Using AI for aurochs reconstructions

AI generated images are not my taste at all, but I have been experimenting with ChatGPT lately. Precisely, I used it on my aurochs reconstructions with the prompt “make it photorealistic”. The results were not always perfect, but some are quite good I think. The AI tends to make the trunk longer and the legs shorter, which is probably because of the cattle photos it was trained with, so I had to correct some of the results for the proportions, but not all. Today I want to share some of the results. Note that I neither claim AI generated images are "art" nor that I acclaim any copyright to these results. I merely attempt to create the perfect aurochs reconstruction. 

 

-) based on my Vig aurochs painting

Readers of my book "Breeding-back wild beasts" will know my painting of the Vig bull. I uploaded it to GPT and prompted it to make it photorealistic. As you see, it made the proportions a bit more domestic and I had to restore the dorsal stripe, but overall I think the result is ok.

-) turning Lamarck into an aurochs

I modified a photo of the Taurus bull Lamarck from 2013 and gave it more aurochs-like horns plus the curly hair on the forehead, then I prompted GPT to make it photorealistic. The result is quite good I think, although I am not 100% satisfied with the horns.

 

-) the “Augsburg aurochs” scenery

I did a painting with acrylics that is supposed to reconstruct the oil-based original of C.H. Smith’s “Augsburg aurochs”, with the landscape based on the 19^th century one. Smith reported the original showed a sooty black colour with a white chin and coarse hair. I uploaded the painting to GPT and prompted it to make it photorealistic, and I think the result is really, really good. It looks a lot like I imagine a European aurochs bull to look like, and captures the “Augsburg” essence well.

 

-) painting of the Stuttgart skull

The skull exhibited at the Stuttgart museum am Löwentor is a very interesting one, as it has quite wide-ranging, almost banana-like horns. I did a 3D reconstruction of those horn cores and sculpted a sheath over them, so that I have an idea what they might have looked in life. I used them as template for a painting, and uploaded the painting to GPT with the usual prompt. I really like the result, I think it gives a good impression of what that skull might have looked like in life.

 

-) Store-damme bull with Holstein bulls as a template

This is based on a sketch of the Store-damme skeleton that I reconstructed in flesh and blood looking at Holstein bulls – whether this breed is the best comparison is very debatable, I was just curious on what it would look like. Then I uploaded it to GPT with the prompt “paint it photorealistical like a Sayaguesa bull” because when I prompt it using the word aurochs it often results in bulls with a colour saddle.

 

-) Store-damme bull with young Lidia/Sayaguesa as a template

This is a sketch from a little animation I did a while ago with the same prompt as above. I am very satisfied with this one, I think the result looks very realistic and is also very aesthetically appealing to me.

 

 

Realtalk: will the aurochs ever come back?

Back in my teenage years, I was very much into Mesozoic dinosaurs. It changed in my late teens and shifted towards recently extinct animals, especially ungulates, because these animals are much closer to our time and thus more graspable. My enthusiasm for aurochs was partly always fueled by the thought that, because of its very recent extinction, it might one day come back with new technology. But will that ever happen?

 

In the past, I made posts on why the aurochs would be the prime candidate for “de-extinction”. You can go here, for example. Having had a deeper look into the genetic framework around CRISPR-Cas9 and what would be necessary for a successful “de-extinction” of the aurochs, I changed my mind fundamentally. The aurochs might actually be a very tricky case, if not technically impossible. Let’s have a look at why that might be the case step by step:

 

1. The reconstructed genome is not the original genome

 

Ancient DNA does not come as a nice, complete DNA strand, but as isolated fragments that are used to try to infer the original genome. Many pieces cover duplications of sequences. For the reconstruction, you need a template, a close relative that is used for puzzling the pieces together. In the case of the aurochs genomes, domestic cattle and related wild bovines are used as a template. The problem: structural variations, such as duplications of genes (that can have a huge impact on the organism), have a high risk of staying undetected in this method. For example, if the aurochs had an additional copy of a gene of the FGF family, or at least an additional enhancer, it would explain why aurochs were taller, larger, had more developed horns, and such a robust skull; yet, the duplication would stay undetected because the template (the genome of the related species) does not have it. Thus, even if scientists manage to create a bovine that has 100% of the reconstructed aurochs genome, there is the risk that this reconstruction is different from the original genome, with those differences having a large impact on the development of the animal – maybe even lethality.

