Did early tetrapods walk on land?

“Fishapods” like Tiktaalik and Ichthyostega are among the iconic transitional fossils that mark part of our own evolutionary journey. Understanding how these intriguing animals lived is key to understanding why they ended up out of water, and how evolution took them there. Of course, one of the most important changes that happened to our fishy forebears is gaining the ability to walk. While “walking” may be less of a challenge to fish than we thought, just how good “classical” proto-tetrapods like Ichthyostega were at it still isn’t entirely clear. (Below: model of Ichthyostega emphasising its aquatic capabilities; photo by Dr Günther Bechly, Wikimedia Commons.)

The recent discovery of tetrapod-like trackways in Mid-Devonian rocks in Poland (Niedźwiedzki et al., 2010) added to the pile of evidence on the “pretty good” side. The footprints appeared to come from feet that looked like those of Ichthyostega, and the trackways betrayed an animal that walked essentially like a modern tetrapod, lifting its body clear off the ground. (The fact that these trackways were left in a different environment and 18 million years earlier than any previously known tetrapod is just the icing.)

As is usually the case in science, the paper wasn’t allowed to be simply right. Using a computerised Ichthyostega skeleton and comparisons to modern tetrapods, Pierce et al. (2012) argue that whatever made the Polish tracks and others like them, it couldn’t have been much like Ichthyostega.

The problem is that Ichthyostega had pretty stiff hip and shoulder joints. The joint surfaces are kind of elongated (and, in the case of the shoulder, also twisted), allowing the bones to hinge in various directions but not to rotate around their long axes. The unrotating hip and rigid knee of the hindlimb made it impossible for the animal to put its soles on the ground. The hindlimb looks very paddle-like with its broad, flattened bones – now it appears that paddling is all it was good for. (Below: a decade of Ichthyostega from Dennis Murphy’s wonderful Devonian Times. See his caption for the sources of the reconstructions.)

By all appearances, Ichthyostega couldn’t rotate its arms and legs enough to walk properly, and couldn’t have left even one of the presumed Polish hindlimb prints without dislocating several joints. The most it could have done is haul itself along like a seal or a mudskipper, which, considering its powerful arms and mobile elbows,might have been exactly what it did. Pierce et al. also argue that other known proto-tetrapods were unlikely to be the trackmakers; while they didn’t examine their limb joints in detail, the skeletons share many of the same features that limit the mobility of Ichthyostega‘s limbs.

That raises plenty of questions. If it wasn’t like any of the known early tetrapods, what sort of creature made those footprints? Were there other ancient tetrapod groups with more limber joints that didn’t leave body fossils because of where they lived? Did the known proto-tetrapods go back to a more aquatic existence and more rigid limbs? (Pierce et al. say that Ichthyostega‘s joints were even stiffer than some of its fishier cousins’!) Which kind begat the lineage that gave rise to living tetrapods? What does all of this say about the significance of walking-like movements we observe in living lobe-finned fish?



Niedźwiedzki G et al. (2010) Tetrapod trackways from the early Middle Devonian period of Poland. Nature 463:43-48

Pierce SE et al. (2012) Three-dimensional limb joint mobility in the early tetrapod Ichthyostega. Nature advance online publication, 23 May 2012. doi: 10.1038/nature11124


Life has gone slightly too crazy for me to embark on a serious meandering, but in the meantime, this face was too funny not to share:

Our bug-eyed beauty is an edible frog (Pelophylax esculentus), and the tactless photographer who captured it from such an unflattering angle is Grand-Duc of the all-knowing Wikipedia.

What use is (not even) half a leg?


(I’m even further behind on things than usual, so this is not that “hot” off the press, but the walking lungfish can’t not be posted on.)

The evolution of new traits serving new functions is always a bit of a chicken and egg problem. Why would you need wings if you don’t fly, and how could you start flying without them? Why would you need legs if you don’t walk, and how would you walk without legs?

Often, as in the case of wings, the most likely answer is that the trait originally had a different function that didn’t necessitate a “perfect” version of it. Wings that are no good for flying could be anything from egg-warmers/shades through mate attraction devices to balancing organs for prey-wrestling predatory dinosaurs (latter idea from Fowler et al., 2011, which by now has probably gone as viral as scientific papers can).

With legs, though, it seems that the chicken really did come first. We’ve known for a long time that coelacanths (which are somewhat distantly related to vertebrates with legs) sometimes move their pectoral and pelvic fins in an alternating rhythm that resembles walking. (IIRC you can find a fair few YouTube videos in which they are filmed doing that.) Nonetheless, coelacanths use this movement for swimming. They don’t actually get down and plod along the bottom.

Lungfish, however, do. King et al. (2011) videoed them doing it.

Just to be clear, the animal in question is the West African lungfish (Protopterus annectens). Unlike the respectable paddles of the Australian species, its spindly paired appendages barely even deserve to be called fins, let alone legs. (Drawing below from King et al., 2011)

Yet this creature uses its pelvic fins to propel itself along the bottom in a variety of ways. It can walk with alternating “steps”, it can bound by moving both fins at once, and sometimes it just ambles along in a slightly irregular way (videos here). If there’s no traction on the bottom of the tank, it slips and can’t get anywhere, which indicates that it does indeed propel itself by pushing against the bottom with its hind fins. And sometimes, when the fins push off, you can see part of the body come clear off the ground.

(Interestingly, the lungfish walks and bounds only with its hind fins. Meanwhile, the pectorals flail around doing other things, but they don’t engage with the floor. The diagram above gives a clue why: the animal has huge, air-filled lungs – the grey blob – that help its front half float. It doesn’t need its forefins to stroll around.)

Given how un-leglike the fins of African lungfish are, it is obvious that walking underwater doesn’t require anything as sophisticated as ankles or toes or, heck, even proper fins. Just about any ancient lobe-finned fish we know could have been capable of it. Could this be how our ancestors took their first unknowing steps towards land? Were they bottom-dwelling fish that patrolled their territories in a stately fin-walk? Did increasingly leg-like fins just help them do that better rather than breaking new ground? As the authors remind us, we already know that many of the earliest tetrapods – creatures with true legs – lived in water. If less tetrapod-like creatures could walk, then the picture fits quite nicely together.

And speaking of chickens and eggs, once again nature proves how much human incredulity is worth. Just because you don’t know what to do with half a wing, just because you don’t think X is possible without Y, doesn’t mean solutions don’t exist. Studying nature is a life-long lesson in humility in that way.



Fowler DW et al. (2011) The predatory ecology of Deinonychus and the origin of flapping in birds. PLoS ONE 6:e28964

King HM et al. (2011) Behavioral evidence for the evolution of walking and bounding before terrestriality in sacropterygian fishes. PNAS 108:21146-21151

The bare bones of fins and limbs

Perhaps the central question in developmental biology is how cells that start out as identical end up making bodies with complex shapes and a multitude of different tissues. And perhaps the central question in evo-devo is how such bodies can change into other bodies during the course of evolution. A really cool paper by Zhu et al. (2010) probes a little bit at both, and shows how relatively simple rules can produce results that are surprisingly similar to what we observe in nature.

The authors modelled the development of limb (or fin) bones in vertebrates. They used a simple model made up of the following:

  1. a virtual limb bud (let me call them “simbuds” hereafter) growing continuously
  2. a signal spreading from the tip of the bud that tells “cells” to keep growing but wanes over time (mimicking the role of the apical ectodermal ridge in real limb buds)
  3. two equations describing the activity of (1) genes that make cells differentiate into bone (“activators”), and (2) genes that prevent cells from doing so (“inhibitors”)

The shape of the simbud could be set at the start, and so could the values of all the parameters in the activator and inhibitor equations.

This is much more simple than real limb develompent. It says nothing about cell movement, and it condenses the effect of genes other than the bone activators and inhibitors into two little parameters in the equations. Yet running it with pretty much any initial settings produces something vaguely limb-like, and some sets of parameters give you simbuds that look eerily like real limbs.

Development of a simbud mimicking a chicken wing, next to drawings of the real thing.

Or fins. Or mutant limbs. Or transitional fossils.

Fully developed simbuds resembling various fossil fins: Brachypterygius was a marine reptile from the Jurassic; the other four are more or less close relatives of tetrapods, among them the famous "fishapod" Tiktaalik.

The similarity is not perfect, of course – but the model is not perfect either. Overall, it’s still pretty amazing what a variety of very realistic limb skeletons you can get out of such a simple setup – and how much you can achieve just by varying small things like how wide the limb bud is to begin with or how strongly two gene networks interact. Evolving fins into limbs should be a piece of cake for a system like that!



Zhu J, Zhang Y-T, Alber MS, Newman SA (2010) Bare Bones Pattern Formation: A Core Regulatory Network in Varying Geometries Reproduces Major Features of Vertebrate Limb Development and Evolution. PLoS ONE 5:e10892

Finger counting

Hello world!

I’m not technically new to the blogging enterprise. I’ve kept a private journal-and-repository-of-ideas for years now, but whenever I felt like I had something to say that was worth sharing, I inevitably changed my mind. But, with so much wonderful, exciting stuff out there, isn’t it practically a crime for a scientist not to share?

I thought it fitting to start my science blogging career with tetrapod limbs, since they played a great part in setting me on the path that led me where I am today. If I had to define my specialisation, I would call myself an evolutionary developmental geneticist. (Add “molecular” in there somewhere to make it more accurate and more cumbersome ;)) This means investigating the genes that control the formation of body parts in various organisms, and trying to figure out what this tells us about the evolution of said body parts. One of the early signs that alerted me to the awesomeness of the field was a paper about limbs (Davis et al., 2007) – more precisely, about how the genetic program that specifies our fingers and toes is present in the fins of fish with nothing resembling digits. That is an intriguing story that I might tell another day, but today, I will discuss something else to do with limbs, apropos a study that came out recently in Science.

Birds have some of the most extremely modified forelimbs among tetrapods. If you squint at the skeleton inside the wing hard enough, you may be able to recognise the stunted remains of three digits. The homology of those three digits to the digits of more conventional forelimbs has been a conundrum ever since someone first examined the limb buds of a bird embryo.

Homology is a tricky concept. It was originally defined before the theory of evolution took off, by a guy who had a rather strange relationship with Darwin and Wallace’s theory later on. Richard Owen (see here for a biography) defined homology as:

the same organ in different animals under every variety of form and function

Later, evolutionary biology embraced the word as a neat shorthand for structures, processes, genes, even behaviours, that evolved from a common ancestor. The obvious problem is that organs and genes don’t come with labels listing what they are homologous to. Therefore, homology has to be inferred. These inferences can be based on a variety of sources such as morphological similarity, fossil evidence, embryonic origin and shared genetic programs. The more lines of evidence converge on the same conclusion, the stronger the conclusion is. But what happens when those different sources contradict?

In the case of avian digits, morphology [1], and later palaeontology, suggested that they were digits I, II and III, that is, a thumb, an index and a middle finger. While it’s a bit hard to say anything about the morphology of those fused and stunted bones in a modern wing with a straight face, the fossil evidence is pretty unambiguous. As you go from the earliest dinosaurs to birds, you can clearly follow the loss of digit V first, followed by digit IV. The remaining three digits are quite clearly I-II-III in animals such as Archaeopteryx.

A comparison of the metacarpals (palms) of dinosaurs. Left to right, in order of increasing relatedness to birds: Herrerasaurus (note the vestigial fifth metacarpal (labelled "V")), Coelophysis (MCV gone), Allosaurus (MC4 gone), Deinonychus, Archaeopteryx, and Nothura (a modern tinamou). Image from Wagner and Gauthier (1999)

However, embryos seemed to tell a completely different story. During development, the skeleton of a limb forms from little condensations of tissue inside the limb bud (which, at that point, looks more like a weird-shaped sausage than a limb). These condensations are first cartilaginous, later laying down bone. Around the forming digits, muscles and connective tissues organise, and finally, the padding between the digits dies away, transforming the paddle-like limb bud into a hand or foot [2]. The forelimbs of birds make four condensations – more than needed to form their three digits, but fewer than five, making it difficult to tell exactly which ones remain.

Nevertheless, regularities observed in limb development gave scientists clues. The condensations that turn into digits in tetrapods don’t  all form at the same time – in fact, they form in a stereotypical order in which CIV forms first and CI last (Burke and Feduccia, 1997). Condensation (and digit) IV is considered part of the main axis of the limb (the so-called metapterygial axis), and its presence is thought to be essential for the other condensations to form. You cannot lose CIV without losing all the rest, so conventional wisdom went. That means that birds must have a CIV. And indeed, the third digit of a wing comes from something that looks suspiciously like a CIV, with two other condensations appearing in front of it and one behind. Thus, embryology would have us think that the digits are II, III and IV.

Diagram comparing the development of the forelimb skeleton in alligators (top) and ostriches (bottom). Image from Wagner and Gauthier (1999).

This presents a dilemma that is a common one in evo-devo. Which kind of evidence do you put more stock in? Is there a way to resolve the conflict, or does it spell doom for the hypothesis that birds evolved from dinosaurs (and then, what’s up with the four- and three-fingered dinosaurs that obviously aren’t birds?)? What’s going on here?

An important paper (Wagner and Gauthier, 1999) over a decade ago suggested a possible resolution called the developmental frame shift hypothesis. They proposed that birds do in fact form condensations II-IV – but then something strange happens. The genetic program that specifies digit identity is not switched on until all pre-digital condensations have formed. Thus, a condensation can become any digit, depending on which genes are expressed at its location. So, Wagner and Gauthier said, at some point in their history, dinosaurs shifted the “normal” forelimb gene expression patterns backwards, so CII now ended up in the genetic environment of DI, and thus formed a DI. That is, the fingers of birds are homologous to different digits of the basic pentadactyl limb at different levels.

Over the years since, strong genetic support has been gathered for this idea. The first digit of the hand of a bird does, indeed, form in an area that bears the genetic hallmarks of DI in more typical tetrapods, including the closest living relatives of birds (Vargas et al., 2008). So, the frameshift hypothesis of Wagner and Gauthier seems to stand on pretty solid legs.

A recent study had a new look at wing development, confirming the identity of wing digits as I-III. Tamura and others (2011) followed the origin of digits at the cellular level. An early limb bud, before it has much obvious internal structure, already contains what developmental biologists call organisers or signalling centres. At the tip of the bud, a narrow ridge (the so-called apical ectodermal ridge, or AER) gives off chemical signals that direct the correct outgrowth of the limb. At the posterior, tailward side of the bud, there is an area called the Zone of Polarising Activity, which defines the anterior-posterior polarity of the limb: the side of the ZPA is the pinky side, and where everything else develops is determined by the levels of morphogens – most famously the protein Sonic hedgehog – that diffuse from the ZPA.

In ordinary, five-fingered tetrapods, the ZPA contributes tissue to the fifth and fourth condensations (and digits). The first thing Tamura and colleagues did, then, was to transplant ZPAs between the fore- and hindlimb buds of chick embryos (remember, chicken legs are four-toed). When you add a ZPA on the wrong side of a limb bud, you get a mirror image duplication of the limb. (How much is duplicated depends on the age of the transplant and the recipient.) In this study, the researchers found that a hindlimb ZPA of the right age grafted onto a forelimb bud often led to the formation of hindlimb digits in the forelimb. The converse was not true: the forelimb ZPA was almost invariably unable to form forelimb digits, though it was able to induce the doubling of the foot. Thus, it seems cells in the forelimb ZPA aren’t good at making digits on their own, only at directing other tissues to form digits.

