Proto-Indo-European from NW Indian subcontinent, Iran or SC Asia - A Proposal based on Genetics

ABSTRACT

qpGraph modeling of the Steppe_Eneolithic samples shows that they received up to 40% of ancestry from the common ancestors of the later Indus Valley Civilization (IVC or SSC). The Steppe_Eneolithic ancestry from the piedmont steppe from 4336-4047 BCE is the major source of subsequent cultures like Yamnaya & Corded Ware, which are widely suggested to be the vector of the language spread of European language families. This genetic connection, therefore, offers a significant genetic clue to the common ancestor of the first Proto-Indo-European language speakers.

THE SOURCE OF THE YAMNAYA ANCESTRY

The Yamnaya culture or culture with an autosomal ancestry profile similar to that of Yamnaya is widely believed to be the source of the Indo-European languages of Europe.

Western and Eastern Europe came into contact ∼4,500 years ago, as the Late Neolithic Corded Ware people from Germany traced ∼75% of their ancestry to the Yamnaya, documenting a massive migration into the heartland of Europe from its eastern periphery. This steppe ancestry persisted in all sampled central Europeans until at least ∼3,000 years ago, and is ubiquitous in present-day Europeans. These results provide support for a steppe origin of at least some of the Indo-European languages of Europe. - Haak et al, 2015 ⁽¹⁾

The source of the Iranian-like ancestry in Yamnaya has always been a mystery. The confusion flows from this formal F3 admixture test where both CHG (Caucasus Hunter-Gatherer, 2 samples found in Georgia) and IranN (Ganj Dareh Zagros herders, 10kya samples from Zagros mountains) along with EHG as source show negative F3 and significant Z-Scores, indicating that both are plausible sources.

Below is a Twitter conversation between Dr. Mathieson from U. Penn and Dr. Laziridis from Harvard University. It is quite clear that neither Caucasus Hunter-Gatherers (CHG) nor Iran Neolithic (IranN) samples from Zagros quite fits the bill by themselves as a second source of the Yamnaya ancestry.

I think can't really distinguish between CHG and Iran Neolithic as Yamanya source.

— Iain Mathieson (@mathiesoniain) September 26, 2017

Neither CHG nor Iran_N is a good source for Yamnaya *on their own*

— Iosif Lazaridis (@iosif_lazaridis) September 26, 2017

I mean Yamnaya=EHG+(Either CHG or Iran_N or both or something similar)

— Iain Mathieson (@mathiesoniain) September 26, 2017

Wang et al, 2019 ⁽²⁾ found 3 samples from the Piedmont Steppe north of the Caucasus mountains dated to 4336-4047 BCE which proved to be very good fits as the ancestors of the Yamnaya culture. However, even this study could not pinpoint the actual source of the CHG/Iran ancestry.

North of the Caucasus, Eneolithic and BA individuals from the Samara region (5200–4000 BCE) carry an equal mixture of EHG- and CHG/Iranian ancestry, so-called ‘steppe ancestry’ that eventually spread further west, where it contributed substantially to present-day Europeans, and east to the Altai region as well as to South Asia. - Wang et al, 2019

Location of the Steppe Eneolithic Samples at Progress & Vonyuchka

The latest research. Chintalapati et al, 2022 preprint ⁽¹³⁾, shows that the admixture between EHG and Iranian-related ancestry in Yamnaya and Afanasievo occurred around 4400-4000 BCE.

To understand the timing of the formation of the early Steppe pastoralist-related groups, we applied DATES using pooled EHG and pooled Iranian Neolithic farmers. Focusing on the groups with the largest sample sizes, Yamnaya Samara (n=10) and Afanasievo (n=19), we inferred the admixture occurred between 40-45 generations before the individuals lived, translating to an admixture timing of ~4,100 BCE . We obtained qualitatively similar dates across four Yamnaya and one Afanasievo groups, consistent with the findings that these groups descend from a recent common ancestor.  

 

Thus, we combined all early Steppe pastoralist individuals in one group to obtain a more precise estimate for the genetic formation of proto-Yamnaya of ~4,400 to 4,000 BCE.

