A genomic and evolutionary analysis of an extinct saber-toothed cat

A recent article in Current Biology, which you should be able get for free by clicking on the screenshot below, describes sequencing the entire genome of an extinct saber-toothed cat, thereby gaining some insight into its evolutionary history. (You can get the pdf here, and the full reference is at the bottom. If you can’t see the piece, make a judicious inquiry.)

The cat is Homotherium latidens, also known as the European saber-toothed cat (it’s also called a “scimitar-toothed cat” because its teeth were smaller than true sabertooths like Smilodon), and it probably lived from a few million years ago until fairly recently (the late Pleistocene, about 12,000 years ago). It may thus have encountered modern humans. It was about the size of a male African lion, and a reconstruction from Prehistoric Fauna looks like this (note the saber teeth and very short bobtail).

The species was also widespread: as the article below notes, “it once spanned from southern Africa, across Eurasia and North America, to South America, arguably the largest geographical range of all the saber-toothed cats.” Although it was clearly a hunter, like all sabertooths, we know nothing about its social life, or whether it was social—nor about whether it hunted by day, night, or twilight (“crepuscular”). Some of these issues were addressed by the authors using the DNA sequence.

Click to read:

The data come from a single specimen found in the Yukon and in the possession of the Yukon Government paleontology program. A section of humerus was used for dating (the fossil was about 47,000 years old), and then was crushed up to extract DNA. The authors were able to get a substantial amount of sequence from the bone, and they compared that sequence to DNA taken from living lions, sand cats, fishing cats, leopard cats, and caracals.

The conclusions about where this cat fits on the evolutionary tree of felids are pretty sound, based as they are on lots of DNA sequence. However, the conclusions about what genes may have propelled its evolution are a lot more speculative. Here are the main conclusions.

1.) The species was long diverged from the lineage that led to modern cats.  The lineage that led to this species diverged from that of all living cats a long time ago: about 22.5 million years. We knew of this substantial age from mitochondrial DNA sequencing in previous work, but it’s dicey to make conclusions about family trees from mitochondrial DNA alone. The date above is one that we can rely on, though, as it’s based on DNA divergence in the whole genome that’s been calibrated from the fossil record.

Here’s the deduced phylogeny, showing where H. latidens (in red) fits in with 17 cats and two hyenas. You can see that it diverged from living cats over 20 million years ago.

2.) There doesn’t seem to have been much hybridization between this species and the ancestors of living cats, which began diverging from each other about 14 million years ago (see phylogeny above). The authors could have detected such hybridization by finding sections of the genome that were discordant in divergence from living cats—perhaps sections of DNA that got into the saber-toothed tiger from species after the divergence of modern cats about 14 million years ago. That is, most genes would show similar amounts of divergence from the same genes in the modern-cat lineage, but a few would be much less diverged, suggesting that those genes got into the H. latidens genome after hybridization with cats that diverged much later.

They didn’t find any such discordance, suggesting that H. latidens simply didn’t hybridize with cats that evolved in the last 14 million years. For some reason this absence caused a lot of consternation for the authors.  I guess they expected to find some evidence of hybridization and “introgression” (transfer of genes between species after speciation had occurred), and they go on at great length to speculate about this absence. They mention things like low population density (so members of different species don’t meet), ecological or behavioral isolation, and so on. But the most obvious possibility, which they don’t mention, is simply that speciation between the scimitar cat and its relatives had been completed by the time they encountered each other, so that no gene flow was possible. Yes, sometimes reproductive barriers are complete, as they are now between our own species and every other species on the planet.  And this is true for lots of species. Just because hybridization is more common than we thought doesn’t mean that nearly every species occasionally exchanges genes with others.

3.) The authors found genes in the H. latidens genome that apparently underwent natural selection. The way geneticists judge this is to look for which regions of a gene have changed relative to the genes of its relatives. This is expressed in what’s called the dN/dS ratio. That ratio gives the frequency of evolutionary changes in “non-silent” parts of proteins (dN: those parts where a mutation changes the protein sequence of the gene) to the changes in “silent” parts of genes (dS: those parts where a mutation is in a noncoding part of the gene or in a third position of a “codon”, where a mutation doesn’t usually change the protein sequence).

If genes just change randomly, without selection, this ratio should be about one. If the ratio is higher than one, protein sequences are changing faster than they would under a “neutral” process in which no changes in the gene alter its effect on reproduction (“fitness”). The authors used a cutoff ratio of dN/dS of between 2 and 5 as a criterion for selection, and they found 31 genes in this range out of the 2,191 analyzed. Eighteen of these genes, potential targets for selection in this cat, are shown in the diagram below (They don’t mention what the other 13 genes do.)

You can see they fall into four general classes, and into subclasses as well, like genes affecting vision fitting into the the “diurnal” class. The authors note that while dN/dS ratios are only suggestions of what genes in the lineage of this species may have been subject to positive natural selection, they do speculate at length about the form of selection. The cats, for example, could have been selected to adapt to daylight hunting (as opposed to most cats), with consequently improved vision. Selection on “endurance” genes may have facilitated “cursorial” hunting (i.e., running down prey). And there may have been positive selection on genes known to involve social behavior—in mice. From that they speculate that this cat may have become more social and thus able to hunt down big prey in groups.

I call this kind of speculation “genomic sociobiology”, because it involves making up “just so” stories about how genetic change impacted an extinct creature. It’s fine to single out genes like this for further examination, but one has to realize that if you see selection acting on a gene affecting social behavior, for instance, it could be reducing social behavior instead of increasing it. How do we know that the ancestors of this cat weren’t social, but then there was selection on those genes to reduce sociality in favor of a more solitary lifestyle? Ditto for all the other genes. That is, showing selection itself, even if these ratios do show selection, doesn’t mean you know the direction of selection. In fact, some media outlets, like this one, have bought uncritically the notion that this study has revealed that the cat evolved to become more social.

4.) This individual, and thus its species, was very genetically diverse. That is, if you looked at the two copies of a gene in the H. latidens genome—remember, we all carry two copies of nearly all our genes except for those on mitochondria and sex chromosomes in the heterogametic sex—there was a high probability that they would be different. This “heterozygosity” would not be the case if the species were in small populations that would lose genetic variation, or in an inbred species. We can conclude that the species was genetically diverse—no surprise given how wide ranging it was.

As to why H. latidens went extinct, well, we just don’t know. Given its genetic diversity, it probably wasn’t inbreeding, and could have been stuff like competition with cats that were better hunters, a disease or parasite, climate change, or any number of things.

Overall, this is a decent paper, and a good one insofar as doing whole-genome sequencing and phylogenetic analysis of a long-extinct species. The conclusions about natural selection are speculative, and the authors realize that. If there’s a flaw in the paper, I think it’s that the authors do go on too long with the natural selection business, especially given that it’s purely guesswork based on ratios of substitutions in DNA, and because we’re totally ignorant about what these genetic changes really meant for the evolution of these cats.

Oh, and I’m disappointed that they didn’t see positive selection in “tooth genes”!


Barnett, R., M. V. Westbury, M. Sandoval-Velasco, F. G. Vieira, S. Jeon, G. Zazula, M. D. Martin, S. Y. W. Ho, N. Mather, S. Gopalakrishnan, J. Ramos-Madrigal, M. de Manuel, M. L. Zepeda-Mendoza, A. Antunes, A. C. Baez, B. De Cahsan, G. Larson, S. J. O’Brien, E. Eizirik, W. E. Johnson, K.-P. Koepfli, A. Wilting, J. Fickel, L. Dalén, E. D. Lorenzen, T. Marques-Bonet, A. J. Hansen, G. Zhang, J. Bhak, N. Yamaguchi, and M. T. P. Gilbert. 2020. Genomic Adaptations and Evolutionary History of the Extinct Scimitar-Toothed Cat, Homotherium latidens. Current Biology.


  1. ThyroidPlanet
    Posted October 18, 2020 at 1:08 pm | Permalink


  2. GBJames
    Posted October 18, 2020 at 1:14 pm | Permalink


  3. cyan
    Posted October 18, 2020 at 1:16 pm | Permalink

    Why are you disappointed that they didn’t see positive selection in “tooth genes”?

    • Posted October 18, 2020 at 1:23 pm | Permalink

      That was a joke! It’s funny because the canine teeth of these cats clearly got bigger compared to some ancestor, but they couldn’t find a signal of that in the DNA. One wouldn’t, actually, given that almost all character changes in evolution are small and probably not detectable using dN/dS ratios of genes.

      It was nerdy humor.

      • jezgrove
        Posted October 18, 2020 at 3:04 pm | Permalink

        A cat with canines! (My stupid and not at all nerdy humo(u)r, sadly.)

      • dom
        Posted October 19, 2020 at 3:41 am | Permalink

        They do not however have Smilodon DNA presumably to compare? They presumably lie closer to the Homotherium than the other cats?

        I do not see that disease or a parasite would be responsible for their extinction. Are there any solid examples of that in any species? Why not lions also? If they had a world-wide distribution, excepting Australia, their chief competitors would be early humans & their dogs. I suggest, in a just-so speculation (& why not as a thought experiment offer those ideas to find a way to a rational explanation?) we probably persecuted them or drove them off their kills. Reduced numbers would make them less effective hunters. If they are right & there are lots of Homotherium fossils hiding in collections under the guise of sabre-toothed cats, that is really interesting.

    • Mike
      Posted October 18, 2020 at 3:09 pm | Permalink

      Jerry is right that it’s a bit of a joke. The ratio dN/dS is a powerful way to detect selection among species – in some ways too powerful because the models detect many false positive signals that are probably not caused by selection.

      In addition to the other weaknesses that Jerry pointed out, the dN/dS ratio is only applicable to the protein-coding sequence of a gene. Lots of adaptive evolution happens in the regulatory regions of genes that affect how much or in which cells the protein is expressed. That’s a likely kind of selection (on gene expression) that leads to larger teeth in saber-toothed cats, but it wouldn’t show up in dN/dS comparisons of protein-coding genes.

      Cool paper, thanks Jerry for the summary.

      • Posted October 19, 2020 at 3:44 am | Permalink

        If we know what genes do what we could experiment with activating such genes, breeding social domestic cats that would hunt dogs in packs!?

        • Posted October 19, 2020 at 7:26 am | Permalink

          Sabre-toothed domestic cats would be really cool 😎

          • gravelinspector-Aidan
            Posted October 19, 2020 at 3:58 pm | Permalink

            Being familiar with the lesser-spotted sabre-tooth midge, I am much less sanguine about how well “sabre-tooth” traits and “domestic” would play together.

  4. cicely berglund
    Posted October 18, 2020 at 1:21 pm | Permalink

    Thank you for a through science post.

  5. Neil Taylor
    Posted October 18, 2020 at 1:25 pm | Permalink

    Great science post. Thank you.

  6. rickflick
    Posted October 18, 2020 at 2:10 pm | Permalink

    I came across this video on Panthera spelaea, the species that was depicted on cave walls in Europe. Several frozen specimens have been found and are being genetically studied.


  7. Steve Gerrard
    Posted October 18, 2020 at 2:11 pm | Permalink

    “known to involve social behavior—in mice”

    There is a Twitter account called @justsaysinmice, which points out all the science reports that neglect to mention in their title that the results are in mice. Sometimes it’s a bit hilarious.

    The point about not knowing the direction of selection is also really important and often overlooked.

  8. Steve Pollard
    Posted October 18, 2020 at 3:36 pm | Permalink

    Masterful. So many thanks.

  9. jbussen
    Posted October 18, 2020 at 3:39 pm | Permalink

    Thanks, rick flick, for the link to the Ben G Thomas cave lion site. Note that it ends with a plug for a trade book, Missing Lynx, about prehistoric animals of Britain, by Ross Barnett. And the lead author of the paper under discussion here is…R. Barnett.

    • rickflick
      Posted October 18, 2020 at 6:13 pm | Permalink

      Thanks for pointing that out. It’s a small world.

  10. alan
    Posted October 18, 2020 at 4:12 pm | Permalink

    They didn’t do a GO analysis on the set of 31? (haven’t checked) I know enrichment of GO terms also gets labeled as “just so” stories (genomic sociobiology is a great term!), but it’s less biased than selective assessment of a subset of genes only. The times I’ve done GO analysis, it made sense, from differentially expressed transcripts in axolotl blastemas to stickleback in temporally varying bar-built estuaries.

  11. Posted October 18, 2020 at 4:54 pm | Permalink

    Homotheres have upper canines far wider, front to back, thinner side to side, sharper-edged and better serrated (saw edge) than those in smilodontines. Front legs far longer and more compact — rather cheetah like, as opposed to short, flexed, jaguar-style smilodontine limbs. Claws were reduced, another cheetah resemblance. Here in Texas, at the State Museumin Austin, we have inferred male and female and cubs from a cave. Mingled w the cat bones were dozens of juvenile mammoths, evidence of prey. How did the cats capture the calves without being attacked by mammoth moms? One wonders.

  12. ladyatheist
    Posted October 18, 2020 at 6:54 pm | Permalink


  13. Posted October 19, 2020 at 5:06 am | Permalink

    Reblogged this on silverapplequeen and commented:
    This is VERY interesting. Take a read!

  14. Charles A Sawicki
    Posted October 19, 2020 at 11:49 am | Permalink

    Nice post!

  15. Torbjörn Larsson
    Posted October 19, 2020 at 4:15 pm | Permalink

    Thanks, I saw the press release but couldn’t take time right then – this was valuable!

    The overall impression seems to be a decent paper – they did a lot of work on having enough sequence coverage, fishing out exomes and resolving the new partition topology – but a bit speculative – a selective assessment (essentially based on a proxy for corrected or uncorrected significance) as alan and Mike also comments on.

    Looking at the methods they may have been computer resource limited, explaining some of the rich hypothesis generation. I’m not sure if using a fast coalescent method like ASTRAL-II against a maximum likelihood tree instead of a slower bayesian method is the best way to go – in most cases the two latter methods balance their results quite well. But that doesn’t touch that they used two methods or that they made the key topology resolution, so just another question on detail.

    And I learned about “cursorial” hunting!

Post a Comment

Required fields are marked *

%d bloggers like this: