Portrait of a Denisovan
DNA methylation data made it possible to recreate the appearance of the Denisov man
When developing the method, information was used on how the failure of a particular gene affects the phenotype of modern people. At the same time, scientists proceeded from the assumption that a sharply increased level of promoter methylation, reducing the level of gene expression, gives approximately the same phenotypic effect as a mutation leading to loss of gene function. Testing the method on species with known morphology (Neanderthals and chimpanzees) showed that the morphological differences predicted with its help from Homo sapiens correspond to reality in about 80% of cases. Applying the method to the Denisovan man, the researchers made 32 reasonable assumptions about the morphology of Denisovans. Until recently, it could only be judged by a few teeth. It turned out that Denisovans resembled Neanderthals in many ways (such as a low forehead, protruding jaws and a large chest), but they also had their own unique features: for example, their skull was wider and the dental arch was longer than that of Neanderthals and Sapiens. The recently described Denisov jaw from a Tibetan cave allowed us to verify 8 out of 32 predictions: 7 out of 8 were confirmed. The study showed that the skulls found recently from Xuchang (central China) aged 100-130 thousand years are very likely to belong to Denisovans.
About 10 years ago, thanks to the achievements of paleogenetics, we first learned about Denisovans – an extinct species of humans, slightly closer to Neanderthals than to Sapiens. Since then, much has become known about the contribution of Denisovans to the gene pool of modern humanity, as well as about episodes of hybridization of Denisovans with Neanderthals and other archaic Homo.
Although the genome of one of the Denisovans, the Denisova Cave girl (Denisova 3), has been read with high accuracy, we still know very little about what the Denisovans looked like and how their morphology differed from ours or Neanderthal. The reason for this, on the one hand, is the extreme scarcity of available bone material, on the other hand, the lack of sufficiently accurate and reliable methods for reconstructing the phenotype by genotype.
Learning to accurately predict the phenotype by genotype is the most important task facing modern biology. Its solution will open up downright fantastic prospects for medical genetics, genetic engineering and other disciplines - both theoretical and the most practical. Unfortunately, science is still far away from this. Even with a complete, qualitatively sequenced genome, we can't say much about the phenotype of its owner. Especially if we are talking about a species for which data cannot be obtained from other sources – as in the case of Denisovans, or, for example, with Asgardarchaeans before they could be cultivated in the laboratory.
In theory, the genome of the girl from Denisova Cave (that is, in the sequence of nucleotides that make up the genome) contains – must contain! – detailed information about the anatomy and appearance of the Denisov man. That's just that we can't decipher it yet, with the exception of certain signs lying, as they say, "on the surface" – such as the color of eyes, hair or skin. Something can be said about these signs by non-synonymous (changing the amino acid in the protein) nucleotide substitutions in the protein-coding regions of genes. However, most of the interesting features (including skeletal features, such as the shape of the skull, the size of the teeth or the proportions of the pelvic bones) are determined not so much by the amino acid sequences of individual proteins, as by the nuances of regulating the activity (expression) of many different genes. These nuances, in turn, depend on complex networks of intergenic interactions, whose intricacies no one yet knows how to decipher reliably, having only the nucleotide sequence of the genome on hand. Theoretically, the task should be solvable – and someday science, hopefully, will come to this. But not today and not tomorrow.
But you can also look for simpler roundabout ways. For example, it would be just fine if it were possible to measure the level of gene expression in the same phalanx of the little finger of Denisova 3, from which the complete genome was extracted. After all, the level of expression is a kind of approximation to the assessment of the integral result of the work of all that tangled tangle of regulatory interactions, which was mentioned above. To evaluate the expression, it would be necessary to isolate ancient RNA from the little finger. That's just RNA, unlike DNA, is destroyed very quickly and has no chance of surviving in ancient bones. Therefore, it is impossible to directly assess gene expression in ancient hominids.
