The complex interplay between tRNA genes and transposable elements in bird genomes

New tRNA genes can emerge through multiplication by transposable elements.

Transfer RNAs (tRNAs) are an underappreciated group of molecules. Apart from transporting the amino acids to the protein-synthesizing ribosomes, they fulfil are range of other important functions in the cell (nicely summarized in this review). During a six-month postdoc at the Karolinska Institute (Stockholm, Sweden), I entered the wonderful world of tRNA genes under the expert guidance of Claudia Kutter. The tRNA-team further consisted of PhD-student Keyi Geng and transposon-guru Alexander Suh. Taking advantage of the increasing number of available bird genomes, we explored the evolutionary dynamics of tRNA genes in our feathered friends. And we were in for some surprises! But before we dive into the results, let’s refresh some basic molecular biology.


The Genetic Code

DNA contains the instructions to make proteins. But the DNA alphabet consists of four letters (A, T, G and C), while the language of the proteins has twenty letters (i.e. amino acids). How does the cell translate DNA language into protein language? The Russian physicist George Gamow looked at the problem through a mathematical lens. How can you combine four letters so that each of the twenty amino acids has a unique DNA code? A code based on two DNA letters does not work, because it only yields sixteen combinations. Not enough for the twenty amino acids. But a code with three DNA letters is possible. That code results in 64 different letter combinations, more than enough for twenty amino acids.

However, this solution led to another problem. Which combination of three DNA letters codes for which amino acid? In 1954 the American biologist James Watson founded the RNA Tie Club with Gamow to crack this genetic code. This club had twenty members (one for each amino acid) and four honorary members. Each member was given a woolen tie with the double helix embroidered on it. One of its members, South African biologist Sydney Brenner, suggested the term “codon” to refer to a combination of three DNA letters. Sixty-four codons and twenty amino acids. Who could solve this puzzle?

The first codons were deciphered with simple experiments. Scientists made long strands containing one DNA letter in the lab. These strands were then translated by a cell into a chain of amino acids. When one strand of As was used, a long chain of the amino acid lysine was formed. And if a strand of C’s was translated, the chain consisted only of the amino acid proline. The conclusion was crystal clear: AAA codes for lysine and CCC codes for proline. With further experiments, researchers deciphered the entire genetic code. The code even turned out to contain start and stop signs. The codon ATG – which codes for methionine – marks the starting point of a protein, while the codons TAA, TGA and TAG signal the end.

The genetic code. Notice that the DNA-letter T (thymine) has been replaced with the RNA-letter U (uracil).


Redundancy and Wobbling

With sixty-four codons and twenty amino acids, it is obvious that multiple codons code for the same amino acid. For example, AAT and AAC both refer to asparagine. This phenomenon, known as the redundancy of the genetic code, provides the cell with some protection against mutations. A mutation in a codon can lead to a different amino acid or stop character in the protein chain. Suppose the codon AAA (which codes for lysine) mutates into TAA (a stop codon). This mutation puts a stop codon in the wrong place in the protein and the production of that protein is stopped prematurely with possible negative consequences for the cell. A detailed look at the genetic code shows that many mutations, however, have no harmful effect. Take the amino acid alanine, which is encoded by the codons GCT, GCC, GCA and GCG. A mutation at the third position of these codons, for example from GCT to GCC, does not lead to a change in amino acid, as both GCT and GCC refer to alanine. After this mutation, the cell still produces the same protein.

Each organism has a certain number of tRNAs per codon available. For example, humans have 44 tRNAs for lysine: 24 for the codon AAG and 20 for the codon AAA. Certain tRNAs are also missing in the human genome. You will not encounter any tRNAs for GGT (glycine), CGC (arginine) or CAT (histidine). Fortunately, there are other tRNAs that provide these amino acids. The absence of certain isoacceptors is explained by wobble base pairing, in which the third anticodon position can deviate from the standard Watson-Crick base pairing, allowing for the translation of multiple synonymous codons by a single tRNA. In addition, modifications at certain positions in the anticodon loop can improve translational efficiency. For example, in the G34 anticodon sparing strategy, an enzyme converts adenine-34 to inosine-34 in specific isoacceptors. This conversion enables position 34 to wobble with adenine, cytosine and uridine. One tRNA molecule can thus be used for multiple codons in the mRNA.

The third position in the anticodon can wobble, allowing it to bind with both C and U on the mRNA molecule. Hence, multiple codons can be translated by a single tRNA.


Genome Size Reduction

Now that we have refreshed our knowledge about the genetic code, we can finally dive in to the results of the paper, which appeared in Genome Biology and Evolution. Comparing the total number of tRNA genes between avian genomes and other vertebrates revealed a striking pattern. On average, birds have about 169 tRNA genes which is significantly less than reptiles (466), amphibians (1229), mammals (579) and fish (813). This reduction could be a by-product of an evolutionary trend towards smaller genomes in birds through deletions of non-coding DNA. Interestingly, the tRNA gene repertoire in birds still contains all necessary tRNA genes that code for all twenty amino acids. Moreover, when we investigated the expression of tRNA genes in chicken (Gallus gallus) and zebra finch (Taeniopygia guttata), we found that all twenty amino acids were represented by at least one tRNA gene. Hence, the reduction in tRNA gene number and complexity in birds occurred within the functional constraints on efficient protein translation mechanisms.

An overview of the total number of tRNA genes in the genomes of birds (brown), reptiles (green), mammals (orange), amphibians (yellow), fish (blue) and yeast (black). From: Ottenburghs et al. (2021) Genome Biology and Evolution.


Transposable Elements

At the very start of the project, we noticed that some bird species had a overrepresentation of certain tRNA genes when we did not apply a quality filter. Why would the Dalmatian pelican (Pelecanus crispus) need almost 600 tRNA genes for isoleucine? And what does the bar-tailed trogon (Apaloderma vittatum) do with 2750 valine tRNA genes? Close inspection of these tRNA genes uncovered the presence of transposable elements: SINEs (short interspersed elements) to be precise. These selfish and highly active genetic elements incorporate themselves into genomes by a copy-paste mechanism, continuously giving rise to new genomic loci. If a tRNA gene is associated with such a transposable element, it can quickly increase in frequency. Our detailed analyses of these SINEs pointed to several known elements, such as TguSINE1 in Eupasseres (i.e. all passerines except the New Zealand wrens) and ManaSINE1 in manakins. But we also discovered some new SINEs, including PeleSINE1 in the Dalmatian pelican and ApalSINE1 in bar-tailed trogon.

And now for my personal favorite fact in our paper: the evolutionary dynamics of transposable elements in the golden-collared manakin (Manacus vitellinus) genome. This small songbird houses two SINEs that have been active at different times during its evolution: TguSINE1 was jumping around about 30 million years ago, while the activity of ManaSINE1 is more recent (about 5 million years ago). When these transposable elements become inactivated by the cell, they get “stuck” in a genomic location and start accumulating mutations. Because TguSINE1 was active millions of years before ManaSINE1, we expected that copies of this transposable element have accumulated more mutations.

We could test this prediction by using the Cove score which is calculated by the tRNAscan-SE progam. This score is based on the ability of a given sequence to form tRNA stem-loop structures and the presence of particular promoter and terminator sequences. Active tRNA genes will have high Cove scores, whereas inactive tRNA genes might be decaying into pseudogenes, leading to lower Cove scores. So, our prediction was straightforward: TguSINEs should have lower Cove scores compared to ManaSINEs. And that is exactly what we found (see figure below)! Isn’t it great when you find support for a hypothesis?

The evolution of transposable elements in the golden-collared manakin genome. Based on past activity patterns, we expected that TguSINEs would have lower quality scores compared to ManaSINEs. And that is exactly what we see. From: Ottenburghs et al. (2021) Genome Biology and Evolution.



Based on the findings of our work (and there is much more that I did not cover in this blog post), we formulated a model for the coevolution of tRNA genes and transposable elements. Some parts of this model are strongly supported by our results, while others still need to be investigated further.

The model consists of three phases. First, a TE recruits a copy of a tRNA gene for its own mobilization and increases its copy number in the genome. Second, the TE is silenced by epigenetic control mechanisms. Third, some TE-associated tRNA genes decay into pseudogenes, while others remain transcriptionally active and become coopted for their original tRNA function.

The last phase in the model – transposable elements that become active tRNA genes – is difficult to prove because the activity of these genes can be due to functioning as an actual tRNA gene or simply as an active TE. In zebra finch, we noticed that the correlation between codon usage in proteins and the available tRNA genes improved if we included transposable elements. This suggests that the activity of transposable elements can help shape the tRNA gene repertoire of an individual and potentially improve the efficiency of protein translation.

The correlation between codon usage and tRNA genes in the zebra finch genome improved if we included transposable elements (notice how rho increases and the p-value decreases). From: Ottenburghs et al. (2021) Genome Biology and Evolution.


An Extra Acknowledgement

I would like to end this long blog post with a personal note. The story behind this paper already starts in January 2016 when I attended the Plant and Animal Genomics (PAG) conference in San Diego. I was invited to give a talk at the “Avian Genomics” workshop (organized by Robert Kraus) where I met another invited speaker: Alexander Suh. Fast-forward to the spring of 2017: I was getting ready to move to Uppsala to start a postdoc, studying the genomics of hybridization in geese. I had already signed the contract and secured an apartment when my postdoc supervisor informed me that the sequencing of the samples was delayed. He wanted me to start a few months later so that I could work on the data during my entire postdoc. I tried to convince him that I could work on other projects while I waited for the goose data, but he would not give in. To fill the unexpected gap in my schedule, I contacted several people that might need help with a small project. Alex replied quickly and told me about a potential collaboration with Claudia Kutter in Stockholm. We discussed the project over Skype and a few weeks later I was on a plane to Sweden. I thoroughly enjoyed my six months in Claudia’s group and I am happy that we managed to turn my work into a nice paper. But more importantly, I made countless new friends during my Swedish adventures.

All invited speakers at the Avian Genomics workshop in San Diego where I (third from the left) met Alexander Suh (second from the left). The workshop was organized by Robert Kraus (right).



Ottenburghs, J., Geng, K., Suh, A. & Kutter C. (2021) Genome size reduction and transposon activity impact tRNA gene diversity while ensuring translational stability in birds. Genome Biology and Evolution. Early Online

Featured image © Claudia Kutter.

Misconceptions and regressions: The evolution of bird brains

Recent study traces the evolution of brain size along the avian tree of life.

Despite the growing number of excellent popular science writers, many misconceptions about evolution are still surviving among the general public. One especially tenacious one is “The March of Progress“, mostly depicted a series of walking primates from an ape-like creature over primitive hominids to modern Homo sapiens. This iconic picture misrepresents evolution as a linear process from primitive creatures to complex organisms. This thinking is often extended to the last universal common ancestor and summarized in the phrase “From monad to man.” In each case, modern humans are considered the ultimate goal of the evolutionary process. Or the pinnacle of evolution. Nothing could be further from the truth. Evolution is not a linear sequence, but an ever-branching tree. Humans are just one of the many twigs on this tree of life. We might be more advanced in terms of brain power compared to a bacterium. But some bacteria are clearly superior to us when it comes to processing methane or living in boiling lakes. Which organism can be considered the pinnacle of evolution is thus a non-sensical question.

Given that most evolutionary biologists are aware of these misconceptions, I was surprised to read the following in a recent Current Biology paper: “two groups—parrots and corvids—independently acquired relative brain sizes, neuronal densities, and sophisticated cognitive potential near the pinnacle of the vertebrate world. [my emphasis]” This sentence seems to suggest that there is an avian March of Progress towards larger brains with parrots and corvids at the finish line. This reasoning is obviously incorrect. Brain size is just one of numerous traits under selection in different species. In one environment a large brain and a small body might be beneficial, while in another environment a small brain and large body have the highest chances of survival and reproduction. Hence, the small-brained, large-bodied species could be considered the pinnacle of evolution in its habitat.

The original March of Progress illustration from Early Man (1965).



Now that I have clarified that misconception and ventilated some frustration, we can finally delve into the Current Biology study on avian brain size evolution. Apart from the hick-up at the end of the paper, this is a beautiful piece of work. The researchers amassed an impressive dataset of brain endocasts of 284 extant bird species, 22 extinct bird species, and 12 non-avian theropod dinosaurs, complemented with data for more than 1,900 extant species from another study.

Next, they investigated the relationship between brain volume and body mass for different bird groups. These relationships can be captured in simple mathematical formula that you probably remember from high school: y = ax + b. The coefficients a and b correspond to the slope of the regression line and the intercept (i.e. the point where the regression line hits the y-axis), respectively. By comparing these coefficient between related bird groups, the researchers were able to reconstruct the evolution of brain size across the avian phylogeny. An increase in the intercept indicates that one bird group increased in brain volume and body mass, but that the relative brain size remained the same. However, a steeper regression line (i.e. an increase in slope) points to an increase in relative brain size in one bird group.

The relationship between body mass (x-axis) and brain volume (y-axis) for different bird groups. Comparing these lines between these bird groups allowed the researchers to study the evolution of relative brain size. From: Ksepka et al. (2020) Current Biology.


Mass Extinction

Detailed analyses of these regressions revealed several evolutionary shifts in relative brain size. Interestingly, most significant changes occurred after the mass extinction at the end the Cretaceous when the non-avian dinosaurs disappeared. This catastrophic event might have set the stage for an adaptive radiation in brain size, a scenario that fits the “cognitive buffer hypothesis”. This hypothesis suggests that large brains provide a buffer against frequent or unexpected environmental changes via enhanced capacity for flexible behavioral responses.

After this adaptive radiation, several bird groups showed independent changes in relative brain size. Birds of prey in the orders Accipitriformes, Strigiformes, Falconiformes, and Cariamiformes, for instance, experienced a small increase in relative brain size. A carnivorous lifestyle probably led to the evolution of a larger body, but was not accompanied by a proportional increase in brain size. In the woodpeckers (order Piciformes), we see the opposite pattern where a decrease in body size was not followed by a decrease in brain volume, resulting in a significant increase in relative brain size. This pattern is also apparent in hummingbirds and swifts (order Apodiformes)m and in sandpipers and buttonquails (order Charadriiformes).

Changes in relative brain size across the avian phylogeny. From: Ksepka et al. (2020) Current Biology.


Different Evolutionary Paths

And that brings us to the “pinnacle” of avian brain size: the parrots and the corvids. The method outlined above allowed the researchers to pinpoint the exact mechanisms behind the high relative brain sizes of these birds. It turns out that they each took a different path to the top: parrots primarily reduced their body size, whereas corvids increased body and brain size simultaneously. A nice example of convergent evolution.

It is no surprise that parrots and corvids have large relative brain sizes. These species are known for their clever tricks, such as mimicking sounds and using tools. But brain size is just one aspect of intelligent behavior. Other studies have found that the structure of the brain and the connections between the neurons are also important. For example, parrots have an additional vocal learning pathway that is absent in songbirds. And both corvids and parrots have the highest cerebral neuronal densities in birds. There is more to avian life than a large brain.



Ksepka, D. T. et al. (2020). Tempo and Pattern of Avian Brain Size Evolution. Current Biology.

Featured image: New Caledonian Crow using a tool © National Geographic

Is prezygotic isolation more important than postzygotic isolation at the onset of speciation?

A simulation study casts doubt on this common statement.

