Evidence for evolution: Hairless animals have dead genes for a full coat of hair

January 15, 2023 • 9:30 am

In Why Evolution is True, one of the most telling pieces of evidence I adduce for evolution is the existence of dead (nonfunctional or “vestigial”) genes found in the DNA in living species. For example, mammals like us carry three dead genes for making egg yolk. Evolution has rendered them nonfunctional, as mammalian embryos are nourished through the placenta, but they’re still there in the genome, rendered useless by mutations.

The genome of nearly all animals we know is a veritable graveyard of dead genes. These, like our egg-yolk genes, constitute irrefutable evidence for evolution. They’re still there because we inherited them from a common ancestor, but evolution usually inactivates unneeded genes not by snipping them out of the genome, but by allowing “inactivation” mutations to kill the genes’s production of protein. (Alternatively, inactivation can occur by killing off a promoter gene that causes a gene to be transcribed.) The genes just sit there, “silent signs of history.”

Both types of dead genes were found in this new eLife paper, and they are genes normally promoting the growth hair in the relatives of species that have lost most of their hair. That the genes are still there, but are nonfunctional, simply can’t be explained by anything other than common ancestry. That’s why creationists, like the chowderhead I’ll highlight in the next post, ignore them. (Similarly, they ignore the evidence for biogeography of oceanic islands—also explained in Why Evolution is True—because there’s no creationist explanation save “Well, God wanted things to look like they’d evolved.”)

This is a long and complicated paper from eLife, but the popular version in Science Alert, shown below that, is not sufficiently detailed. I’ll try to simplify the eLife paper but give more information than the popular precis.

The pdf for the eLife paper is here, and the reference is at the bottom.

The short take: the authors sequence a handful of relatively hairless species that descended from ancestors that had hair, looking for genes in common to these set that a. were likely involved in producing hair, but b. had been inactivated in these species by “relaxation of selection”. That is, there was no longer natural selection in these species to maintain a coat of hair (and good reason not to), and so mutations inactivating the genes–and their controlling elements–accumulated.  Further, natural selection can accelerate this trend by favoring gene variants that reduce hair, either because regular genes that make hair use up metabolic energy that isn’t needed and, more likely, that hair is an impediment to their lifestyle.

The interesting thing about the paper is that, by sequencing the DNA of relatively hairless species, they found sets of genes in common among the hairless species, implying that there were common evolutionary-genetic pathways for hair reduction. This is what’s called convergent evolution, which usually refers to similar appearances of organisms that have similar lifestyles but aren’t closely related—like the marsupial mole and the placental mole—but in this case it’s convergence at the level of genes.

Here are the species they looked at:

naked mole rat

. . and a subset of all species studied showing their evolutionary relatedness. I love the example they use for humans:

(from the paper): Hairless species show an enrichment of hair-related genes and noncoding elements whose evolutionary rates are significantly associated with phenotype evolution. (A) Phylogenetic tree showing a subset of the 62 mammal species used for analyses. Note that all 62 species were included in analyses and only a subset are shown here for visualization purposes. Foreground branches representing the hairless phenotype are depicted in orange alongside photographs of the species.

Most of these animals have some hair, but the authors conjecture, with reason, that their ancestors were much hairier. This is likely to be true, though the elephant and manatee had a recent common ancestor and it’s not clear whether their hairlessness evolved twice. The authors support this by adducing the existence of the hairy mammoths as animals more closely related to the modern elephant, implying that the ancestral pachyderm was hairy. But hairy elephants could have represented the re-evolution of hair in a relatively hairless ancestor. Likewise with the dolphins and orcas; I’m not sure how these two, which are fairly closely related marine mammals, could be taken as independent losses of hair. (On the other hand, the walrus, less closely related, could have lost its hair independently).

They also looked at 52 other species, for you need a comparison of DNA sequences in hairy animals. The figure above shows some of the hairy species whose DNA was sequenced (they looked at a lot of genome: nearly 20,000 coding genes and 350,000 regulatory regions).

Surprisingly, they found a fair number of genes that lost function (or had “relaxed selection”) in all of the hairless species. Not all of the genes had a known function, but most were associated with hairs themselves, the hair follicles,or the dermal papillae, the crucial structures that allow hair to grow.  Here’s a list of five genes and a table of the likelihood that they would have changed so rapidly in all the species. The colors show where the genes act.

