A Song of Sheep and Horses, revised edition, now available as printed books


As I said 6 months ago, 2019 is a tough year to write a blog, because this was going to be a complex regional election year and therefore a time of political promises, hence tenure offers too. Now the preliminary offers have been made, elections have passed, but the timing has slightly shifted toward 2020. So I may have the time, but not really any benefit of dedicating too much effort to the blog, and a lot of potential benefit of dedicating any time to evaluable scientific work.

On the other hand, I saw some potential benefit for publishing texts with ISBNs, hence the updates to the text and the preparation of these printed copies of the books, just in case. While Spain’s accreditation agency has some hard rules for becoming a tenured professor, especially for medical associates (whose years of professional experience are almost worthless compared to published peer-reviewed papers), it is quite flexible in assessing one’s merits.

However, regional and/or autonomous entities are not, and need an official identifier and preferably printed versions to evaluate publications, such as an ISBN for books. I took thus some time about a month ago to update the texts and supplementary materials, to publish a printed copy of the books with Amazon. The first copies have arrived, and they look good.


Corrections and Additions

I have changed the names and order of the books, as I intended for the first publication – as some of you may have noticed when the linguistic book was referred to as the third volume in some parts. In the first concept I just wanted to emphasize that the linguistic work had priority over the rest. Now the whole series and the linguistic volume don’t share the same name, and I hope this added clarity is for the better, despite the linguistic volume being the third one.

Uralic dialects
I have changed the nomenclature for Uralic dialects, as I said recently. I haven’t really modified anything deeper than that, because – unlike adding new information from population genomics – this would require for me to do a thorough research of the most recent publications of Uralic comparative grammar, and I just can’t begin with that right now.

Anyway, the use of terms like Finno-Ugric or Finno-Samic is as correct now for the reconstructed forms as it was before the change in nomenclature.


The most interesting recent genetic data has come from Iberia and the Mediterranean. Lacking direct data from the Italian Peninsula (and thus from the emergence of the Etruscan and Rhaetian ethnolinguistic community), it is becoming clearer how some quite early waves of Indo-Europeans and non-Indo-Europeans expanded and shrank – at least in West Iberia, West Mediterranean, and France.

Some of the main updates to the text have been made to the sections on Finno-Ugric populations, because some interesting new genetic data (especially Y-DNA) have been published in the past months. This is especially true for Baltic Finns and for Ugric populations.


Consequently, and somehow unsurprisingly, the Balto-Slavic section has been affected by this; e.g. by the identification of Early Slavs likely with central-eastern populations dominated by (at least some subclades of) hg. I2a-L621 and E1b-V13.

I have updated some cultural borders in the prehistoric maps, and the maps with Y-DNA and mtDNA. I have also added one new version of the Early Bronze age map, to better reflect the most likely location of Indo-European languages in the Early European Bronze Age.

As those in software programming will understand, major changes in the files that are used for maps and graphics come with an increasing risk of additional errors, so I would not be surprised if some major ones would be found (I already spotted three of them). Feel free to communicate these errors in any way you see fit.

European Early Bronze Age: tentative langage map based on linguistics, archaeology, and genetics.

I have selected more conservative SNPs in certain controversial cases.

I have also deleted most SNP-related footnotes and replaced them with the marking of each individual tentative SNP, leaving only those footnotes that give important specific information, because:

  • My way of referencing tentative SNP authors did not make it clear which samples were tentative, if there were more than one.
  • It was probably not necessary to see four names repeated 100 times over.
  • Often I don’t really know if the person I have listed as author of the SNP call is the true author – unless I saw the full SNP data posted directly – or just someone who reposted the results.
  • Sometimes there are more than one author of SNPs for a certain sample, but I might have added just one for all.
More than 6000 ancient DNA samples compiled to date.

For a centralized file to host the names of those responsible for the unofficial/tentative SNPs used in the text – and to correct them if necessary -, readers will be eventually able to use Phylogeographer‘s tool for ancient Y-DNA, for which they use (partly) the same data I compiled, adding Y-Full‘s nomenclature and references. You can see another map tool in ArcGIS.

NOTE. As I say in the text, if the final working map tool does not deliver the names, I will publish another supplementary table to the text, listing all tentative SNPs with their respective author(s).

If you are interested in ancient Y-DNA and you want to help develop comprehensive and precise maps of ancient Y-DNA and mtDNA haplogroups, you can contact Hunter Provyn at Phylogeographer.com. You can also find more about phylogeography projects at Iain McDonald’s website.

I have also added more samples to both the “Asian” and the “European” PCAs, and to the ADMIXTURE analyses, too.

I previously used certain samples prepared by amateurs from BAM files (like Botai, Okunevo, or Hittites), and the results were obviously less than satisfactory – hence my criticism of the lack of publication of prepared files by the most famous labs, especially the Copenhagen group.

Fortunately for all of us, most published datasets are free, so we don’t have to reinvent the wheel. I criticized genetic labs for not releasing all data, so now it is time for praise, at least for one of them: thank you to all responsible at the Reich Lab for this great merged dataset, which includes samples from other labs.

NOTE. I would like to make my tiny contribution here, for beginners interested in working with these files, so I will update – whenever I have time – the “How To” sections of this blog for PCAs, PCA3d, and ADMIXTURE.

