accounts for this bias, from which they conclude that there are no more nonsynonymous substitutions than expected under neutral evolution.

The problem, however, is in what d_N/d_S is expected to look like under neutral evolution. Plikat et al. state that the signature of neutral evolution is an equal rate of nonsynonymous and synonymous substitutions per site (d_N/d_S = 1) and, because this is observed in the nef gene, claim that they have documented evolution by genetic drift. If this were true, however, nef sequences would gradually lose their similarity over time, becoming difficult to recognize in a very short time given the mutation rate of HIV-1, and generally behave like “junk” DNA. To put it another way, d_N/d_S = 1 is characteristic of nonfunctional DNA (perhaps pseudogenes) and is never expected to be observed in a functional protein like nef (4). What Plikat et al. have misunderstood is that neutral theory also allows for purifying (negative) selection: amino acids that are functionally constrained will evolve at lower rates than those that are less functionally important. Therefore, the proper neutral model for a functional gene such as nef is one in which nonsynonymous substitutions accumulate at a lower rate than synonymous substitutions, because the former are more likely to be deleterious and so are subject to purifying selection (d_N/d_S < 1).

To test whether nef genes are subject to purifying selection we examined the spatial distribution of amino acid changes in 51 nef genes from HIV-1 subtype B taken from the Los Alamos database (3). If all amino acid changes are completely neutral, we should expect to see an even distribution of variation across the protein: because there is no selective constraint, all sites should evolve at the same rate. This is clearly not the case for nef (Fig. 1) for which there is substantial variation in rates of amino acid change across the protein. Furthermore, in our 202-amino-acid sequence alignment, 89 residues (44.1%) are invariant across all 51 sequences, implying that they are under especially strong functional constraint.

FIG. 1.

Spatial distribution of amino acid variability in 51 HIV-1 subtype B nef protein sequences extracted from the Los Alamos HIV database (3). Variability is plotted as the percent normalized index of variation (v%) obtained according to the following formula: v% = [(RD/N)100], where RD is the raw diversity, i.e., the number of distinct character states (20 amino acids plus 1 for gaps) observed at an amino acid site divided by the total number of sequences (51), and N (0.412) is the maximum possible diversity per site (21/51). For ease of analysis, two regions of high variability due to frequent insertion and deletion events were removed (codons 24 to 39 and 78 to 85). Sequences with stop or ambiguous codons were not included in the analysis.

Because nef produces a functional protein, a model of genetic drift would not predict equal numbers of nonsynonymous versus synonymous substitutions per site. How then do we interpret the finding of Plikat et al. that this is the case for nef? To us, the most reasonable explanation is that the evolutionary dynamics of this gene are shaped by the conflicting processes of negative and positive selection. The low rates of evolution at those sites that are functionally important will pull d_N/d_S to <1, while positive selection at a limited number of residues will push the ratio to >1. The result will be a fuzzy d_N/d_S ratio of around 1, which could lead to a misinterpretation of complete neutrality. That the d_S/d_N ratio is a compound measure is clear from many other studies. For example, in the classic work on adaptive evolution in the major histocompatibility complex, most sites in these sequences were subject to negative selection, with positive selection only localized to the antigen recognition site (1).

We also believe that the model of random evolution of HIV-1 proposed by Plikat et al. has its roots in some ill-conceived evolutionary theory. While we fully agree that there is a strong chance element in HIV dynamics, as there is no deterministic mechanism by which an antigen activates an HIV-infected cell, this does not mean that immune selection does not take place: once replicated, an HIV molecule will only infect another cell once it has successfully negotiated the immune response. It is at this stage that immune selection comes into play.

Evolution is a blend of both chance and necessity and it is only by considering both that we will fully understand the processes shaping genetic variation in HIV.