However, a roundabout path is also possible here, which paleogeneticists have recently begun to explore. One of the ways to regulate gene activity is the methylation of cytosines in CD dinucleotides in the promoter regions of genes. As a rule, if a gene has a strongly methylated promoter (that is, a methyl group is attached to many cytosines), then the activity of the gene decreases sharply. Exactly which parts of the genome, in which cells, at what stages of development and under what conditions will be methylated – all this, or almost everything, must be very cleverly, in a roundabout way encrypted in the nucleotide sequence of the genome. We don't know how to decipher these hereditary instructions yet, but we can just see which cytosines in the genome are methylated and which are not. This will already give important information about the regulation of gene activity.
Fortunately, cytosine methylation leaves discernible traces in ancient DNA. The fact is that ordinary, unmethylated cytosines in the course of postmortem DNA degradation tend to turn into uracils, and methylated ones into thymines. On this basis, David Gokhman from the Hebrew University in Jerusalem and his colleagues from Israel, Germany and Spain five years ago developed a method for reconstructing "methylomes" – methylation profiles of fossil hominid genomes (D. Gokhman al., 2014. Reconstructing the DNA Methylation Maps of the Neandertal and the Denisovan).
In a new study, the results of which were published on September 19 in the journal Cell, scientists used the methylome of a Denisovan man to reconstruct his appearance. The method they developed for this purpose (Fig. 1) is based on the assumption that the phenotypic changes caused by strong methylation of the promoter are similar to those that occur as a result of mutations that disrupt the operation of this gene or completely disable it (because in both cases the functionality of the corresponding protein is sharply reduced).
Fig. 1. Graphical summary of the article under discussion in Cell. First, genes are identified in the promoter regions of which the species of interest (for example, the Denisovan) has significantly changed the level of methylation compared to its closest relatives. Then, based on data on modern humans – carriers of mutations that disable this gene – conclusions are drawn about what phenotypic changes could lead to a change in methylation. It is taken into account that hypermethylation of the promoter usually leads to a decrease in gene expression, so the phenotypic effect may be similar to the effect of a mutation leading to loss of function. The developed method is tested on species with known morphology (Neanderthals and chimpanzees) and, finally, is used to reconstruct the Denisovan genome.
The work used the methylomes of a Denisovan (the same Denisova 3), two Neanderthals (Altai from Denisova Cave and European from Vindia Cave in Croatia), five ancient Sapiens who lived from 45 to 7.5 thousand years ago, as well as bone methylomes (this is important, because methylation profiles may differ in different tissues!) five chimpanzees and 55 modern humans.
The first stage of the study consisted in identifying genome regions, the level of methylation of which clearly differs in different species (differently methylated regions, DMRs). Only those DMRs in which the level of methylation does not depend much on age, gender, health status and bone type, but depends only on species affiliation, were taken into account. Such DMRs are likely to reflect precisely evolutionary (and not age, sex, tissue-specific or environmental) changes in the level of methylation. In addition, only the strongest changes were considered. The DMRs selected for analysis had to include at least 50 CpG sites and differ in different species by at least 50 "percentage points": for example, 80% of methylated CpG sites in one species and only 30% in another. For comparison, fluctuations in environmental conditions usually change the level of methylation of certain parts of the genome by no more than 10%. Finally, of all DMRs satisfying these conditions, those located in the promoter regions of protein-coding genes (at a distance of 1 to 5 thousand base pairs from the transcription start point) were selected for further analysis, because it is known that it is in these regions that cytosine methylation correlates most strongly with gene expression.
After applying all the filters, 154 "promoter" DMRs unique to modern humans remained in the studied sample (that is, sites whose methylation changed in Sapiens compared to other hominids), 171 DMRs whose methylation changed in the common ancestors of Neanderthals and Denisovans after their separation from Sapiens, 113 promoter DMRs specific to Neanderthals, 55 Denisovans-specific, as well as 2,031 promoter DMRs by which chimpanzees differ from all human species. For the vast majority of genes whose promoters contain selected DMRs (the authors designate these genes by the abbreviation DMG – differently methylated genes) and for which there is evidence of a correlation between methylation and expression, this correlation is negative, that is, hypermethylation corresponds to reduced expression or complete shutdown of the gene, and hypomethylation corresponds to high gene activity. There are exceptions to this rule, and they were taken into account by the authors when developing the method.
In order to predict morphology based on DMR data, the authors used the HPO (Human Phenotype Ontology) database, which contains the most reliable and complete information to date on how mutations that disrupt the operation of a particular human gene (mutations leading to loss of function, loss-of-function mutations) affect on the phenotype. From the HPO database, only those phenotypic changes were selected that, firstly, affect the skeleton, and secondly, have a certain orientation. For example, "pelvic bone development disorder" does not fit this condition, because it has no orientation, and "shortened iliac bones" are suitable. In total, the HPO database managed to find data on 815 directional changes in the skeleton, which are observed in humans when a particular gene is broken.
These data were used to "predict" the phenotype of two species with already known morphology: Neanderthals and chimpanzees using the methyl. It was assumed that hypermethylation of the promoter changes the phenotype in the same direction as the mutational breakdown of the corresponding gene. Only qualitative assessments were made: which feature was changed and in which direction, and how much it was changed, the authors did not even try to predict. The direction of changes was also not always possible to predict. In many cases, not one, but several DMG turned out to be associated with the same sign (according to HPO). If, at the same time, different DMGs "predicted" multidirectional changes in the trait, then the final prediction was that the trait had changed, and in which direction is unknown.
Reconstructions of chimpanzees and Neanderthals turned out to be surprisingly close to reality. So, for Neanderthals, according to the methylome data, 64 signs were predicted that distinguish Neanderthals from their common ancestors with modern humans. At the same time, judging by paleoanthropological data, 53 out of 64 traits (82.8%) really changed in the Neanderthal evolutionary line. For 33 changes, it was possible to predict the direction of the methyl, and these predictions turned out to be correct in 29 cases (87.9%). If we compare Neanderthals not with common ancestors, but with modern humans, then here the authors counted 107 known skeletal differences. Of these, 75 have corresponding phenotypes in the HPO database, that is, these 75 differences could in principle be predicted by the methyl if the method had one hundred percent sensitivity. In fact, 62 differences out of 75 (82.7%) were predicted, and the direction of the differences was predicted for 46 signs, and in 36 cases (78.3%) the prediction turned out to be correct. Approximately the same results were obtained for chimpanzees. However, for the differences between humans and chimpanzees, adequate data were found in the HPO database not in 70%, but only in 41% of cases. But for those signs that were still found there, the accuracy and sensitivity of the method of "morphological predictions by methyl" turned out to be about the same as in the case of Neanderthals.
Thus, the method turned out to be quite effective, which is even surprising, given that the methylation of promoters is, although important, but far from the only way to regulate gene activity. In part, this may be explained by the fact that different methods of such regulation (including DNA methylation, histone modifications and attachment of regulatory proteins – transcription factors to regulatory DNA sites) often act in concert, so that by one sign indicating increased or decreased gene activity, others can be predicted. In other words, if we see that a gene has a hypermethylated promoter, then we can expect that other signs of a decrease in the activity of the gene will be found if we search.
After making sure that the method works, the authors used it to reconstruct the morphology of the Denisov man (the same girl from Denisova Cave, Denisova 3 – there are no methylomes for other Denisovans yet). It was possible to predict 56 differences between Denisovans and Neanderthals or modern humans (or both at the same time). The directionality of the differences was predicted for 32 traits. The results for 18 skull features are summarized in Fig. 2.
Fig. 2. Reconstruction of the skull of the Denisov man. The reconstructed sections of the skull and the corresponding parts of the skulls of modern man and Neanderthal are shown in different colors. MH is a modern man, N is a Neanderthal. Blue arrows pointing upwards mean that Denisovans are superior to Sapiens or Neanderthals on this trait (for the time of eruption and loss of teeth, this means an earlier time). Brown arrows pointing downwards mean that Denisovans have a lower value of this trait than Sapiens or Neanderthals. The circles show no differences. For example, the forehead of a Denisovan was lower than that of a Sapiens, and there were no differences from a Neanderthal on this basis. A drawing from the discussed article in Cell.
It turned out that in most of the signs that distinguish Denisovans from modern humans, Denisovans were similar to Neanderthals (powerful jaws, a low skull with a wide base, a low forehead, thick tooth enamel, a wide pelvis, a large chest, expanded fingertips). Denisovans surpass both Sapiens and Neanderthals in three ways: this is the size of the head of the lower jaw, the width of the skull in the parietal part and the length of the dental arch (the latter fact is consistent with the fact that the Denisov teeth found are really very large).
When the article under discussion had already been written and was being reviewed, a message appeared about the identification of the lower jaw of a Denisov man from a Tibetan cave. Before that, of the morphologically informative Denisov samples, scientists had only teeth at their disposal. The Tibetan find made it possible to verify the predictions associated with the four signs of the lower jaw. Since the authors compare the Denisovan with two other species (Neanderthals and Sapiens), a total of eight predictions are obtained. Seven of them were confirmed, which can be considered a very good result. The only prediction that has not been confirmed concerns the width of the "chin" (the lower jaw in its front part): according to the methyl, it turned out that Denisovans do not differ from Neanderthals on this trait, and the Tibetan find showed that Denisovans have a wider jaw.
Fig. 3. Portrait of a Denisov girl based on the predicted features of the skull structure. A drawing by Maayan Harel from a popular synopsis published in the journal Nature for the article under discussion in Cell. You can watch a video showing all the stages of creating this reconstruction.
The authors emphasize that their reconstruction is very similar to some controversial Middle and Upper Pleistocene finds from China. There are especially many "Denisovian" features in two recently described skulls from Xuchang aged from 100 to 130 thousand years (Zhan-Yang Li et al., 2017. Late Pleistocene archaic human crania from Xuchang, China). For those parts of the skull that were preserved in the specimens from Xuchang, the authors of the work under discussion predicted 7 differences between Denisovans and Sapiens. All these 7 distinctive features (including the width of the parietal region, larger than that of Sapiens and Neanderthals) are present in the skulls from Xuchang. So these skulls most likely belong to Denisovans (Fig. 4).
Fig. 4. Skulls from Xuchang. Image from the article by Z. Li et al.
The work represents an important step towards the development of methods for accurate prediction of the phenotype from genomic data. A significant limitation of the proposed method is that it gives only qualitative, but not quantitative predictions: it is possible to predict that some feature will be expressed stronger or weaker, but it is impossible to say how much. In addition, in many cases, even the direction of the differences cannot be predicted, because the nature of methylation of different genes affecting the same trait points in different directions, and no one knows which of the effects is stronger. Nothing can be said about those signs concerning which it is not known how the loss of function of a particular gene affects them.
As for the reliability of the reconstruction of the Denisovan morphology, it should be remembered that the data on the methyl is still available for only one Denisovan. Therefore, it is impossible to say for sure which of the reconstructed features of the skeleton characterize only the Denisova 3 girl, and which ones characterize the entire Denisova population. The authors give a number of arguments in favor of the fact that most of the predictions are probably true for all Denisovans. Firstly, it is known that most of the signs by which a randomly selected Neanderthal differs from all Sapiens make it possible to distinguish all other Neanderthals from Sapiens. Secondly, the analysis was based on promoter DMRs, the level of methylation of which in the bones of modern people practically does not depend on age, gender, health status and external influences and is approximately the same in all people. Therefore, we can hope that the methylation pattern of these sites in Denisovans also did not depend on all of the above, and that in other Denisovans it was about the same as in Denisova 3.
Another interesting observation is that among the genes in which mutations are associated with known hereditary human diseases, there are unexpectedly many that somehow affect phenotypic traits that have recently undergone evolutionary changes. It is significant that for morphological differences between Sapiens and Neanderthals, "suitable" genes can be found in medical databases in 70% of cases (for 75 traits out of 107), whereas for differences between humans and chimpanzees – only in 41% of cases (for 83 traits out of 201). Perhaps this means that malfunctions in genes that have recently been subjected to selection are more likely to lead to medical problems compared to genes that have not been subjected to evolutionary changes for a long time. This is consistent with the idea that stabilizing selection needs time to ensure high stability of the subsystems exposed to the action of driving selection.
Source: Gokhman et al., Reconstructing Denisovan Anatomy Using DNA Methylation Maps // Cell. 2019.
Portal "Eternal youth" http://vechnayamolodost.ru