Detail of the PCA of European Iron Age populations. See full versions.

For unsupervised ADMIXTURE in the maps, a K=5 is selected based on the CV, giving a kind of visual WHG : NWAN : CHG/IN : EHG : ENA, but with Steppe ancestry “in between”. Higher K gave worse CV, which I guess depends on the many ancient and modern samples selected (and on the fact that many samples are repeated from different sources in my files, because I did not have time to filter them all individually).

I found some interesting component shared by Central European populations in K=7 to K=9 (from CEU Bell Beakers to Denmark LN to Hungarian EBA to Iberia BA, in a sort of “CEU BBC ancestry” potentially related to North-West Indo-Europeans), but still, I prefer to go for a theoretically more correct visualization instead of cherry-picking the ‘best-looking’ results.

Since I made fun of the search for “Siberian ancestry” in coloured components in Tambets et al. 2018, I have to be consistent and preferred to avoid doing the same here…

In the first publication (in January) and subsequent minor revisions until March, I trusted analyses and ancestry estimates reported by amateurs in 2018, which I used for the text adding my own interpretations. Most of them have been refuted in papers from 2019, as you probably know if you have followed this blog (see very recent examples here, here, or here), compelling me to delete or change them again, and again, and again. I don’t have experience from previous years, although the current pattern must have been evidently repeated many times over, or else we would be still talking about such previous analyses as being confirmed today…

I wanted to be one step ahead of peer-reviewed publications in the books, but I prefer now to go for something safe in the book series, rather than having one potentially interesting prediction – which may or may not be right – and ten huge mistakes that I would have helped to endlessly redistribute among my readers (online and now in print) based on some cherry-picked pairwise comparisons. This is especially true when predictions of “Steppe“- and/or “Siberian“-related ancestry have been published, which, for some reason, seem to go horribly wrong most of the time.

I am sure whole books can be written about why and how this happened (and how this is going to keep happening), based on psychology and sociology, but the reasons are irrelevant, and that would be a futile effort; like writing books about glottochronology and its intermittent popularity due to misunderstood scientist trends. The most efficient way to deal with this problem is to avoid such information altogether, because – as you can see in the current revised text – they wouldn’t really add anything essential to the content of these books, anyway.

Continue reading

Official site of the book series:
A Song of Sheep and Horses: eurafrasia nostratica, eurasia indouralica

More Hungarian Conquerors of hg. N1c-Z1936, and the expansion of ‘Altaic-Uralic’ N1c

Open access Y-chromosomal connection between Hungarians and geographically distant populations of the Ural Mountain region and West Siberia, by Post et al. Scientific Reports (2019) 9:7786.

Hungarian Conquerors

More interesting than the study of modern populations of the paper is the following excerpt from the introduction, referring to a paper that is likely in preparation, Európai És Ázsiai Apai Genetikai Vonalak A Honfoglaló Magyar Törzsekben, by Fóthi, E., Fehér, T., Fóthi, Á. & Keyser, C., Avicenna Institute of Middle Eastern Studies (2019):

Certain chr-Y lineages from haplogroup (hg) N have been proposed to be associated with the spread of Uralic languages. So far, hg N3 has not been reported for Indo-European speaking populations in Central Europe, but it is present among Hungarians, although the proportion of hg N in the paternal gene pool of present-day Hungarians is only marginal (up to 4%) compared to other Uralic speaking populations. It has been shown earlier that one of the sub-clades of hg N – N3a4-Z1936 – could be a potential link between two Ugric speaking populations: the Hungarians and the Mansi. It is also notable that some ancient Hungarian samples from the 9th and 10th century Carpathian Basin belonged to this hg N sub-clade: Three Z1936 samples were found in the Upper-Tisza area (Karos II, Bodrogszerdahely/Streda nad Bodrogom) and two in the Middle-Tisza basin cemeteries (Nagykörű and Tiszakécske). The haplotype of the Nagykörű sample is identical with one contemporary Hungarian sample from Transylvania that tested positive for B545 marker downstream of N3a4-Z193632. Similar findings come from the maternal gene pool of historical Hungarians: the analyses of early medieval aDNA samples from Karos-Eperjesszög cemeteries revealed the presence of mtDNA hgs of East Asian provenance.

A commenter recently wrote that in a study by Fehér (probably this one) two Hungarian conquerors, from Ormenykut and Tuzser, will be of hg. N1c-2110. Assuming no other lineages will appear, this would leave the proportion of N1c-L392 vs. R1a-Z280/Z93 closer to the reported proportion of hg. N vs. R1a (5 vs. 2) among Sargat samples, and is thus compatible with a direct migration of Hungarians from around the Urals.

However, the sampling of Iron Age populations around the Urals is scarce, and we don’t know what other lineages these studied Magyars will have, but – based on the known variability of the published ones, and on the ca. 50-60 early Magyar males available to date in previous studies to obtain Y-chromosome haplogroups – I would say these reported N1c lineages are just a tiny proportion of what’s to come…

“Altaic-Uralic” N1c

Phylogenetic tree of hg N3a4. Phylogenetic tree of 33 high coverage Y-chromosomes from
haplogroup N3a4 was reconstructed with BEAST v.1.7.5 software package.