“Ethological barriers to random mating constitute the largest and most important class of isolating mechanisms in animals.” If you like bold, in-your-face proclamations, you can always turn to the work of German biologist Ernst Mayr. He never shied away from strong statements, such as the one at the start of this blog post. In his book Animal Species and Evolution, he argued that speciation in animals usually begins with behavioral differences. Members from different populations prefer familiar faces and rarely interbreed. This behavior – assortative mating – generates a first genetic barrier between the populations. Later in the speciation process, selection against the occasional hybrid can arise in the form of sterility or unviability.

To put this argument in the modern jargon of speciation research: prezygotic isolation is more important than postzygotic isolation in the initial stages of speciation. If you are not familiar with these terms, I will briefly define them. Luckily, I can quote from the introduction of my PhD thesis for a concise explanation.

Prezygotic isolation mechanisms act before fertilization, whereas postzygotic isolation mechanisms act after fertilization and can be either intrinsic or extrinsic. Intrinsic postzygotic isolation mechanisms lead to sterility or unviability of the offspring, while extrinsic postzygotic isolation mechanisms encompass lower fitness of the offspring for ecological or behavioral reasons, not developmental defects.

The idea that prezygotic isolation is more important than postzygotic isolation makes intuitive sense. First, there is a reduction in gene flow between populations due to behavioral differences. Next, the populations follow different evolutionary paths and accumulate genetic incompatibilities. When the populations establish secondary contact, potential hybrids are sterile or unviable due to the many genetic mismatches. Makes sense, right? But if there is one thing I have learned during my short scientific career, it is to be careful with intuitive ideas. Evolutionary biology is often counterintuitive.


Simulating Clines

A recent study in the journal The American Naturalist put this idea to the test. Darren Irwin performed some simulations comparing different levels of assortative mating and postzygotic isolation (i.e. hybrid fitness). To assess the impact of these processes on speciation, he turned to cline theory. Loyal readers of this blog might be familiar with this mathematical framework, but to get everyone on the same page I will outline the main concepts of cline theory (based on a previous blog post).

Let’s say you have a white species and a black species that produce gray offspring in a hybrid zone. You observe birds along a transect and note down their plumage color. When you put the data in the graph, you will see a transition from white birds (when you were in the habitat of the white birds) through grayish birds (in the hybrid zone) to black birds (in the black bird habitat). The useful aspect of cline theory is that the shape of a cline can tell you something about the biology of the birds. For example, if the gray hybrids interbreed with their parental species, there will be a variety of backcrosses of different colors. Some more white and some more black, depending on the species they crossed with. This will result in a smooth transition from white through different (perhaps 50) shades of gray to black. In other words, a wide cline. However, if gray birds cannot find a mate, there will be mostly gray hybrids in the contact zone. This will result in a rapid transition from white to black plumage, a steep cline.

Darren Irwin applied this approach to his simulations. A wide transition with many backcrosses indicates that reproductive isolation is weak and there will be gene flow between the pure populations. A steep cline, however, points to strong reproductive isolation. What kind of cline does assortative mating produce?

An example of a cline plot. There are only white birds to the left of the hybrid zone and black ones on the right side. In the hybrid zone there are different shades of gray birds.


Genetic Bridge

Let’s start by comparing two simple scenarios: (1) modest selection against hybrids (10% reduction in hybrid fitness) and (2) modest assortative mating (female is 10 times more likely to pick a partner from her own species). Running the first scenario produces a steep cline because hybrids are less fit than their parents and rarely backcross. This result confirms the so-called tension zone model that was described by Nick Barton and Godfrey Hewitt in the 1980s. The narrow hybrid zone is maintained by a balance between pure individuals moving into the contact zone and selection against unfit hybrids.

The second model – with modest assortative mating – produces a much wider cline. Because there is no selection against hybrids, they can interbreed with each other and their parental species. These dynamics result in a variety of hybrids and backcrosses that form a genetic bridge between the initial parental populations. If gene flow is sufficiently high, the two populations might even merge into one large population.

The outcomes of simulating hybrid zone dynamics. A reduction in hybrid fitness (i.e. postzygotic isolation) results in a steep cline (in green), while modest assortative mating produces a wide cline (in purple). Adapted from: Irwin (2020) The American Naturalist


Search Costs

These initial findings indicate that modest assortative mating cannot jumpstart speciation. However, further analyses uncovered some conditions in which assortative mating leads to a narrow, steep cline, namely (1) when assortative mating is encoded by a single locus, (2) when assortative mating is very strong, and (3) when the cost of searching for a mate is high.

The first condition – assortative mating is encoded by a single locus – is not very realistic. Genomic studies have shown that reproductive isolation is mostly encoded by multiple genes (but see this case of single gene speciation in snails). The other two conditions – strong assortative mating and high mate searching cost – are actually a form of postzygotic isolation because they impact the mating chances of hybrids. Indeed, if assortative mating of the parental populations is strong, the hybrids won’t be able to find a mate.

The effect of search time is more complex. In the simulations, search time was modelled as the cost a female pays when she rejects a potential mate. A high search cost thus leads to females that quickly choose a mate. This is at the disadvantage of hybrids because they are rare in the population. In addition, hybrid females in the center of the hybrid zone will most likely mate with other hybrids. This lowers the likelihood of backcrossing with the parental populations and the consequent formation of a genetic bridge.

Results of simulations with and without mate search costs. Each graph shows the relationship between assortative mating (on the x-axis) and hybrid fitness (on the y-axis). The colors indicate the width of the cline (on the left) and the probability of a bimodal hybrid zone (i.e. two distinct populations, on the right). Including mate search costs leads to narrower clines (more yellow in the left figure) and a higher probability of distinct populations (more yellow in the right figure).


Postzygotic Isolation

These simulations suggest that assortative mating is less important in speciation than we think. But it’s just a modelling exercise, some of you might say. How realistic are these results? Well, several studies have tried to estimate the strength of assortative mating in wild populations. The highest estimate by Christophe Randler – based on 58 avian hybrid zone studies – points to  an assortative mating strength of 2.6. This is considerable lower than the modest strength of 10 used in the simulations. Of course, this strength might vary between hybrid zones, but it seems that assortative mating is rarely strong enough to keep populations separate.

This study thus indicates that assortative mating on its own cannot prevent populations from merging, some form of postzygotic isolation is needed (e.g., lower mating chances for hybrids). The intuitive idea that “prezygotic isolation is more important than postzygotic isolation” does not hold here. Why did this statement become so popular? The main reason is probably the focus on extreme forms of postzygotic isolation (i.e. hybrid sterility and unviability) that take long to evolve. We should not forget other postzygotic isolation mechanisms, such as sexual selection against hybrids. This insight opens up new research opportunities, nicely summarized at the end of the paper: “While assortative mating, unless perfect or very nearly so, is ineffective on its own in maintaining isolation of two species, the effects of sexual selection and sexual signals on postzygotic isolation are likely strong and worthy of renewed research focus.”



Irwin, D. E. (2020). Assortative mating in hybrid zones is remarkably ineffective in promoting speciation. The American Naturalist, 195(6), E150-E167.

Featured image © Zorba the Geek | CC-BY-SA-2.0 Wikimedia Commons


Non-coding DNA drives the evolution of avian beak morphology

“Living things do not inherit skulls, backbones, or cell layers from their ancestors—they inherit the processes to build them.”

– Neil Shubin (Some Assembly Required)

A few days ago, I finished the book “Some Assembly Required” by Neil Shubin. It is a great read about the big changes during evolution. Combining insights from paleontology, genomics and development biology, Shubin explains how major evolutionary transitions took place. One key idea that I took away from this book is illustrated by the quote above: if we want to reconstruct the evolution of complex traits, we need to understand how they are build. In other words, we have to unravel the genetic and developmental underpinnings of these traits.

One of the most important traits in avian evolution is the shape of the bill. Variations in bill morphology – from the sharp beaks of predators to the long bills of avocets – have allowed birds to exploit a myriad of ecological niches and diversify into the many species we see today. Surprisingly, little is known about the genetic and developmental processes underlying this variation. A recent study in the journal Genome Research took a macroevolutionary perspective on the avian beak.

The diversity of bird beaks – Adapted from L. Shyamal | Wikimedia Commons


Candidate Genes

To be honest, we do know quite a bit about the genetics of beak morphology. Studies on particular species have uncovered several candidate genes. For example, the beak shapes of Darwin’s finches are determined by, among others, the genes BMP4 (depth and width) and CALM1 (beak length). Recent genomic work suggested additional roles for ALX1 (craniofacial development) and HGMA2 (beak size). Interestingly, in great tits (Parus major) another gene – COL4A5 – was linked to variation in beak morphology. These examples indicate that beak shape is probably influenced by many genes and that different genes are under selection in different bird groups.

Leeban Yusuf and his colleagues compared the genomes of 72 bird species to see whether there is a common genetic mechanism underlying the evolution of beak morphology. They divided the species into separate bins based on the rates of beak shape evolution. Next, they estimated the rate at which different protein-coding genes evolved (using the dN/dS approach). If certain protein-coding genes are involved in the evolution of beak morphology, you expect a positive correlation between the bird-bins and the rate of molecular evolution. This analysis uncovered 1434 candidate genes of which several are part of developmental pathways that are involved in beak morphology (namely the Wnt signalling pathway and the ESC pluripotency pathways). There was, however, no positive correlation between the rates of beak shape change and molecular evolution. This result shows that the evolution of beak morphology is more complicated than a few mutations in protein-coding genes.

An example of how rates of beak morphology change are divided into different bins (six in this case) from slowest to fastest. From: Yusuf et al. (2020) Genome Research.


Non-coding DNA

Next, the researchers turned to non-coding genomic regions, which make up the majority of the genome. Neil Shubin formulated it nicely in his book: “Gene sequences that code for proteins compose less than 2 percent of the human genome. That leaves some 98 percent with no genes at all in it. Genes are but islands in a sea of DNA.” These non-coding regions were initially seen as “junk DNA” without a function, but we now know that some of these regions play crucial roles in regulating gene expression. The search for significant non-coding regions involved in beak morphology evolution resulted in no less than 39,806 of these genetic elements. But which genes were they regulating?

These regulatory regions come in two main types: cis-regulatory elements that are linked to nearby genes and trans-regulatory elements that affect distant genes (millions of DNA-letters apart). Analyses focused on cis-regulatory elements pointed to 884 genes of which most are involved in early craniofacial development, as shown in mice (including the the ESC pluripotency pathways that were also found for the protein-coding genes). Similarly, detailed analyses of trans-regulatory elements identified genes associated with the development of beak morphology.

The difference between cis- and trans-regulatory elements. Cis-regulatory elements affect nearby genes, while trans-regulatory elements code for proteins (e.g., transcription factors) that influence distant genes. Adapted from: Signor & Nuzhdin (2018) Trends in Ecology & Evolution.


Endless Forms

These findings highlight that fundamental developmental pathways underlie the evolutionary changes in beak morphology. Mutations in the non-coding elements that regulate these pathways can result in novel beak phenotypes, providing the raw material for natural selection. In addition, changes in protein-coding genes – which seem to be specific to particular bird lineages – add more variation to this pool of possibilities. Together, coding and non-coding genomic regions drive the spectacular diversification of avian beaks. To end with a famous quote from Charles Darwin, who would have been delighted by these findings: “From so simple a beginning endless forms most beautiful and most wonderful have been, and are being, evolved.”



Yusuf, L., Heatley, M. C., Palmer, J. P., Barton, H. J., Cooney, C. R., & Gossmann, T. I. (2020). Noncoding regions underpin avian bill shape diversification at macroevolutionary scales. Genome research30(4), 553-565.