Thursday, October 12, 2017

The Modern Genetics Of Island Southeast Asia

Bernard's blog has a nice summary of a new paper by Georgi Hudiashov, et al. entitled "Complex Patterns of Admixture across the Indonesian Archipelago", which examines the autosomal genetics of 498 individuals from 25 populations in the region.

Despite the title, some of the patterns and trends are very straight forward as figures 1A and 1B from the paper, below, illustrate:



There is a stark division between the populations to the west of the Wallace line, which line up along a mostly north to south axis along the far left side of the PCA analysis, and the populations to the east of the Wallace line which stretch in a west to east cline along the first principal component.

For the most part, the east of the Wallace line populations cluster with Austronesian populations on the second principal component. But, a pull driven by Austronesian admixture may be problematic, because individuals from the Papuan highland and coast with almost no Austronesian ancestry in ADMIXTURE are also found at that point on the second principal component. Oceania, however, looks more like Eastern Indonesia than it does like Melanesia.

Bernard, as translated from the French by Google, summarizes the ADMIXTURE analysis that was done as follows (emphasis added):
In the region there are two main components: a continental component in yellow and a Papuan component in violet. A third component often referred to as Austronesian appears in light blue from K=5. There is a clear difference between Western Indonesia which includes a high proportion of Austronesian ancestry and virtually no Papuan ancestry, and Eastern Indonesia which has little Austronesian ancestry and a high proportion of Papuan ancestry.
Basically, this suggests that while there were modern humans in Sumatra shortly after the Toba eruption, that there was a major migration by a genetically distinct population from mainland Southeast Asia, resulting in near total population replacement of what may have been a Papuan-like, Denisovan admixed population, around the time of the Last Glacial Maximum when low sea levels created a land bridge to the mainland, ca. 20,000 years ago. 

This would have happened despite the fact that both populations would have been modern human hunter-gatherers, perhaps because the early Indonesians, protected from invaders by the sea, didn't need to master the art of war until it was too late to catch up with the men of the mainland who hadn't had that luxury. 

This also corroborates other evidence that few mainland Asians are not the first wave of modern humans in the region.

Anomalies

Nias and Mentawai

The most "southerly" populations on the second principal component actually from small islands in western Indonesia, rather than up against the Wallace line. This is true despite the fact that they appear almost completely Austronesian, which is more "northerly" than these populations, in an ADMIXTURE analysis at K=5 to K=8 (after which they start to appear as distinct populations of their own). 

These populations actually have more of the "Austronesian" component than populations from the Taiwanese point of origin of the Austronesians or populations form the Philippines that appear to be closer to the original Austronesians than modern indigenous Taiwanese populations, which suggests that something is backwards in how ADMIXTURE generated its populations.

In other words, only part of the "Austronesian" component in the ADMIXTURE analysis is probably actually Austronesian, and the balance is probably what is giving these populations their "southerly" pull on the second principal component in the PCA chart.

Baining, Nasioi and Vanuatu 


Some of the Baining peopleNote that, despite superficial appearances, the Baining people are no more closely related to Africans than people from China or England or Hawaii.

Even more surprising are the Baining people who live in the Baining mountains the esatern part of the island of New Britain off the coast of Papua New Guinea who appear to be a relict population with a distinct language small language family. According to Wikipedia: 
They currently inhabit the Baining Mountains into where they are thought to have been driven by the Tolai tribes who migrated to the coastal areas in comparatively recent times. Another factor that might have influenced their migration inland was major volcanic activity that took place over centuries. (As recently as 1994, the nearby town of Rabaul was almost completely destroyed by two volcanoes, Tavurvur and Vulcan).
The Baining people (there are about 12,000 of them based upon the number of people who speak the five or six living Baining languages) are subsistence farmers with few animals who have been historically described as "the dullest people on Earth" for the very low levels of play, myths, festivals, religious traditions, or even interest in sex in their lives, and are compared to the early American "Puritans" in their lifestyle. They have one notable, somewhat all purpose ritual, a costumed fire-dance performed by young men which women and children do not participate in or watch. They are also notable for a very high level of adoption, about 36% of children, something promoted as "unnatural".
[T]he Baining shun the bush, which they view as chaotic and dangerous, and they derogate play, especially that among children.

According to Fajans, the Baining eschew everything that they see as “natural” and value activities and products that come from “work,” which they view as the opposite of play. Work, to them, is effort expended to overcome or resist the natural. To behave naturally is to them tantamount to behaving as an animal. The Baining say, “We are human because we work.” The tasks that make them human, in their view, are those of turning natural products (plants, animals, and babies) into human products (crops, livestock, and civilized human beings) through effortful work (cultivation, domestication, and disciplined childrearing).
They do not justify their dances or their life choices with religion or symbolic meaning.

The fifteen to nineteen languages of the general vicinity of the East Papuan area have some sub-groups with some similar pronouns, but many of these languages have almost no more similarity in vocabulary than would be expected from random chance, despite the fact that each is spoken by only a tiny number of people and many of the languages are spoken by people who live quite close to each other geographically and have done so for 35,000 years or so.

The Baining people are surprising, genetically, given their location, because they are very northerly indeed on the second principal component of the PCA chart, more so than the Han Chinese and Japanese, despite being also very Papuan on the first principal component. Similar, but much less strong impacts in that direction are seen on the adjacent island of Nasioi and in the near Oceanic island of Vanuatu, which probably derived from Baining admixture revealed in an ADMIXTURE analysis.

A naive explanation for the northerly anomaly would be that they arose during Japanese occupation. The effect seems very large for a fairly short duration event, but the base populations were small and there is no way to be sure how representative these samples are of these small populations. Cryptic maritime trade links to Japan or Northern China could also provide an explanation. 

But, ADMIXTURE analysis strongly disfavors this interpretation. Even at K=3, Baining Island has none of the predominant component found in the Japanese and Han Chinese, so the recent admixture or trade linkage theory is pretty definitively shot down. A distinctive Baining component emerges at K=6 making up almost all of the ancestry from that island of the northern coast of Papua New Guinea, and present in lower proportions in neighboring Naisio and nearby Vanuatu, and even even lower proportions elsewhere in Oceania (i.e. Tonga, Samoa and Tahiti), where a distinct Oceanian variant of this component emerges at K=8. 

The Clade based analysis below also shows the Baining people as sharing a clade with Papuans and in admixture at K=3 to K=5 they are almost indistinguishable from Papuans.

Thus, it seems likely that the divide between the Baining people and Papuans on the northern-southern second principal component, rather than being attributable to late Holocene admixture, probably has deep roots indeed. But, as a farming culture (presumably of indigenous Papuan crops as the Papuans were an independent Neolithic center), their ethnogenesis can't be much older than the Holocene (i.e. 10,000 years ago) either. It could be that the Baining people, given their small population and long standing isolation from the larger populations of Papuan New Guinea in the pre-Austronesian era (and apparently even after that development) have simply been more susceptible to genetic drift than people on Papua New Guinea itself who may have had low levels of bride or husband exchange sufficient to harmonize the genetic subpopulations somewhat, and that this is manifesting in what looks like a great northerly divergence on the second principal component in this study. 

A Clade Based Analysis

While we know that a branching clade model is not a great description of the historical process by which Island Southeast Asia acquired its genetic diversity, the investigators did generate a clade based analysis with their data that captures the broad outlines of the conclusions reached by other means.



10 comments:

Ryan said...

"This would have happened despite the fact that both populations would have been modern human hunter-gatherers, perhaps because the early Indonesians, protected from invaders by the sea, didn't need to master the art of war until it was too late to catch up with the men of the mainland who hadn't had that luxury."

I don't think we have to peer too deeply for reasons for this turnover. Europeans today only get about 4% of their ancestry from the first wave of modern humans to show up in Europe. These sort of massive turnovers seem to be normal for humans even before the Neolithic.

"Basically, this suggests that while there were modern humans in Sumatra shortly after the Toba eruption, that there was a major migration by a genetically distinct population from mainland Southeast Asia, resulting in near total population replacement of what may have been a Papuan-like, Denisovan admixed population, around the time of the Last Glacial Maximum when low sea levels created a land bridge to the mainland, ca. 20,000 years ago."

I'm not so sure about that. I think there's merit in John Hawks' suggestion that the major Denisovan admixture found in Papuans today. Flores is east of the Wallace Line but shows evidence of human habitation for the last 800,000 years, and is far enough east to have not been devastated by Toba.

Keep in mind too that even in Reich's 2011 paper, Papuan/Australian Denisovan admixture occurred in a population unrelated to both mainland East Asians and other Oceanians like the Onge and Jehai. I guess what I'm getting at is that there may never have been a Papuan-like population west of the Wallace line, and that Papuans themselves descend from at least two distinct streams of modern humans.

http://www.sciencedirect.com/science/article/pii/S0002929711003958#fig3

As we discussed on Razib's blog before too, based on the contrast between Oceanian ancestry in the Americas and Denisovan ancestry in the Americas, there must have been at least two different admixture events with Denisovans.

andrew said...

"Europeans today only get about 4% of their ancestry from the first wave of modern humans to show up in Europe."

About 95% of Europe was uninhabitable because it was buried under a glacier during the LGM. The first wave population in Europe was eliminated except for three refugia (Franco-Cantrabrian, and one in Italy, one in the Caucasus). Much of Europe remained uninhabitable until well into the Mesolithic era.

"Flores is east of the Wallace Line but shows evidence of human habitation for the last 800,000 years, and is far enough east to have not been devastated by Toba."

There is evidence for hobbit habitation for a very long time on Flores, there is evidence of modern human habitation for 50-60 kya at most. The is not evidence of other hominins there.

"there may never have been a Papuan-like population west of the Wallace line, and that Papuans themselves descend from at least two distinct streams of modern humans."

Any population that had maritime capabilities sufficient to reach islands east of the Wallace line necessarily had the capabilities to cross that line by sea. There would have been no reason for first wave modern humans to stop at the Wallace line, particularly post-Toba, which would have decimated any H. erectus population in most of Island Southeast Asia to the east of the Wallace line, leaving a vacuum.

A don't see how it matters that they descend from one or more than one distinct stream of modern humans, given that both would have to far pre-date the LGM and the land bridge to mainland Asia given their genetics.

"Based on the contrast between Oceanian ancestry in the Americas and Denisovan ancestry in the Americas, there must have been at least two different admixture events with Denisovans."

There is no pre-Columbian Denisovan ancestry in the Americas. There was pre-Columbian American ancestry in Oceanians on Easter Island, but there is no evidence yet of gene flow in the other direction that survived to the post-Columbian era or that is found in ancient DNA. Presumably, a few Oceanians left some offspring in the Americas, but their descendants either died out, or were diluted beyond recognition. Contact with Oceanians cannot explain the tiny smidgen of Paleo-Asian ancestry found in one or two tribes in the deep Amazon and no where else in the New World, because Oceanians don't and didn't have Paleo-Asian DNA.

While I wouldn't rule out more than one admixture event with the Denisovans, it isn't necessary to explain the data. It would be plausible that the Negritos of the Philippines, Papuans, Australians, and the residual very low levels of Denisovan ancestry in mainland Asia arose from more than one pulse of admixture. But, I don't see how that would change the analysis much. Ancient DNA data combined with archeology and linguistics and modern population genetics has given us a precise and vivid idea about when and how Papuan and Denisovan ancestry made its way into Oceanians.


Ryan said...

"There is evidence for hobbit habitation for a very long time on Flores, there is evidence of modern human habitation for 50-60 kya at most. The is not evidence of other hominins there."

We don't know the genetic affinities of the hobbits though, or if they were alone. They could be the source of Denisovan ancestry in Papuans, or the source of homo erectus ancestry in Denisovans.

"Any population that had maritime capabilities sufficient to reach islands east of the Wallace line necessarily had the capabilities to cross that line by sea."

Some sort of Denisovan ancestor or homo erectus or even australopithecus or whatever the hobbits really were did have the maritime prowess to colonize islands east of the Wallace line - namely Flores (at least 800,000 years of habitation) and Sulawesi (at least 200,000 years of habitation), but did not manage to get past the Weber Line or Lydekker Line into the Sahul.

"There is no pre-Columbian Denisovan ancestry in the Americas."

This is incorrect. See for yourself.

https://academic.oup.com/mbe/article/32/10/2665/1210305

"There was pre-Columbian American ancestry in Oceanians on Easter Island, but there is no evidence yet of gene flow in the other direction that survived to the post-Columbian era or that is found in ancient DNA."

I'm not referring to this. The Oceanian ancestry in the Americas is not from recent admixture but from deep population structure, probably dating back to the first settlement of the Americas or even earlier. It's not recent Oceanian ancestry, but an ancient affinity to someone like the Onge that is unevenly distributed across the Americas.

https://www.nature.com/articles/nature14895.epdf?referrer_access_token=jLOJcw9ZkEUkitvKpb4icdRgN0jAjWel9jnR3ZoTv0N6yB-nEyCdRoL51ykMO5E91orKw8mXMzuxoORJJBQhEhMryvaarH9hFKiGoJjc1JvX7zNtkTMbwuOPicFdsM1y7uzyyWsvTeUZY4iN3Kq60ZhvW781KSFuU4fHlV0EhySXsodIVFx76RjUcfg3OxAfFtAfWeS579uPXTb94CfsRhYDaIFWuW50EJJArdUBaLgEMHNKjN0PA9CwAMcJjrQ0&tracking_referrer=www.nature.com

So the neat thing is when you take the two together, the indigenous Americans with the most Oceanian ancestry have the least Denisovan ancestry.

It's not just one or two tribes in the Amazon, but also in Central America, Argentina and the Andes.

South Asia has a similar issue, with Denisovan ancestry substantially exceeding what one would expect based on affinity with Papuans.

"While I wouldn't rule out more than one admixture event with the Denisovans, it isn't necessary to explain the data. It would be plausible that the Negritos of the Philippines, Papuans, Australians, and the residual very low levels of Denisovan ancestry in mainland Asia arose from more than one pulse of admixture. "

So what these papers find is that Denisovan and Papuan on islands more or less east of the Wallace Line are strongly are proportional to one another, but that same does not hold if you replace Papuans with Australian Aborigines. So they're saying all these island peoples get their Denisovan ancestry from a population related to Papuans from after they split with Australian Aborigines, and not from a single common ancestor. That may be from late settlement of these islands or from subsequent migrations from New Guinea - either directly or mediated by Polynesians.

The Mamanwa in the Phillipines do not follow this trend though IIRC, and probably represent a deeper split from the populations of the Sahul.

terryt said...

Responses to both, in no particular order:

"there is evidence of modern human habitation for 50-60 kya at most".

Ryan is correct in that human tools have been found on Flores dating back 800,000 years. Of course it we don't know whether that occupation was continuous or not.

"there may never have been a Papuan-like population west of the Wallace line, and that Papuans themselves descend from at least two distinct streams of modern humans".

I agree with the two steam model however it must be said that both streams must have come via Sundaland, and also must have mixed there even if that was after the two streams had reached New Guinea.

"Any population that had maritime capabilities sufficient to reach islands east of the Wallace line necessarily had the capabilities to cross that line by sea".

No so. Sundaland was at least to some extent continental at times of low sea level. It is therefore quite possible that a population had arrived in Sundaland before they had the ability to cross Wallace's Line. What I find interesting is that Y-DNA K2b forms separate clades in the island string. To me that suggests they spread along what is now the island chain at a time of lowered sea level and became isolated when sea level rose. Just one clade crossed the Line: K2b1. That may have been the first to adopt effective boating technology.

"A don't see how it matters that they descend from one or more than one distinct stream of modern humans, given that both would have to far pre-date the LGM and the land bridge to mainland Asia given their genetics".

But humans may have been in Australia by 70,000 years ago, another period of low sea level. But they must have crossed Wallace's line in boats to get there! Interesting that Y-DNA C1b divides into two clades: C1b1 west of the Line and C1b2 east of the line. To me that indicates that C1b spread from somewhere near Wallace's Line.

"There is no pre-Columbian Denisovan ancestry in the Americas".

Ryan is correct. There is Denisovan in America. And there is no reason to believe it is the product of movement from Polynesia. The Denisovan admixture presumably goes a very long way back in human history and may have been carried to America by the very first immigrants.

"Contact with Oceanians cannot explain the tiny smidgen of Paleo-Asian ancestry found in one or two tribes in the deep Amazon"

Exactly.

terryt said...

"The Oceanian ancestry in the Americas is not from recent admixture but from deep population structure, probably dating back to the first settlement of the Americas or even earlier. It's not recent Oceanian ancestry"

Correct.

"We don't know the genetic affinities of the hobbits though, or if they were alone. They could be the source of Denisovan ancestry in Papuans, or the source of homo erectus ancestry in Denisovans".

I doubt that very much. We know Homo erectus lived on Java and that could easily be the source of both genetic strains.

"Some sort of Denisovan ancestor or homo erectus or even australopithecus or whatever the hobbits really were did have the maritime prowess to colonize islands east of the Wallace line - namely Flores (at least 800,000 years of habitation) and Sulawesi (at least 200,000 years of habitation), but did not manage to get past the Weber Line or Lydekker Line into the Sahul".

Could have been accidental. And don't forget that elephants made it to Flores and several Eurasian species are found in Sulawesi.

"So the neat thing is when you take the two together, the indigenous Americans with the most Oceanian ancestry have the least Denisovan ancestry".

Very interesting.

"So what these papers find is that Denisovan and Papuan on islands more or less east of the Wallace Line are strongly are proportional to one another, but that same does not hold if you replace Papuans with Australian Aborigines. So they're saying all these island peoples get their Denisovan ancestry from a population related to Papuans from after they split with Australian Aborigines, and not from a single common ancestor. That may be from late settlement of these islands or from subsequent migrations from New Guinea - either directly or mediated by Polynesians".

To me I think your very first statement is relevant here:

"I don't think we have to peer too deeply for reasons for this turnover. Europeans today only get about 4% of their ancestry from the first wave of modern humans to show up in Europe. These sort of massive turnovers seem to be normal for humans even before the Neolithic".

I think it is quite possible that the reality fits the data as we have it at present. The Denisovan admixture happened in the Altai and later expansions have obliterated through much of its original spread, leaving the largest remaining remnant in the most isolated region: New Guinea and Australia.

andrew said...

@Ryan @terryt

I will review your citations and arguments and decide what I think about those arguments in a future post. I'm keeping an open mind.

Ryan said...

@terry - agreed on the rafting issue. Orangutans didn't make it past the Wallace Line though. Elephants I would not read as much into - they can swim very long distances. I saw an article recently about the Sri Lankan navy having to rescue two elephants that had been swept 10 miles out to sea lol. I think in order to have enough people / hobbits raft over they would need to regularly along the coast, exploring marine resources and maybe using reed rafts. Just exploiting marine resources is generally considered upper Paleolithic behaviour. These are the first humans anywhere to cross open water. They were ahead of their time. The barrier the sea represented is a good explanation for the elevated Denisovan ancestry because we would expect it to slow the migration of modern humans, allowing more demographic parity.

The Altai Denisovans are known to be a very very poor fits for the Denisovan admixture into us modern humans so I think we can actually rule there out as a major site of admixture. Our methods are biased to cold and dry sites that preserve bones and DNA well, but those are not areas that could support many people. More likely that the relevant events of our pre-history occurred in warmer, moisture climates. Those climates unfortunately also destroyed the evidence we would look to today. Coastal and estuarian sites of habitation would also be under water today, which potentially creates big blind spots in our view of prehistory.

terryt said...

"Elephants I would not read as much into"

True.

"I think in order to have enough people / hobbits raft over they would need to regularly along the coast, exploring marine resources and maybe using reed rafts".

That may be true, however it's strange that they were unable to move further along the island chain to the east of Wallace's Line if that were the case.

"These are the first humans anywhere to cross open water. They were ahead of their time".

I agree that it appears efficient boating may
have originated in the region and spread from there.

"The barrier the sea represented is a good explanation for the elevated Denisovan ancestry because we would expect it to slow the migration of modern humans, allowing more demographic parity".

It is perhaps the barrier that Wallace's Line represented that is the significant aspect here. The first people to cross the line had a higher level of Denisovan admixture than did later arrivals in the region.

"The Altai Denisovans are known to be a very very poor fits for the Denisovan admixture into us modern humans so I think we can actually rule there out as a major site of admixture".

The old story: we need more ancient DNA!

Ryan said...

@Terryt - Well I'm just pointing out we know for a fact the first to cross the line weren't modern humans. We just don't know if they're the source of the archaic ancestry or not.

"That may be true, however it's strange that they were unable to move further along the island chain to the east of Wallace's Line if that were the case. "

Currents? I have no idea what the currents would have been like there but if they couldn't get past the Weber line that could be why. Or just not enough people on the coast to accidentally raft any further (ie the accidental rafting to Flores and Sulawesi may have happened from a much more heavily populated Sundaland).

terryt said...

Good points there.