Archaeogenetic studies based on mtDNA haplotypes have shown that ancient Hungarians were relatively close to contemporary Bashkirs who are a Turkic speaking population residing in the Volga-Ural region. Another study reported excessive identical-by-descent (IBD) genomic segments shared between the Ob-Ugric speaking Khantys and Bashkirs but a moderate IBD sharing between Turkic speaking Tatars and their neighbours including Bashkirs.

Phylogenetic tree of hg N3a4 has two main sub-clades defined by markers B535 and B539 that diverged around 4.9 kya (95% confidence interval [CI] = 3.7–6.3 kya). Inner sub-clades of N3a4-B539 (defined by markers B540 and B545) split 4.2 kya (95% CI = 3.0–5.6 kya). (…) The phylogenetic tree reveals that all five Hungarian samples belong to N3a4-B539 sub-clade that they share with Ob-Ugric speaking Khanty and Mansi, and Turkic speaking Bashkirs and Tatars from the Volga-Ural region. Hungarian and Bashkir chrY lineages belong to both sub-clades of N3a4-B539.

Modern distribution of the “Ugric N1c”

To test the presence and proportions of hg N3a4 lineages in a more comprehensive sample set and with a higher phylogenetic resolution level compared to earlier studies, we analysed the genotyping data of about 5000 Eurasian individuals, including West Siberian Mansi and Khanty who are linguistically closest to Hungarians

Map of the entire hg N3a4.

There is a clear difference in geographic distribution patterns of these two hg N3a4 sub-clades. Hg N3a4-B535 (Fig. 3b) is common mostly among Finnic (Finns, Karelians, Vepsas, Estonians) and Saami speaking populations in North eastern Europe. The highest frequency is detected in Finns (~44%) but it also reaches up to 32% in Vepsas and around 20% in Karelians, Saamis and North Russians. The latter are known to have changed their language or to be an admixed population with reported similar genetic composition to their Finnic speaking neighbors. The frequency of N3a4-B535 rapidly decreases towards south to around 5% in Estonians, being almost absent in Latvians (1%) and not found among Lithuanians. Towards east its frequency is from 1–9% among Eastern European Russians and populations of the Volga-Ural region such as Komis, Mordvins and Chuvashes (…)

Map of N3a4 subclades defined by B535.

Hg N3a4-B539, on the other hand, is prevalent among Turkic speaking Bashkirs and also found in Tatars but is entirely missing from other populations of the Volga-Ural region such as Uralic speaking Udmurts, Maris, Komis and Mordvins, and in Northeast Europe, where instead N3a4-B535 lineages are frequent. Besides Bashkirs and Tatars in Volga-Ural region, N3a4-B539 is substantially represented in West Siberia among Ugric speaking Mansis and Khantys. Among Hungarians, however, N3a4-B539 has a subtle frequency of 1–4%.

Map of N3a4 subclades defined by B539, with a local snapshot showing the N3a4-B539 distribution among Hungarian speakers.

The battle to appropriate N1c-L392

So, basically, the team of Kristiina Tambets is arguing that N1c-VL29 expanded Finnic to the East Baltic (hence from a common Finno-Mordvinic dialect splitting ca. 600 BC on?) because, you know, apparently the agreed separation of known Uralic dialects from ca. 2000 BC, and their Bronze Age presence around the Baltic, is not valid when you follow haplogroups instead of languages or archaeology.

But now this other group of Tambets (co-author of this paper) considers that hg. N1c-Z1936 – which is probably behind the N1c-L392 samples from Lovozero Ware in the Kola Peninsula – represent either the True Uralic-speaking Palaeo-Arctic peoples, or else merely Ugric-speaking peoples which happened to expand to Fennoscandia but left no trace of their language…

To accept this identification you only have to NOT ask why:

  • N1c is first found in ancient cultures close to Lake Baikal.
  • N1c-L392 appears in ancient East Asian populations speaking completely different languages, with Altaic and Uralic being just some among many Palaeo-Siberian populations where the haplogroup will pop up.
  • Turkic populations like Bashkirs and Tatars (who expanded to the Volga through the southern Urals before the expansion of Hungarians) show a shared distribution of the B539 haplotype with Hungarians.
  • The phylogenetic tree and areas of N1c-L392 expansions don’t make any sense in light of the known linguistic and cultural expansions of Uralic-speaking peoples.

In fact, the Hungarian research group of Neparáczki – publishing the recent paper on Hungarian Conquerors – was apparently looking for a connection with Turkic peoples to support some traditional Turanian myths, and they found it in some scattered R1a-Z93 samples which supposedly connect Hungarian Conquerors to Huns (?), instead of looking for this closer link through N1c-Z1936 (especially haplotype B539)…

Also, is it me or are there two opposed trends with completely different interpretations among researchers publishing papers about hg. N1c: one systematically arguing for Altaic origins, and another for Uralic ones?

