Did GDF6 “gene tweak” allow humans to become upright?

The short answer is, “Not really.” But as is often the case, the real story behind so many headlines last week is a bit more complicated.


smh. Links to the first, second, third, and fourth stories.

What are they talking about, Willis?

These headlines, each saying something slightly different, are referring to a study by Indjeian and colleagues published in Cell.  Researchers identified a stretch of DNA that is highly conserved across mammals, or in other words, it is very similar between very different organisms. In humans, however, this conserved region is actually missing (“hCONDEL.306”):

Fig. 4A from Indjeian et al. 2016. A stretch of DNA, "hCONDEL.306" is completely missing in humans (as is another stretch, hCONDEL.305) but otherwise very similar between chimpanzees, monkeys and mice.

Fig. 4A from Indjeian et al. 2016. A stretch of DNA on Chromosome 8, “hCONDEL.306,” is very similar between chimpanzees, macaque monkeys, and mice, but is completely missing in humans (as is another stretch, hCONDEL.305).

That a stretch of DNA should be highly conserved across diverse animal groups suggests purifying natural selection has prevented any mutations from occurring here – alterations to this stretch of DNA negatively affected fitness. But that humans should be missing such a highly conserved region suggests that this deletion came under positive natural selection at some point in human evolution. This strategy, of seeking stretches of DNA that are similar between many animals but very different in humans, has led to the identification of hundreds of genetic underpinnings of human uniqueness. Some of these, such as the case in question, involve deleted sequences and have been termed “hCONDELs,” for “regions with high sequence conservation that are surprisingly deleted in humans” (McLean et al., 2011: 216). Others involve the accumulation of mutations where other animals show few or none (e.g., HACNS1; Prabhakar et al. 2008). In many (most?) cases these are “non-coding” sequences of DNA.

How can “non-coding” DNA help make humans upright?

As was predicted 30 years ago (King and Wilson, 1975), what makes humans different from other animals isn’t so much in the protein-coding DNA (the classical understanding of the term, “genes”), but rather in the control of these protein-coding genes. “Non-coding” means that a stretch of DNA may get transcribed into RNA but is not then translated into proteins. But even though these sequences themselves don’t become anything tangible, many are nevertheless critical in regulating gene expression – when, where and how much a gene gets used. It’s wild stuff. Indeed, “Many human accelerated regions are developmental [gene] enhancers” (Capra et al., 2013).

In the present case, hCONDEL.306 refers to the human-specific deletion of a developmental enhancer located near the GDF6 gene, which is a bone morphogenetic protein. The major finding of the paper, as stated succinctly in the Highlights title page, is that “Humans have lost a conserved regulatory element [hCONDEL.306] controlling GDF6 expression…. Mouse phenotypes suggest that [this] deletion is related to digit shortening in human feet.”

How do they link this “gene tweak” to digit shortening?

Since humans have lost this gene enhancer that is highly conserved in other mammals, Indjeian and team reasoned that the chimpanzee DNA sequence associated with this deletion, retaining the enhancer sequence, is likely the ancestral condition from which the human version evolved. They inserted the chimpanzee version into mouse embryos and watched what happened as they developed. The enhancer was only active in the mice’s back legs, specifically in the cartilage that would later become the lateral toe bones and cells that would become a muscle of the big toe (abductor hallucis). These are areas where humans and chimpanzees differ: our lateral toes are shorter than chimps’, and we only have one abductor hallucis muscle whereas chimpanzees have an additional, longer abductor hallucis  (Aiello and Dean, 2002). So, we’re on our way to seeing how hCONDEL.306 might relate to our big toe or upright walking, as the headlines say.

But this still doesn’t explain how this deletion affects GDF6 gene expression, and therefore what this does for our feet. Pressing onward, the scientists compared the size of certain bones in mice with a normal Gdf6 gene, and those in which the Gdf6 gene was completely turned off (or “knocked out”).  The Gdf6 knock-out mice had shorter lateral toe bones than regular mice, but they also had shorter big toes as well – the previous experiment staining mouse embryos showed the ancestral enhancer was expressed more in the latter toes, not so much the big toe.

Figures 5-6 from Indjeian et al. (2016) sum up the findings. Figure 5 (left) shows that the ancestral version of the GDF6 enhancer (blue staining) is most strongly expressed in the lower limb, especially the fifth toe bone. Figure 6 (right) shows that a lack of GDF6 expression (black bars) results in shorter skull and toe bones. Combining these findings, humans lack a gene enhancer associated with the development of long lateral toes.

