How to make a bat

The relative length of bat forelimb digits has not changed in 50
million years. (a) Icaronycteris index, which is a 50-million-year-old bat fossil. (b) Extant adult
bat skeleton. The metacarpals (red arrows) of the first fossil bats are already
elongated and closely resemble modern bats. This observation is confirmed by
morphometric analysis of bat forelimb skeletal elements.

or•gan•ic | ôr’ganik | adjective. denoting a relation between elements of something such that they fit together harmoniously as necessary parts of a whole; characterized by continuous or natural development.

One of the wonderful things about how development works is that organisms function as wholes, and changes in one property trivially induce concordant changes in other properties. Tug on one element, changing it’s orientation or size, and during embryogenesis any adjacent elements make compensatory adjustments, so that the resultant form flows, fits, and looks organic. This isn’t that surprising a feature of development, though, unless you have the mistaken idea that the genome encodes a blueprint of morphology. It doesn’t; what it contains is a description of interacting agents that work together in a process to produce a complex result. Changes in genes and regulatory elements can essentially produce changes in rules of development, rather than crudely specifying blocks of morphology.

What does this mean for evolution? It means that subtle changes to the rules of development can be caused by small changes to genes (and especially, to regulatory regions of genes), and that the resulting morphological changes may be dramatic, but are still integrated organically into the form of the organism as a whole. Our understanding of how development works is making it clear that large scale macroevolutionary change may be much easier than we had thought.

Here’s an example where this insight is clarifying the evolution of an organism: the fossil record of bats shows an abrupt appearance of fairly sophisticated creatures with elongated digits, clearly capable of gliding or powered flight, with no known intermediates. We expect there were less fully flight-ready predecessors, but fossil preservation is not kind to small, delicate boned animals. It’s also possible that the transitional period was fairly brief; it looks like turning a paw into a long-fingered membranous wing may be a fairly simple change on a molecular level.

This is a story about how short, stubby finger bones are turned into long, thin finger bones, so let’s start with how the bones of the hand form. Like all the long bones of the body, it begins with a condensation of mesenchymal tissue (this just tissue where the cells are connected in a loose network, swimming in a soup of extracellular proteins, and are relatively free to move) to form knots of cartilaginous cells where each digit will form. These cells proliferate and organize themselves into rods of a rubbery extracellular matrix. Continued growth is by continued proliferation of these cells, which extends the length of the rod. Later, other cells invade the cartilaginous matrix and replace the more flexible early cartilage with bone. When the digit is completely ossified, growth stops (we keep bones growing for a long time by setting aside specific regions, growth plates, that retain their proliferative cartilaginous nature.)

The proliferation of the early cells is a key control point. Keep them dividing longer before replacing them with bone, and the digit grows longer. Start the ossification process earlier, and you end up with short digits by truncating the proliferation process. How is that period of growth regulated?

There is a whole family of proteins called the BMPs—Bone morphogenetic proteins—that modulate bone growth, as you might expect from the name. In particular, one member of the family, Bmp2 is expressed specifically in hypertrophic cells of the growth plate in mice. As it turns out, if you take an embryonic mouse limb and culture it in a dish soaked with Bmp2, the digits grow longer; put an embryonic mouse limb in a dish that contains the protein Noggin, which antagonizes the function of Bmp, and the digits are stunted. This is suggestive: Bmp2 concentration could be the vernier that regulates digit length. Add more, fingers grow long, and add less, they grow short and stubby.

That work is all done in mice…so what about bats? There has been some fascinating work on bat embryology lately, and here’s one relevant result: in the early bat embryo, the proportions of the limbs and digits are indistinguishable from the proportions of the mouse. That means the organization of the tissue is initially similar, and only later, as the bones of the limb develop in that process regulated by the Bmps, do the proportions change.

Developmental elongation of bat digits occurs after the initial cartilage condensations are formed. (a and b) Cartilage revealed by Alcian blue staining of stage-16 bat forelimb and E12.5 mouse forelimb.

These embryonic bat limbs can be snipped off and grown in culture as well, in the presence of either Bmp2 or Noggin. To no one’s surprise, the digits grow longer in Bmp2, shorter in Noggin.

But there’s more! Bats have elongated digits in their forelimbs, but not in their hindlimbs. Quantitative analysis of the distribution of the Bmp proteins in the limbs shows that levels of Bmp2 in the hindlimb was roughly equivalent to what’s found in the mouse, but that Bmp2 levels were about 35% higher in the bat forelimb than either in mouse limbs or the bat hindlimb. Other members of the Bmp family do not show any correlated differences.

What this means is that the authors have identified a discrete molecule with a difference in its expression in the bat, and that this molecule is specifically involved in regulating digit length. This is a significant step in figuring out the evolutionary pathway that led to the bat wing.

Together, our results indicate the up-regulation of the Bmp
pathway as a major and fundamental (although not necessarily
the only) mechanism responsible for the developmental elongation of bat forelimb digits. Based on our results, we raise the
intriguing possibilit y that a similar up-regulation of the Bmp
pathway had a role in the evolutionary elongation of bat forelimb
digits, which is an event that was critical to the achievement of
powered flight in bats. Recent studies have suggested
that modifications to the cis-regulator y elements of developmental genes have central roles in the evolutionary diversifica-
tion of morphology. Our evidence that Bmp2, but not Bmp4 or
Bmp7, is differentially expressed in the bat wing digits is suggestive of a cis-regulatory change that affects the level, but not
the temporal or spatial regulation, of Bmp2 expression. By
linking a simple change in a single developmental pathway to
dramatically different morphologies, we provide a potential
explanation as to how bats were able to achieve powered flight
soon after they diverged from other mammals nearly 65 million
years ago.

Like the authors say, it’s not likely to be the only step involved. I would imagine that mutations in the Bmp2 pathway would have led to long-fingered ancestral Chiropteran, and another change might have suppressed apoptosis in the finger webbing to produce a large flat membranous wing-hand; this might have been a the first step in producing an animal capable of simple gliding. Coordinated, powered flight would have required a suite of genetic changes…but that first step that would have launched an animal in the direction of flight could have been very easy to achieve.

Sears KE, Behringer RR, Rasweiler JJ, Niswander LA (2006) Development of bat flight: Morphologic and molecular evolution of bat wing digits. Proc. Nat. Acad. Sci. USA 103(17):6581-6586.

The SciAm blog also has a summary.


  1. Stuball says

    Yay for Evo Devo! This is particularly good because I seem to recall reading somewhere that info on bat evolution is particularly scarce. As such, they were often used in creationist arguments.

  2. Karl says

    Does this inply potential treatment for children who appear to be growing more slowly than average (at, say, 10 years old)?

  3. Bruce Thompson says

    The locus for stature, specifically leg length in humans maps to the region containing BMP2 . The origins of Batman and the National enquirer’s Bat boy and the Native American Cochiti Bat boy legend may be explained by this observation.

    I’ve always wanted to use that reference for something.

  4. fnxtr says

    Would the concentration of bmp2 have increased gradually in the most succesful lineages, or was there a sudden ‘bump’ in bmp2? Would it be just a matter of scale or reiteration once the tendency was in place, or was it a single event?

  5. ermine says

    I would suppose that there would be an initial mutation that upped the bmp2 level enough to give the firt enlarged ‘gliding’ handwing, and subsequent mutations would increase more gradually the area, now that the ‘wing’ was large enough to be selected for. Remember the heat-shocked caterpillars article a month or so ago? The initial mutation turned them black, the shock caused the latent color-changing genes to express themselves. (a drop in pigment-stimulating hormones, as I recall?) All future generations were selected for color, allowing those with the largest color change to breed over and over, while selecting out those that didn’t change as much.

    In only 13 generations, the color-changeing mutation was fixed in the population, and -every- caterpillar went from black to wholly green when heated above a certain temperature during one phase of life.