If somebody sees some complex reasoning behind the discussions of all these recent papers, beyond the simplest “let’s follow N for Uralic/Altaic”, feel free to comment below. Just so I can understand what I might be doing wrong in assessing Neolithic and Bronze Age migrations in linguistics and archaeology with help of ancient haplogroups coupled with ancestral components, but these researchers are doing right by playing with obsessive ideas born out of the 2000s coupled with phylogenetic trees and maps of modern haplogroup distributions…

This is probably going to be this blog’s most used image in 2019:



Mitogenomes suggest rapid expansion of domesticated horse before 3500 BC

Open access Origin and spread of Thoroughbred racehorses inferred from complete mitochondrial genome sequences: Phylogenomic and Bayesian coalescent perspectives, by Yoon et al. PLOS One (2018).

Abstract (emphasis mine)

The Thoroughbred horse breed was developed primarily for racing, and has a significant contribution to the qualitative improvement of many other horse breeds. Despite the importance of Thoroughbred racehorses in historical, cultural, and economical viewpoints, there was no temporal and spatial dynamics of them using the mitogenome sequences. To explore this topic, the complete mitochondrial genome sequences of 14 Thoroughbreds and two Przewalski’s horses were determined. These sequences were analyzed together along with 151 previously published horse mitochondrial genomes from a range of breeds across the globe using a Bayesian coalescent approach as well as Bayesian inference and maximum likelihood methods. The racing horses were revealed to have multiple maternal origins and to be closely related to horses from one Asian, two Middle Eastern, and five European breeds. Thoroughbred horse breed was not directly related to the Przewalski’s horse which has been regarded as the closest taxon to the all domestic horses and the only true wild horse species left in the world. Our phylogenomic analyses also supported that there was no apparent correlation between geographic origin or breed and the evolution of global horses. The most recent common ancestor of the Thoroughbreds lived approximately 8,100–111,500 years ago, which was significantly younger than the most recent common ancestor of modern horses (0.7286 My). Bayesian skyline plot revealed that the population expansion of modern horses, including Thoroughbreds, occurred approximately 5,500–11,000 years ago, which coincide with the start of domestication. This is the first phylogenomic study on the Thoroughbred racehorse in association with its spatio-temporal dynamics. The database and genetic history information of Thoroughbred mitogenomes obtained from the present study provide useful information for future horse improvement projects, as well as for the study of horse genomics, conservation, and in association with its geographical distribution.

Bayesian skyline plot (BSP) based on mitochondrial genome sequences from 167 modern horses.
The dark line in the BSP represents the estimated effective population size through time. The green area represents the 95% highest posterior density confidence intervals for this estimate.

Interesting excerpts:

We carried out a Bayesian coalescent approach using extended mitochondrial genome sequences from 167 horses in order to further assess the timescale of horse domestication. Here, we first calculated the time of the most recent common ancestor of Thoroughbred horses. Our analysis revealed the age of the most recent common ancestor of the racing horse to be around 8,100–111,500 years old. This estimate is much younger than that of the most recent common ancestor of the global horses, which has been estimated at 0.7286 Mys old.

Bayesian maximum clade credibility phylogenomic tree on the ground of the mitochondrial genome sequences of 167 modern horses.
The data set (16,432 base pairs) was also analyzed phylogenetically using Bayesian inference (BI) and maximum likelihood (ML) methods which showed the same topologies. 95% Highest Posterior Density of node heights are shown by blue bars. Groups are marked by a “G”. Numbers at the nodes represent (left to right): posterior probabilities (≥0.80) for the BI tree and bootstrap values (≥70%) for the ML tree. The racing horses were revealed to have multiple maternal origins and to be closely related to horses from one Asian, two Middle Eastern, and five European breeds. Results of phylogenomic analyses also uncovered no apparent association between geographic origin or breed and heterogeneity of global horses. The most recent common ancestor of the Thoroughbreds lived approximately 8,100–111,500 years ago, which was significantly younger than the most recent common ancestor of modern horses (0.7286 My).

On the domestication time of modern horses, there have been several publications derived from both archaeological [49–51] and molecular [11–12, 23, 48] evidences. D’Andrade [49] reported that the origin of domestic horses was around 4,000 years ago. Ludwig et al. [50] stated the domestication time to be about 5,000 years ago, while Anthony [51] noted that horse rearing by humans may have occurred approximately 6,000 years ago. Subsequently, on the basis of mitochondrial genome sequences, Lippold et al. [11] and Achilli et al. [12] postulated domestication time to be about 6,000–8,000 and 6,000–7,000 years ago, respectively. Warmuth [48] dated domestication time to 5,500 years ago based on autosomal genotype data, while Orlando et al. [23] claimed that Przewalski’s and domestic horse populations diverged 38,000–72,000 years ago based on analysis of genome sequences. In contrast to the previous hypothesized date of horse domestication, the results of our Bayesian skyline plot (BSP) analysis depict a rapid expansion of the horse population approximately 5,500–11,000 years ago, which coincides with the start of domestication.

It seems that we will not have an update on horse aDNA from the ISBA 8, so we will have to make do with this for the moment.


Bantu distinguished from Khoe by uniparental markers, not genome-wide autosomal admixture


The role of matrilineality in shaping patterns of Y chromosome and mtDNA sequence variation in southwestern Angola, by Oliveira et al. bioRxiv (2018).

