November 29, 2010

Sex bias in the prehistory of Homo sapiens?

There is a somewhat interesting new paper that finds some gender bias irregularities in the prehistory of our species:

Emery et al., Estimators of the Human Effective Sex Ratio Detect Sex Biases on Different Timescales, The American Journal of Human Genetics (2010), doi:10.1016/j.ajhg.2010.10.021

Fig. 3
They argue that there is a "female bias" (greater female transmitted diversity) in Africans and a "male bias" in non-Africans (Eurasians hereafter).

I broadly agree. But I think it needs a qualification: Africans in particular must be always studied with their whole structure, so any study that ignores the branches best represented by Khoisan and Pygmies is itself biased (it applies to many other papers, not just this one).

Similarly Eurasians are studied only in the two standard HapMap samples, whose representativeness for all non-Africans is more than questionable (South Asian and Negrito/Papuan/Australian Aboriginal, as well as Native American samples are also needed).

I'll explain here the bias in my own terms (all the necessary simplifications should be taken with the  lassitude they deserve):

First bifurcation:
  • Southern branch (Khoisan-plus): no bias
    • Y-DNA A
    • mtDNA L0
  • Main branch (all others): no bias (yet)
    • Y-DNA B'CDEF, aka Y(xA)
    • mtDNA L1"6
Second bifurcation (main group):
  • Western branch (Pygmy-plus): no bias
    • Y-DNA B
    • mtDNA L1
  • Main branch (all others):no bias (yet)
    • Y-DNA CDEF, aka Y(xA,B)
    • mtDNA L2"6
Third bifurcation (main group):
  • Major African branch: female bias
    • Y-DNA DE (E)
    • mtDNA L2"6 (again, all sublineages in Africa)
  • Eurasian branch: no clear bias (possible male bias)
    • Y-DNA CF and DE (D)
    • mtDNA M and N (< L3 < L3'4 < L3'4'6 < L2'3'4'6 < L2"6)

I understand that what is apparent in this third bifurcation, where there is a concentration of a particular male lineage, E (this is called female bias). In parallel there is a possible (but rather unclear, also rather mild a signal in the paper, see fig. 4).

It is very intriguing but hard to explain how this extinction of all male branch lineages between CDEF and DE (or rather E) in Africa happened, because there are nothing less than 16 coding region mutations between L2"6 and L3, what implies a lengthy period (but see PS below), and four clear successive branches in the line to L3 (L5, L2, L6 and L4).

We can maybe discard L5, L6 and L4 because they are small but L2 is not small at all, so there must have been some structure already in Central/East Africa in this period within the L2"6 population. 

Even if we totally ignore the Eurasian branches, there is something odd with that African male-only bottleneck.

The authors explain this by polygyny, while Dienekes rejects this and offers in turn the same explanation by another name: 

... an alternative explanation is that the higher female/male ratio in Africans is due to the fact that they are descended from a relatively small number of males who overwhelmed the pre-existing African gene pool. 

Hmmm, how is that different from polygyny?

I'm not really sure, sincerely, but what about the L2"6 population coalescing in a relatively small area (roughly Chad-Sudan-Ethiopia in my reconstruction) and this allowing for a case of male-biased drift that did probably have some component of that polygyny/overwhelming macho element in it, along with drift/founder effects?


PS - While there are 16 mutations between L2"6 and L3, most of them are upstream of the L2'3'4'6 node (i.e. after tiny lineage L5 branched out). Between L2'3'4'6 and L3 there are "only" 7 coding region mutational steps, less than half, the 16 mentioned above, and therefore also less than half the time. The branch leading to L5 can surely be safely ignored for this purpose, it is L2, L4 and the internal diversity of L3 which really cause the contradiction: the female gender bias.

18 comments:

  1. "'descended from a relatively small number of males who overwhelmed the pre-existing African gene pool.' Hmmm, how is that different from polygyny?"

    You could have something like the demographics of early colonial America (or modern Afghanistan) where there were many widowers because many women died in childbirth. This leaves some men as serial husbands with many descendants. This allows anything that favors one lineage over another to have a big impact without polygyny.

    Alternately, suppose that some E lineage men were better food producers, and that various mtDNA L2"L6, already widely mixed in the population, was distributed more or less randomly among available women and was not a survival factor. In bad years, poor food producers and their families died. Voila, a few male lineages thrive while the female lineage mix doesn't change much.

    ReplyDelete
  2. I'm inclined to accept Dienkes 'alternative explanation is that the higher female/male ratio in Africans is due to the fact that they are descended from a relatively small number of males who overwhelmed the pre-existing African gene pool'. This is easily explained if Y-hap E is an immigrant to Africa from Eurasia, and spread through Africa from the north, replacing many indigenous African haplogroups. I have assumed that to be the case since I first began studying the distribution of human haplogroups.

    ReplyDelete
  3. Not really persuaded by either. In general, foragers tend to serial monogamy (but are open-minded to polygyny and maybe other possibilities), their marriages are lax and free, divorce is easy.

    This really allows for for some female bias alone, as women have opportunity to choose. But this would not explain the CDEF-E gap, not easily at least.

    There's no reason to think that E would be from elsewhere and it does not make explaining the phenomenon easier at all.

    I don't think that dying women help explaining anything either, if anyone died here, it was men: all the cousins of E (and C, D and F) in Africa.

    But I don't have a good explanation of my own anyhow either. An epidemic does not seem to make any sense and the macho rapist horde behavior seems totally anachronistic.

    ReplyDelete
  4. Expansion of E at the time of Sahel agriculture or proto-agriculture (probably in one wave, rather than two like mtDNA L2/L3) in the early African Neolithic, outnumbering or killing off, or driving out the cousins of other Y-DNA clades leaving relict populations in jungles and East African and Southern African hunting grounds seems like the easiest explanation.

    Since Sahel agriculture arose independently and probably before contact with Near Eastern agriculturalists, no back migration is required. (Or if not with agriculture, with a more effective means of hunting and gathering).

    Notably, six of the eight discovered DE* individuals are in West Africa (the other two are in Tibet). Perhaps E arose from one community in a larger DE* population in West Africa, with some DE* exiles joining the Out of Africans (who appear to have been paired initially with mtDNA M women in Asia) where most descend from a Y-DNA D man.

    Since the YAP v. not YAP divide is so ancient, the original DE* population may have been West African bound hunter-gatherers breaking from A/B/F populations in East Africa very early on, and only happening to be the group that develops Sahel agriculture or whatever else their competitive edge was much later.

    ReplyDelete
  5. I understand that:

    1. The residual DE* found in West Africa (several locations) and in Tibet (probably a single location) tell us only that DE branched out in the Asian and African populations before D and E coalesced. But IMO Asian DE* is probably just pre-D, while the various African DE* may or not be pre-E. Further research would be required to clarify this matter anyhow and specially in the Red Sea area, so important for understanding the OoA and so limitedly researched, specially in the African shores.

    2. I do not think it is possible at all to relate the expansion of Y-DNA E overall with Neolithic. If we agree that E1b1b1b is related to the Afroasiatic expansion and this one is of Late Paleolithic (???) time frame, we are forced to consider a much older time frame (surely Middle Paleolithic) for Y-DNA E as a whole, specially if we are to related it with the L3 and L2 expansions (or at least part of them). Let's not forget that its, probably younger, "brother" D is generally accepted to be from at least some 40 or 50 Ka ago.

    ReplyDelete
  6. On what grounds do you claim that haplogroup D is probably haplogroup E's "younger brother"? This is precisely the opposite of what published data on these haplogroups have shown. Y-DNA haplogroup E is much, much less diverse than Y-DNA haplogroup D; in fact, what is now known as haplogroup E was originally proposed as a subclade of what is now known as haplogroup D (East Asian YAP+). Haplogroup E was elevated to a status parallel to haplogroup D only after the announcement of the discovery of M174.

    ReplyDelete
  7. "if anyone died here, it was men: all the cousins of E (and C, D and F) in Africa".

    As far as I know there is no evidence at all that C, D or F were ever in Africa. Their ancestor may have come out in the form of CDEF.

    "Expansion of E at the time of Sahel agriculture or proto-agriculture (probably in one wave, rather than two like mtDNA L2/L3) in the early African Neolithic, outnumbering or killing off, or driving out the cousins of other Y-DNA clades leaving relict populations in jungles and East African and Southern African hunting grounds seems like the easiest explanation".

    And that seems the most likely explanation for me as well.

    "no back migration is required. (Or if not with agriculture, with a more effective means of hunting and gathering)".

    I strongly suspect the latter. To me it seems that E must have been already present in Africa before the Neolithic. I agree with Maju: 'I do not think it is possible at all to relate the expansion of Y-DNA E overall with Neolithic'.

    "Perhaps E arose from one community in a larger DE* population in West Africa, with some DE* exiles joining the Out of Africans"

    That isn't an adequate explanation as I see it. D has the characteristics of having been overcome by later expansions of C and F-derived Y-haps. Perhaps DE* emerged from Africa earlier than did CF or, more likely for me, expanded first through regions of Eurasia that C and F had not yet been able to expand into.

    "(who appear to have been paired initially with mtDNA M women in Asia) where most descend from a Y-DNA D man"

    Again that doesn't really add up. MtDNA M seems certainly to have expanded first in South Asia, where Y-hap D is virtually unknown. And I'm reluctant to pair specific mtDNAs with specific Y-haps, except where we're dealing with expansion into virgin territory.

    ReplyDelete
  8. @Ebizur:

    It is my opinion that D could be younger than E because E correlates with L2"6 and D only with some sublineages of sublineages of L3.

    It's an opinion rather than a strong claim, because I do not think you can estimate ages for Y-DNA at this stage of knowledge yet.

    I don't think there's any reason to consider DE "Asian" by origin. But anyhow that's a matter of opinion as the hard evidence is inconclusive. In any case E is an African haplogroup very clearly.

    "Haplogroup E was elevated to a status parallel to haplogroup D only after the announcement of the discovery of M174".

    So? What matter is what it is, not what geneticists once (but not anymore) thought. That has only an interest as matter of history of genetics.

    ReplyDelete
  9. @Terry:

    "As far as I know there is no evidence at all that C, D or F were ever in Africa. Their ancestor may have come out in the form of CDEF".

    I'd say that as pre-CF and as DE. This correlates well with matrilineal diversity.

    What does not correlate well is the pruning of the stem between B'CDEF and E in Africa, including the CDEF and DE stages. Here is where the problem seems to be.

    "D has the characteristics of having been overcome by later expansions of C and F-derived Y-haps".

    I understand that there are other possibilities, for example that DE, pre-C and some minor F (notably K leading to MNOPS) arrived to Eastern Eurasia more or less together, scattering around in a series of founder effects. Similarly mtDNA pre-N and M-derived would have arrived more or less at the same time to the East.

    This still allows for a secondary expansion of Y-DNA O maybe (with mtDNA R primarily?) but the displaced ones would be all the rest, not just D (C also).

    ReplyDelete
  10. ***"(who appear to have been paired initially with mtDNA M women in Asia) where most descend from a Y-DNA D man"

    Again that doesn't really add up. MtDNA M seems certainly to have expanded first in South Asia, where Y-hap D is virtually unknown. And I'm reluctant to pair specific mtDNAs with specific Y-haps, except where we're dealing with expansion into virgin territory.***

    The correlation of Y-DNA D and mtDNA M is simply a piece of evidence that is other there. Puzzling or not, it is the case. If you sift through each of the populations that have Y-DNA D now, and parse out the effects of later migrations from non-Y-DNA D populations, and then remove the mtDNA haplogroups that arrive with the later non-Y-DNA D populations, you are left with mtDNA haplogroup M women each and every time.

    This observation is coroborated by the Andaman Islands population where all the men are Y-DNA D and all the women are mtDNA M, which hasn't had any subsequent non-Y-DNA D populations arrive later.

    ReplyDelete
  11. I'm rather with Andrew in this, with all the lassitude needed, as all clades are intertwined to some extent (IMO sign of rather simultaneous expansion or multiple criss-crossing flows at least), by best guess for East Eurasia only and in the context of earliest human colonization of that area is as follows:

    mtDNA M with Y-DNA D and some C
    mtDNA N with Y-DNA C and some MNOPS (K, F)
    mtDNA R with MNOPS (K, F)

    Very much grosso modo always.

    ReplyDelete
  12. "On what grounds do you claim that haplogroup D is probably haplogroup E's "younger brother"?"

    I agree with Maju on this score. The fact that D is found almost exclusively out of Africa powerfully suggests that it arose no sooner than the out of Africa event, which is realistically 50,000 to 100,000 years ago (conceiveably even later for Y-DNA D Out of Africa if it was not the first wave out). The earliest point in time that we can definitively place Y-DNA D is in the Jomon settlement of Japan ca. 30,000 years ago, and Y-DNA D is notably absent from Australia and Papua New Guinea. We can also say with considerable confidence that Y-DNA D was in Tibet pre-LGM.

    We can also say based on phylogeny, that the Japanese, New World, and Andaman/Tibeto-Burman branches of Y-DNA D probably arose from a common source but are not descendants of each other. This suggests a single pretty rapid expansion from a common source rather than a serial founder effect expansion that occured in widely separated in time stages. And, of course, we can say with some certainty (reinforced by whole genome evidence) that the divide between the CF branch and the DE branch is a very ancient one -- not as old as the breaks of Y-DNA A and Y-DNA B from CF and DE, but almost as deep.

    Likewise, we can safely assume that CF became defined as a separate branch roughly contemporaneously with the main Out of Africa wave (thus not much sooner than 100,000 years ago), since CF is exclusively Eurasian (with the exception of back migrations of R1b and T), and that CF can be no younger than about 50,000 years ago, because it is found in Australia and New Guinea which were settled around then, have very thin effective founer populations, and were largely isolated thereafter as the passage to them became harder with rising sea levels.

    It is also entirely reasonable to assume that CF breaks away from the A, B, pre-CDEF Y-DNA haplogroups contemporaneously with the break of mtDNA M and N (the most basal Eurasian mtDNA haplogroups) from L3, their parent, and that this M and N break (which were parallel, not sequential, in the phylogeny) was from the original ancesteral version of L3, which would be before other subhaplogroups of L3 emerged.

    The mutation clock data is suggestive of a date around 67,000 years ago for these events, but the archaeology showing early modern humans outside Africa as much as 100,000 years ago seems to be pointing to many thousands of years older than that.

    The conventionally presumed date for separations of A and B from the rest of the Y-DNA lineages is on the order of 70,000+ years ago, i.e. earlier than the CF split, but not necessarily all that much earlier than Out of Africa, relatively to each other, regardless of the accuracy of the actual specific estimated date in absolute terms. It isn't out of the range of the margins of error to see the Khoisan, Pygmy and Eurasian branches of humanity breaking out of East Africa and separating from the West and East African branches of humanity in a single episode of expansion over a few thousand years.

    The DE split from the CF branch had to have taken place earlier than the departures of the pre-Eurasians from Africa, but after the splits off of A and B, so that puts the YAP division is a fairly narrow time window. Yet, D was widespread in its ancestral form from the Andamans to the New World at least as late as 30,000 years ago. (Although the Japanese version of Y-DNA D might very well have evolved somewhere else, like Southeast Asia, and its bearers might then have been forced out to Japan when Y-DNA C people moved in, and been wiped out without a trace in their original homeland.)

    ReplyDelete
  13. I'd say the main issue here is basal diversity, which appears highest for Y(xB,A) in Eastern Eurasia and lowest in Africa.

    However, as the branches involved are not many (only two) and the results are in clear contradiction with mtDNA, my conclusion is that the Y(xA,B) concentration into E in Africa (and only there) is a phenomenon that needs to be explained in African terms.

    There's nothing in the mtDNA phylogeny suggesting any such sort of back-migration from Asia. So, yeah, most probably both CDEF and DE branched out in Africa; the contradiction exists ONLY because of the hyper-concentration of African Y-DNA at the E node... but that is precisely the African-specific phenomenon we are trying to address. So what happens in Africa, stays in Africa, so to say.

    In fact I'd dare say that Y-DNA D (or oriental DE) may well represent a secondary flow into Arabia and that's why we do not see it taking any major role in the Eurasian colonization: it arrived in time to catch up with the main wave upon arrival to East Asia but not earlier.

    In fact D is very curious because it's like clinging to the corners of the region (C does too but it's more exaggerated in D I believe). Nothing to do with the solid hegemony of its African "brother", which in Eurasia is only paralleled by F.

    I'd sincerely propose a colonization of East Asia in a rapid sequence:

    1. Clan D (with M primarily)
    2. Clan C (with N primarily)
    3. Clan MNOPS (with R primarily)

    Alternatively they all arrived more or less together but a second wave lead by NO is required in any case, I think.

    ReplyDelete
  14. "The DE split from the CF branch had to have taken place earlier than the departures of the pre-Eurasians from Africa"

    Not necessarily so. Could have been after.

    "I'd say that as pre-CF and as DE".

    Possibly. But I still see no reason at all why CDEF could not have moved out of Africa, with E (or ED, I suppose) moving back in from somewhere nearby.

    "What does not correlate well is the pruning of the stem between B'CDEF and E in Africa, including the CDEF and DE stages. Here is where the problem seems to be".

    Exactly. A 'pruning' within Africa is difficult to explain. However a pruning somewhere outside Africa is quite easy to envisage. We know there was some sort of hiatus, perhaps as much as 50,000 years, between the original OoA and the subsequent one. And both N and M mtDNAs have a 'tail' before they expand. In other words the second OoA may not actually have been from Africa at all if M and N came out with CDEF. Or perhaps one with DE and the other with CF.

    "you are left with mtDNA haplogroup M women each and every time".

    Possibly so. But in many regions where you have what is almost certainly an 'original' M you don't have D at at all. For example in South Asia and in New Guinea. So the correlation is hardly rigid.

    "This observation is coroborated by the Andaman Islands population where all the men are Y-DNA D and all the women are mtDNA M, which hasn't had any subsequent non-Y-DNA D populations arrive later".

    But that in no way proves they moved all the way from Africa together.

    "It isn't out of the range of the margins of error to see the Khoisan, Pygmy and Eurasian branches of humanity breaking out of East Africa and separating from the West and East African branches of humanity in a single episode of expansion over a few thousand years".

    Except that doesn't take into account the various haplogroups that had become spread round Africa well before the OoA. In fact haplogroup L0 is common in Khoisan and is presumably much older than L3.

    "I'd sincerely propose a colonization of East Asia in a rapid sequence"

    I'd tend to put Y-hap F in there with mtDNA M. In fact I suspect F is the haplogroup that has been the most drifted out through South, southeast and east Asia. Even its descenadant haplogroup K is far from common, but is more widespread than is D. The big Y-chromosome replacement is with the expansion of the various MNOPS clades.

    "In fact D is very curious because it's like clinging to the corners of the region (C does too but it's more exaggerated in D I believe)".

    And the MNOPS expansion would explain that.

    ReplyDelete
  15. "A 'pruning' within Africa is difficult to explain. However a pruning somewhere outside Africa is quite easy to envisage".

    I see where you're coming but the problem is still in the pruning in Africa: why nothing equivalent to L2, L3 and the many branches that hang from these nodes survived in the Y-DNA side in Africa while the earlier branches did? This is first and foremost an African issue, regardless of the origin of DE. Would it have been any other lineage we'd be in the same situation (and it was going to be always a relative of the Eurasian CDEF).

    "in many regions where you have what is almost certainly an 'original' M you don't have D at at all".

    So? You may have C, right?

    "So the correlation is hardly rigid".

    Not rigid, can't be rigid.

    "And the MNOPS expansion would explain that".

    Maybe but that's as good as saying the NO expansion, right?

    ReplyDelete
  16. Andrew said,

    "We can also say based on phylogeny, that the Japanese, New World, and Andaman/Tibeto-Burman branches of Y-DNA D..."

    What is this "New World" branch of Y-DNA haplogroup D that you have mentioned?

    ReplyDelete
  17. "So? You may have C, right?"

    Perhaps. But C is hardly common in India either. And mtDNA M is particularly common in India. What isn't M is mostly R-derived, and so comes in later from somewhere in SE Asia. The most common Y-haps in India are F-derived: L, H and the Fs and some Ks. That's why I strongly suspect mtDNA M was originally associted with F.

    "Maybe but that's as good as saying the NO expansion, right?"

    Partly, especially through East Asia, and even then the O expansion was fairly obviously much more extensive than the NO one. But the NO expansion does not seem to have involved India at all. Just some downstream O haplogroups, especially O2a. That might explain the paucity of C in India but, as you are aware, I suspect that C is actually a later immigrant to India, not indigenous to it.

    "why nothing equivalent to L2, L3 and the many branches that hang from these nodes survived in the Y-DNA side in Africa while the earlier branches did?"

    L2 and L3 are both basically Sahelian haplogroups, exactly the region where E is so dominant. I presume that any Y-hap originally associated with those two mtDNAs has been replaced by E, just as has happened with certain Y-haps in East and Southeast Asia. Y-haps A and B tend to be more southerly in their overall distribution.

    ReplyDelete
  18. "Perhaps. But C is hardly common in India either".

    Precisely, as it's obvious that mtDNA M did not expand to East Asia with F (or mostly not) but with the more pioneering D and C lineages. And that's why the RapidCoastalMigration is a loved model: it allows the various lineages to have been there right in time, producing a number of distinct founder effects.

    For our purpose it'd be the same if C and D would be subclades of F or if N was hanging from M: they all took part in the same mini-bang: but F took the ride with 30 Ms and C and D did only with a dozen or so (plus N).

    "The most common Y-haps in India are F-derived: L, H"

    Actually the most common Y-DNA in India is R, I believe and, like L, it is part of K, something you seem to be forgetting. So you'd be more clear if you said "F-ferived: IJK, H".

    So sure: all West and South Eurasian Y-DNA is within F (with lesser exceptions). Instead in the East, while F is numerically important, it is restricted to quite downstream sublineages: MNOPS and particularly O.

    While the MNOPS/O spread seems to require a second-layer explanation, the C and D founder effects do not seem to require anything other than an open mind.

    "But the NO expansion does not seem to have involved India at all".

    It does if you go one step upstream to MNOPS: P is very common in South Asia and IMO is originary from there. I think you can track P via the Ganges in Westward and then Northward direction pretty clearly.

    It's anyhow part of the MNOPS expansion.

    "I presume that any Y-hap originally associated with those two mtDNAs has been replaced by E, just as has happened with certain Y-haps in East and Southeast Asia".

    In East Asia you can at least track up to some point the associated mtDNA flows. It's not at all a purely male replacement thing. D and C are mixed with NO and NO quite clearly brought Oriental R (macro-F and B) with it.

    In Africa, you need to associate E with all L3 or L3'4'6 (what already implies a pruning in relation with the diversity we find in the (related) Eurasian Y-DNA... and then you need to explain how E also managed to hijack L2 as a whole, without leaving any remnants, except rare A and B clades, which is not what we could expect (I'd expect CDEF* or even pre-CDEF*, really).

    This is really strange. And the Sahelian and likely neighbor nature of both L3'4'6 and L2 only helps somewhat with this explanation, as we would still expect two or more distinct, structured, populations to have existed, for mtDNA as for Y-DNA.

    ReplyDelete

Please, be reasonably respectful when making comments. I do not tolerate in particular sexism, racism nor homophobia. Personal attacks, manipulation and trolling are also very much unwelcome here.The author reserves the right to delete any abusive comment.

Preliminary comment moderation is... ON (your comment may take some time, maybe days or weeks to appear).