Australia: The Land Where Time Began
Denisovan Anatomy - Reconstructing by Use of DNA Methylation Maps
Denisovans are an extinct group of humans whose morphology has remained unknown. In this paper Gokhman et al. present a method for reconstructing skeletal morphology by the use of DNA methylation patterns. The basis of their method is unidirectional methylation changes to loss-of-function phenotypes. By reconstructing Neanderthal and chimpanzee skeletal morphologies they tested performance and obtained >85% precision in the identification of divergent traits. Then they applied this method to the Denisovan and offer a putative morphological profile. They suggest that it is likely the Denisovans shared traits such as an elongated face and a wide pelvis with Neanderthals. They also identify changes that are Denisovan-derived, such as an increased dental arch and lateral cranial expansion. Their predictions match the only Denisovan bone that is morphologically informative known to date, as well as the Xuchang skull, which was suggested by some to be Denisovan. Their conclusion was that DNA methylation can be used to reconstruct anatomical features, which include some that do not survive in the fossil record.
Not much is known about the anatomy of Denisovans. The first specimen, Denisova 3, comprises a manual phalanx that was recovered from Denisova Cave in Siberia, which dated to between 74,000 BP and 82,000 BP (ka) (Krause et al., 2010). It was indicated by DNA that was extracted from this bone that this individual belonged to a sister group of Neanderthals (Meyer et al., 2012), thereafter called Denisovans. At 390-440 ka these 2 groups separated (Prüfer et al., 2017), and the ancestors of this group split from our lineage between 520 and 630 ka (Prüfer et al., 2017), though these dates are still being debated (Mafessoni & Prüfer, 2017; Rogers et al., 2017). Denisovan ancestry, based on this genome, up to 6% was detected in Melanesians and Aboriginal Australians and to a lesser level in East Asians, Native Americans, and Polynesians (Meyer et al., 2012; Prüfer et al., 2014; Racimo et al., 2015; Reich et al., 2010; Skoglund & Jacobssen, 2011). Gokhman et al. suggest that some introgressed Denisovan haplotypes might have conferred an adaptive advantage to modern humans (MHs) [anatomically modern humans (AMH)] in high altitude (Baell et al., 2010) and cold climates (Racimo et al., 2017).
Findings that provide information on the anatomy of Denisovans remain scarce, in spite of increasing understanding of their genetics. To date the only Denisovan samples that have been confirmed are the aforementioned mentioned Denisova 3 phalanx, from which a 30x gnome was sequenced (Meyer et al., 2012), A lower jawbone (Chen et al., 2019), and several teeth (Sawyer et al., 2015; Slon et al., 2017). It has been revealed by anatomical studies of the teeth that the molars of Denisovans differ in their cusp and root morphology, and the range of modern AMHs, as well as being mostly outside the range of Neanderthals (Chen et al., 2019; Sawyer et al., 2015; Slon et al., 2017). It was shown that the jawbone was robust, protruding, with a long dental arcade and no chin (Chen et al., 2019). A study of anatomical difference between human groups is critical to an understanding of adaptations that are human-specific, selective pressures; and developmental trajectories as well as the effects of phenotype of introgression events.
Current ability to decode DNA sequences of Denisovans is very restricted, while these data potentially bear ample information on its anatomical features. An examination of the biological consequences of substitutions that alter protein sequence is a direct approach. Less than 100 fixed nonsynonymous substitutions distinguish modern humans from Denisovan and Neanderthal, whereas about 30,000 fixed changes are noncoding of synonymous (Prüfer et al., 2014). Though it is likely that many of the noncoding changes are neutral (or nearly so), many others probably alter gene activity and may be anatomically highly informative. Pinpointing such variants, however, is notoriously difficult.
Gokhman et al. suggest that a possible approach to circumvent this is to predict the combined effect of SNPs that are known to be associated with various traits. Accuracy of prediction for traits such as skin, hair and eye pigmentation is >80% in Europeans (Walsh et al., 2013), though for the vast majority of traits, genome-wide association study (GWAS)-based predictions reach accuracy levels that are substantially lower (Price et al., 2015), including facial morphology (Brinkley et al., 2016; Cole et al., 2016; Erlich et al., 2017; Liu et al., 2012; Shaffer et al., 2016). The ability to extrapolate GWASs that were European-based to non-European populations was shown, however, to be very limited (Martin et al., 2017). Possibly most importantly, GWASs are based on within-population variability, which usually reflects variants which more recently emerged. Older variants that separate lineages and variants that diverged more deeply with considerable phenotypic effects are more likely to reach fixation and therefore are not likely to be pinned down in GWASs, even if they have a substantial effect (Martin et al., 2017; Price et al., 2015). These factors together limit the applicability of variants that are detected by GWASs of morphological analyses of groups that diverged deeply, such as the Denisovan.
Ideally, attempts should be made to measure directly gene expression, which is more easily interpretable than noncoding sequence changes. In ancient samples RNA molecules degrade more rapidly, however, and are not available for sequencing. Gokhman et al. therefore used DNA methylation, which is a key regulatory layer of the genome, as a proxy for gene activity. For this study Gokhman et al. developed a method that compares methylation patters of Denisovan DNA to those of modern humans, Neanderthals, and chimpanzees and infers which genes may have become upregulated or downregulated along each lineage. Then they linked these changes to potential phenotypic alterations. They did this by analysing phenotypes that are known to be incurred by loss-of-function mutations in these genes and therefore could be roughly paralleled with reduced activity. Importantly, their aim was considerably more modest, which differed from previous efforts to make quantitative morphological estimations (Claes et al., 2014; Erlich, 2017; Lippert et al., 2017), as they strove to reconstruct a qualitative skeletal profile by predicting traits that are divergent between human groups and, when possible, to determine in which direction their changes occurred. Their rationale was that providing accurate magnitudes of anatomical changes is not feasible, mainly because precise activity levels of archaic human genes cannot be determined, and the quantitative contribution of each gene to the trait, even in samples from the present, is not possible to predict at present. They quantified the accuracy of their method by applying it to Neanderthal and chimpanzee then compared their predictions with the known morphology of these groups.
They found that they reached prediction precision of 82.8% in reconstructing traits that separate Neanderthals and modern humans and 87.9% in predicting their direction of change. They reached a similar performance in the chimpanzee, with a precision of 90.5% at predicting which traits are divergent and 90.9% in predicting their direction of change. They propose a methylation-based profile of Denisovan morphology by applying their method to the Denisovan morphology.
It is of interest that many of the Denisovan traits that were reconstructed by Gokhman et al. were identified in fossils from China dating to the Middle Pleistocene. Various characters that are Neanderthal-like are displayed by these fossils, tough their phylogenetic classification is yet to be determined (Bae et al., 2010; Li et al., 2017). According to Gokhman et al. probably the most Neanderthal-like are the crania from Xuchang in eastern China that date to 100,000-130,000 BP. Together with their location in eastern China, the similarity of these crania to Neanderthals, that they might be Denisovans. This could not be confirmed without DNA. Included among these bones are the skull cap and base, though not the face or jaws, and they exhibit the following 10 directional morphologies:
1. Lateral expansion of the temporal bones;
2. Low cranial vault;
3. Lateral expansion of the parietal bones, outside the range of Neanderthals and modern humans (Suzuki & Takai, 1970);
4. Wide cranial base;
5. Cranial gracility;
6. Prominent supraorbital tori;
7. Reduced thickening (restricted nuchal torus) of the occipital bone;
8. Sagittal flatness;
9. Mastoid process that is short and inwards sloping;
10. Small anterior semicircular canal radii and lateral versus posterior canals that are more superior.
There are equivalent phenotypes on HPO for traits 1-8 which could therefore be examined against the reconstruction profiles of Gokhman et al. strikingly, 7 of them were identified as divergent traits in the reconstructed profile of Denisovans by Gokhman et al. Traits are linked to methylation changes that are unidirectional and therefore could be predicted. The directionality reported by Gokhman et al. for all these traits matched the directionality that was observed in the Xuchang fossils. A fact that is even more outstanding is that Gokhman did not predict any other divergent traits that were related to the vault of the crania (i.e., the region that was preserved in the Xuchang crania) with the exception of the 7 that are observed in the Xuchang fossils (Li et al., 2017). The first genetic support for the notion that the Xuchang skulls are related to Denisovans is provided by the almost complete overlap between the Xuchang crania and the reconstructed profile.
It has been observed that the molars of Denisovans are significantly broader than those of modern humans and are mostly outside the range of Neanderthals molars (Chen et al., 2019; Sawyer et al., 2015; Slon et al., 2017). Gokhman et al. did not predict larger molars, though it is possible these might be related to the fact that, though the HPO database includes some phenotypes it does not include phenotypes that that are specific for the size of the molars or width. However, Gokhman et al. do predict a longer dental arch, which could potentially be linked to large molars.
There is only limited ability to extract phenotypic information from a single gene. In this paper it is demonstrated by Gokhman et al. that by:
1. Looking at the set of genes that underlie a trait,
2. Considering only marked changes in promotor methylation,
3. Adding unidirectionality filters high reconstruction accuracy of phenotype can be achieved.
Also, as the direction predictors pass all 3 unidirectionality filters, it is likely they represent either higher-level regularity changes that cascade to affect several loci (e.g., silencing of a transcription factor that affects the promotor methylation of multiple target genes) or alternatively, the polygenic adaptation processes, where it is advantageous to adapt in a certain phenotypic direction, therefore resulting in coordinated changes in several loci (Fraser et al., 2010). Pronounced phenotypic alterations are likely to be driven by both options and, therefore, potentially explain why the predictive power increases when looking at the collective effect of unidirectional changes.