Figures 5-6 from Indjeian et al. (2016) sum up the findings. Figure 5 (left) shows that the ancestral version of the GDF6 enhancer (blue staining) is most strongly expressed in the lower half of the body, especially the fifth toe bone. Figure 6 (right) shows that a lack of Gdf6 expression (black bars) results in shorter skull and toe bones. Combining these findings, humans lack a gene enhancer associated with the development of long lateral toes.

hCONDEL.306 doesn’t completely turn off GDF6, so this second experiment doesn’t really tell us exactly what the hCONDEL does. But the results are highly suggestive. Indjeian and team showed that Gdf6 affects toe length, among other skeletal traits, in mice. The ancestral enhancer that humans are missing seems to affect GDF6 activity in the leg/foot only. This illustrates a mechanism of modularity – as the authors state, “Loss of this enhancer would thus preserve normal GDF6 functions in the skull and forelimbs, while confining any … changes to the posterior digits of the hindlimb.” In other words, developmental enhancers allow different parts of the body to evolve independently despite being made by some of the same genes (such as GDF6).

As with any good study, results are intriguing but they raise more questions for future studies. The researchers conducted two experiments to investigate the function of hCONDEL.306: first putting the chimp version in mouse embryos to see where the ancestral enhancer is expressed, and then turning off Gdf6 completely in mice to see what happens. A more direct way to see what hCONDEL.306 does might be to put a longer stretch of the human sequence surrounding GDF6 containing (or rather missing) the ancestral enhancer into mouse embryos. I’m not a molecular biologist so maybe this isn’t possible. But this is important because the ancestral (chimpanzee) enhancer appeared to be most strongly expressed in the little toe, but of course this isn’t our only toe that is short compared to chimps. Similarly, how hCONDEL.306 relates to the abductor hallucis muscle remains in question – does it reduce the size of the intrinsic muscle present in both humans and chimps, or does it prevent development of the longer muscle that chimps have but we lack? We can expect to find hCONDEL.306 in the genomes of Neandertals (and Denisovans?), since they also have short toes, but what would it mean if they retained the ancestral enhancer?

So how does this gene tweak help with upright walking?

This is a really cool paper with important implications for human evolution, but something seems to have been lost in translation between the paper and the headlines (the news pieces themselves are good, though). The upshot of the study is that humans lack a stretch of non-coding DNA, which in chimpanzees (or chimp-ified mice) promotes embryonic development of the lateral toes and a big toe muscle. This may be a genetic basis for at least some aspects of our unique feet that have evolved under natural selection for walking on two legs.

But the headlines misrepresent this result, with words like “led to,” “allowed,” and “caused,” especially when these are followed by “big toe” or “upright walking.” hCONDEL.306 doesn’t really have anything to the big toe bone itself, although it might relate to a muscle affecting our big toe. The only sense in which the “Gene tweak led to humans’ big toe” (first title above) is that hCONDEL.306 might be responsible for our short lateral toes, which make our first toe look big by comparison. The other headlines are misleading since we know from fossil evidence that hominins walked upright long before we have evidence for short toes:

These little piggies get none. Fourth toe bones of living apes and humans (left) and possible hominins from 3-5 million years ago (right).

These little piggies get none. Fourth toe bones of living apes and humans (left) and (probable) hominins from 3-5 million years ago (right). I did my best to get all images to scale.

“Epigenetic,” from the fourth article headline, is simply wrong. Modern day epigenetics is a field concerned with the chemical alterations to the structure of DNA. Even the broad concept of epigenetic as originally devised by Conrad Waddington was about how environments (cellular or outside the body) influence development.

ResearchBlogging.orgIt’s hard to fit all the important and interesting information from scientific papers into news headlines. Still, it would be good if headlines more accurately portrayed scientific findings, especially avoiding such definitive verbs as “caused.” Especially in the realm of biology, people should know that there’s a lot that we still don’t know, that there’s lots more important work left to be done.


Aiello and Dean, 2002. Human Evolutionary Anatomy. Academic Press.

Capra et al., 2013. Many human accelerated regions are developmental enhancers. Philosophical Transactions of the Royal Society B 368: 20130025.

Indjeian et al. 2016. Evolving new skeletal traits by cis-regulatory changes in bone morphogenetic proteins. Cell http://dx.doi.org/10.1016/j.cell.2015.12.007

King and Wilson, 1975. Evolution at two levels in humans and chimpanzees. Science 188: 107-116 DOI: 10.1126/science.1090005

McLean et al., 2011. Human-specific loss of regulatory DNA and the evolution of human-specific traits. Nature 471: 216-219.

Prabhakar et al., 2008. Human-specific gain of function in a developmental enhancer. Science 321: 1346-1350.

eFfing #FossilFriday: toolmakers without tools?

Matt Skinner and colleagues report in today’s Science an analysis of trabecular bone structure in the hand bones of humans, fossil hominins and living apes. Trabecular bone, the sponge-like network of bony lattices on the insides of many of your bones, adapts during life to better withstand the directions and amounts of force it experiences. This is a pretty great property of the skeleton: bone is organized in a way that helps withstand usual forces, and the spongy organization of trabeculae also keeps bones fairly lightweight. Win-win.

An X-ray of my foot. Note that most of the individual foot bones are filled with tiny 'spicules' (=trabeculae) of bone. Very often they have a very directed, or non-random, orientation, such as in the heel.

An X-ray of my foot. The individual foot bones are filled with narrow spicules (=trabeculae) of bone. Very often they have a directed, or non-random, orientation: in the calcaneus, for instance, they are oriented mostly from the heel to the ankle joint.

This adaptive nature of trabecular bone also means that we can learn a lot about how animals lived in the past when all they’ve left behind are scattered fossils. In the present case, Skinner and colleagues tested whether tool use leaves a ‘trabecular signature’ in hand bones, looking then for whether fossil hominins fit this signature. Their study design is beautifully simple but profoundly insightful: First, they compared humans and apes to see if the internal structure of their hand bones can be distinguished. Second, they tested whether these differences accord with theoretical predictions based on how these animals use their hands (humans manipulate objects, apes use hands for walking and climbing). Third, they determined whether fossil hand bones look more like either group.

Comparison of first metacarpals (the thumb bone in your palm) between a chimpanzee (left), three australopithecines, and a human (right). In each, the palm side is to the left and the wrist end of the bone (proximal) is down. Image by Tracy Kivell, and found here.

Looking at the image above, it’s difficult to spot trabecular differences between the specimens with the naked eye. But computer software can easily measure the density and distribution of trabecular bone from CT scans. With these tools, researchers found key differences between humans and apes consistent with the different ways they use their hands. Neandertals (humans in the past 100 thousand years or so) showed the human pattern, not unexpected since their bones look like ours and they used their hands to make tools and manipulate objects like we do.

What’s more interesting, though, is that the australopithecines, dating to between 1.8-3.0 million years ago, also show the human pattern. This is an important finding since the external anatomy of Australopithecus hand bones shows a mixture of human- and ape-like features, with unclear implications for how they used their hands. Their trabecular architecture, reflecting the forces their hands experienced in life, is consistent with tool use.

This is a very significant finding. Australopithecus africanus fossils from Sterkfontein aren’t associated with any stone tools; bone tools are known from Swartkrans, though it is unclear whether Australopithecus robustus or Early Homo from the site made/used these. In addition, in 2010 McPherron and colleagues reported on a possibly cut-marked animal bone from the 3.4 million year old site of Dikika in Ethiopia, where Australopithecus afarensis fossils but no tools are found. Skinner and colleagues’ results show that at the very least, South African Australopithecus species were using their hands like tool-makers and -users do.

This raises many fascinating questions – were australopithecines using stone tools, but we haven’t found them? Were they using tools made of other materials? What do the insides of Australopithecus afarensis metacarpals look like? What I like about this study is that it presents both compelling results, and raises further (testable) questions about both the nature of the earliest tools and our ability to detect their use from fossils.

Results of the toe-tally easy lab activity

Alternate title: Dorsal canting in primate PPP4s

Earlier this year I suggested a classroom activity in which students can scrutinize the evidence used to argue that the >5 million year old (mya) Ardipithecus kadabba was bipedal. To recap: Ar. kadabba is represented by some teeth, a broken lower jaw, and some fragmentary postcrania. The main piece of evidence that it is a human ancestor and not just any old ape is from a single toe bone, and the orientation of its proximal joint. In Ar. kadabba and animals that hyperdorxiflex their toes (i.e., humans and other bipeds when walking), this joint faces upward, whereas it points backward or even downward in apes. This “dorsal canting” of the proximal toe joint has also been used as evidence that the 4.4 mya Ardipithecus ramidus and 3.5 mya owner of the mystery foot from Burtele are bipedal hominins. A question remains, though – does this anatomy really distinguish locomotor groups such as bipeds from quadrupeds?

Use ImageJ to measure the canting angle between the proximal joint and plantar surface. Proximal to the right, distal to the left.

STUDENT SCIENTISTS TO THE RESCUE! Use ImageJ to measure the canting angle between the proximal joint and plantar surface, as I’ve done on this Japanese macaque monkey (they are not bipedal). Proximal to the right, distal to the left Note I changed the measured angle from the March post.

I sicked my students in Ant 364 (Human Evolutionary Developmental Biology) here at NU on this task. I had students look at only 11 modern primates from the awesome KUPRI database. Most groups are only represented by 1 (Homo sapiens, Hylobates lar and Macaca fuscata) or two (Pongo species and Gorilla gorilla) specimens, all adults. For chimpanzees (Pan troglodytes) there is one infant and four adults. The database has more individuals, and it would be better to include more specimens to get better ideas of species’ ranges of variation, but this is a good training sample for a class assignment. The fossil group includes one Ardipithecus ramidus, one Ar. kadabba, one Australopithecus afarensis, and the PPP4 of the mystery foot from Burtele. The human and all fossils except Ar. kadabba are based off of lateral photographs and not CT scans like for the living primates, meaning there may be some error in their measurements, but we’ll assume for the assignment this is not a problem. Here are their results:

Dorsal canting angle of the fourth proximal pedal phalanx in primates.

Dorsal canting angle of the fourth proximal pedal phalanx in primates. The lower the angle, the more dorsally canted the proximal joint surface. The “Fossil” group includes specimens attributed to ArdipithecusAustralopithecus and something unknown.

Great apes have fairly high angles, meaning generally not dorsally canted proximal joint surfaces. The two gorillas fall right in the adult chimpanzee (adult) range of variation, while chimp infant and orangutans have much higher angles (≥90º means they’re actually angled downward or plantarly). The gibbon (Hylobates) is slightly lower than the chimpanzee range. The macaque has an even more dorsally canted joint, and the human even more so. The fossils, except the measurement for Ar. ramidus (see note above), have lower angles than living apes, but higher than the human and the monkey. If dorsal canting really is really a bony adaptation to forces experienced during life, then the fossil angles suggest these animals’ toes were dorsiflexed more so than living great apes (but not as much as the single monkey and human).

This lab helps students become familiar with CT data, the fossil record, taking measurements (students also measure maximum length of the toe bones and look at the relationship between length and canting), analyzing data, and hypothesis testing. You can also have fun exploring inter-observer error by comparing students’ measurements.

Here’s the full lab handout if you want to use or modify it for your own class: Lab 5-Toe instructions and report

Friday excitement: Panoramic data inspection

I teach Tuesdays and Thursdays this year, leaving Fridays welcomely wide open for non-teaching related productivity. Today’s task is arguably the most exhilarating aspect of doing Science – inspecting raw data to make sure there are no major errors or problems in the dataset, so I can then analyze it and change the world. The excitement is truly hard to contain.

Delectable dog food is the dataset; I’m the dog.

No, it’s not the funnest, but it’s an important part of doing Science. To make your life easier, you should inspect data daily as you collect them. This way, you can identify mistakes and make notes about outliers early on, so that you are not stupefied and stalemated by what you see when you sit down to begin analysis.

You (corgi) are getting ready to analyze and you find an anomalous observation (door stop) you didn’t notice when you were collecting data.

Today I’m looking at measurements I took from ape mandibles housed in an English museum last summer; I inspected data before I left the UK for KZ, so today should be a breeze. But no matter how meticulous you are in the field/museum, you still need to inspect your data before analyzing them, just to be safe. If you’re as disorganized as I am, there will be lots of programs each with lots of windows. Here’s a tip: plug into multiple monitors (or at least one big ass monitor), so you can easily espy all open windows and programs in prodigious panorama.

Using two monitors helps when checking data for errors and patterns

Using two monitors helps when checking data for errors and patterns. On my left screen I’m using R to visualize and examine the raw data open in Excel on the right screen. If something seems off on the left screen, I can quickly consult the original spreadsheet on the right.

Barely visible in the above screenshot, these are chimpanzee (red) and gorilla (black) mandible measurements plotted against a measure of body size, preliminarily described in this post from last August. I’m looking at whether any mandibular measurements track body size across the subadult growth period, in hopes that bodily growth can be studied in fossil species samples dominated by kid jaws. As you can (barely) see, some jaw measurements correlate with body size better than others, and sometimes the apes follow similar patterns but other times they don’t.