(From the paper): Diagram of hair shaft and follicle with shading representing region-specific enrichment for coding and noncoding sequence. Both coding and noncoding sequence demonstrate accelerated evolution of elements related to hair shaft (cortex, cuticle, and medulla). Noncoding regions demonstrate accelerated evolution of matrix and dermal papilla elements not observed in coding sequence. All compartment genesets were compiled from Mouse Genome Informatics (MGI) annotations that contained the name of the compartment except the arrector pili geneset (Santos et al., 2015).

Note that both coding (genes) and noncoding (controlling-element) DNA was involved; in fact, among all the genes identified as likely contributors to hairlessness, there were more noncoding changes than coding changes, which is often what we find when either new structures evolve or old structures are lost. I used to think—and wrote a controversial paper about this with Hopi Hoekstra—that structural (coding) genes were more important in evolutionary change, but the data show that it might be the other way around. In other words, Hopi and I may have been wrong.

Now the paper is long and complicated, and bits of it are beyond my pay grade, but I do have a few comments. First, the significance levels they use to ascertain common evolution of genes among the set of relatively hairless species are not that small. They even highlight genes, as you can see above, with adjusted probability values above 0.05; conventionally these would be considered “nonsignificant”. I’m not sure why they did that. However, as you can see from the table above, some of the adjusted probabilities were very, very small: the p value for noncoding sequences in the hair cortex is, for instance, 0.000003. I’m confident that they did at least find some genes that changed rapidly in the entire group of hairless species.

Second, they’re not sure in some cases that the rapid gene evolution was indeed associated with loss of gene function. You can tell this for coding genes because there will be a “stop codon” or a “nonsense codon” in the DNA sequence that will code for mRNA that makes a nonfunctional protein.  They don’t talk about this in detail, but simply use “rapid evolution” as an index of nonfunctionality. (I may have missed something.)

Finally, for pairs like the elephant and manatee on one hand and the dolphin and orca on the other, I don’t have a lot of confidence that their loss of hair occurred independently.

Nevertheless, we can have confidence, given the low probability values, that some structural and controlling DNA has evolved independently in a group of hairless species, causing them to lose hair. That’s a case of convergent evolution of genes that is quite novel.

Oh, I forgot to mention why these species lost hair. In our species, it probably happened to promote easier cooling of our bodies via sweating as we evolved into upright creatures on the savanna. This probably also holds for rhinos and elephants, especially because elephantine species in northern climes, like mammoths, were hairy. In marine mammals it’s obvious: hair is useless for insulation, and is just an impediment to swimming. As for armadillos and pigs, it’s anybody’s guess. Wild pigs are pretty hairy (at least the ones I’ve seen), but armadillos have shells, and that serves to insulate the animal (they do have hair on their bellies, but it’s sparse).

An armadillo’s belly from Flickr:


THE UPSHOT:  These are likely cases of “vestigial genes,” though the cases would become textbook examples if they knew exactly what the genes did and, importantly, could show without doubt that they have been inactivated in the hairless species. Those data will come some day, but in the meantime I prefer to cite the broken egg-yolk genes in mammals: remnants of genes that produced nutrients for the embryos in our reptilian, fishy, and amphibian ancestors. That is a very solid case.

You can read the “popular” take below:

h/t: Barry


Kowalczyk, A., M. Chikina, and N. Clark. 2022.  Complementary evolution of coding and noncoding sequence underlies mammalian hairlessness. eLife 11:e76911https://doi.org/10.7554/eLife.76911

41 thoughts on “Evidence for evolution: Hairless animals have dead genes for a full coat of hair

  1. I cannot help but think of the differences in humans, as in why some people have so little body hair, or go bald very young, or like myself, have more than their fair share of “fur” (as my son used to call it) on the arms, legs, chest, and *ahem* other places, or one man in particular I knew, whose back hair erupted out the collar of his shirt like a squirrel trying to escape up his neck.