Interesting excerpts (emphasis mine):

The origins of NRY diversity in SW Angola

In accordance with our previous mtDNA study9, the present NRY analysis reveals a major division between the Kx’a-speaking !Xun and the Bantu-speaking groups, whose paternal genetic ancestry does not display any old remnant lineages, or a clear link to pre-Bantu eastern African migrants introducing Khoe-Kwadi languages and pastoralism into southern Africa (cf. 15). This is especially evident in the distribution of the eastern African subhaplogroup E1b1b1b2b29, which reaches the highest frequency in the !Xun (25%) and not in the formerly Kwadi-speaking Kwepe (7%). This observation, together with recent genome-wide estimates of 9-22% of eastern African ancestry in other Kx’a and Tuu-speaking groups35, suggests that eastern African admixture was not restricted to present-day Khoe-Kwadi speakers. Alternatively, it is likely that the dispersal of pastoralism and Khoe-Kwadi languages involved a series of punctuated contacts that led to a wide variety of cultural, genetic and linguistic outcomes, including possible shifts to Khoe-Kwadi by originally Bantu-speaking peoples36.

Although traces of an ancestral pre-Bantu population may yet be found in autosomal genome-wide studies, the extant variation in both uniparental markers strongly supports a scenario in which all groups of the Angolan Namib share most of their genetic ancestry with other Bantu groups but became increasingly differentiated within the highly stratified social context of SW African pastoral societies11.

Y chromosome phylogeny, haplogroup distribution and map of the sampling locations. The phylogenetic tree was reconstructed in BEAST based on 2,379 SNPs and is in accordance with the known Y chromosome topology. Main haplogroup clades and their labels are shown with different colors. Age estimates are reported in italics near each node, with the TMRCA of main haplogroups shown with their corresponding color. A map of the sampling locations, re-used with permission from Oliveira et al. (2018) 9, is shown on the bottom left, and the haplogroup distribution per population is shown on the bottom right, with color-coding corresponding to the phylogenetic tree.

The influence of socio-cultural behaviors on the diversity of NRY and mtDNA

A comparison of the NRY variation with previous mtDNA results for the same groups 9 identifies three main sex-specific patterns. First, gene flow from the Bantu into the !Xun is much higher for male than for female lineages (31% NRY vs. 3% mtDNA), similar to the reported male-biased patterns of gene flow from Bantu to Khoisan-speaking groups33, and from non-Pygmies to Pygmies in Central Africa 37. A comparable trend, involving exclusive introgression of NRY eastern African lineages into the !Xun (25%) was also found. (…)

Secondly, the levels of intrapopulation diversity in the Bantu-speaking peoples from the Namib were found to be consistently higher for mtDNA than for the NRY, reflecting the marked association between the Bantu expansion and the relatively young NRY E1b1a1a1 haplogroup, which has no parallel in mtDNA25,39. (…)

In the context of the Bantu expansions, these patterns have been mostly interpreted as the result of polygyny and/or higher levels of assimilation of females from resident forager communities38,40. However, most groups from the Angolan Namib are only mildly polygynous11 and ethnographic data suggest that the actual rates of polygyny in many populations may be insufficient to significantly reduce Nem2,41. In addition, the finding of a large Nef/ Nem ratio in the Himba (Fig. S5), who have almost no Khoisan-related mtDNA lineages9, indicates that female biased introgression cannot fully explain the observed patterns.

An alternative explanation may be sought in the prevailing matrilineal descent rules, which might have created a sex-specific structuring effect, similar to that proposed for patrilineal groups from Central Asia (…)

Bayesian skyline plots (BSP) of effective population size change through time, based on mtDNA (red) and the NRY (black). Thick lines show the mean estimates and dashed lines show the 95% HPD intervals. The vertical line highlights the 2 ky before present mark. Effective sizes are plotted on a log scale. Generation times of 25 and 31 years were assumed for mtDNA and the NRY, respectively32.

The third important sex-specific pattern observed in this study is the much lower amount of between-group differentiation for NRY than for mtDNA among Bantu-speaking populations (4.4% NRY vs. 20.2% mtDNA), in spite of the patrilocal residence patterns of all ethnic groups (Table S5). This difference can hardly be explained by unequal levels of introgression of “Khoisan” mtDNA lineages into the Bantu, since the percentage of mtDNA variation remains high (18.8%) when the Kuvale, who have high frequencies of “Khoisan”-related mtDNA, are excluded from the comparisons. It therefore seems more plausible that differentiation is higher in the mtDNA simply because there is more ancestral mtDNA than NRY variation that can be sorted among different populations (see 45). Moreover, due to the matriclanic organization of all Bantu-speaking communities, factors enhancing inter-group differentiation, like kin-structured migration and kin-structured founder effects46, would have been restricted to mtDNA. Finally, it is also likely that the discrepancy between among-group divergence of mtDNA and NRY might have been influenced by higher migration rates in males than females. In fact, although all Bantu-speaking populations have patrilocal residence patterns, the observance of endogamy rules severely constrains the between-group mobility of females. In this context, the children from extramarital unions involving members from different populations tend to be raised in the mother’s group, effectively increasing male versus female migration rates. Moreover, it is likely that, in the highly hierarchized setting of the Namib, most intergroup extramarital unions would involve men from dominant groups and women from peripatetic communities. This hypothesis is indirectly supported by the finding that in NRY-based clusters (but not in mtDNA) pastoralist populations are grouped together with peripatetic communities that share their cultural traits (Figs. S6 and 3b), suggesting that migration of NRY lineages follows a path that is similar to horizontally transmitted cultural features.