 

2. Creating a bovine with 100% of the reconstructed genome is technically basically impossible

 

Domestication changed the hormonal and developmental system of cattle dramatically, with the morphological changes being a consequence of that. Not only were probably hundreds or even thousands of genes determining the hormonal and developmental regime of the animals affected by domestication, but also the accompanying regulatory sequences and epigenetic programs. So if we were to change a cattle genome into a (reconstructed) aurochs genome, we are probably talking about hundreds or thousands of edits, with a technique that only allows one change at a time and might also produce errors. It is highly impracticable.

 

3. Even exchanging a few important alleles would be problematic

 

Many of the changes during domestication did probably not change the proteins produced by the genes directly, but when, where and how long the genes are expressed. The typical domestic paedomorphy is basically a prolonged juvenile gene expression. “Removing” paedomorphy would make the bovines already a lot more aurochs-like. But developmental genes are part of a highly nuanced system with complex interactions between the individual genes, and changing a few “key genes” can lead to wrong expressions, distorted development or even lethality. And even in more discrete traits, such as horn shape, even if we identify a gene that regulates the classic aurochs horn shape and we were to insert the allele into a cattle genome, there is no guarantee the allele would work the way it would work in an aurochs, because the genetic, epigenetic and developmental environment is different. It would have to be executed by “trial and error”, which would require a large number of cattle individuals and many breeding generations. And if the inserted aurochs allele is to work in a proper, aurochs-like fashion, it would require the developmental program of an aurochs and not of domestic cattle, which brings us to the problem illustrated in point 2 again. Just inserting an aurochs allele into a domestic developmental framework might lead to distorted development and traits that are nothing like an aurochs’ at all.

 

4. In the same time span and at much lower cost, good “breeding-back” cattle could be released into a suitable reserve and natural selection does the rest

 

Implanting a few “key” alleles from ancient DNA into “breeding-back” cattle would require research on which alleles are to be inserted and a lot of trial and error to not disrupt the animal’s development, and it would probably work only with one allele per generation. Considering this, breeding for a more aurochs-like phenotype might actually have been the more efficient way all along. It selects for aurochs-like traits without disrupting the developmental pathways – surely, it takes long, but so would genome editing. And both strategies will not result in an original aurochs.

Therefore, my take on the subject currently is take the best “breeding-back” cattle, release them in a suitable area and let natural selection do the rest. According to my “dedomestication hypothesis”, that I am sure many of my readers are familiar with, wildtype traits (be it morphological, ethological, hormonal/developmental) will have a selective advantage over the domestic counterparts and the cattle will become continuously more wildtype-like with time. The cost of letting natural selection do the rest is virtually zero compared to highly effortful genome editing, which has no guarantee that it will work as planned. Therefore, I think that before we try to create an “aurochsified” cattle in the lab, we should concentrate our resources on finding or creating an area where a large-enough population of “breeding-back” cattle can live under natural circumstances, ideally with natural and sexual selection, and also pressure from predators. 

 

5. Are modern breeding-back cattle “ready to be released”?

 

This brings us to the next question, namely if modern “breeding-back” cattle are already suited for such an endeavour. My answer is a clear and strong yes. I think that modern “breeding-back” is very close to the pinnacle of what can be achieved with domestic cattle, apart from the fact that they are still very heterogeneous. But stabilizing the phenotype might actually take centuries, considering the large number of genes involved and the slow reproduction of cattle. From the ecological standpoint, I think the cattle are very fit – cattle tend to feralize rather easy, and so should “breeding-back” cattle, which descend from hardy and robust landraces. Of course, the currently running projects can still strive for even better animals, but I am very satisfied with the current results. I think it is time to look for an area large enough and isolated from humans enough so that the cattle can develop freely and unhindered from any human interference – at the same time, it would be very interesting to document phenotypic changes of these released cattle over decades in a long-term study.

 

6. Conclusion: Will the aurochs ever come back?

 

If you strictly mean the original animal in all aspects like it was 100.000, 10.000 or 1000 years ago, we can effectively almost rule that out with certainty. No matter which route we take, there will be genetic and phenotypic differences. But in a more sensu lato way of the term, we can produce a type of bovine that is barely distinguishable from the aurochs (apart from vestiges of domestication that will appear in several individuals for a very long time, such as small white spots, recessive diluted coat colours etc.) that also functions ecologically the same way, shows the same social behaviour patterns (as all domestic cattle do) and is also at least partly shaped by natural selection, by taking the best of the best of “breeding-back” cattle and releasing them into a suitable area.

