Evolution of the cichlid mandible


When we look at the face of another person, we can recognize specific features that have familial resemblances. In my family, for instance, I can recognize a “Myers nose” that my grandmother and my father and some of my siblings and kids have, and it’s different than my wife’s or my mother’s nose. These are subtle differences in shape—a bit of a curve, a knob, a seam—and their inheritance suggests that these differences are specified somehow in the DNA. If you think about it, though…how can whether the profile of a nose is straight or curved be encoded in a linear stretch of nucleotides? The complicated answer is that it isn’t—morphology is a consequence of epigenetic interactions during development—but we know that the alleles present in the genome do contribute in some significant way to three-dimensional form. How?

We don’t know all the details. This is one of those huge research problems that has gaping holes, full of promise and interest, where we don’t understand exactly how all the pieces fit together. However, here’s an important point that is relevant to other, larger issues in evolution: even where we lack full information about mechanisms, we can roughly perceive the shape of the answer, and that helps us rule out many alternative explanations and guides us in the general direction of a more complete understanding.

People’s noses are a difficult subject for research; we don’t get to define human crosses, people tend to be a little snippy about telling them who to breed with and taking their genes apart, and humans are awfully slow to breed. Fish are better experimental animals, much more pliable and faster and more prolific in their breeding. Some fish, such as the African cichlids, also have highly diverse populations and species with easily recognized and often quite dramatic morphological differences—and we can explore how those differences are generated by genetic and molecular differences in development. In particular, we can start to figure out how fish jaws are shaped by developmental processes.

African cichlids are a popular aquarium fish, and they’re also extremely popular in evolutionary circles. The reason is that they come from a chain of lakes with known and relatively recent times of origin, and the original stock has rapidly diversified into over a thousand new species. Evolution has shaped this animals in various ways that we can measure.

A proposed phylogenetic history of Lake Malawi’s rock-dwelling cichlids based on several molecular phylogenies of Lake Malawi cichlids. Lake Malawi is presumed to have been invaded by a riverine generalist closely allied with Lake Tangyanika’s haplochromine tribe approximately 700 000 years ago. This common ancestor subsequently diverged during the primary radiation into the sand-dwelling and rock-dwelling lineages. The rock-dwelling lineage diverged during the secondary radiation into the 10-12 currently recognized mbuna genera. These genera are distinguished primarily on the basis of trophic morphology suggesting the importance of trophic competition during this period of the radiation. The spectacular species richness of the mbuna principally arose during the tertiary radiation. During this period, as many as 25 species per genus diverged presumably in response to sexual selection via female choice for male secondary sexual characteristics such as colour pattern.


One of the key parameters of cichlid success is the morphology of the jaw. Different species have acquired different feeding strategies with different jaw shapes: short stout jaws for species that gnaw algae off of rocks, for instance, and long slender jaws for species that suck up free floating plankton. A short-jawed gnawer would be Labeotropheus fuelleborni (LF from now on), and a more general feeder would be Metriaclima zebra (MZ). The jaw itself can be characterized as a simple mechanical lever. Muscles pull on a structure called the coronoid process (marked by the purple line to the right, called the closing-in lever) to close the jaw, and on the retroarticular process (the green line, or opening in-lever) to open it. These all act to move the axis of the jaw itself, the out-lever.


Changing the length of these levers changes the action of the jaw. Long closing-in levers, or elongating the coronoid process, makes jaw closure slower but more forceful; shortening it reduces the leverage and makes closure weaker, but faster. Lengthening the jaw itself makes it weaker and faster, while shortening it makes it slower and more forceful. We can see these differences in the fish; MZ has a longer jaw and shorter coronoid process, while LF has a short jaw and long coronoid process. The problem is to determine what genes are responsible for these differences.

There is a linkage analysis technique to do this which identifies something called quantitative trait loci, or QTL. This procedure identifies chromosome regions associated with particular traits. The way it works is to first find a range of recognizable genetic markers scattered throughout the genome; these will typically be somewhat obscure but easily measured elements like restriction fragment length polymorphisms, tandem repeats, and single-nucleotide polymorphisms. There is no assumption that these features have anything at all to do with the morphological trait of interest—they are nothing but convenient markers for a region in the genome.

The next step after identifying a suite of markers is to cross an individual carrying a trait of interest and the markers to another individual lacking both; thinking back to your old Mendelian genetics, the result is a collection of progeny that are heterozygous, carrying the markers and traits on one set of chromosomes, carrying different alleles and markers on another. When those are crossed back with the markerless parent or each other, the grandchildren (or F2s, to use a little genetics jargon) will express a mix of traits, some showing the paternal phenotype, others the maternal.

The F2s are analyzed for their traits and the presence of the markers. If the presence of a marker is correlated with a particular trait, then it is likely that that marker is somewhere near a gene or genes that affects the trait. By analyzing the relationship of all of the markers that are expressed in organisms that exhibit a particular trait, regions of the genome that are correlated with positive or negative effects on the trait can be identified…such as the regions in the diagram below, which map out the QTL, or rough locations of genes, that are associated with each of the levers of the jaw.

Genetic basis of jaw opening and closing in cichlids. (a) Lever systems of the lower jaw. The out-lever is shown in blue, the closing in-lever is shown in purple, and the opening in-lever is shown in green. (b) Cichlids linkage groups that show significant associations with functional morphology of the jaw. Bars indicate regions exceeding the 95% genome-wide significance threshold for the corresponding QTL. Colors represent QTL affecting different traits. (ce) Genetic correlation of traits in the F2 population. (c) The out-lever is negatively correlated with the closing in-lever. (d) The out-lever is also negatively correlated with the opening in-lever, although the correlation coefficient is significantly less than that for the closing in-lever (q = 4.21; P < 0.01; ref. 40). (e) Closing and opening in-levers are not correlated in the F2 population.
Venn diagram depicting regional differences in the level of integration of the cichlid mandible. The out-lever and closing in-lever share 2/3 QTL and exhibit high levels of integration. The out-lever and opening in-lever only share 1/5 QTL and exhibit low levels of integration. The closing and opening in-levers have no QTL in common (0/8) and are genetically decoupled.

Another property of the data is that it can be examined for correlations between the QTL. The scatterplots in the lower half of the above diagram are showing the relationships between pairs of parameters. For instance, the out-lever is negatively correlated with the closing-in lever, which means that genes that make the jaw shorter also make the coronoid process longer, and vice versa.

The diagram to the right illustrates the push and pull of genes on each other. Two of the three loci that affect the jaw length also affect the coronoid process, and the formation of these two structures are closely integrated. The loci that influence the out-lever have little effect on the opening in-lever (1 out of 5 loci), and the closing and opening in-levels are completely independent of each other. What we’re seeing here is a quantitative measure of rules of correlated growth in the jaw—that some features, when changed, will automatically induce related changes in other structures. These relationships are not arbitrarily imposed by selection, but are consequences of the necessary developmental interactions between genes.

Development, of course, is where all the interesting action is. The authors have also examined the early development of the jaw in the two species, LF and MZ, and found that the differences in shape are apparent as early as 7 days after egg fertilization. Look at C, below, which shows the alcian blue stained cartilages of the larval fish, and in particular, the Meckel’s cartilage (jaw) lower center in each panel. There are differences emerging at very early ages.

Ontogeny of biomechanical lever systems in cichlids. (a) Adult LF and MZ (>1 year) show significant differences in the mechanical advantage of closing and opening (n = 29). (b and c) This difference is evident in juveniles at 21 dpf (n = 6) (b) and in larvae as early as 7 dpf (n = 10) (c).

One gene that is known to have craniofacial effects in other organisms, including cichlids (which in fact exhibit a pattern of rapid evolution within the lineage) is bmp4. The bmp4 gene also turns up in a QTL that affects the out-lever and the closing in-lever—if you look closely at chromosome 19 in the linkage map up above, you’ll find it there. What’s also interesting is that if you look at the expression pattern of bmp4 in MZ (the slender-jawed fish) and LF (the stout-jawed fish), you also see a marked difference in expression: more bmp4 is associated with sturdier jaws.

LF and MZ embryos exhibit different levels of bmp4 expression in the mandibular arch. (a) At the high-pec stage, MZ pharyngula express bmp4 at the distal tip of the first arch (red arrowhead). In Left, the black arrows indicate the first and second pharyngeal arches. In Right, the black arrow indicates the mandibular arch. (b) Similarly staged LF embryos express bmp4 throughout the mandibular mesenchyme (red arrowhead). In Left, the black arrow indicates the second arch. In Right, the black arrow indicates the mandibular arch. Differences in bmp4 expression were not size-dependent. At no stage did MZ embryos show a level of mandibular bmp4 expression comparable to what was observed in LF. (Scale bars, 20 µm.)
Bmp4 expression and jaw morphogenesis in the zebrafish, D. rerio.(a) Expression of zebrafish bmp4 at 36 h postfertilization. Throughout much of the pharyngula period of embryonic development, bmp4 was expressed in the developing heart (h), ear (otic vesicle, ov), and pectoral fin (pf). However, very little bmp4 was detected in the mandibular arch (black arrowhead). (b) In 6-dpf wild-type embryos, Meckel’s cartilage (Mk) possessed a distinct retroarticular process (rp) but lacked a discernable coronoid process (cp).

That’s cool. The authors have identified banks of genes that interact with one another and are involved in patterning the jaw, and furthermore have found one particularly promising candidate gene, bmp4, that may be directly involved. Now to take it a step further and look at another teleost, the zebrafish, Danio rerio.

That picture to the right is awfully familiar—that’s my experimental animal! You can find bmp4 expressed early, at 36 hours post fertilization (zebrafish develop very rapidly, and that guy is actually only a day and a half old), and a few days later the Meckel’s cartilage can be stained with alcian blue. The jaw is a simple bar of cartilage cells, with two small bumps, the coronoid and retroarticular processes.

Why work with zebrafish? Because there are some extremely useful laboratory techniques that can be carried out in this model organism. In particular, you can take the one-celled embryo and inject it with vectors carrying the zebrafish bmp4 gene, effectively causing the embryo to overexpress bmp4—you can change the dosage of the gene, and then look at what effect the gene has on morphology.

Distinct morphological transformation of Meckel’s cartilage (Mk) by bmp4 overexpression. Landmark-based morphometric analysis revealed concomitant growth of the coronoid and retroarticular processes in 6-dpf zebrafish larvae when embryos were injected with 100 or 150 ng/µl translation-competent bmp4 mRNA at the one- to two-cell stage, compared with gfp-injected negative control embryos.

Excess bmp4 in the zebrafish causes exactly the set of correlated changes in jaw shape predicted from the observations in the cichlid: the jaw gets shorter and thicker, and the coronoid process becomes more prominent. I particularly like the grids drawn below them. Those are standard grids to illustrate the morphometric transformation of the structure, as were used by Albrecht Dürer and D’Arcy Thompson—a lovely bridge between historical embryology and modern molecular/genetic developmental biology.

This is a wonderful paper that illustrates what we can expect for the future. We’re starting to see research that is tying variations in the genetics of populations and species to deeper molecular mechanisms of shape and form. We’re beginning to tease apart the developmental rules that define morphology, and furthermore, see how those structure evolutionary trajectories and are modified by evolution.

Albertson RC, Streelman JT, Kocher TD, Yelick PC (2005) Integration and evolution of the cichlid mandible: The molecular basis of alternate feeding strategies. Proc.Nat.Acad.Sci. USA 102(45):16287-16292.

Danley PD, Kocher TD (2001) Speciation in rapidly diverging systems: lessons from Lake Malawi. Molecular Ecology 10(5):1075-1086.


  1. Scrofulum says

    Amazing stuff. Gives an insight into the mechanisms behind much more rapid mechanisms of adaptation than a layperson like myself had realised.


  2. Bad Albert says

    This is one of those huge research problems that has gaping holes

    Ah hah! So those creationists were right. There are some problems with evolution theory. :)

  3. Paul Lurquin says

    A somewhat similar type of regulation was discovered earlier in Darwin’s finches (long beak vs. short beak). Yes, those finches! Overexpression of the CAM gene is at work there.

  4. noncarborundum says

    Bah. I found a post over at UD the other day that lays out in just this much exquisite detail all the evidence that’s been amassed to prove that each fish species has jaws carefully designed by an unnamed intelligence in a such a way that their shape can’t be influenced by simple physical processes such as changes in gene expression. Wait just a sec and I’ll post a link.


    Oh. Never mind.

  5. Odonata says

    Wow. That is really fascinating. Thanks for explaining the cichlid mandible PZ. I never know what I’m going to learn on Pharyngula.

  6. don kane says

    One thing to think about is that many of the traits that shape the face (and nose for that matter) are semi or co dominant and not recessive. Eye color would be an example of a recessive trait, and it seems that most genetic alleles act more like recessives than dominants. Anyway, saw a bit ago some idea that facial differences in people might have to do with helping to connect men to their biological childern in groups where matings are promiscuous. Guess that’s so you know who to teach clubbing to help insure the suvivial of your genome. Maybe the part of the evolutionary change in people was the addition of this varibility. Afterwards, social selection acting on co dominant traits would develop differences in subpopulations at a rather rapid pace.

  7. procyon says

    In primitive hominids facial recognition between father and children as a means of kin recognition wouldn’t work cause they didn’t have mirrors. No one knew what they really looked like. So, maybe if the children looked like their grandfather so the father could see the resemblance to his own father it might work. Sounds like an area that needs to be researched.

  8. Peter Ashby says

    Congrats PZ, you managed to get that across very well without even once mentioning convergence-extension. Well done.

  9. R. Paul Wiegand says

    This is a very interesting post. Thank you for a clear and fascinating explanation.

    I’m not a biologist, so my question will probably be silly, but hopefully it makes enough sense that it can be properly untangled.

    Can there not be inheritable traits that are transmitted indirectly by way of development? What I mean is that a gene might pass along the female side of a species that affects say the morphology of the birth canal, the presence of certain chemicals during development that affect the process, etc., that lead to traits in the offspring.

    Rephrased: Because you see a common physical trait in a family line, does that always imply a direct DNA encoding for that trait? Or can there be a one-off level of indirection because of development? If “yes”, what are some examples? And what are some of the tests that help distinguish those from more explicitly encoded traits?

    Just curious.

    (My vote, if it matters, is to have more posts like this one and the moray one, please!)

  10. Peter Ashby says

    Paul the problem here is that for environment to have an effect there must be a responsiveness in the embryo/foetus that is set up by its genetics. They cannot be so clearly disentangled wrt cause and effect. Also since we are not yet very good at growing placental mammals ex utero we have no means of examining such things.

    At the moment much attention is placed instead on what the offspring do to the mother with cells from male babies found in the blood of their mothers. Cells from female babies will be there too, but they are not so easily spotted. This very probably underlies the statistic that women are much more likely than men to get Lupus, which is an autoimmune disease. This leakage across the placenta seems one way, or is tolerated only in one direction as to my knowledge no maternal cells have been found in us males.

    Now that I think of it, there have been experiments where fertilised eggs from one strain of mice are transplanted into the uteri of other strains of mice. However such experiments are interested in metabolic set points and again to my knowledge nobody has reported morphological changes. We cannot do this in humans as we cannot reliably induce monozygotic twins in tissue culture. It works with the mice as they are so inbred all are effectively twins.

  11. SeanH says

    In primitive hominids facial recognition between father and children as a means of kin recognition wouldn’t work cause they didn’t have mirrors. No one knew what they really looked like.

    Primitive hominids probably got a good look at their own reflections every time they had a drink of water. I can’t remember details at all, but I seem to remember an article somewhere a few years ago saying adolescent chimps examine their own reflections in water much more frequently than adults and that may be evidence that teen image conciousness has basically always been wired into us.

  12. R. Paul Wiegand says

    the problem here is that for environment to have an effect there must be a responsiveness in the embryo/foetus that is set up by its genetics. They cannot be so clearly disentangled wrt cause and effect

    I hadn’t considered that. Interesting…

    I suppose at some point one is waxing semantic about what is “direct” and what is “indirect”? In other words, if an explicitly encoded trait in the mother dictates a certain developmental context, and an explicitly encoded trait in the offspring predisposes it respond to that context in a particular way, then the distinction between a directly encoded morphological trait and that may be merely technical. Morphology is (no doubt) governed by fairly complicated dynamics (genetically and developmentally) in any event.


  13. caynazzo says

    Great z-fish plug. They’re also my research model. I study hematapoetic tumorigenesis, and know little about fish development before day seven. But now I know a little more.

  14. says

    Presumably fathers could see that their children resembled each other and also the father’s siblings, if any, especially younger ones.

    OTOH, you’ll remember that in fairy stories a wicked uncle always seemed to be in charge. That seems to come from an assumption that a mother’s brother was her children’s closest male relative.

  15. katie says

    Um…that graph (diagram D) up there doesn’t look like much of a correlation to me. It looks like one outlier is pushing most of the correlation…

  16. Galbinus_Caeli says

    I think my family nose is a result of a tendency to fade into a right hook.

    Thanks for the accessible science PZ. I also just read your cervical vertebrae article in the August Seed. Great stuff! But how does it explain all these subluxations my chiropractor keeps finding at $50 each.