Phylogeny of leprosy, relevant for prehistoric Eurasian contacts


Some interesting studies were published at roughly the same time as Damgaard et al. (Nature 2018 and Science 2018), and that’s probably why they got little attention (at least by me).

Monica H. Green (also in Academia.edu), specialized in History of Medicine, summed up their relevance in Twitter quite well (her text is edited here for clarity):

I’ve been disappointed that three recent exceptional studies of one of the world’s most historically important diseases, leprosy, have gotten so little notice from the science communication. It will take me a few hours to lay out their significance. But I think it’s important to do so.

So, here are the new studies on historical distribution and evolutionary development of Mycobacterium leprae, one of two organisms that causes leprosy (fourth study dropped yesterday!).

  1. Phylogenomics and antimicrobial resistance of the leprosy bacillus Mycobacterium leprae, by Benjak et al., Nature Communications (2018) 9:352.
  2. Abstract:

    Leprosy is a chronic human disease caused by the yet-uncultured pathogen Mycobacterium leprae. Although readily curable with multidrug therapy (MDT), over 200,000 new cases are still reported annually. Here, we obtain M. leprae genome sequences from DNA extracted directly from patients’ skin biopsies using a customized protocol. Comparative and phylogenetic analysis of 154 genomes from 25 countries provides insight into evolution and antimicrobial resistance, uncovering lineages and phylogeographic trends, with the most ancestral strains linked to the Far East. In addition to known MDT-resistance mutations, we detect other mutations associated with antibiotic resistance, and retrace a potential stepwise emergence of extensive drug resistance in the pre-MDT era. Some of the previously undescribed mutations occur in genes that are apparently subject to positive selection, and two of these (ribD, fadD9) are restricted to drug-resistant strains. Finally, nonsense mutations in the nth excision repair gene are associated with greater sequence diversity and drug resistance.

  3. Ancient DNA study reveals HLA susceptibility locus for leprosy in medieval Europeans, by Krause-Kyora et al., Nature Communications (2018) 9:1569
  4. NOTE. I referred to this study in this blog.

  5. Ancient genomes reveal a high diversity of Mycobacterium leprae in medieval Europe, by Schuenemann et al., PLOS Pathogens (2018)
  6. Abstract:

    Studying ancient DNA allows us to retrace the evolutionary history of human pathogens, such as Mycobacterium leprae, the main causative agent of leprosy. Leprosy is one of the oldest recorded and most stigmatizing diseases in human history. The disease was prevalent in Europe until the 16th century and is still endemic in many countries with over 200,000 new cases reported annually. Previous worldwide studies on modern and European medieval M. leprae genomes revealed that they cluster into several distinct branches of which two were present in medieval Northwestern Europe. In this study, we analyzed 10 new medieval M. leprae genomes including the so far oldest M. leprae genome from one of the earliest known cases of leprosy in the United Kingdom—a skeleton from the Great Chesterford cemetery with a calibrated age of 415–545 C.E. This dataset provides a genetic time transect of M. leprae diversity in Europe over the past 1500 years. We find M. leprae strains from four distinct branches to be present in the Early Medieval Period, and strains from three different branches were detected within a single cemetery from the High Medieval Period. Altogether these findings suggest a higher genetic diversity of M. leprae strains in medieval Europe at various time points than previously assumed. The resulting more complex picture of the past phylogeography of leprosy in Europe impacts current phylogeographical models of M. leprae dissemination. It suggests alternative models for the past spread of leprosy such as a wide spread prevalence of strains from different branches in Eurasia already in Antiquity or maybe even an origin in Western Eurasia. Furthermore, these results highlight how studying ancient M. leprae strains improves understanding the history of leprosy worldwide.

  7. The genome sequence of a SNP type 3K strain of Mycobacterium leprae isolated from a seventh‐century Hungarian case of lepromatous leprosy, by Mendum et al., International Journal of Osteoarchaeology (2018).
  8. Abstract:

    We report on a Mycobacterium leprae genome isolated from the remains of an individual with lepromatous leprosy that were excavated from a seventh‐century Hungarian cemetery. We determined that the genome was from a single nucleotide polymorphism (SNP) type 3K0 M. leprae strain, a lineage that diverged early from other M. leprae lineages. This is one of the earliest 3K0 M. leprae genomes to be sequenced to date. A number of novel SNPs as well as SNPs characteristic of the 3K0 lineage were confirmed by conventional polymerase chain reaction and Sanger sequencing. Recovery of accompanying human DNA from the burial was poor, particularly when compared with that of the pathogen. Modern 3K0 M. leprae strains have only been isolated from East Asia and the Pacific, and so these findings require new scenarios to describe the origins and routes of dissemination of leprosy during antiquity that have resulted in the modern phylogeographical distribution of M. leprae.

A fifth study can be added to the list, which, though not as extensive, is significant because it validates findings of others: Mycobacterium leprae genomes from naturally infected nonhuman primates, by Honap et al. PLOS Neglected Tropical Diseases (2018).