 

Two new aurochs head busts based on actual skulls

Recently I finished two aurochs head busts I have been working on over the last couple of weeks. They represent the Ilford (= London) skull and a skull from the Natural history Museum whose age and location I do not know. Both skulls are available as 3d scans on sketchfab, which was great for my reconstructions.

 

based on the NHM skull

based on the Ilford skull

How I reconstructed them

 

As usual, I started with trying to accurately reproduce the skulls. I decided to make the skull 20 cm long, which is quite a handy size for the busts that still enables much detail. Then I started sculpting the eyes, nasal cartilage, lips and facial muscles around the skulls. Next step was to accurately reproduce the shape of the horn cores, and then adding the keratinous sheaths, which involves guesswork. While the horn sheaths of the Ilford skull were pretty straightforward to infer from the cores, that of the NHM skull was more tricky because the tips of the cores face forwards, not inwards as usual in aurochs. It could be the case that the keratinous sheaths were not long and did not create inwards-curving horn shapes in life, or that the sheaths were rather long, long enough to create the inwards curve. I chose latter option because the relation between the solidly keratinous part of the horn and the horn core seems to have been highly variable in aurochs and seems to be generally that way in bovines.

After sculpting the horns, the next step was to sculpt the skin and fur.

 

Why two of them

 

I chose to reconstruct two aurochs bulls to show the intraspecific variation. The Ilford skull is very massive and robust in morphology, perhaps from a very old bull, while the NHM skull is more filigree in comparison. It shows that the aurochs was quite variable in some aspects of its morphology and by fabricating two busts I wanted to show that. To add a female, I am going to make a bust based on the Sassenberg cow in the future.

 

 

How did the results work out?

 

I am quite satisfied with the results. I think I did not underestimate the soft tissue this time, in the case of the nose of the Ilford specimen I might even have overestimated it. The Ilford reconstruction has eyes that appear surprisingly large compared to one would guess from the skull alone. I think the curly hair on the forehead turned out to be quite realistic, although I could have made the hair even longer and shaggier in at least one of them considering what historic texts say.

 

What remains a speculation

 

The exact extent of the keratinous sheath in life cannot be derived from the bones and therefore remains a speculation. The extent of the white muzzle ring probably also varied among aurochs bulls, with some having virtually none and some having it fully expressed as in my reconstruction of the NHM skull.

 

Here are my head busts next to each other:

 

 

I call them Ozzy and Lemmy.

 

 

 

 

The wisent hybrid origin hypothesis is outdated

Traces of hybridization among closely related species have been found everywhere it was looked for, including the genus Homo. Bovines are no exception to this. About ten years ago, it was proposed that the wisent, which I refer to as Bos bison bonasus here, originated as a hybrid of aurochs and steppe bison, Bos bison priscus, and inherited around 10% of its nuclear genome from aurochs. Now, ten years later, it has been demonstrated that this hypothesis was based on premature conclusions and partly also the result of methodological problems. However, the rebuttal of the hybrid origin hypothesis did not nearly attract as much attention as the paper by Soubrier et al., 2016, so that many still think this is the current state of research. Considering that I wrote a post when the paper came out titled “Confirmed: the wisent is an aurochs hybrid” back then in 2016 that is still online, I feel obliged to present a (hopefully) up-to-date picture of the current research on the origin of the European bison. Let’s dive into the literature.

 

A possible hybrid origin for the wisent was first proposed in 2004 by Verkaar et al., when it was surprisingly revealed that the wisent does not cluster with American bison and yak on the mitochondrial genome but with taurine and indicine cattle instead [1]. This would have suggested that bison bulls repeatedly mated with cows of the cattle lineage, which, given the time and region, would have been aurochs, giving rise to a bison that has “cattle-line mitochondria”. Then, more than ten years later, this hypothesis was endorsed by Soubrier et al., who executed a test of 10.000 single nucleotide polymorphisms and concluded that 10,9% of the nuclear genome of the wisent stem from aurochs [2]. At that time, I found that convincing, given that wisent sometimes clearly show the primigenius spiral in their horns and have a shorter body length than American bison, traits that I associated with the aurochs. But, as we will see, the results of Soubrier et al., did not remain uncontested for long.