    I’m not a biologist I’m afraid to say, but I would think it would be a similar process. Once the initial mutation made a selectable difference in a population, the genes for extra bmp2 production would concentrate in the population, and selection for ever-longer and more adept ‘wings’ would go smoothly from there.

    It might have been a single mutation that brought about the first extended phalanges, but positive selection from then on would provide constant improvement. I can’t bring myself to say ‘steady’, as it was likely not steady at all!


  6. says

    I suspect this is a major conceptual block for the Intelligent Designers: They think of designed items (such as a watch or a dolphin) as having the blueprint drawn up, and via a watch factory or “poof” made tangible. Which is to say, they don’t appreciate how embryological development leads to the individual they see as “designed” and how perturbations in the developmental factors can produce something that differs from the Blueprint. You’ve explained very clearly how this can be, but alas, the Creationists will likely only respond by ranting about Haeckel’s pictures.

  7. Steviepinhead says

    I can’t think of a science post of PZ’s that I haven’t really appreciated.

    And this is yet another of those, “Wow, ain’t Evo-Devo cool” moments!

  8. says

    Are there similar evo-devo compaisons with gliding mammals? It just seems to me that long fingers would be more beneficial to an animal with existing gliding capabilities than to an otherwise terrestrial creature.

  9. Paul W. says

    Bmp2, but not Bmp4 or Bmp7, is differentially expressed in the bat wing digits is suggestive of a cis-regulatory change that affects the level, but not the temporal or spatial regulation, of Bmp2 expression.

    I’m probably misunderstanding “spatial” or “regulation” or something, but it sounds to me like there must be a difference in the spatial regulation of Bmp2—the level in the forelimb region is elevated, but the level in the hindlimb region is not.

  10. says

    Looks like I have something to keep me from reading about HDACs tomorrow. Darn. :)

    Geek moment. My HS senior yearbook has the laminated and preserved bones from the wing of a bat I disected in AP biology. I was amazed by the thin bones of the skull and the tremendous brain to body ratio. Our disection rat was preggers and we disected the fetal rats as well.

  11. G. Tingey says

    Thee two pictures remind me of diagrams in a truly wonderful engineering primer/textbook.

    Published in England as:
    Structures – or why things don’t fall down.
    (US title may be different)
    by Prof. J. E. Gordon – who, if he’s still alive, will be 94.

    If you haven’t read it, go out and buy a copy …..

  12. Ian says

    You have a picture of a baseball player at the head of your page, you post an excellent article about bats right under it – how did you resist not titling it “Case Study at the Bat”?!

  13. chuko says

    That’s so amazing! Maybe I should quit this programming thing and be a biologist…

  14. says

    Maybe I should quit this programming thing and be a biologist…

    Just make sure you do it for the right reasons:

    • Money
    • Fame
    • Talking to dolphins
    • Bat (and squid, and zebrafish) embryological development
  15. speedwell says

    On the advice of my vet, who is also a research biologist, I had my kitten spayed very early (it’s shown they heal much faster fom the surgery and are not plagued by hormonal issues). As a result, she did not ever go through puberty. An interesting side effect of this is that her bone growth (which usually stops at puberty) did not stop until after the usual time, so she became a tall, long, wispy cat. (Her activity level, mobility, and bone strength are not affected in the slightest.)

    I wonder if any analogous expression mechanism is at work in her case. I would guess it is.

  16. paul says

    While it is easy to see that every intermediate step in the development of a light-sensitive organ would be an advantage to its posessor, and thus be selected for, (as Dawkins), it is not so clear how longer fingers would have selective advantage to a mouse-like creature, while still too short for a wing.
    Don’t get me wrong – I am a firm adherent of evolution by natural selection. I suspect that there were other intermediate advantageous functions between paws and wings which we have not yet discovered.

  17. Paul W. says


    My own guess would be that the easiest evolutionary pathway would be webbedness first, big hands second. A gliding mammal could more easily take advantage of big hands, and it looks to me like bats have apoptosis suppressed fairly generally—not just between the fingers or toes, but between the limbs and body, and the hind limbs and the tail. That would work for limited gliding, without the big hands.