Leprosy is caused by the bacterial pathogens Mycobacterium leprae and Mycobacterium lepromatosis. Apart from humans, animals such as nine-banded armadillos in the Americas and red squirrels in the British Isles are naturally infected with M. leprae. Natural leprosy has also been reported in certain nonhuman primates, but it is not known whether these occurrences are due to incidental infections by human M. leprae strains or by M. leprae strains specific to nonhuman primates. In this study, complete M. leprae genomes from three naturally infected nonhuman primates (a chimpanzee from Sierra Leone, a sooty mangabey from West Africa, and a cynomolgus macaque from The Philippines) were sequenced. Phylogenetic analyses showed that the cynomolgus macaque M. leprae strain is most closely related to a human M. leprae strain from New Caledonia, whereas the chimpanzee and sooty mangabey M. leprae strains belong to a human M. leprae lineage commonly found in West Africa. Additionally, samples from ring-tailed lemurs from the Bezà Mahafaly Special Reserve, Madagascar, and chimpanzees from Ngogo, Kibale National Park, Uganda, were screened using quantitative PCR assays, to assess the prevalence of M. leprae in wild nonhuman primates. However, these samples did not show evidence of M. leprae infection. Overall, this study adds genomic data for nonhuman primate M. leprae strains to the existing M. leprae literature and finds that this pathogen can be transmitted from humans to nonhuman primates as well as between nonhuman primate species. While the prevalence of natural leprosy in nonhuman primates is likely low, nevertheless, future studies should continue to explore the prevalence of leprosy-causing pathogens in the wild.

These five studies are doing whole-genome sequencing on either modern isolates of M. leprae, or genomic fragments retrieved from buried remains (aDNA). The main objective of all the studies is to understand the diversity of M. leprae, both in terms of its history and in terms of its present-day distribution. (Benjak et al. 2018 are especially concerned to study possible reasons for variance in multiple drug resistance).

The following comments are concerned only to discuss leprosy’s history.

So, let’s start with a common claim of the science communication pieces on Schuenemann et al. 2018, which was published last week. A common formula: “New Study Suggests Leprosy Began To Spread From Europe To The World“. Is it plausible that Europe was where leprosy originated as human disease?

The answer, actually, is no. There’s two reasons for this, one having to do with chronology, the other with geography.

For chronology, these studies cumulatively suggest we are looking at a bottleneck. The current Time to Most Recent Common Ancestor (TMRCA) suggested for the divergence of M. leprae from its closest known “cousin,” M. lepromatosis (which also causes leprosy in humans) is estimated to be ca. 13.9 million years. There were no humans around 13.9M ya. So we cannot have been M. leprae‘s original host. All studies being discussed here agree on a consensus phylogeny, which puts the origin of all known strains of M. leprae at about 4-5K ya. So when we talk about the “origin” of M. leprae, we only talking about those lineages formed after this bottleneck.

Next we have to look at geography. Let’s start with this statement from the most recent study, Mendum et al. 2018, which is discussing a genome sequenced from an individual in Hungary from the 7th c. CE: “Modern 3K0 M. leprae strains have only been isolated from East Asia and the Pacific and so these findings require new scenarios to describe the origins and routes of dissemination of leprosy during antiquity that have resulted in the modern phylogeographical distribution of M. leprae.”

Okay, so stop and consider the implications of this. We have someone from 7th c. Hungary with leprosy. The strain of M. leprae that he has is not most closely related to strains sequenced earlier in Denmark or Sweden or England (see Schuenemann et al. 2018, refs. 9, 20, & 21). Rather, the strain he has (3K0) is most closely related to modern strains currently documented on the Pacific Rim, a very, very long way from Hungary. Here are the summary reflections of Mendum et al. 2018:

The global distribution of 3K0 and 3K1 strains is today restricted to regions of the Western Pacific such as Japan (except Okinawa), Korea, China, The Philippines, New Caledonia and Indonesia amongst others (Kai et al, 2013; Avanzi et al, 2015; Monot el al, 2009; Weng et al, 2013; Honap et al, 2018). This could indicate that the 3K lineage originated in Northern or Eastern Asia. The presence of two type 3K cases (KD271 and 222) in early medieval Hungary would then suggest a route of dissemination from Asia to central Europe, perhaps via trade links or migrations. This would be consistent with what is known of the origins of the Pannonian Avars, who are believed to have reached the Hungarian plain from the Eurasian steppe in the late 6th to early 9th centuries (Curta, 2006). The other possibility is that Europe was a centre of dissemination of the ancestral 3K0 and related strains, some of which later became less common or even absent from Europe but persisted in East Asia and the Pacific. Determining the likelihood of each of these scenarios will require more sampling and characterisation of both ancient and modern strains.

Two things bear stressing:

  1. The lineage in which the Hungarian sample has been placed, Lineage 0, has now been documented in historical remains from Denmark, too. (Schuenemann et al. 2018) So whatever transmission routes are postulated to connect the Pacific Rim to Hungary, we will also need to postulate routes to connect the Pacific Rim to Denmark.
  2. Schuenemann et al. 2018 document four of the five known M. leprae lineages in medieval western Europe. (Mendum et al. 2018 now declare the existence of 6 lineages; see tree.)
Phylogenetic relationships between selected modern (regular text) and ancient (bold text) M. leprae strains. The phylogeny was inferred by the Maximum Likelihood method of MEGA7 (Kumar et al, 2016) and the Tamura-3-Parameter model. The tree with the highest log likelihood value is shown. Bootstrap percentages from 1000 replicates are shown next to the branches. The scale indicates the number of substitutions per site. All positions with less than 90% site coverage were eliminated. M. lepromatosis was used as an outgroup (not shown). CM1 and Br15-1 are derived from a cynomolgus macaque and a red squirrel respectively.