The data look good, so now I can go on to examine relationships between mandible and body size in more detail. Stay tuned for results!

Genes and culture in animals

Recent UM Ph.D. Kevin Langergraber and others (including UM primatologist John Mitani; 2010) recently reported on a high correlation between genetic relatedness and ‘cultural’ behavioral repertoire in wild chimpanzees.

Chimpanzees, and other animals, have been observed to display behaviors that appear ‘cultural,’ since the behaviors are variously a) learned from other individuals, b) specific to certain chimp populations, and/or c) not recognizably adaptive. Such behavioral variants include, for example, how hands are clasped during grooming (photo from Whiten, 2005), or whether/how insects are acquired.

Researchers have debated whether these behavioral variants actually represent culture, in the sense that humans have culture. This itself is tough because anthropologists have had a helluva time defining what ‘culture’ is simply for humans, let alone animals. I’m a bit anthropocentric myself, and I’m wont to view culture as something uniquely human, the adaptation (or set of adaptations) that has essentially shaped our evolution for over a million (2 million?) years.
Anyway, back to Pan, Langergraber and colleagues set out to test whether genetic variation may help explain some of the behavioral variation between different chimp populations. Lo and behold! there was a significant correlation between groups’ genetic dissimilarity and behavioral dissimilarity. This isn’t at all to say the authors have found the genetic basis for cultural behaviors, but rather that some genetic variation may underlie some behavioral variation we see in chimpanzees. Indeed, the authors note that the mtDNA used in the study doesn’t ‘code for’ any of the putatively cultural behaviors; it’s a proxy for genetic relatedness. However, there was no clear pattern of which types of behaviors (e.g. grooming- vs. feeding-related) correlate with genetic relatedness.
The results are a bit tough to interpret. The authors state that the finding of a correlation does not mean that many chimp behaviors analyzed are not cultural. But it doesn’t necessarily mean that the behaviors are cultural, either. This gets really tricky for a number of reasons.
First, identifying “the” or “a” genetic basis for phenotypes is difficult, and it’s especially difficult for complex phenotypes like behaviors (in general, if you ever hear about a “gene for” some behavior, immediately disbelieve it). The analysis uses an allegedly neutral DNA marker, that admittedly does not ‘code for’ any of the behaviors in question. All the DNA can do here is attempt to indicate relatedness among groups. To say that “genetic differences cannot be excluded as playing a major role” in patterning behavioral variation (p. 7), basically means that some unexamined genetic region may be patterned among populations the same way as the mtDNA marker, and might be responsible for specific, fine-tuned, non-adaptive aspects of their behavior. The authors discount the possibility of the link being due to kin teaching behaviors to kin, but I would suppose a higher resolution (like looking at relatedness and behavior between individuals rather than groups) would put that matter to rest.
Next, how much of a correlation is biologically (and here culturally?) meaningful? In various permutations of their analysis, the correlations between the behavioral and genetic dissimilarity matrices ranged from r = 0.37 – 0.52, most of which were significant. “Significant” here means that the correlation coefficients, r, are different enough from zero – there isn’t no relationship between the variables (I mean to say the double negative). Put another way, we can square the r coefficients to get the ‘amount of variance explained’: 13.7 – 27.0% of the behavioral dissimilarity can be ‘explained’ by genetic dissimilarity. What if the correlation coefficients had been higher – would this be better evidence for some genetic basis for chimp behavioral variants? I love correlation as much as the next guy, but aside from significance level, variation in linearity is not always completely understandable.
So, regardless of the results of the analysis, do apes (or other non-human animals) have culture? An interesting conundrum is that when people describe the subtle variants of behavior as cultural, they’re assuming the variation itself is non-adaptive, while the grand behavior itself purportedly is. Can things that are readily adaptive (ecological explanation) not also be cultural? Moreover, how widespread in a population must a behavioral variant be to be cultural? How many variants on a theme are permissible within a population? Questions like these are why I tend to shy away from the topic of culture, in humans and animals.
Langergraber K, Boesch C, Inoue E, Inoue-Murayama M, Mitani JC, Nishida T, et al. Genetic and ‘cultural’ similarity in wild chimpanzees. Proc R Soc B in press. Proc. R. Soc. B doi:10.1098/rspb.2010.1112 (2010)
Whiten A. 2005. The second inheritance system of chimpanzees and humans. Nature 437: 52-55.