The second experiment involved tracing cell lineages with a clever method that’s widely used in similar experiments due to its relative simplicity. They stained cells with a dye that sticks to the cell membrane. This dye is inherited by the descendants of the mother cell: when it divides, the daughter cells will split the original membrane – and the dye – between them. So, if the ZPA contributes to the last digit of the wing, we should see staining in the digit if we’d dyed the ZPA before digits begin to condense. In the foot, that’s exactly what happens. Not so in the wing: the last digit forms outside the ZPA, with no contribution from the labelled cells.

And lastly, the researchers went further back in time, using the same cell labelling method to trace the fate of different parts of the limb bud from an even earlier stage. What they found is that the region that gives rise to the third finger does lie inside the ZPA at this point – but it leaves the zone very soon after, unlike the progenitor of the fourth toe. Thus, well before any condensation is apparent, this bit of tissue is different from conventional fourth digits.

Relationship of ZPA and digit identity according to Tamura et al. (2011). Top: regions that will give rise to condensations and digits. Middle: these regions are specified before condensation begins. Bottom: identities of the forming digits. Light blue region = ZPA, SHH = Sonic hedgehog

Interestingly, what this study identifies as a digit III based on its origin appears to lie on the metapterygial axis – a place thought to be the privilege of fourth digits. What’s more, removing the posterior portion of the limb bud relocates the second digit to the main axis. Thus, it appears that limb development is a lot more flexible than Burke and Feduccia had concluded over a decade earlier.

So, in a sense, Wagner and Gauthier were wrong – but in a different sense, they were more right than they thought. It appears that the frame shift they hypothesised doesn’t happen after condensations appear: it happens well before. What was previously identified as “CIV” already bears characteristics of CIII when the first lumps of cartilage begin to form.

In a way, the story of avian digits is a beautiful illustration of the scientific process. From a controversy sparked by a seemingly insurmountable contradiction, we have moved to a synthesis that accounts for all available evidence. Scientists did not dismiss the contradiction, they worked to make sense of it. They called on new lines of evidence to resolve what the old evidence could not. Some ideas proposed in the process – such as the rigidity of the digit developmental program – turned out largely wrong. Others – the original frame shift hypothesis – still seem somewhat wrong, but their essence carried over into the newest picture. It’s entirely possible that this isn’t the last word on bird fingers either. But at this point, I am reminded of Asimov’s essay The Relativity of Wrong (Asimov, 1989). Whatever the future brings, it’s a good bet that we’re not nearly as “wrong” as we were before.

– – –

[1] Comparing ordinary, five-fingered forelimbs, some characteristics are pretty consistent across the different groups of tetrapods. Most importantly, thumbs have only two phalanges, fewer than all the other digits.

[2] This isn’t universally the case. When the interdigital tissue doesn’t disappear, you get webbed feet.

– – –


Asimov I (1989) The Relativity of Wrong. The Skeptical Inquirer 14:35-44

Burke AC and Feduccia A (1997) Developmental patterns and the identification of homologies in the avian hand. Science 278:666-668

Davis MC, Dahn RD and Shubin NH (2007) An autopodial-like pattern of Hox expression in the fins of a basal actinopterygian fish. Nature 447:473-476

Tamura K et al. (2011) Embryological evidence identifies wing digits in birds as digits 1, 2, and 3. Science 331:753-757

Vargas AO et al. (2008) The evolution of HoxD-11 expression in the bird wing: insights from Alligator mississippiensis. PLoS ONE 3:e3325.

Wagner GP and Gauthier JA (1999) 1,2,3 = 2,3,4: A solution to the problem of the homology of the digits in the avian hand. PNAS 96:5111-5116