Cryptic Divergence Occurs When Compensatory Changes Occur within Genes

How frequent are the incompatibilities assumed by the Dobzhansky–Muller model? Kondrashov et al. (2002) have addressed this question by asking how often an amino acid that is present in one species would be deleterious when present in another. They identified 32 well-studied proteins, for each of which at least 50 different mutations are known that cause severe defects in humans. Although proteins that mutate to cause human disease tend to evolve more slowly than average, reflecting greater functional constraint, they nevertheless differ extensively in amino acid sequence from their homologs in other species (ranging from primates to invertebrates).

Remarkably, approximately 10% of these amino acid differences that have evolved in other species would cause severe defects if present in humans. These potentially deleterious effects must have been compensated by other changes. (Such substitutions are called compensated pathogenic deviations [CPDs].) Still more surprising, this fraction does not increase with evolutionary divergence. Even among primates (Fig. 22.9A), about 10% of amino acid substitutions would cause disease if present in humans (Fig. 22.9B).

At first sight, the high proportion of compensated pathogenic deviations conflicts with the slow rate of speciation. We have seen that species can differ at tens of thousands of amino acids and yet still produce viable and fertile hybrids. Kondrashov and colleagues argue that CPDs are almost always compensated by changes within the same gene. Thus, hybrids would suffer only if intragenic recombination separates the two coadapted amino acids—an extremely rare event. They identify candidates for the compensatory changes by looking for amino acids that always accompany the CPDs and that occur in the same part of the protein (e.g., Fig. 22.9B).

The scenario here is that a substitution occurs at one site, because of either selection or drift. That makes change at a second site selectively advantageous, and so this change occurs soon afterward. We have, indeed, already seen evidence for this kind of compensatory evolution in RNA sequences, where the need for complementary base pairing leads to the same strong selective interactions as those proposed in this example (pp. 544–546).

These observations do not directly address the cause of reproductive isolation. These tightly linked changes cause negligible fitness loci in hybrids because they are rarely separated by recombination. However, if compensatory substitutions were to occur in different genes, then reproductive isolation would result, just as in the Dobzhansky–Muller model. (It is the short interval between the two substitutions that explains why the fraction of CPD is so high, even between closely related species [Fig. WN22.9A]. If pairs of compensated changes occur together, and about 10% of substitutions consist of such paired substitutions, then we expect to see the same fraction of CPD regardless of relatedness.)