Evolutionary tinkering with transposable elements.

I t was almost 30 years ago when François Jacob declared that evolutionary innovation (the emergence of novel form and function over time) occurred primarily via a process of ‘‘tinkering’’ (1). By tinkering, Jacob essentially meant the creation of novelty through random combinations of preexisting forms. Two fundamental and countervailing notions are implicit in this view of evolution: optimality versus constraint. Were evolution to perform optimally, a more apt metaphor might be that of an engineer. An engineer works according to a plan, with a precise goal for the desired end, and uses material designed specifically toward that end. Evolution, on the other hand, must work without the benefit of foresight and is subject to very real constraints with respect to the material at its disposal; as such, evolutionary biology is replete with examples of suboptimal solutions to functional challenges (2). Similarly, a tinkerer works without a clear plan by using anything and everything at his disposal to produce an entity that possesses some kind of (unanticipated) functional utility. In this issue of PNAS, Cordaux et al. (3) explore an example of tinkering along the human evolutionary lineage, whereby an existing host gene merged with a transposable element (TE) to create a primate-specific chimeric gene. In the decades since Jacob’s exposition, molecular biology studies have produced a deluge of primary data (tens of thousands of three-dimensional protein structures and literally billions of nucleotides of gene sequences, including hundreds of complete genomes in the past few years alone). Comparative studies of the resulting data have underscored the extent to which genome evolution is indeed characterized by tinkering. There are a discrete and finite number of structural folds, protein sequence domains, and gene families (4); new genes evolve through slight modifications and or recombinations of these preexisting forms. The actual de novo evolution of protein coding sequences is exceedingly rare. For instance, despite the 80–100 million years that have elapsed since the human and mouse lineages diverged, the genomes of these two species share 99% homologous genes (5). Clearly, however, mammalian evolution has been marked by substantial functional innovation, and so it must be that the genome-level dynamics underlying this innovation are dominated by creation through rearrangement. One of the largely unanticipated results of mammalian genome sequencing efforts was the revelation of the extent to which these genomes are made up of sequences derived from TE insertions. The human genome sequence was found to consist of 45% TE-derived sequences (6), and this figure is certainly a vast underestimate because many TEderived human sequences have diverged beyond recognition. In addition to being ubiquitous genomic elements, TEs are also autonomous in the sense that they carry the regulatory and protein coding sequences necessary to catalyze their transposition. The ubiquity of TEs, along with the functional machinery that they encode, makes them ideal genetic building blocks that evolution can tinker with to create novel forms. Indeed, despite the early notion of TEs as being strictly selfish (parasitic) elements that serve no function for their hosts (7), there now exist numerous examples of formerly mobile TE sequences that have been ‘‘domesticated’’ (8) to serve some functional role for the host genomes in which they reside (9, 10). However, there is still a relative paucity of detailed studies that address both the evolutionary dynamics of TE-derived host genes as well as the functional roles of the proteins that they encode. The work of Cordaux et al. (3) on the SETMAR gene represents an important step toward alleviating this knowledge gap. SETMAR, originally discovered by Robertson and Zumpano (11), is a chimeric gene made up of a SET histone methyltransferase transcript fused to the transposase domain of a formerly mobile TE sequence. The transposase domain in question comes from a member of the Hsmar1 mariner-like family of elements. Mariner-like elements are more commonly found in insects, and Hsmar1 was the first TE of this type found in the human genome. Hsmar1 elements are class II, or DNA elements, that have terminal inverted repeats (TIRs) flanking an ORF that encodes a transposase. Class II elements transpose via a cut-and-paste mechanism catalyzed by the transposase, which binds to the TIRs, excises the element, and then inserts it in a new location. Class I elements, or retrotransposons, which transpose via the reverse transcription of an RNA intermediate, are actually far more common than DNA elements in the human genome. The so-called longand short-interspersed nuclear elements, LINEs and SINEs, respectively, make up 25% of the human genome. However, for as yet unknown reasons, DNA elements like Hsmar1 are overrepresented among host genes with TE-derived coding sequences. This overrepresentation may be because of the broad utility of the DNA-binding properties encoded by the transposase ORF. In fact, there is a distinct possibility that, as these kinds of chimeric genes are born, they are able to bind to multiple dispersed sites around the genome (those occupied by their cognate TIRs), resulting in the