As you almost certainly know, animals from many groups have colonized caves, and more often than not they evolve to lose or reduce their eyes in the Stygian environment. But why? It’s hard to tell, for losing eyes takes thousands of generations, and we’re not around long enough to do experiments. I seem to recall an experiment in which Drosophila workers kept flies in the dark for years and observed no reduction of eye size, but they didn’t test their vision (this can now be done). At any rate, that experiment wasn’t long enough.

A new paper in The American Biology Teacher by Mike U. Smith, a Professor of Medical Education at Mercer University School of Medicine (Macon, Georgia), goes over the various theories for eye loss, with the piece aimed at biology teachers, suggesting how this subject should be taught and how to avoid misconceptions. (Reference below; access is free.) Smith gives three theories, but I think he gets it a bit wrong, and I wanted to give my take. I’ll ignore the stuff about teaching, as I want to concentrate on the biology.

First,  an example: Smith’s is the classic case of Astyanax mexicanus, the Mexican tetra or “blind cave fish” found in caves in the southwest U.S. and northern Mexico.  It is in fact considered the same species as its surface-dwelling form. Here are photos of each form and the range of the blind fish (from the paper). There are 26 known populations of the blind variant, representing at least five independent cases of evolutionary eye loss. Breeding experiments show that the cave and surface forms are interfertile, and that the loss of eyes in the cave form involves several genes, not just one:

The cave form eats mostly the bacteria film on the water that results from the breakdown of bat and cricket feces. The eyes are still there as vestigial remnants below the surface of the skin, but begin development as normal eyes and then regress as the fish grows up. (That itself is evidence for evolution.) Fish from at least one cave have an ability to detect light, but others have no such ability; this probably reflects different evolutionary stages of eye loss (or perhaps differential light levels in the caves). The fish find their way around via vibrations detected in their lateral lines. As Smith notes, “In fact, scientists capture these fish simply by putting a net in the water and vibrating it.”

Here are Smith’s three ideas for the evolution of eye loss. His words are indented (my emphasis). I maintain that two of the hypotheses are conflated, one is largely incorrect, and he’s neglecting another hypothesis.

According to the first hypothesis, eye loss is indeed caused by direct natural selection because there is an advantage to being eyeless in the dark. Studies have shown that maintaining eye tissue, especially the retina, and the related neural tissue comes at a high metabolic cost (Moran et al., 2015; Protas et al., 2007). Therefore, cavefish without eyes are at an advantage in this environment where energy sources (food) are scarce, because blind fish do not waste energy on these useless structures.

This is a reasonable hypothesis, and one my students used to always think of first when I asked them. It applies to the disappearance of any non-used structure, like the tiny nubbins that are the vestigial “wings” of the kiwi. The “not wasting energy”, of course, implies that that energy be directed towards other structures or functions that enhance reproduction, for that’s implicit in saying that reduces eyes give cave fish an advantage via natural selection.

A second hypothesis employs the phenomenon of pleiotropy, that is, cases in which multiple phenotypic effects are caused by the same mutation in a single gene. There is, for example, evidence that one of the genes responsible for eye loss in cavefish also increases the number of taste buds on the ventral surface of the head, which helps cavefish find food more effectively (Gross, 2012). Natural selection for this increase in taste buds would, therefore, also promote blindness.

I would argue that this second hypothesis isn’t substantively different from the first. After all, if resources are redirected from inactivated eye genes to other structures or functions that enhance reproduction, those other features would reflect pleiotropic effects of the mutations that reduced the eyes. I don’t see a material difference between a). An eye-reducing gene increasing the number of taste buds (the “pleiotropic” theory) or b). An eye-reducing mutation making more nourishment available for other structures by reducing the energy requirement for building an eye. In both cases, the mutation reducing eye formation has beneficial effects on other aspects of development. Those are both instances of “pleiotropy”.

The third hypothesis is based on neutral mutation and genetic drift. All too often textbooks use the terms “evolution” and “natural selection” interchangeably, ignoring the importance of genetic drift. Genetic drift is “the process of change in the genetic composition of a population due to chance or random events rather than to natural selection, resulting in changes in allele frequencies over time” (Biology Online, 2008). Genetic drift differs from natural selection because observed changes in allele frequency are completely at random, not the result of natural selection for a trait. Genetic drift can have a relatively larger impact on smaller populations such as a typical population of cavefish. According to the neutral mutation and genetic drift hypothesis, therefore, normal mutation processes in a small population of cavefish sometimes produce neutral mutations (mutations that lead to phenotypic changes that natural selection does not act on), and in the absence of natural selection, totally random events can sometimes result in the increased frequency of such mutations over time. Such changes could include eye degeneration.