  2. Very nice post. Thanks!
    If pigs are included (I assume hair loss was lost during domestication) one could just as well include hairless breeds of rabbits, d**s and cats, no? Of course, at least in d**s the genetic basis for hair loss is well characterized already…

    1. At least one, Jeff Tomkins (of ICR), starts by asserting that the vitellogenin pseudogene is actually a functional enhancer of a gene involved with neurons (IIRC).

      His analysis was horribly flawed and distorted (synteny), ignored (pseudogenes get repurposed so having function means little), or exaggerated (significance of ubiquitous, low level transcripts of long noncoding RNAs) everything that could be twisted to support his creationist belief.

      Paper likely sounded great to anyone with a similar belief system and not enough technical knowledge. And from there, not a big leap for them to accept the early ENCODE hype of virtually all DNA being functional (“science has proven gods perfection again” type stuff).

      1. So are you saying that the ICR article is wrong when it states:
        “the vtg pseudogene is the presence of a 150-base human DNA sequence that shares a low level of similarity (62%) to a tiny portion of the chicken vitellogenin (vtg1) gene 8”

        The chicken vtg1 gene is 42,637 bases long, so the evidence is a 150-base fragment of 62% similarity seems rather inconclusive.

        1. As I recall the 150-base sequence portion was very misleading in that it only discussed that one fragment, leaving the impression there were no other fragments to be found (which there definitely are).
          So yes, the ICR article by Tompkins is wrong that the 150-base sequence is the only remnant of the gene present. The other remnants are neither mentioned nor addressed.

          1. Could you give a link to the publication with this information.
            Failing that what size fragments and what percentage similarity.
            I wonder what the probably of getting a 62% match for a 150 base pair sequence in a sequence of 42000 bp selected at random?

  3. Aka pseudogenes. In the American Chestnut genome, a gene called Germin exists that is strongly homologous to wheat oxalate oxidase – something like (by memory here) 70% positional identity. There are some indels (insertion/deletions) but the three catalytic histidine residues are present and in register, but expression of the gene gives a product with no oxalate oxidase activity.

    So why is this noteworthy? The current effort in the long history of efforts to rescue the tree from functional extinction by the blight fungus that kills the tree by secreting oxalic acid is by genetic modification. The gene for wheat oxalate oxidase has been successfully spliced into the chestnut genome, and the results are highly encouraging. The tree is referred to as “Darling 58” The USDA, EPA and FDA need to weigh in on whether the tree can be released. USDA has issued a favorable draft and EPA has concurred. The USDA decision is currently pending a public comment period that ends Jan 25 or so.

    Most people who the tree means something to are on board, but of course the anti-GMO cabal is apoplectic, wailing among other things that it will extinct the native tree (Hello? It’s already functionally extinct.) Lots of experimentation had been done showing that the GMO tree poses no risks. My argument has been that the tree already had the enzyme in earlier evolution but it was lost as being superfluous, but now it’s needed.

    This is of course very brief, but if you would care to send a favorable comment to the USDA, here is the link.

    1. FWIW, here’s the comment I submitted, which gives the correct identity/similarity percentages, and also notes the analogy in us to our deficiency in the Vitamin C synthetic pathway, also highlighting some of the other issues in the arguments, too.

        1. Yep, that argument hadn’t occurred to the folks at SUNY-ESF, but they were happy for it when I mentioned it to them. I knew of it from long back, but permanent residence in my head was probably boosted by WEIT.

    2. I have made a favorable comment, for what it’s worth. I of course lack the education and intelligence to make a truly mean full comment, but I made one just the same. I wish I had had your example comment to draw from but hopefully those from the great unwashed such as I are also worthwhile. Considering that around 33 million hectares of GM corn are planted in the US annually (as far as I could tell from a quick search) I really don’t see what all the fuss is about, except that it proves yet again that “money talks”. If the chestnut had the weight of agribusiness behind it, they’d be planted in a heartbeat.

      1. Thanks! Many comments in favor are simply things like “Just do it.” and “What are we waiting for?” And as further evidence that agrobusiness is not behind the chestnut effort, Darling 58 is expressly NOT patented.

        To see other comments, just advance the number at the end of my link. (To considerable irritation, there is no NEXT button on the pages.) The last number is currently 15472. There’s a raft of boilerplate negatives between 15000-15100 before things settle back into people personally enthusiastic by the effort.

    3. I can’t imagine why putting a gene back, especially what I guess is some kind of ordinary immune defense gene, could possibly have negative repercussions. The anti-GMO crowd there seem more like the sort that would wear tinfoil hats.

      1. They’re like a religious cult, mostly posting the same lame boilerplate, and those posts come in waves. I think they have little wine & cheese parties – everyone that comes gets some crackers, brie and chardonnay after they go over to the laptop and post the boilerplate.

        The gene is for an enzyme that converts oxalic acid to CO2 and H2O2. Oxalic acid is what the fungus secretes, essentially as a metabolic waste product I think, and the tree can’t tolerate that in its cambium layer, so the fungus basically girdles the tree. However, fortunately, it can’t survive below ground so for many decades, at least, the roots can send up new shoots from the root collar. Sometimes these survive for a few years after nut-bearing age before getting hit by the fungus again, which is what has allowed for the preservation of the germplasm.

        1. I think you and I discussed this before – going from hazy memory :

          Rhubarb, maybe another plant, oxalic acid crystals _in_the_plant_, as a defense mechanism…. for something…

          1. Yep, rhubarb and spinach – high in oxalic acid. Many more plants/foods are relatively high in it, including, sadly, tea & chocolate. And most kidney stones are composed of calcium oxalate.

  4. Somehow I think baldness illustrates the illusion of Free Will : they have the genes, so why can’t Free Will conjure up the hair? Oh, Free Will doesn’t work that way. Ah.

  5. Great post– I’ll have to read the paper now! Four comments, the first three of which point out areas for future research:

    1. The killer whale and the dolphin are not only in the same (sub)order (the whales), all of which are essentially hairless, but they’re even in the same family– the chances of the two having evolved hairlessness independently is zilch.
    2. Jerry is right that wild boar are well-haired. Hair reduction must have evolved recently in domestic pigs.
    3. In humans, the persistence of thick hair on various parts of the body raises interesting questions of the genetic and developmental basis of this within-body variation. (There is regional variation in other species, too.)
    4. Also in humans, the substantial geographic variation in hairiness raises interesting microevolutionary questions.


    1. Maybe I am wrong, or misremembering, but don’t domestic pigs, when they’ve gone feral, resume their hirsute ways? I don’t know by what genetic mechanisms would account for this (but then what I understand of genetics couldn’t fill a thimble) Or I could be wrong, so take my comment as just another goober from the peanut gallery.

      1. Good point. Feral domestic pigs do resume their hirsute ways– think “Arkansas Razorbacks”. A few possible explanations, all needing further study: i) the particular (and early) breeds of hog that have gone wild were not so hairless to begin with; ii) rapid re-evolution of hairiness (turning the hairy genes back on); iii) phenotypic plasticity (i.e. developmental, but not genetic, changes).


    2. I am curious as to why Jerry is confident that hair loss in killer whales and dolphins (and probably other cetaceans?) did not happen independently. It does seem much more probable that the two emerged from common ancestors that already had lost hair or were in the process of doing so, but it must also be possible that the modern dolphins and killers did indeed complete their haircuts down independent evolutionary paths – as perhaps false killers, belugas etc did too. Convergent evolution or perhaps a mix of convergent evolution and sharing common ancestors that were already part-way through their haircuts?
      David Lillis

      1. Since every descendant of the ancestral cetacean is hairless (see here), it’s reasonable to assume that the ancestor was hairless too. You could be right, and say that all those existing species lost their hair, but then you shouldn’t use phylogenies as the authors did–you should look at the DNA of EVERY cetacean. The conservative thing to do in a paper like this is to assume that if every descendant of a common ancestor has a trait, then the ancestor also had that trait.

      2. Sorry to pose another basic question – to which I do not demand an answer from busy people. For a non-biologist (like myself) the question is whether any evolutionary advantage of hair loss (e.g. more efficient locomotion) was significant enough to play out from generation to generation and not masked by other factors that might favour the presence of body hair (e.g. insulation)? OK – I guess that the same question could be asked of any evolutionary process.

        Let’s assume divergence of two species from a common ancestor over a period of ten million years and that each generation takes up ten years. That’s one million generations for the effect to play out. That number of generations (or much more) is surely(?) enough to complete the haircut, but the process can only play out to the extent of complete loss of hair if the effect is significant enough and not masked by other factors that favour hair.

        Another one from a non-biologist – was hair loss likely to have resulted from direct advantage across generations or because the genes that produce body hair are not being actively selected for? May be those possibilities are actually the same thing?!

        1. Looking back – I did not phrase my last post very well. We know that the cetaceans lost hair but was it locomotion that did it? For a non-biologist like me, it seems that various factors could have been at play and that several could have been competing over the hundreds of thousands of generations; e.g. locomotion vs. insulation. We know that there was an endgame (a haircut) but was the tiny evolutionary advantage of a very small improvement in locomotion enough? I guess so.

          For example, evolution of shark teeth from mako-like sharks about ten million years ago to those of the great white of today was also a very gradual process. See:


          Other evolutionary processes must have been very gradual too (taking millions of years) but yet others relatively rapid (just a few generations).

  6. Very interesting. Much of it is above my pay grade, but I would appreciate a couple gene mapping pictures that side by side compare one of the genes for hairs in a hairy mammal to the same, putatively dead gene in a related hairless mammal. Then we ordinary folks could see if the latter gene is really and truly dead.

  7. Jerry you said: “They even highlight genes, as you can see above, with adjusted probability values above 0.05; conventionally these would be considered “nonsignificant”. I’m not sure why they did that.”

    A 2019 special issue in The American Statistician covers the controversy over the use of p < 0.05 in scientific research. The editorial Moving to a World Beyond “p < 0.05” is worth reading:
    Ronald L. Wasserstein, Allen L. Schirm & Nicole A. Lazar (2019)
    Moving to a World Beyond “p < 0.05”, The American Statistician, 73:sup1, 1-19, DOI:

    1. Even though this paper by Wasserstein et al came out after I retired, the theme of getting rid of the tyranny of p values is well taken. And yes, even confidence intervals don’t answer completely.

      There was an increasing tendency in the biomedical literature to report the results of statistical inference testing as what the p value actually is, instead of just saying “p < 0.05" or "p = N.S." Modern computer packages easily allow reporting of whatever the p value comes out as. When I took stats as an undergrad, by the time you had worked out with pencil and paper and tables of square roots what your chi-square or t-test test statistic was, you just wanted to know if the p associated with it it was 0.05, it seems to honour the “ATOM” spirt of Open-ness (as advocated by Wasserstein et al) to report the p value of every comparison you made.

      There are two reasons to do this:

      1) It defends you against accusations of occult cherry-picking, that you reported only the comparisons with p values less than the null rejection criterion.

      2) It discloses that you made multiple comparisons and therefore had to consider that some comparisons might give p < 0.05 simply on the basis of chance alone and, in the language condemned by Wasserstein, "reject the null hypothesis" falsely. After all, at p = 0.05, 1 in 20 will, on average. People gamble on odds of 19:1, if the payoff is 19:1 or better. (Pro tip: it never is.) For people for whom this paper is well within their pay grade, the appropriate correction method for multiple comparisons will be obvious. To me, a p value in the range of 10^-6 with multiple comparisons is pretty darn good for biology. Not bad even for physics. But then, genetics is almost physics.

      1. Something (probably me) made a hash of that second paragraph. What I wanted to say was,
        ” . . .you just wanted to know if the p associated with it was less than 0.05. If it was greater, working out what it actually was (0.09, 0.23, 0.72) was too late at night. Yet there is insight to be gained from “non-significant” p-values as later work on the correct interpretation of “negative” studies made clear, particularly sample size. It seems to honour the “ATOM” spirit . . .”


  8. Looking at the diagrams, I’m always amazed at how something like a hair has so many working parts (and of course, all the parts are made-up of thousands/millions? of specialized cells). It takes all that for a strand of hair to grow and function. Wha? Where would we be without microscopes in studying life. Nowhere, it seems.

    I know armadillos are leprosy carriers, but I find them adorable. Don’t think I’ve seen their somewhat hairy belly, which is another boost of endearment.

    Thanks for another riveting post explaining an aspect of life here on Earth that once again answers Why Evolution is True.

  9. I’m sure I’m not the only person thinking about how we could be less depending on Russian natural gas if we could turn these genes on!

Leave a Reply