Sibling rivalry among paralogs promotes evolution of the human brain

Abstract

Geneticists have long sought to identify the genetic changes that made us human, but pinpointing the functional-relevant changes has been challenging. Two papers in this issue suggest that partial duplication of SRGAP2, producing an incomplete protein that antagonizes the original, contributed to human brain evolution.

Humans differ from the other great apes in many ways, one of the more obvious being our larger, more complex, brains. These human-specific characteristics have evolved in the last six to seven million years since we split from our common ancestor with chimpanzees and bonobos. We can catalogue the phenotypic differences between humans and other great apes, and in some cases, we can determine when they evolved, using evidence in the fossil record. Brain size increased throughout this period, although unevenly (Figure 1A and 1B), but fossil details of internal brain structure are sparse. In a complementary way, we can catalog the genetic differences between humans and chimpanzees from their genome sequences (Chimpanzee Sequencing and Analysis Consortium 2005) and identify which differences arose on the human lineage by comparison with an outgroup, a species external to the chimpanzee-human branch, such as gorilla (Scally et al. 2012). But there are around 20 million genetic changes that are specific to humans, most of which probably have no functional impact whatever. How do we pick out the few functionally-relevant changes and link them to their phenotypic consequences? Two papers in this issue of Cell now show one way that this can be accomplished (Charrier et al. 2012, Dennis et al. 2012). The authors characterize a series of partial duplications of the single ancestral SRGAP2 gene. They make a strong case for the duplicates leading to a higher density of spines on dendritic cells in the brain and contributing to the neotenous development characteristic of humans.

Evolutionary context of SRGAP2 duplications on the human lineage.

In A-D, the vertical axis shows evolutionary time from the present (0, bottom) to seven million years ago (top). Horizontal dotted lines represent times of SRGAP2 duplications with their uncertainty indicated by grey shading. (A) Brain volume estimates of fossils and living humans, colored according to genus. (B) Timescales of genera thought to include human ancestors. (C) Sequence of duplication events of SRGAP2 copies. (D) Inferred SRGAP2 activity as a consequence of duplication of antagonistic paralogs and the decay of SRGAP2B. (E) Levels of protein (schematic) and copy number of paralogs in modern humans.

How do you sift through 20 million genetic differences? A good starting point is the subset of changes that have major effects on protein-coding genes, which includes duplications and deletions. A previous systematic survey of such copy number changes in great apes had identified 140 events specific to the human lineage (Fortna et al. 2004), including several implicated in neuronal function. These should be enriched for functionally-important changes. Among them was SRGAP2, which was shown in mice to regulate neuronal migration and morphology (Guerrier et al. 2009). SRGAP2 was therefore an excellent candidate for more detailed investigation of the link between a genetic change and evolution of the human brain.

Investigations of gene duplicates (i.e., paralogs) can face substantial obstacles. Copies recently duplicated remain similar in sequence–almost as similar as alleles–and often confuse genome assemblies; indeed, the SRGAP2 gene family was grossly misassembled in the human reference sequence. Now Dennis et al. (2012) present an approach to characterize the paralogs that should be widely applicable: use DNA from a haploid source, a complete hydatidiform mole (the product of fertilization of an enucleated oocyte by a single sperm), in which there are no alleleic variants to confound assembly, and sequence large-insert clones. This allows the authors to identify four copies: the parental SRGAP2A and its three duplicates SRGAP2B-D, and also to infer the order of the duplication events, A>B>C>D, with C and D independently derived from B (Figure 1C). These findings are fully consistent with the less detailed conclusions of the accompanying study (Charrier et al. 2012). Furthermore, comparisons of the SRGAP2 sequences suggest that the duplication events occurred ~ 3.4, 2.4 and 1.0 million years ago. The first event duplicated only nine of the 22 exons, truncating SRGAP2 in its F-BAR domain; therefore, all the other duplicated copies are also truncated, with key functional consequences.

The duplicates are all expressed, and, given their similarity, sorting out their specific roles is challenging. The first simplification comes because there is an additional deletion within SRGAP2D, removing exons two and three; this, together with its absence from some individuals in the general population, suggests it is unlikely to play any important role (Figure 1E). Interest thus focuses on SRGAP2B and C, which are extremely similar in sequence. Their expression patterns are also similar, but again there are simplifying factors: the level of SRGAP2B transcripts is low, and it also is absent from some normal individuals (Figure 1E). Thus, the main player is SRGAP2C, and its interaction with SRGAP2A. This cannot have always been true during the evolution of humans; later in this Preview, we will discuss the more important role that SRGAP2B must have played earlier in evolutionary history.

Previous work on the mouse ortholog of SRGAP2A (srGAP2 or Srgap2) had shown that it induced filipodia formation in the developing cortex through its F-BAR domain; in addition, decreasing the levels of SRGAP2A reduced axonal and dendritic branching and increased the rate of neuronal migration (Guerrier et al. 2009). Now Charrier and colleagues (2012) use mouse and cultured cell models to understand the functional consequences of the human-specific SRGAP2 duplications. The authors further characterize the phenotypes resulting from Srgap2 knock-down or knock-out and compare them with the effects of SRGAP2C expression. Srgap2 knock-down leads to neurons with increased densities of immature-looking dendritic spines in juveniles. Knock-out mice are viable even as homozygotes, retaining ~10% of Srgap2 expression and show continued growth of spine heads during development with the result that spine head size in adults are close to wild-type, but spines are more numerous and necks are longer. SRGAP2C can dimerize with SRGAP2A through its truncated F-BAR domain and decrease SRGAP2A activity. Strikingly, the simple conclusion is that SRGAP2C expression closely mimics the Srgap2 knock-down and knock-out phenotypes in almost all the characteristics examined. In summary, the functional studies suggest that SRGAP2C, by reducing SRGAP2A activity, contributes to human-like features, including extended brain development – neoteny – and cell structure in the neocortex.

In humans, the phenotypes associated with natural loss-of-function or duplication variants of SRGAP2A and C are of great interest. Particularly relevant is a balanced translocation disrupting one copy of SRGAP2A in a five-year-old girl with symptoms including intellectual disability and seizures (Saitsu et al. 2011). Loss or gain specific to SRGAP2C has not yet been reported, but Dennis and colleagues find large duplications affecting numerous genes including SRGAP2C – predicted to increase SRGAP2A antagonism –in both one control and three patients with intellectual disability and/or autism spectrum disorder. Further surveys of human variants and their detailed phenotypes, particularly SRGAP2C deletions, should be highly informative.

These conclusions have several implications for our thinking about human evolution. The duplications would have had immediate and perhaps substantial phenotypic effects (Figure 1D). SRGAP2B, the progenitor of SRGAP2C, must have been an active antagonist at the time of it duplication 3.4 million years ago, and SRGAP2 activity would have reached a minimum after the SRGAP2C duplication 2.4 million years ago (Figure 1D). These duplications would have occurred in Australopithecus species (Figure 1B,C). Did they have consequences for gross brain anatomy that might be recognized in rare fossil endocasts (e.g. Dart 1925) – mouse models might be informative here – and did they contribute to the development and behavior of these species documented by palaeontologists? Intriguingly, the use of recognizable stone tools began about 2.5 million years ago (Jobling et al. 2004) and brain size started to increase soon after (Figure 1A), but it is difficult to test for a direct link. Neoteny has long been recognized as a human characteristic (Bufill et al. 2011), and now we can begin to understand its genetic and developmental basis.