    (Even a little bit of gliding ability can be advantageous to some animals, e.g., ants with no webs at all that use it to get back to the same tree in case they fall off, because if they fall to the ground they’re likely never to get back. And a really crappy glide ratio is okay for some others, like flying snakes that just flatten their bodies a bit, and flying squirrels—as long as they can get from one tree to the next, better than just jumping, they do okay.)

    Based on flight physics alone, I’d think that tiny mammals would find it easier to evolve flight. Mass is proportional to volume, lift is proportional to area. At a really small scale, you don’t need much area, or elongated limbs.

    Maybe the first gliding mammals were tiny little things which you wouldn’t easily recognize as gliders at all, unless the fossils were very well-preserved and showing the delicate webs. They wouldn’t need very long arms or fingers to make it work, but could easily take advantage of big hands once they got them.

    That makes me wonder if we do actually have fossils of gliding proto-bats, but don’t recognize them as such.

  18. Ktesibios says

    To follow up on G. Tingey’s comment:

    J. E. Gordon, Structures or Why things don’t fall down
    ISBN 0-306-82183-5
    Da Capo Press, second edition 2003

    Still in print- I found my copy at Borders.

    A really cool book. Who’da thought that the wings of bats and pterosaurs share a structural principle with the sails of a Chinese junk?

  19. Linda says

    Wow, how cool! This is such an elegant concept.

    It’s things like this that tempt me to drop out of chem and come to the dark side of biology..

  20. says

    I have to admit to only relatively recently becoming aware of evo-devo and the implications of this field. PZ, I think this item explains very powerfully the importance of evo-devo.

    It just dawned on me that this may mean that the old ‘novel features evolved once’ parsimonious approach to phylogenetics may not be the best way to look at things. For example (and this is probably a poor one) could ‘snakes’ have evolved twice – once for the boids and again for the colubrids – if the legless state was due to the production or limitation of a certain crucial protein like a bmp? It’s been a while since I studied the mechanics of evolution after getting sidetracked into ecology – so this might be obvious to everyone else already – but this would seem to be to have some pretty big implications for cladistics.

  21. PaulC says


    This isn’t that surprising a feature of development, though, unless you have the mistaken idea that the genome encodes a blueprint of morphology. It doesn’t; what it contains is a description of interacting agents that work together in a process to produce a complex result.

    I’m probably too late to this topic, but the observation above is one of the main reasons I find “intelligent design” absurd to begin with.

    When we claim to find a design following in the footsteps of Paley, we’re invariably saying that the phenotype was designed. But what good is that in the context of living things?

    Say I’m a non-omniscient designer (since the design analogy is invariable to human designers) and I’ve worked out what my creature should look like–the colors of its feathers, the cutting surface of its teeth, the length of its snorkel appendage, etc. Wonderful, now like any honest engineer, I have to implement my creature.

    The only way to implement it is to put the right DNA sequence into a eukaryotic cell and let that cell develop. But the function from DNA to phenotype is what cryptographers call a “trap door function.” You can get phenotype from DNA by letting the cell develop, but there is no known method of generating corresponding DNA given a phenotype. There is no natural process that can do this, and there certainly does not seem to be a feasible computational process.

    So how might I go about coming up with some DNA to at least approximate my feathered snorklebeast? Probably I would have to do some kind of search of the genotype space by guessing at the sequence in some fashion and testing it against my phenotype. I would have little to guide this search, because the connection between DNA and phenotype appears to be a trap-door function. But a long enough search could conceivably give me the right DNA eventually.

    OK, fine, but what good did it even do me to design the phenotype? Wouldn’t it make more sense to do the same search starting from DNA and try to optimize the survivability in its environment rather than the degree to which it matches my preferred phenotype?

    So in short, the entire design step is a waste of time for non-ominscient designers if the end result is to be an embryo with the appropriate DNA encoding. You might as well just have started searching for DNA to optimize fitness–which is more or less what evolution does.