Wednesday, June 3, 2009

FoxP2: Evolving Human Intelligence

Like an answer floating to the window of a magic 8-ball, Foxp2 has floated back to the top of the attention heap with a just-published paper in Cell: A Humanized Version of Foxp2 Affects Cortico-Basal Ganglia Circuits in Mice by Enard et al. This gene/enzyme/Transcription Factor (TF) is an item of extreme interest in the study of human evolution because it's highly conserved across tetrapods, with just a one-amino-acid difference between chimpanzees and mice, but humans have managed to acquire two changed amino acids since our lineage split from chimpanzees and gorillas.

I'm going to refer to this change (both amino acids) as an "improvement", although I recognize it's somewhat chauvinistic of me. It will make it easier to distinguish from other mutants in the discussion.

The FoxP2 protein is an enzyme that, at least, participates in the complex dance that is gene activation. Defective mutant alleles of this gene in humans cause cause a "severe disorder, involving profound deficits in the control of complex coordinated face and mouth movements, resulting in disrupted speech ([ref's])."11 The combination of this with the fact that humans have a recent improvement in this gene8 is what makes it so interesting.

Biochemical Effects of Foxp2 Change

What are the effects of this improvement? There are several ways to approach this question, and I'm going to start with biochemistry: I've discussed the fact that the cell is like a powerful analog computer with a large collection of enzymes that can be phosphorylated, which changes their behavior, or can phosphorylate other proteins, or, most importantly, both. ...

Each enzyme in this system may be envisioned as a transistor in a large network of transistors wired together (along with appropriate resistors according to the type of transistor and how they're connected). Each transistor also has a "knob" for programming its contribution to the network, which corresponds to the total quantity of enzyme in all phosphorylation states, a function of its gene expression level (and other mechanisms that control overall quantities of proteins: see my How Smart is the Cell? Part II: The Gene Activation network as an Analog Computer).

Note that I said all states rather than both. Many proteins, including many enzymes, can be phosphorylated in more than one place -- often by different kinases. Whether a protein can be phosphorylated at one place can depend on whether it's phosphorylated at another (or not), or more than one other place, creating a complex logic in its phorphorylation. And this logic can differ for different kinases. The same is potentially true for dephosphorylation, although research is just beginning into the specificity of phosphatases.

Now, the reason this is important is that one of the changes in human Foxp2 involves potential phosphorylation sites. One change, at position 303, replaces a threonine with an asparagine. Evidently, the original threonine was not a site for phosphorylation, although this residue can be phosphorylated under certain circumstances.

The other change, at position 325, replaces an asparagine with a serine. One study3 showed that "the human-specific change at position 325 creates a potential target site for phosphorylation by protein kinase C together with a minor change in predicted secondary structure."3 It goes on to say:
Several studies have shown that phosphorylation of forkhead transcription factors can be an important mechanism mediating transcriptional regulation. Thus, although the FOXP2 protein is extremely conserved among mammals, it acquired two amino-acid changes on the human lineage, at least one of which may have functional consequences.3

This may well be true, but even if this effect is not present, the results are extremely important.

How does this relate to our transistor network analogy? The first change probably has no effect, but the second may be compared to adding a whole new set of connections, from previously unconnected transistors to this one (the one that represents the foxp2 enzyme).

(In fact, it's more than this: our analogy is simplistic in that a transistor always (AFAIK) has a scalar output while an enzyme with multiple sites for phosphorylation, or any other type of activation, will output an n-dimensional vector where n is one less than the total number of possible states involved, including the completely inactivated one. Adding a site for phosphorylation, then, adds an entire dimension to the output as well as adding the new connections from kinases capable of phosphorylating it.)

At this point we're ready to consider the new paper.

Enard, W., Gehre, S., Hammerschmidt, K., Hölter, S., Blass, T., Somel, M., Brückner, M., Schreiweis, C., Winter, C., & Sohr, R. (2009). A Humanized Version of Foxp2 Affects Cortico-Basal Ganglia Circuits in Mice Cell, 137 (5), 961-971 DOI: 10.1016/j.cell.2009.03.041

What these researchers did was to induce an improvement similar to the normal human type into mice, and compare the results with both the wild type and some mice with defects like those of the human defective mutants. They demonstrated that the results were totally unlike those of the defective allele, and influenced vocalization, the dopamine system, exploratory behavior, and dendrite growth and neural plasticity in the striatum, "a part of the basal ganglia affected in humans with a speech deficit due to a nonfunctional FOXP2 allele".1 The last four are probably related, and the first may be.

What we should not assume is that this mutation makes mice "more like humans", and especially that they will be able to speak. As it happens, adult male mice actually have "songs" they use during courtship,5, 6 although this fact wasn't mentioned in this paper. I would be interested how the adult male mice in their study performed in experiments recording their songs.

The most important implication of this research is that, when applied to an arbitrary mammal, it produces a viable animal with a consistent set of changes. Subsequent evolution can take it from there, and probably did in humans.

Network Implications

To understand the importance of this improvement, we need to make a short excursion into network theory. This field of study maps various objects and the relationships among them into "nodes" and "edges" respectively: the nodes are points on a graph, the edges are lines connecting them. The resulting network can then be analyzed without reference to how those objects relate, considering only which objects relate (are connected by an edge) to which, how many relationships (edges) each object has, looking at averages, loops, path lengths between nodes without direct connections, and so on. Obviously, creating a bunch of new "edges", connecting previously unconnected nodes, will change the characteristics of the network. It will be more "tightly" connected, and potentially modules that were only distantly connected will now be cross-talking.

In How Smart is the Cell? Parts II and III I discussed how the gene activation system can "program" the enzyme system by "twisting knobs", that is varying the total amount of each enzyme created. Mutations to the control sequences (TF binding sites) can potentially make fine adjustments to gene expression levels, as well as major changes. Other types of mutation can add whole new logic modules, duplicating an existing one, which can then begin to diverge (evolve away), or grafting in logic from the control regions(s) of some other gene(s). With the new version of Foxp2 in place, protohuman lineages could then begin accumulating, and selecting, such mutations.

What makes Foxp2 particularly interesting is that it is a TF itself, which means mutations could arise not only to its control sequences, but to those it interacts with. Thus Foxp2 is right at the intersection of a complex phosphorylation network and a complex gene activation network. Adding a new site for phosphorylation adds considerable complexity to those networks, making it an even more important participant in the "cellular brain" that controls development, especially in the "neural brain" of amniotes.

Foxp2 Across the Amniotes

Although Foxp2 is generally conserved, it has seen some changes, and since there's considerable interest in its relationship to human vocalization, these changes have been studied and correlated with such things as learned vocalization.

Foxp2 is present in birds, although they lack the changes that constitute the human "improvement".4

Figure 1: Alignment of deduced amino acid sequences from the zebra finch FoxP2 cDNA with three mammalian sequences. Click on image to see original caption. (From Ref 4.)

A more complete study7 analyzed 18 animals spread through the amniotes, and determined that the single change present in songbirds occurred independently in the mouse lineage (see Ref 7, figure 1.) It's interesting that mice also have a "song" somewhat reminiscent of birds, although much higher pitched.4, 5, 6 Although it's been a popular idea that the human improvements might contribute to "song learning" no such correlation has been found,8 only the one between mouse "songs" and bird songs.

While the human improvement hasn't been found to correlate to any sort of "song learning", important discoveries have been made regarding the times and places of Foxp2 expression in such animals.6, 7, 10

Another important discovery involves bats: although none of them seem to share the human improvement, they have seen an impressive number of changes to this gene.2 Instead it has been speculated:
[... T]hat observed variation in bats might be associated with aspects of echolocation. As mentioned, amino acid variation at both exons 7 and 17 in bats corresponds well to echolocation types/phylogenetic boundaries, with almost complete conservation across groups of confamilial species but contrasting signatures between families. Such high sequence diversity at exons 7 and 17 in bats relative to other mammals, including echolocating cetaceans, indicates that FoxP2 plays a role in the sensorimotor demands that are peculiar to bat echolocation rather than echolocation in general. Indeed while all cetaceans shared three amino acid substitutions (Pro302, Ala304 and Met316) no differences were observed between echolocating and non-echolocating baleen cetaceans, also supporting an earlier comparison of one example from each suborder [ref].2

I should note, however, that the closest non-cetacean relative, the hippopotamus, does not share their changes.8

Another interesting fact is that the carnivores share one of the human changes, the important one that changes the network characteristics.8, 9

Figure 2: Table of various animals and their status in terms of the human improvement. Click on image to see original capation. (From Ref 9.)

The Human Improvement

In view of all the above, we can make a hypothesis regarding the mutation that led to the improvement in human language ability. The original mutation, like the changes in mice,1 led to a creature with a few changes to neural development, but a more complex network in the "cellular computer" involved in controlling development. There were changes to vocalization and learning-related mechanisms, perhaps directly freeing up vocal behavior to be learned in a way that even chimpanzees don't seem to be able to. After this, a series of more minor mutations to the control sequences (and, perhaps, to non-coding RNA sequences) worked over the new, tighter, network, adapting it and its phenotype to fit the new behavioral and ecological niche made possible by the improvement.

Human language, then, represents the end (so far) of a long process of small mutations and selection, building on the original enabling improvement. Not a "big bang", but more than a simple progression of small mutations.

Links: (Not all of these are called out in the text. Use the back key if you came via one of the footnotes.) (I've included only the links referenced in this leader.)

1. A Humanized Version of Foxp2 Affects Cortico-Basal Ganglia Circuits in Mice

2. Accelerated FoxP2 Evolution in Echolocating Bats

3. Molecular evolution of FOXP2, a gene involved in speech and language

4. Parallel FoxP1 and FoxP2 Expression in Songbird and Human Brain Predicts Functional Interaction

5. Ultrasonic Songs of Male Mice

6. Singing Mice, Songbirds, and More: Models for FOXP2 Function and Dysfunction in Human Speech and Language

7. An evolutionary perspective on FoxP2: strictly for the birds?

8. FoxP2 in Song-Learning Birds and Vocal-Learning Mammals

9. Accelerated Protein Evolution and Origins of Human-Specific Features: FOXP2 as an Example

10. FoxP2 Expression in Avian Vocal Learners and Non-Learners

11. FOXP2 expression during brain development coincides with adult sites of pathology in a severe speech and language disorder


  1. FOXP2 is the first gene relevant to the human ability to develop language2. Really human FOXP2 contains changes in amino-acid coding and a pattern of nucleotide polymorphism, which strongly suggest that this gene has been the target of selection during recent human evolution. Nice to see your site. Thanks!

  2. My daughter used to work at a lab focused on the FOXP2 gene. I will pass this site on to her. I am certain she will enjoy reading what you have to say on this.