Now, remember that we also need to keep chronology in mind: Lineage 0 is thought to have diverged from the common ancestor of Lineages 1-4 at least 3.5K ya. (Here’s the phylogenetic tree from Benjak et al. 2018, which I have marked with time divisions for emphasis.)

Modified image by Monica H. Green. Phylogeny of M. leprae. Bayesian phylogenetic tree of 146 genomes of M. leprae calculated with BEAST 2.4.4. Hypermutated samples with mutations in the nth gene were excluded from the analysis. The tree is drawn to scale, with branch lengths representing years of age. Samples were binned according to geographic origin as given in the legend. Posterior probabilities for each node are shown in gray. Location probabilities of nodes were inferred by the Discrete Phylogeny model

So what we need to explain is how a strain (Lineage 0, or 3K0 as Mendum et al. 2018 call it) can be found all the way from Denmark to New Caledonia. An “Out of Europe” narrative isn’t really helpful, any more than the earlier “Out of Africa” narrative worked.

Given the extreme amount of suffering leprosy has caused, and continues to cause around the world, and given the extraordinary investigative power that paleogenetics has now developed, it’s really time that we did a better job pulling these global narratives together.

If you have Twitter, be sure to retweet this thread!

NOTE. Another (probably also interesting) article was published recently, Digging up the plague: A diachronic comparison of aDNA confirmed plague burials and associated burial customs in Germany, by Gutsmiedl-Schümann, Praehistorische Zeitschrift (2018) 92:2, but sadly my university does not have access to it.


Plague outbreaks in the past are mainly known from written sources; in particular, the Justinianic Plague of the Early Middle Ages and the Black Death of the Late Middle Ages have been described in vivid detail. Yet prior to the introduction of aDNA analysis, it was often quite difficult to associate burials with plague beyond doubt – especially in areas where written evidence of the plague is scarce. As analysis of ancient DNA now allows the detection of plague victims in the archaeological record, new ways are being developed for combining archaeological, historical and ancient DNA research. In this paper we would like to present and compare known examples of plague graves from the Early Middle Ages, the Late Middle Ages and the Thirty Years’ War in Germany that have also been confirmed by ancient DNA analyses. We would like to argue for a differentiated view of the burial customs, especially when more than one plague victim shared a grave, and would like to show possible conclusions, drawn from the aDNA-confirmed plague burials, that can indicate the different strategies adopted by ancient societies to deal with catastrophic events like a pandemic disease.


Yet another Bayesian phylogenetic tree – now for Dravidian


Open access A Bayesian phylogenetic study of the Dravidian language family, by Kolipakam et al. (including Bouckaert and Gray), Royal Society Open Science (2018).

Abstract (emphasis mine):

The Dravidian language family consists of about 80 varieties (Hammarström H. 2016 Glottolog 2.7) spoken by 220 million people across southern and central India and surrounding countries (Steever SB. 1998 In The Dravidian languages (ed. SB Steever), pp. 1–39: 1). Neither the geographical origin of the Dravidian language homeland nor its exact dispersal through time are known. The history of these languages is crucial for understanding prehistory in Eurasia, because despite their current restricted range, these languages played a significant role in influencing other language groups including Indo-Aryan (Indo-European) and Munda (Austroasiatic) speakers. Here, we report the results of a Bayesian phylogenetic analysis of cognate-coded lexical data, elicited first hand from native speakers, to investigate the subgrouping of the Dravidian language family, and provide dates for the major points of diversification. Our results indicate that the Dravidian language family is approximately 4500 years old, a finding that corresponds well with earlier linguistic and archaeological studies. The main branches of the Dravidian language family (North, Central, South I, South II) are recovered, although the placement of languages within these main branches diverges from previous classifications. We find considerable uncertainty with regard to the relationships between the main branches.

MCC tree summary of the posterior probability distribution of the tree sample generated by the analysis with the relaxed covarion model with relative mutation rates estimated. Node bars give the 95% highest posterior density (HPD) limits of the node heights. Numbers over branches give the posterior probability of the node to the right (range 0–1). Colour coding of the branches gives subgroup affiliation: red, South I; blue, Central; purple, North; yellow, South II.

With every new paper using these revamped pseudoscientific linguistic methods popular in the early 2000s, including glottochronology, Swadesh lists, phylogenetic trees, mutation rates, etc. I feel a little more like Sergeant Murtaugh…

Featured image, from the article: “Map of the Dravidian languages in India, Pakistan, Afghanistan and Nepal adapted from Ethnologue [2]. Each polygon represents a language variety (language or dialect). Colours correspond to subgroups (see text). The three large South I languages, Kannada, Tamil and Malayalam are light red, while the smaller South I languages are bright red. Languages present in the dataset used in this paper are indicated by name, with languages with long (950 + years) literatures in bold.”

See also: