diet

This week, Thure Cerling and colleagues report in PNAS [1] carbon stable isotope data from 24 specimens of Australopithecus boisei. This is a huge sample as fossil hominins go, and they give a very consistent picture about the diet of this most robust of the australopithecines. These 24 individuals got between 61 and 91 percent of their carbon from grasses.

My 2005 explainer on stable isotope chemistry and early hominin diets fills in the details about carbon-12, carbon-13 and their relationship to 3- and 4-carbon photosynthetic cycles. The salient aspect of the comparisons involving A. boisei here is that C₄ plants, mostly grasses, incorporate relatively more carbon-13 than do other plants, and herbivores assimilate this carbon-13 into their bones and teeth.

The high ratio of grass-derived carbon in A. boisei is fundamentally different from all living and fossil apes, and it is far higher than the values found for other early hominins. The only other primate that comes close is the fossil giant gelada Theropithecus oswaldi, a savanna-living species.

What were these extinct species really eating? Was grass the food? For living geladas, grass consumption includes seeds -- a fact that led Clifford Jolly to suggest that early hominins might also have specialized on seeds [2]. Of course, humans today also specialize on grass seeds. We call them grains, eat them in bread and drink them in soda. And beer.

But what about A. boisei? The large, thick-enameled premolars and molars, with their low cusps, seem well suited to grinding small hard objects and resisting the resulting wear. But Cerling and colleagues devote a good chunk of their discussion to the description of molar wear in A. boisei and other early hominins. Their argument is that the teeth of A. boisei show no signs of "hard object" feeding:

Of perhaps greater moment than its potential specific simila- rities, the microwear of P. boisei molars, which shows remarkable uniformity over time from about 2.3 Ma to about < 1.4 Ma (9, 24), stands in stark contrast to the wear fabrics exhibited by primate hard-object consumers. Indeed, there is no evidence beyond the anecdotal [e.g., the broken left first permanent molar crown in the KNM-ER 729 P. boisei mandible (8) and the observation that a couple of P. boisei molars show antemortem enamel chipping (25)] that these food items were hard.

These observations are not new, but putting them together with the evidence of grass consumption makes it pretty clear that seed eating was not a predominant source of dietary carbon. The "Nutcracker Man" sobriquet, applied to A. boisei because of its powerful jaw mechanics, must be false. No significant hard object feeding, very low dietary carbon from trees and non-grassy (or sedgy) plants.

Instead, Cerling and colleagues propose that both A. boisei and other early hominins wore their teeth on the, well, grassy parts of grass.

P. boisei cheek teeth display notable gradients of gross wear, resulting in large, deeply excavated dentine exposures, and in this regard, they are similar to other australopith species (e.g., A. afarensis and A. africanus) that also possess low tooth cusps with thick enamel. Thus, like other australopiths, P. boisei undoubtedly had a diet that consisted of foods with abrasive qualities—the gross wear is as likely due to repetitive loading of phytolith-rich tough foods as exogenous grit. Thus, either grass or sedge consumption and/or exogenous grit might well have contributed to P. boisei’s notable wear gradient.

And:

Recent dental microwear studies suggest that the mechanical properties of A. afarensis (and A. anamensis) diets were nearly identical to those of P. boisei (9, 24, 40–42). If this is so, could it be that the australopith masticatory package represents an adaptation to C4 resources such as grasses or sedges? The similarity in dental microwear fabrics among the eastern African australopiths, all of which lack any evidence for hard-object food consumption (9, 24, 40–42), is consistent with the notion that their craniodental morphology could reflect “repetitive loading” rather than hard-object consumption (7, 8, 43).

Grit might get in from eating underground parts like rhizomes. Phytoliths are small, hard silicate structures in the green parts of plants, including the stems and leaves of grass.

Last year I wrote about carbon isotope analysis of two specimens of Australopithecus boisei, the famous OH 5 "Zinj" specimen, and the Peninj mandible. Both specimens show evidence of a high consumption of grass-derived carbon -- estimated at 77% and 81% grass-derived carbon, respectively. Those levels are characteristic of grazing animals. Cerling and colleagues show that these values are right in the middle of the range among specimens of A. boisei that cover a half million years in Kenya and Tanzania.

In the paper reporting the carbon stable isotopes of OH 5 and Peninj, van der Merwe and colleagues [3] suggested that A. boisei may have relied on papyrus as a staple. The culms and rhizomes of papyrus both have substantial nutritional content but are very fibrous and require much chewing and spitting out fiber at intervals. The hypothesis would imply that A. boisei relied on these foodstuffs for the majority of its calories.

Cerling and colleagues do not mention papyrus, and take a much more direct approach on grass-eating. But they do report data on oxygen stable isotopes from the specimens that may be relevant to the ecological context of grass (or sedge) consumption. Oxygen isotopes in bone and teeth reflect the pattern of water consumption by an animal. Oxygen-16 evaporates and transpires preferentially from leaves, so an animal living in an arid environment that gets most of its water from plants will be relatively enriched for the heavier oxygen-18. An animal that depends on drinking water from lakes or rivers will tend to have lower oxygen-18. A. boisei is almost as low in oxygen-18 composition as hippopotamus, suggesting they were strongly dependent on water sources.

A highly water-dependent grass-eating A. boisei is a very different picture of the biology of this robust species. The South African robust species, A robustus, is very different in this regard. These two species are often lumped together, but this is unfair in many ways to their distinctive anatomical patterns. Knowing that their dietary adaptations were very distinct, we should be more inclined to focus on the details where they differ.

Bottom line: A. boisei represents a highly distinctive dietary pattern, not present in any living ape, that no longer exists. At least the giant gelada, T. oswaldi, may also have exploited similar resources. Some grass resources, including papyrus corms and rhizomes, have high caloric and nutritional value, but require adaptations to deal with the fibrous content.

References

Dental plaque is a biofilm made up of bacteria adhering to the enamel surface of the teeth. Plaque is soft but over many days can gradually calcify. The hardened plaque, called calculus (or dental tartar) can build up in layers. This forms an ideal surface for further plaque formation and can damage the connective tissue between the teeth and gums -- so dentists and dental hygienists work hard to remove tartar.

Despite these risks to dental health, calculus formation is a natural process in populations of humans and animals. Teeth from archaeological sites often have calculus adhering to them. This calculus contains partially mineralized bacteria, organic material, epithelial cells and fragments of foods ingested by the individual during her life.

Recently, archaeologists and paleontologists have begun to examine calculus samples microscopically to identify the traces of ancient foods. Phytoliths are microscopic structures made of silica or calcium oxalate, many of which are distinctive to species or genera of plants. These durable inorganic structures can persist for hundreds of thousands of years. Many kinds of plants store their starches in granules that can also persist over long periods of time. These granules differ in form among families of plants. Starch granules also display characteristic changes when they are heated or cooked in liquid. Hence they can provide evidence of cooking practices by ancient people.

Amanda Henry and colleagues [1] scraped a bit of calculus off the teeth of Shanidar 3, from the north of Iraq sometime around 50,000 years ago. This skeleton is housed at the Smithsonian National Museum of Natural History, and figured in the 2009 story about a projectile wound between its ribs ("Real stories of the Neandertal CSI"). The Shanidar skeletal remains are generally called Neandertals on the basis of their morphology. This case is better for Shanidar than for near-contemporaries in the Levant such as Amud or Kebara, and may reflect connections with populations to the north in the Caucasus. No genetic sampling has been done on the Shanidar sample.

Neandertals are known for a diet stereotype -- they ate a very high proportion of meat. This stereotype is rooted in fact: the majority of Neandertal sites show a clear reliance on large mammal acquisition. Bison, horse, red deer and other large mammals are represented, often with a statistical preference toward one of these species at a given site. The faunal remains from many Neandertal sites are consistent with an ambush hunting strategy, with a higher proportion of prime age adult animals than found among persistence hunters or scavengers. The stable isotope record in Neandertal teeth so far seems consistent with an estimated 90% or more meat consumption, leaving relatively little dietary intake from plant foods.

This is an odd picture for a hunter-gatherer: all hunting and little gathering. Most living hunter-gatherers rely on plant foods to buffer the risks of hunting. Many eat far more calories from plants than from meat. Plant processing in the archaeological record is well-known from among the earliest archaeological traces ("Plant processing with early Oldowan tools") and continued throughout the Paleolithic. The question at hand is specific to Neandertals -- how dependent were they on plant foods, and how much would they have been specifically adapted to meat acquisition and consumption? Some high-latitude hunting groups, such as the Inuit, do maintain very high meat consumption, and the Neandertals may have relied on this strategy in Europe.

In contrast to Europe, a pattern of plant exploitation has long been known in the Middle Paleolithic of the Levant, at sites like Amud and Kebara that many argue were occupied by Neandertals. For example, Marco Madella and colleagues [2] describe phytoliths from the soils of Amud Cave. They demonstrated that these people were probably gathering seeds from grasses, systematically enough to concentrate the phytoliths from mature grass panicles in the sediments. Efraim Lev and colleagues [3] described charred plant remains including legumes and pistachio nuts from the Mousterian levels of Kebara Cave. Middle Paleolithic people in the Levant were using plants in quantities as great as can be shown at any contemporary sites anywhere in the world.

Henry and colleagues help to put the seeds in the mouths of the Shanidar Neandertals. The Shanidar 3 calculus samples yielded substantial evidence of barley consumption. Many of the starch granules were clearly cooked:

The overall pattern of damage to the starch grains matches most closely with that caused by heating in the presence of water, such as during baking or boiling, rather than “dryer” forms of cooking like parching or popping (38). The finding of cooked Triticeae starches on the Shanidar teeth reinforces evidence from other studies (13) that suggest that Near Eastern Neanderthals cooked plant foods.

They report that 42% of the starch granules on these teeth are consistent with damage from cooking, suggesting that cooking was a systematic strategy for plant exploitation in these people. They also find other plants besides grains -- including legume seeds and tubers of some kind, and phytoliths from date palms.

In addition to the Shanidar 3 skeleton, Henry and colleagues examined calculus samples from the two skeletons from Spy, Belgium. This gave them the opportunity to examine plant consumption in a European context. These teeth included starch molecules in relatively large numbers, mostly derived from some kind of plant underground storage organs (USOs) which the authors tentatively identify as a water lily. At least one starch granule of a sorghum or related grass seed is also present, along with a few other unidentifiable starches and no phytoliths.

The authors cannot conclude much about the importance of these plant foods to the overall diet. The remains of starch grains and phytoliths tell us about diet breadth but not the proportions of different foods. They do note that nitrogen stable isotopes are most informative about protein-rich food sources, so that a substantial consumption of starchy plants such as grains and USOs might be hidden by isotope analysis. Their main conclusion is about dietary flexibility and the sophistication of Neandertal foraging strategies:

These lines of evidence [cooking and processing of grains] indicate Neanderthals were investing their time and labor in preparing plant foods in ways that increased their edibility and nutritional quality (24, 45). It should also be noted that date palms and possibly other un- identi␣ed plants have different harvest seasons than barley and legumes, a factor that may suggest that the Shanidar Neanderthals practiced seasonal rounds of collecting and scheduled returns to harvest areas. Overall, these data suggest that Nean- derthals were capable of complex food-gathering behaviors that included both hunting of large game animals and the harvesting and processing of plant foods.

I expect that quite a bit more evidence from dental calculus will be forthcoming. The study of microfossils from a broader range of sites will help to give a picture of the local resource exploitation. It may not be possible to get an estimate of dietary proportions from these kinds of evidence, but I imagine that similar comparisons of calculus samples from other animals will provide some useful context for the human numbers. In addition, calculus has become a promising source of DNA recovery, from the epithelial cells trapped in its calcified matrix. This has a good chance of recovering ancient DNA from specimens that have not previously yielded any successful extraction.

References

When I wrote earlier in the week about the 1000 Genomes Project results, I mentioned that a second paper was being published in Science. That paper, by Peter Sudmant and colleagues [1], works to quantify the amount of copy number variation of genes in the genomes of the study participants.

It can be challenging to study copy number variation using shotgun sequencing methods, because each duplicated part of the genome creates multiple alignment targets for short reads. One way to deal with this problem is to use the drawbacks of shotgun sequencing as an advantage: Look for template regions of the genome that have much higher read depth than others. These places include many where a gene has been duplicated in the target genome, giving one-and-a-half or twice the number of reads for each duplication. Looking at read depth genome-wide is a quick way to assess copy number variation at sites where it was previously unknown. Once these are ascertained in a sample of genomes, they can be targeted for further study, including characterizing the boundaries of the duplicate region.

The paper describes this methodology in some detail, with various embellishments to get more precise answers to certain kinds of structural questions. They developed a large set of SNPs that differentiate paralogous gene copies, among other things allowing them to examine which members of various gene families had been duplicated, and whether events were shared between populations.

Through our analysis, we identified that duplicated regions are more likely to be stratified between human populations when compared with copy number variation within unique regions of the genome. For example, 59 (92%) of the top 64 stratified gene families overlap segmental duplications (P –16). Remarkably, many of these highly polymorphic genes map to duplications that promote recurrent rearrangements associated with intellectual disability, autism, schizophrenia and epilepsy. We hypothesize that the extreme polymorphism may contribute to genomic instability associated with disease and may predispose certain populations to different chromosomal rearrangements (30).

Segmental duplications can be relatively effective ways to change the amount of gene product without changing the gene product. In other words, a duplication can increase the dosage of a particular gene product. That can sometimes be very useful. For example, salivary amylase production varies among people due to the number of duplicate copies of the gene [2]. The copy number variation is related to population history of agricultural subsistence -- old agricultural populations have more amylase copies. It's a simple case where the dietary ecology favors a dosage increase for an enzyme.

Gene duplications and other structural changes to the genome are rare events -- any particular kind of change is substantially less likely than a single nucleotide mutation at a given point in the genome. So it is of some interest to consider which regions are actually invariant in copy number -- duplications that occurred on the human lineage but have been conserved in more recent populations -- because these may reflect old adaptations essential to the evolution of hominins. Here's what the paper concludes:

We have also defined the ~49% of gene duplicates that are largely invariant in copy among humans. Although this is based only on an assessment of 159 genomes from select populations, the fact that this fraction of genes remains copy number invariant in a milieu of recurrent unequal crossover suggests functional importance. Among these, we find a number of genes involved in neurological development and disease. We note that many of these duplicated genes are themselves incomplete and may represent nonprocessed pseudogenes, which may modulate the expression of the ancestral gene. The characterization of the most recently duplicated genes should facilitate identification of those that acquired new functions (neofunctionalization) versus those that have become pseudogenes or have partitioned their function among duplicate copies (31).

I was going to write that there's not much analysis in the paper and let it go at that. But the paper has a 108-page supplement.

I know I write this like once a week, but what the heck is the point of a 4-page paper with a 108-page supplement? Granted, 7 of the supplement pages are the author list (!!), but I view the whole thing mainly as a rip-off for the people who did the analyses in the supplement. Why don't they get their own first-authored publications? Are other journals satisfied to accept first-authored versions of analyses that have already been in a supplement in Science?

The supplement lists 64 gene families including segmental duplications that differ substantially in average copy number among the CEU, YRI and CHB/JPT samples to which the low-coverage whole-genome sequencing has been applied thus far. The table (S8) lists the mean copy number in the three populations and the total variance in copy number; the key statistic is a value called Vst, which is analogous to F_ST for length variations.

These are not generally duplications of whole genes, and their boundaries don't generally correspond to the boundaries of coding regions or exons. Without further analysis, it is not clear which of these duplicated regions may have functional import. Many of the additional copies may be inactive, either because of pseudogenization or because the duplication may not include the promoter/enhancer elements needed for gene expression. Some of the duplications occur in regions with known pseudogenes. The "involvement" of some genes in these regions with neurological development and disease is interesting, but the paper attempts no statistical assessment of this. It's a list of candidates, with some interesting ones that are obviously worth further examination, but without a clear story for any of them.

It is maybe interesting that salivary amylase didn't make the list. It's not clear from the supplement whether that is an omission or whether its population differentiation, great as it is, is not as high as the lower cutoff. The greatest differentiation for amylase copy number is between populations that are not yet represented in the 1000 Genomes whole-genome sequencing.

That raises an interesting question: What if we applied the same methods to the read data from some of the other public genomes? The Bushman genomes from earlier this year are an especially interesting sample because they are notably not drawn from a long-time agricultural population. In which areas would they score atypical copy number variation compared to the 1000 Genomes samples?

References