These admixture dates give another line of evidence to the theory that Steppe_eneolithic were the ancestors of Yamnaya, and were the first in the region to possess this unique EHG + Iranian ancestry.

THE IndiaN ANCESTRY

Shinde et al, 2019 ⁽³⁾ concluded that the 'Iranian like' ancestry which gives between 60-87% ancestry to the Indus Periphery samples is not the same ancestry as that of IranHG or IranN, but rather they share a common ancestor deep in time (split before 10000BCE). Since the study does not name this ancestry, I have decided to name it IndiaN, which will be used in my subsequent models. The location of these IndiaN people prior to mixing with AASI people of the Indian subcontinent is unknown, but they could well be inhabitants of NW India since their split with IranN 12000 ya, or they could be more recent migrants to NW India around 7000 ya (5000 BCE) where they mixed with AASI. We just don't know yet due to a lack of relevant samples. The IndiaN ancestry is so named not because we know it's the original location, but because it provides the largest chunk of the ancestry in modern people of the Indian subcontinent today.

Fig3 from Shinde et al 2019 shows IndiaN split first from a common ancestor

Using qpGraph(Patterson et al., 2012), we tested all possible simple trees relating the Iranian-related ancestry component of these groups, accounting for known admixtures (Anatolian farmer-related admixture into Hajji Firuz and Tepe Hissar and Andamanese hunter-gatherer-related admixture in the IVC Cline)(Figure S3), using an acceptance criterion for the model fitting that the maximum Z scores between observed and expected f-statistics was <3 or that the Akaike Information Criterion (AIC) was within 4 of the best-fit (Burnham and Anderson, 2004). The only consistently fitting models specified that the Iranian-related lineage contributing to the IVC Cline split from the Iranian-related lineages sampled from ancient genomes of the Iranian plateau before the latter separated from each other.

This qpGraph method ⁽⁵⁾ is what we will use to build our tree for Steppe_Eneolithic, which helps us overcome the limitation of the absence of IndiaN samples from the aDna record.

 

BUILDING THE TREE MODEL

The following sample labels from the available ancient and modern DNA datasets will be used to build the tree.

1. Mbuti.DG - Pygmy group from Modern Congo as Outgroup

2. China_Paleolithic - 2 samples, one from Tianyuan ~40kya and one from Amur River China ~33kya

3. Ong.SG - Modern Onge Andamanese

4. Yana_UP.SG - 2 samples from NE Siberia Yana RHS region, ~32kya, ancestry label ANE

5. Tarim_EMBA1 - 12 samples from Tarim Basin China, ~4-3.5kya, related to ANE ancestry

6. IronGates_Mesolithic - 29 samples from Serbia, ~9-8kya, also labeled WEHG

7. EHG_Karelia - 2 Samples from Western Russia, ~8.5kya, also labeled EHG

8. CaucasusHG or CHG - 2 samples from Georgia, ~13.5 - 9.5kya

9. GanjDareh_N - 10 IranN herder samples from Zagros, W Iran, ~10kya

10. IVC - 6 outlier samples from Shahr_i_Sokhta which are migrants from IVC, ~4.5kya

11. Steppe_Eneolithic - 3 samples from Piedmont steppe, ~6.4-6 kya

Model Parameters and Acceptance Rules

The common parameter file for qpGraph is pasted below. Allsnps: YES is used along with Inbreed: NO 

1. Z-Score threshold is set at 3, bad fits will be reported if the model has F Stats with Z-Scores above +3 or below -3, and we will look for a better model. 

2. All drift edges must be of non-zero drift length.

3. Sometimes, a 0 drift length edge as seen on the graph is actually non-zero (but minuscule) when the output file is inspected.

Starting from Scratch

1. Adding Mbuti, China_Paleo & Onge 

2. Adding Yana_UP

3. Adding Tarim_EMBA1

4. Adding IronGates_Mesolithic

5. Adding Karelia_EHG

6. Adding CHG_Georgia

7. Adding GanjDareh_N

8. Adding IVC

Some Notes on the findings of the preceding qpGraphs

1. Shahr_I_Sokhta IVC or Indus_Periphery is modeled as 29% AASI and 71% of IndiaN1 which itself is an admixture of IranN related ancestry (IndiaN) and 11% Tarim Basin related ancestry (related to ANE and WSHG). This is consistent with the conclusions of Shinde et al, 2019 & Narasimhan et al, 2019.