 

The first critical response to mention would be Wang et al from two years later [3]. They find no evidence of 10% nuclear aurochs ancestry, and also found that parts of the nuclear genome show a similar phylogeny as the mitochondrial genome. The authors consider it more likely that the evolutionary phenomenon called incomplete lineage sorting is responsible for the weird mitogenomic position of the wisent on the phylogenetic tree.

ILS is a rather cool phenomenon. It happens when there are polymorphic genes (with more than one allele) in an ancestral population and the population splits up and undergoes cladogenesis. One branch would inherit only one allele of the polymorphic gene (let’s call it “A”), while the other population still has the alleles A and B. Then, this other population splits up too and one filial population retains only A, and the other one only B. What would the result on a phylogenic tree be if you look at this gene? The population that branched off first would group together with the population that retained A, while B would result as the outgroup to the other two – although this is not what happened in reality. This is incomplete lineage sorting. In fact, it is not rare that mitogenomes show phylogenies that are inconsistent with those inferred from nuclear DNA [3]. Wang et al. find that ILS is a more plausible explanation for the mitochondrial position of the wisent and more consistent with the genomic structure of the bovine [3]. The authors also caution that phylogenies based on single genes can be misleading because of ILS [3].

 

What about the 10% aurochs ancestry on the mitochondrial genome? It seems that this number is erroneous because an inadequate method was used [4]. Soubrier et al. looked at fewer than 10.000 SNP, which is a rather scarce sample, and did an f4-statistics analysis with them, which is how they arrived at the 10,9% aurochs ancestry. Grange et al. point out that this might be problematic because f4-statistics assume an ancestral population that is polymorphic for the SNPs, while the sample the authors used were from the BovineSNP50 chip that contains SNPs for differentiating individuals of modern cattle breeds [4]. With another approach, namely the Approximate Bayesian Computation (ABC), Grange et al. arrive at a 97,2% probability that there is at least one percent aurochs ancestry in the wisent genome, which is significantly lower than the results of Soubrier et al. [4]. The method employed by Soubrier et al. also results in ancient wisents, wisents from 1911 and Pleistocene priscus being closer to each other than the wisent from 1911 to modern wisents, indicating that something is methodologically wrong here [4].

Weçek et al. did a whole genome sequence analysis with modern wisent genomes as well as four historical samples and found that only 2,4 to 3,2% of the wisent genome seem to come from the aurochs/cattle lineage and the genomic structure suggests that it occurred a long time ago, therefore they consider it possible that this reflects aurochs introgression but they note it requires more research [5].

 

If the wisent is not an aurochs hybrid, where does it come from? One hypothesis that was postulated is that it is a descendant of late surviving B. b. schoetensacki. Bone remains from the Sirejol cave from the Upper Pleistocene have been tentatively morphologically assigned to this form previously, and a phylogenetic analysis revealed that the material belonged to the Clade X proposed by Soubrier et al. and Bb1 proposed by Massilani et al. (2016) [6]. The authors of Palacio et al. therefore conclude that schoetensacki corresponds with this mitochondrial clade [6]. The presence of schoetensacki in the Late Pleistocene would imply a ghost lineage spanning over a 500.000-year period, since this variant disappears from the fossil record shortly after aurochs enter the European continent in the Middle Pleistocene.

However, the morphological assignment of the Sirejol material to schoetensacki was likely premature, as a morphometric analysis taking sexual dimorphism into account rejects that this material was of a late surviving schoetensacki bison [4]. The much more basal B. b. schoetensacki clusters morphologically with the earlier B. b. menneri, but not with steppe bison and wisent.

 

So the wisent is not a descendant of the European Bos bison schoetensacki either. What we can say is that molecular clock data suggests that the lineages of B. b. bonasus and B. b. bison diverged around 215kya (roughly the same time when B. b. priscus first entered America) and gene flow ceased at 102kya [4]. 57 ancient genomes suggest that late Pleistocene steppe bison originated from the north while wisent originate from a refuge in the southern Caucasus after the most recent glacial maximum [7]. The American bison is nested in the priscus tree on the mitogenome [4], so we can be rather confident that B. b. bison descends from B. p. priscus in some way (either directly or over B. b. latifrons).