The main strengths and weaknesses of the method that was proposed by Gokhman et al. were summarised.
1. A comparative direction of change is offered by its predictions, instead of precise quantitative evaluation of the extent of phenotypic change.
2. The direction of change cannot be determined in instances where there is no unidirectionality.
3. Morphologies cannot be reconstructed when there is no equivalent term on HPO.
4. Reconstructed traits that are based solely on Denisovan-specific DMRs (3 traits predicted out of 56 in total) could represent only the Denisova 3 individual, and not the Denisovan population.
5. The method is more accurate at identifying traits that emerged earlier that became fixed or nearly fixed, while ignoring traits with high Intrapopulation variability, as a result of the strict directionality filtering, and the varying number of individuals used for DMR detection along each branch. At the same time, such changes that are unidirectionally fixed are more likely to be driven by selection (Fraser et al., 2010) and therefore could represent a subset that is more interesting of morphological alterations.
6. The precision and sensitivity levels reported in this paper are dependent partly on the way traits are discretised, as Gokhman et al. have clustered together overlapping traits, thereby forming morphological units (See STAR Methods). Yet there are many morphologies that could be intertwined developmentally, as reflected by the overlapping set of genes that underlie them.
7. Finally, it is yet to be determined which of the Denisovan morphologies that were reported by Gokhman et al. are confined to Denisova 3 and which reflect the Denisovan population, as Denisovan-specific DMRs are based on a single sample. In regard to this there are several observations that support the notion that the majority of traits that are reconstructed are shared throughout the Denisovan population.
1. It has been shown in the fossil record that traits that separate a single Neanderthal fossil from modern humans tend to be shared among all Neanderthals (Aiello & Dean, 2002).
2. Of the reconstructed traits roughly half are based on DMRs that emerged along the lineage of modern humans. As in modern humans such traits are derived, Denisovans are expected to share the ancestral form of the trait.
3. The entire analysis is based on DMRs that are not affected by age, sex or type of bone, and therefore, a Denisovan from a different age and sex and where the sample was obtained from a different type of bone is expected to display similar methylation patterns (Gokhman et al., 2017a).
According to Gokhman et al. it is enlightening that the fraction of derived traits that have equivalent HPO phenotype are markedly different for the Neanderthals and chimpanzees, though the reconstruction accuracy levels for them are very similar. 75 out of 107 derived traits (70%) for the Neanderthal have a parallel HPO phenotype, whereas divergence is ~10x deeper for the chimpanzee, the fraction is significantly lower (83 out of 201, 41%, p = 1.5 x 10-6. Χ2 test. The possibility is raised by this that along short timescales, the genes underlying disease may also be those that underlie evolutionary phenotypic divergence, which would possibly account for the overlap between the 2 lists. A study that suggested that the loci involved in craniofacial disorders are also likely to underlie normal variation, within humans and between humans and chimpanzees (Claes et al., 2018). Gokhman et al. suggests it is likely that the fraction of Denisovan traits that can be predicted by using HPO phenotypes is high and similar to that in Neanderthals, given that divergence time between Denisovans and modern humans equals that between Neanderthals and modern humans.
The approach presented in this paper relies on 2 basic hypotheses:
1. Pronounced methylation changes, i.e. statistically significant unidirectional promoter methylation changes of >50% that extend across at least 50 Pg positions, are more likely to be associated with phenotypic effects than subtle changes, and
2. The direction of a phenotype that is driven by downregulation of a gene is expected to be similar to the direction of a phenotype that is driven by loss-of-function mutations.
It was demonstrated by this study that using these assumptions, it was possible to reconstruct dozens of the traits that differ between human groups with more than 80% accuracy. At first glance it might be surprising that the analysis of DNA methylation alone is sufficient to reach such accuracy. It was shown, however, that there was a high correlation and interplay between DNA methylation and other methylation layers, such as binding of transcription factor and modifications of histones (Banovich et al., 2014). The information that was extracted from it might, therefore, reflect a sizeable portion of the regulatory landscape, though this analysis was based on a single regulatory layer. Furthermore, as the extent of each phenotypic change was not reconstructed, rather only its direction of change, it is likely that the other regulatory layers, as well as protein changes and non-promotor methylation changes, contribute further quantitatively to the phenotypes that are observed.
Gokhman et al. concluded that the unidirectional promoter methylation changes can be used to identify phenotypic divergence between organisms that are closely related. The ultimate test of this approach would be to match the reconstructed profile to a more complete collection of Denisovan samples, when they are discovered, even though they validated this approach on Neanderthals and chimpanzees, and also on the Denisovan jawbone.
Gokhman, D., et al. (2019). "Reconstructing Denisovan Anatomy Using DNA Methylation Maps." Cell 179(1): 180-192.e110.
|Author: M.H.Monroe Email: firstname.lastname@example.org Sources & Further reading|