2. The hypothetical component AASI shares a common ancestor with Onge Andamanese deep in time, consistent with Narasimhan et al, 2019.

3. China Paleolithic and Onge share a common East Eurasian ancestor, corroborating with He et al ⁽⁶⁾, 2021. Minor archaic Human admixture in China_paleolithic is not seen as we havent included Denisovan or Neanderthal in our tree. 

4. CHG and IranN don't form a clade, consistent with Laziridis, 2018 preprint ⁽⁷⁾. CHG and IranN obtain 60% and 50% ancestry from a West Eurasian source respectively, consistent with Laziridis 2018 preprint. We also detect ~8% and ~14% basal/African ancestry in CHG and IranN respectively, which is close to the estimate of Laziridis 2018 preprint. The remaining ancestry in our tree is East Eurasian, unlike Laziridis 2018 which gives a mixture of ANE (MA1 related) and east Eurasian. This could be because Laziridis 2018 used the actual Dzudzuana sample in their paper, whereas we use a proxy WE node as the Dzudzuana sample has not been published yet.

5. EHG is 40% West Eurasian + 60% ancestry related to ancestors of Tarim_EMBA1. This is close to Laziridis et al., 2016 ⁽⁸⁾ model of 25% West Eurasian + 75% ANE related to AG2 (or MA1) as Tarim_EMBA is more east Eurasian Shifted than MA1 or AG2.

6. Tarim_EMBA is modeled as 70% ANE (related to Yana and 30% East Eurasian (related to the deep ancestor of Onge). This is close to Zhang et al, 2021 ⁽⁹⁾ model of 72% AG3 + 28% Northeast Asian ancestry.

7. This is the basic tree that we will work on in our next step. Please note that this is not the only possible tree for fitting all the samples, and neither can we say that this is the best fitting model. However, it passes all our criteria and we can proceed with the analysis using this as a baseline. The worst fit F stat has a Z-Score of 2.03 <3, and all the drift edges are non-zero. The edges that have a 0 label on them are rounded off by the program, the actual non-zero values of which can be seen in the respective .ggg files (supplement).

FITTING STEPPE_ENEOLITHIC

Now that we have the base model setup, it is a matter of adding Steppe_eneolithic to the graph and seeing what sources give the best fits. We will test various models one by one - choosing between EHG and CHG/IranN/IndiaN as sources, as a 2 source, 3 source, and 4 source model.

1. Steppe = EHG + CHG

A bad fit with worst ZScore of 6.6 for f4 (Yana, IVC; CHG, Steppe_En) indicating that Steppe needs ancestry from a source close to IVC. There are total 91 outlier F stats, each signaling a need for either IVC or IranN ancestry.

2. Steppe = EHG + IndiaN1

Since the previous model showed the need for IVC-related ancestry, we will try this model.

The model fails with the worst Z Score of -10 for f4 (Irongates, Steppe; CHG, IVC), signaling that CHG ancestry is needed for steppe. There are 100+ outlier F4 Stat, most of which signal a need for CHG ancestry in the steppe.

 

3. Steppe = EHG + IranN

The model fails with the worst F Stat Zscore = 11.15 for f4(Irongates, CHG; IraN, Steppe) signaling a need for CHG. 

Outlier Z Scores

With the above (failed) models, what we see is that apart from EHG, at least 2 sources for Iranian components are needed in Steppe_EN, CHG & IndiaN. 

4. Steppe = EHG + CHG + IranN

The model fails with worst F4 Z-Score of 4.64 for f4(Yana, IVC; IranN, Steppe) signaling that there is still a need for IVC-related ancestry. All the outlier F4 Z-scores also signals the same need. They are pasted below.

5. Steppe = EHG + CHG + IndiaN

Model is a success with the worst f4 Z-Score of 2.52. IranN ancestry is not needed, unlike in the above model where IVC-related ancestry was needed.

Steppe_Eneolithic = 50%EHG + 44% IndiaN1 + 6% CHG +- standard errors

FINAL MODEL

From the above, we can see that the that the only working model for Steppe_Eneolithic is a 3 source model of EHG + CHG + IndiaN with EHG and IndiaN being the major components.

However, the model whihc provides the best fit is a 4 source model which also includes IranN. It gives no F4 outliers, the lowest final score (best fit) as well as all non zero drift edges.

Worst F4 Z-Score 2.7. Steppe = 51% EHG + 40% IndiaN1 + 6% CHG + 3% IranN

DISCUSSION

No study so far has delved into the nature of the source of the Iranian related ancestry in the Steppe Eneolithic. The above qpGraph models conclude that the only models which fit for Steppe_en have to include EHG, CHG, and IndiaN as a source. The western Iranian Zagros herder-related ancestry has little part to play in the genesis of the Steppe ancestry. The graphs show a high prevalence (40%) of ancestry related to the ancestors of IVC people in the steppe profile. The usual caveats apply - we need to find the actual samples corresponding to the 5000-4000 BCE time period from the NW Indian subcontinent, SC Asian, and Eastern Iranian regions. qpGraph outputs may also change with the inclusion of other reference groups.

Narasimhan et al., 2019 provide the admixture date between AASI and IndiaN components in IVC samples from Shahr-i-Sokhta as 4483-3811 BCE. This could have occurred in two ways:

1. The IndiaN ancestry in IVC (node IndiaN3) resided near the IVC region and it was the AASI ancestry that moved to NW Indian subcontinent for the admixture.

OR

2. The AASI ancestry resided near the IVC region and IndiaN3 ancestry admixed with this ancestry from the west of it.

Ancient DNA from the NW Indian subcontinent region from 5000 BCE should give a definitive answer to this question. What is clear is that the same ancestors of IVC people gave ancestry to both IVC and Steppe post 6000 BCE, which provides evidence to explain the common source of Indo & European languages. 

THE INDO IRANIAN BRANCH OF IE

The linguistic case for a steppe homeland (Sintashta/Andronovo Horizon) of Proto-Indo-Iranian language assumes a (one or more) non-IE language substrate in NW of the Indian subcontinent. The hypothesis assumes a complete replacement of extant languages by the incoming people (predominantly male) from the steppes. However, there are many inconsistencies with this hypothesis as has been laid out in a comprehensive article by Jaydeepsinh Rathod ⁽¹¹⁾, using tens of references from the work of linguists. The article also argues for a much older presence of the IE language family in the Indian subcontinent than is proposed by the steppe theory (1500BCE). There have been other historians, linguists, and archaeologists who have also argued for a much older presence of IE languages in the subcontinent. This paper agrees with the assertion of the older presence of IE languages in the NW Indian subcontinent given the genetic contact between ancestors of IVC and Steppe_eneolithic.

The answer to the question regarding the linguistic nature of contact between the bronze age Steppe people and Indo Iranian speaking people of the Indian subcontinent & Iran remains unclear because of the lack of any extant texts from the steppe. It is clear that the steppe bronze age ancestry appeared in the Indo Iranian speaking regions post-2000 BCE, but the nature of linguistic contact, loanword exchange, etc needs more study.

THE PROTO-INDO-EUROPEAN HOMELAND

The crux of the above analysis is that the same source that provides the maximum amount of ancestry to modern Indians (especially north Indians & Pakistanis), also provides a big chunk of ancestry to the Steppe component. This steppe component is believed to be the vector of language spread to the ancestors of most IE-speaking Europeans today.

This validates the theory of an Iranian PIE homeland, also supported by Johannes Krause of Max Planck Institute in his 2021 book ⁽¹⁰⁾. However, given that the bulk of the Iran-like ancestry in the Steppe is related to IVC ancestry rather than CHG or western Iranian herder ancestry, the locus of PIE must be shifted to the east of what has been suggested in the book.  Another reason why a northwest Iranian PIE is unlikely is that the region had a big amount of Anatolian farmer ancestry already by 6000 BCE, which is missing from the Steppe Eneolithic. The high in Anatolian Farmer ancestry 6000 BCE Hajji Firuz chalcolithic samples are evidence of this. The location of these samples is actually inside the locus proposed by Krause (in the graphic below)

This eastern locus supports the hypothesis by the anthropologist from St. Petersburg State University, Alexander Kozintsev ⁽¹²⁾ among others, who proposed east of Caspian sea origin of PIE.

From Chapter 6:  A Short History of Humanity by J. Krause, Max Planck Institute

THE PROPOSAL FOR PIE

THE BIG PICTURE: Genetic admixture events post 6000BCE can explain the IE Language dispersal

TOOLS, DATA, AND OUTPUT FILES

1. The latest version of ADMIXTOOLS was used for qpGraph and convertf. Available here.

2. Plink 1.9 was used to make only a subset of required samples from the large eigenstrat database.

3. The genotype files and qpGraph parameter/input/output files are uploaded here, which can be used for verification and rebuilding the models.

REFERENCES

1 Haak, W., Lazaridis, I., Patterson, N. et al. Massive migration from the steppe was a source for Indo-European languages in Europe. Nature 522, 207–211 (2015). https://doi.org/10.1038/nature14317

2 Wang, CC., Reinhold, S., Kalmykov, A. et al. Ancient human genome-wide data from a 3000-year interval in the Caucasus corresponds with eco-geographic regions. Nat Commun 10, 590 (2019). https://doi.org/10.1038/s41467-018-08220-8

3 Shinde V, Narasimhan VM, Rohland N, et al. An Ancient Harappan Genome Lacks Ancestry from Steppe Pastoralists or Iranian Farmers. Cell. 2019;179(3):729-735.e10. doi:10.1016/j.cell.2019.08.048

4  Narasimhan VM, Patterson N, Moorjani P, et al. The formation of human populations in South and Central Asia. Science. 2019;365(6457):eaat7487. doi:10.1126/science.aat7487

5 Patterson N, Moorjani P, Luo Y, et al. Ancient admixture in human history. Genetics. 2012;192(3):1065-1093. doi:10.1534/genetics.112.145037

6 He G, Wang M, Zou X, et al. Peopling History of the Tibetan Plateau and Multiple Waves of Admixture of Tibetans Inferred From Both Ancient and Modern Genome-Wide Data. Front Genet. 2021;12:725243. Published 2021 Sep 3. doi:10.3389/fgene.2021.725243

7 Paleolithic DNA from the Caucasus reveals core of West Eurasian ancestry

Iosif Lazaridis, Anna Belfer-Cohen, Swapan Mallick, Nick Patterson, Olivia Cheronet, Nadin Rohland, Guy Bar-Oz, Ofer Bar-Yosef, Nino Jakeli, Eliso Kvavadze, David Lordkipanidze, Zinovi Matzkevich, Tengiz Meshveliani, Brendan J. Culleton, Douglas J. Kennett, Ron Pinhasi, David Reich

bioRxiv 423079; doi: https://doi.org/10.1101/423079

8 Lazaridis, I., Nadel, D., Rollefson, G. et al. Genomic insights into the origin of farming in the ancient Near East. Nature 536, 419–424 (2016). https://doi.org/10.1038/nature19310

9 Zhang, F., Ning, C., Scott, A. et al. The genomic origins of the Bronze Age Tarim Basin mummies. Nature 599, 256–261 (2021). https://doi.org/10.1038/s41586-021-04052-7

10 Krause J, 2021 A Short History of Humanity- How Migration Made Us Who We Are Penguin Books. Chapter 6 Europeans Find a Language

11 Rathod J., 2021 Can Linguistics prove AMT & reject OIT ?

12 Kozintsev A, 2019 Proto-Indo-Europeans: The Prologue Journal of Indo-European Studies, vol. 47 (3-4), pp.293-380

13 Chintalapati M., Patterson N., Moorjani P. et al. Reconstructing the spatiotemporal patterns of admixture during the European Holocene using a novel genomic dating method

bioRxiv 2022.01.18.476710; doi: https://doi.org/10.1101/2022.01.18.476710