This discussion is confusing. Even if the eye-reducing genes were neutral, and didn’t give eyeless fish a reproductive advantage, genetic drift (the random fluctuation of eyeless and eyed forms couldn’t by itself contribute to pervasive eye loss in caves, for the caves contain only fish without eyes. Drift would produce a “random” effect: varying mixtures of eyed and eyeless fishes in different caves. We don’t see that.

Now drift may play a slight role in eye loss (slightly deleterious mutations are more likely to persist in small populations), but I think what Smith is neglecting here is a non-random phenomenon: directional mutation. By that I don’t mean that somehow there is an increased frequency in the caves of mutations that inactivate eyes compared to the surface populations—that would be a Lamarckian or teleological process—but that random mutation applies to both cave and surface populations.  In surface populations those mutations that reduce or inactivate the eyes are weeded out by selection, and these mutations are more numerous than those creating better eyes. Remember that in the genes for eye formation, as in all genes, a random hit in a complex and evolved DNA sequence is more likely to damage the gene than improve its effect on reproduction.

Therefore, with a rain of mutations affecting eyes in both populations, and in general degenerating the eyes, the more numerous “bad” mutations will be selected out of the surface populations, but, with no selection against them in the cave populations, will tend to accumulate—perhaps aided by natural selection (hypotheses 1 and 2 above). Look at it this way: if you have a fleet of cars that are never driven, and people randomly adjust the engines of those cars without knowing anything about them, all the non-used cars will eventually lose their ability to run. That’s because a random adjustment of an engine is more likely to hurt it than to improve its function.  The engines are the eyes of cave fish, and the adjustments are mutations. The adjustments accumulate because the cars don’t need to run. I think this is a more plausible explanation than simple genetic drift, which seems implausible anyway because eye-reducing mutations aren’t likely to be “neutral”, for reasons given above but also because of what I say just in my fourth hypothesis below.

Smith says this:

. . . studies of the sequences of other genes related to the cavefish eye show high frequencies of substitutions in both coding and noncoding regions, which would support the genetic drift hypothesis (Retaux & Casane, 2013).

But that seems to be wrong for several reasons.  First a high frequency of substitutions in coding regions can be due to any of the forms of natural selection discussed above. Second, non-coding regions (parts of the DNA that do not code for proteins) can sometimes affect gene expression and regulation. More important, I couldn’t find any data in the Retaux and Casane paper suggesting an increased frequency of truly neutral non-coding mutations in these cavefish. (I may have missed it, but it doesn’t seem to be there.) What I see is this paragraph (note: this is for evolutionary geneticists)

The reports cited above only concern the evolution of the coding sequences. However, phenotypic evolution (including the loss of structures) can also occur through changes in non-coding, cis-regulatory sequences. Famous examples include the loss of the pelvic spine in freshwater sticklebacks through deletion of a Pitx1 enhancer [98, 99], or gain or loss of pigmentation patterns in Drosophilae through co-option or mutation of regulatory elements in the pigmentation gene yellow [100]. Although the exact mechanism is unknown, this happened for crystallin αA in cave Astyanax [55, 101]. This chaperone and anti-apoptotic crystallin whose coding sequence is almost identical in surface fish and cavefish (one amino-acid difference only) is strongly downregulated in the cavefish lens during development and was suggested as a potential major player in the onset of cavefish lens apoptosis. In the naked mole rat Heterocephalus glaber, gamma-crystallins are turned off after birth [46]. In the mole rat Spalax ehrenbergi, the αB-crystallin promoter and intergenic regions have selectively lost lens activity after 13.5 days of embryogenesis [102, 103]. These examples show that changes in regulatory sequences also occurred in cave and other underground animals.

Note that there are no data here on “high frequencies of substitutions in noncoding regions” of cavefish eyes. We see a change in gene regulation without accompanying changes in the sequence of the regulated genes, but that’s probably due to “coding” changes in other regulatory genes or substitutions in regulatory regions that are not “neutral” because they affect eye formation. (Note that Smith emphasizes “neutral” mutations in his third hypothesis.) These regulatory regions are thus subject to natural selection, and are not “neutral” changes acted on solely by genetic drift, even if they’re noncoding. We would in fact expect that selection would produce that observation: more substitutions accumulating in regulatory regions in cave fish than in surface fish! No need for drift here.

Coyne’s fourth hypothesis (not really mine but neglected by Smith). Eyes are delicate organs, easily damaged and prone to infection. If you reduce the eyes when you don’t need them, you’re less prone to this kind of environmental damage, and so the genes reducing the eyes make their bearers more likely to live and reproduce. Yes, this is a form of eye loss promoted by selection, but is conceptually different from hypotheses 1 and 2 above. I wish Smith had mentioned this idea as well.

At the end, Smith says that all his suggested processes might act together: