john hawks weblog

During the past couple of years, new evidence has really shifted our view of Neandertal diet. Even three years ago, it was not unusual to hear Neandertals described as "hypercarnivores", more heavily reliant upon meat than any living hunter-gatherers, except possibly for Inuit who live on seal meat and whale blubber.

The idea that Neandertals had diets with a very high fraction of meat -- maybe as high as 90-95% meat -- came from analyses of stable isotopes. I reviewed some of the stable isotope work on Neandertal diet in 2005 - "Neandertals noshed on mammoth meat?", "Neandertals: gone fishin' or not?". Here at the beginning of 2013, stable isotopes are well worth another review here on the blog.

The extreme view of Neandertals as hypercarnivores has been softened by new evidence from several sources. Phytoliths and starch grains from Neandertal dental calculus have shown a wide variety of plants were consumed by Neandertals at least occasionally. Meanwhile, the starch grains have not only documented consumption of grains and tubers, but have also shown that Neandertals were cooking those plant foods. I wrote about the phytolith and starch granule discoveries by Amanda Henry and colleagues [1] last year ("Tartar control and Neandertal plant use").

A new article by John Speth in Before Farming reconsiders the archaeological record of game exploitation by Neandertals and early modern humans in the Near East [2]. Speth begins with a short review of how Neandertals gradually came to be known as hypercarnivores -- in spite of many archaeologists' insistence that they must have been incompetent in various ways. After some discussion of the limits of the archaeological record, he notes that the zooarchaeological record doesn't tell us about the quantitative contribution of meat to the diet. In short:

Lots of gazelle bones doesn’t necessarily mean lots of gazelle meat per capita per day.

He illustrates this point with a historical case, the excavation of trash heaps from Fort Ligonier, Pennsylvania, occupied by the British during the French and Indian War. There, the total meat yield represented by animal bones was estimated at only 4,000 pounds, a tiny fraction of the meat ration known to have been issued to soldiers. The point of the example is that many biases prevent the accumulation and discovery of animal bone, even in historic contexts. The Paleolithic record of faunal exploitation can represent only the merest fraction of animal carcasses that were actually handled or consumed by ancient peoples. Biases guarantee that this record will be unrepresentative in ways that we may be poorly able to assess.

Speth addresses the idea that Middle Pleistocene people consumed a very high fraction of meat by emphasizing that a diet of lean meat is unsustainable at such a level. If Neandertals' animal consumption was as high as Inuit peoples, then they must have been eating a high fraction of fat somehow:

The Inuit or Eskimos provide a classic example of peoples whose traditional sustenance was provided almost entirely by meat, the diet commonly envisioned for cold-climate Neanderthals. But when looked at quantitatively, Inuit diet was actually composed primarily of fat, not lean meat, with the protein contribution seldom surpassing about 35 per cent of their calories, and usually lower, closer to 25 per cent. Pemmican, the traditional mainstay of Native Americans and First Nation peoples (‘Indians’) inhabiting the Great Plains of mid-continental North America, was a mixture of rendered fat and dried, pulverized lean meat, the mix carefully prepared so that the pro- tein component did not exceed 25–30 per cent of total energy (eg, Stefansson 1956; Speth 2010). In habitats where plant foods are neither abundant nor available for long periods of the year, and particularly for foragers in such habitats who do not store foods, fat becomes the principal non-protein macronutrient for much of the year. Foragers in the northern latitudes did obtain some carbohydrates by consuming fermented stomach contents of reindeer and ptarmigan, and sometimes inner bark (cambium), as well as small quantities of berries during the summer months (Eidlitz 1969; Gottesfeld 1992; Östlund et al 2009; Sandgathe & Hayden 2003; Zackrisson et al 2000). Until fairly recently, stomach contents were actually considered a delicacy (often referred to as ‘Eskimo ice cream’), not an emergency resource resorted to only when all else failed (Starks 2007; Speth 2010). Unfortunately, we lack quantitative data on the actual amounts that were consumed, how those amounts varied over the year, and whether men and women had comparable access. Did Neanderthals also con- sume fermented stomach contents? If so, would such a practice have had any detectable impact on their unusually high nitrogen isotope values?

Through the middle of the article, Speth provides a detailed account of the biases due to taphonomy and ancient behavior that apply to faunal collections in Middle Paleolithic contexts. Many of these factors, such as biases in transport of different size animals, are well-known to archaeologists, but Speth's review will be useful for those who may not have studied the issue. The value of this part of the article is in its application of prey transport and landscape use to the unique geography of the Near East. Here, Middle Paleolithic peoples hunted amid water scarcity and temperature regimes that were very different from those found in Southwestern Europe. Yet by several indicators, the Middle Paleolithic population in both areas was relatively dense and successful.

Speth reminds us that ancient hunters were active agents who made choices in their hunting strategies. Some of those choices may have been influenced by landscape use and prey abundance, but others are less easily predictable in such terms:

The Hadza, one of the most thoroughly documented modern foraging populations, offer another interesting example. Wildebeest are one of the most abundant prey available to Hadza hunters, but they commonly avoid wildebeest in favour of zebras. Why? According to Hadza informants, the fat from wildebeest is hard and sticks to one’s teeth and palate,while zebra marrow and back-fat, especially the yellow subcutaneous deposits near the rump, are far more desirable (Oliver 1993:217; Selous 1907:220; Speth 2010:66–70). Were we to assume that Hadza hunters took prey in direct proportion to their availability on the landscape, our conclusions would be very wide of the mark.

Back to the problem of lean meat: Hunter-gatherers in ethnographic and historical records have used boiling to degrease bone. This allows the use of the fat from inside the cancellous structure of the bone, which is a key resource supporting the use of lean wild animal meat. Boiling or slow-cooking using heated stones has been applied by many peoples around the world, and tends to leave a very distinctive archaeological trace -- the heated rocks, lined pits dug to enclose the slow-cooking mass, all show up in the archaeology. These techniques were not used by Middle Paleolithic people, or if such people used heated rocks, they did not use them terribly extensively. Stone boiling became common only later in the Upper Paleolithic of Europe.

But Speth discusses other means of boiling, including the use of skin and bark containers. These are expedient and perishable, yet filled with water will effectively contain boiling liquid over hot coals or indirect flame. Whether such techniques were used by Neandertals remains speculative. The suggestion is latent in the identification of cooked starches within Neandertal dental calculus. If they were capable of cooking grains in moist heat, they must at least have been using bark packets or some other style of slow-cooking. The rendering of fat from bone by boiling in perishable containers would not take much additional innovation, and would have been energetically and nutritionally very advantageous.

As I was discussing this with friends a couple of weeks ago, it occurred to me that the combination of cooked grains and meats within an animal bladder is a recurrent feature of the cuisine of Northern Europe. Neandertal haggis.

References

How did I get myself quoted in a story as the skeptic about recent human evolution? ("Human Evolution Enters an Exciting New Phase"). After all, I've been a huge advocate of the idea that recent human evolution was a lot faster and more interesting than anthropologists used to think ("Why human evolution accelerated").

The story, by Brandom Keim, is a good account of a new paper in Nature by Wenqing Fu and colleagues, "Analysis of 6,515 exomes reveals the recent origin of most human protein-coding variants" [1]. It's a pretty cool study, which has identified protein-coding alleles in large samples of European-American and African-American individuals.

Fu and colleagues compared all the coding variants they found in large samples of European-Americans and African-Americans, and discovered that the European-ancestry people have a higher fraction of rare coding variants. They propose that the rate of new coding variants entering and persisting within the population actually accelerated in the ancestral European population. Why would this happen? In their view, demography is the most likely explanation. As European populations expanded during the Neolithic and later time periods, the rate by which new mutations are lost by genetic drift began to decline. These new mutations have pooled up within the European population, giving them a glut of new changes to protein-coding sequences. Many of these mutations may be deleterious, just not bad enough for natural selection to have weeded them out in the growing ancient population.

I think in large part this explanation is correct. In some ways it is incomplete.

The effect of population history on our evolution was the theme of our 2007 paper on positive selection in recent humans [2]. We relied on exactly the same mathematical relations used in this new paper: More people means more different mutations entering the population. In our case, the increase in the total number of mutations meant that we could expect more potential adaptive mutations to be selected within a growing population. In this case, the increase in the total number of mutations means more mutations remain to be picked up by resequencing rare neutral or deleterious variations in present samples.

One of the senior authors of the study, Joshua Akey, commented:

Most of the mutations that we found arose in the last 200 generations or so. There hasn’t been much time for random change or deterministic change through natural selection. We have a repository of all this new variation for humanity to use as a substrate. In a way, we’re more evolvable now than at any time in our history.

(this is quoted by Punnett Square, not sure about the original source)

That's a cool concept. These rare protein-coding variations may be mostly unimportant to fitness today, and many are slightly deleterious. Still they provide a store of variability that increases the potential range of responses to future adaptive challenges. Or, they give us room to examine the effects of small differences, which will help us to understand better how genes work. For the past few thousand years, a small proportion of those have come under positive selection, the part that we have been studying in my lab since 2007.

The current study has some drawbacks. For one, it isn't evident from the results how these new coding mutations are distributed among individuals. Under population growth alone, we should expect that the number of these new coding variants carried by any one individual should be approximately the same as any other individual, regardless of the population size. Where a big population differs from a small population is in the variety of mutations carried by different individuals, with the average number per individual being equal. That may be true in this study, but it isn't possible to tell from the results presented.

To the extent that some of these mutations are deleterious, their distribution matters. In Europeans, there may be a greater number of deleterious mutations that are on average more rare; all things being equal, this pattern should make it harder to find statistical evidence for association of these rare variants with complex disorders. By contrast, in Africans, the higher average frequencies of such variants should make them easier to tie to phenotypic variation. All this can be concluded from frequencies alone, without a need to relate frequency to age.

Probably the biggest shortcoming of the paper is in its estimation of ages for these rare mutational variants. Estimating the ages of mutations in human populations has been a real problem for those of us working with genotyping or sequencing data from small samples. When we depend on the linkage between a rare allele and nearby genetic loci, we run into a sampling problem: Estimating the proportion of recombinants in a population fundamentally has a lot of error when you are working with a sample of 10 copies of the rare allele.

Estimating dates by LD is bad enough, but this paper doesn't even go that far. Instead, it estimates the ages of alleles from their frequency.

Frequency estimation of age is OK if the genome sequences have come from a Wright-Fisher population (that is, a random-mating, constant size population). More common alleles tend to be older, new alleles tend to be very rare. This isn't a very accurate means of dating any particular mutation, because the relationship of age and frequency under genetic drift has a tremendous variance. But when pooling large sets of alleles into frequency classes, the age-by-frequency approach gives a rough idea of whether mutations have accelerated or stayed at a constant rate over time.

But there's one obvious thing missing from the model that may have a large effect on the frequencies of rare coding variants: Introgression from Neandertals! If we want to know why Europeans have a large store of rare coding variants relative to Africans, their ancient mixture of a small fraction of a very divergent human population is one obvious reason. None of the Neandertal alleles in Europeans today are new, they are all old. But a method that estimates their ages by allele frequency alone will always conclude that these rare Neandertal alleles are very young.

In the current paper, the relation of frequency and age is derived from simulations that are based on a model of human population history. Like all recent papers that apply a model of human population history, this one is both overcomplicated (lots of parameters to which we have no good estimates) and oversimplified (too few events to accommodate known historical phenomena). Here's the population model used to derive allele ages in the paper:

Population model from Figure S5 in the supplementary information from Fu et al. 2012

The parameters for population divergence times and ancient population sizes are estimated from genetic data, so any systematic error will propagate through to the estimation of allele ages. The exclusion of Neandertal introgression in the model really does bias the allele age estimates badly, as Neandertal genes today are mostly rare, and mostly very old. This year's shift in our assumptions about mutation rates (to a much slower rate than previously assumed) will also affect the estimates of the demographic parameters in the model. An older coalescence time for most genes means a larger ancestral effective size for these populations, and much older allele ages when frequency is the estimator.

Our lab is working very hard on allele ages, and I hope to be able to share some of that work soon.

This study is not alone in demonstrating the real importance of rare coding variation in human populations. This line of research has substantial value, as it helps to show why so much of the additive genetic variation underlying variation in human phenotypes has not yet been assigned to genes. We know that many traits are heritable by comparing genetic relatives with each other. Finding the genetic loci that explain similarity among relatives is relatively easy when the genes involved are common, because the same gene variants will be shared across many families. But pooling many families doesn't help us find very rare mutations, as these are likely carried only by a few pedigrees even in a very large sample. By showing the large store of rare coding variation, these studies help to establish that much of the genetic variation underlying disease may be there for us to discover, if we change our discovery approach.

References