 

The two mitochondrial clades Bb1 and Bb2 are present in the late Pleistocene in Western Europe and the Caucasus, they are sister clades to the modern wisent clade [7,8]. It seems that while aurochs retreated southwards during glacial periods, ancient bison stayed in refuges dispersed over Europe [7,8]. This fragmentation and isolation could explain how the ancestors of the wisent split off from the rest of late Pleistocene bison around 215kya. While the more cold-adapted ancestors of the steppe bison/American bison clade probably remained in the north, the ancestors of the wisent might have stayed in their refuges in the Caucasus and parts of Europe.

 

As you see, the situation is much more nuanced than my post from 2016 suggests. It is a reminder for me that the last word is never spoken in science, contrary to what the title of my 2016 post suggests (“confirmed”, a word I should not have used).

If I missed any relevant papers in this post, I'd be grateful if it could be pointed out to me. 

 

Literature

 

[1] Verkaar et al.: Maternal and paternal lineages in cross-breeding bovine species. Has wisent a hybrid origin? 2004.

[2] Soubrier et al.: Early cave art and ancient DNA record the origin of European bison. 2016.

[3] Wang et al.: Incomplete lineage sorting rather than hybridization explains the inconsistent phylogeny of the wisent. 2018.

[4] Grange et al.: The evolution and population diversity of bison in Pleistocene and Holocene Eurasia: sex matters. 2018.

[5] Weçek et al.: Complex admixture preceded and followed the extinction of wisent in the wild. 2016.

[6] Palacio et al.: Genome data on the extinct Bison schoetensacki establish it as a sister species of the extant European bison (Bison bonasus). 2017.

[7] Massilani et al.: Past climate changes, population dynamics and the origin of Bison in Europe. 2016.

[8] wisenthybrid3

[9] Zver et al.: Phylogeny of Late Plesitocene and Holocene Bison species in Europe and North America. 2021.

 

Did the wisent evolve because of humans?

For this post, I created the new post category “wild speculations”, because it is indeed a wild speculation. Some species even evolved as a result of human activities. One example for human-induced speciation would be the plant Erythranthe peregrina, which is a hybrid of two species introduced to Britain, E. guttata and E. lutea, which hybridized, created a sterile hybrid which experienced allopolyploidization (doubling of the genome) and produced a fertile hybridogeneous species. It was discovered relatively recently (2011). Without human intervention, this species would not exist. It is also possible that other kinds of human intervention, such as wiping out species, can result in a speciation event over a longer period of time in evolutionary terms because the niches that become vacant have potential for the evolution of new species. I have the suspicion that it is possible that the wisent might be such a case.

The story begins around 600kya, when there was only one Bos species in Europe, Bos (bison) schoetensacki. Around that time, a second bovine species migrated to Europe, either by island hopping from Africa or via the continental route from western Asia, the aurochs. Only a few millennia after the aurochs arrived in Europe, B. schoetensacki disappeared from the fossil record. I think it is quite possible that B. schoetensacki was outcompeted by the newly arrived aurochs as both bovines must have occupied a similar niche and the extinction shortly after the arrival of the aurochs is just as suspicious as the fact that bison subsequently were absent from interglacial faunal assemblages of Europe until the very late Pleistocene around 13kya, when bison (now in the form of the wisent) re-entered the interglacial megafauna assemblage. I call that the “bison gap” (roughly 600kya-13kya). If the aurochs outcompeted bison in Europe, why did they rejoin the interglacial fauna and why only so recently? B. bison priscus, which was most likely the ancestral form of the wisent, was present in Europe during all of the recent glacial periods. So, it could have evolved into an interglacial bison type much earlier, right after the extinction of B. schoetensacki if the aurochs did not outcompete bison in Europe – if it did, it is an open question why this competition was not a factor anymore at the end of the Pleistocene and during the Holocene.

Several thousand years earlier, there was another event that had an impact on the European megafauna: Palaeoloxodon antiquus and Stephanorhinus hemitoechus died out, most likely because of hunting from humans. Elephant extinction reduced competition for other grazers, at the same time forest growth became more common. Aurochs could not invade denser forest while bison are adapted to consume more wooden vegetation and can invade denser forest. This might have created new ecological space for bison to re-enter the interglacial faunal assemblage.

 

This guess is impossible to verify, but I think it must have one or several reasons why bison were absent from the European interglacial faunal assemblages for over 500.000 years.

 

If this guess is true, it would illustrate how nature is dynamic and might respond to anthropogenic impacts not necessarily always in a diversity-reducing way but sometimes also in a diversity-increasing way.