The review assumes that readers of this journal are familiar with the actions of mobile DNA and other genome restructuring functions. It will try to integrate that familiarity into the historical development of evolutionary concepts and incorporate recent discoveries from genome sequencing. Just as our knowledge of mobile DNA has introduced new ways of thinking about hereditary change, the results of sequence analysis have documented several types of genome alterations at key places in evolutionary history, alterations which are notable because they happened within a single generation and affected multiple cellular and organismal characters at the same time: horizontal transfers of large DNA segments, cell fusions and symbioses, and whole genome doublings (WGDs). These rapid multi-character changes are fundamentally different from the slowly accumulating small random variations postulated in Darwinian and neo-Darwinian theory.

Cell mergers and WGDs are the kinds of events that activate mobile DNA and genome restructuring. In order to fully integrate the genomic findings with our knowledge of mobile DNA, we have to make use of information about the molecular regulation of mobile DNA activities as well as McClintock's view that cells respond to signs of danger, frequently restructuring their genomes as part of the response [1]. This regulatory/cognitive view of genome restructuring helps us to formulate reasonable hypotheses about two unresolved questions in evolutionary theory: (i) the connections between evolutionary change and ecological disruption; and (ii) the origins of complex adaptive novelties at moments of macroevolutionary change.

All evolutionists accept descent with modification as fundamental to the evolutionary process, but views towards issues (ii) and (iii) have depended on the existing state of biological knowledge in each historical period.

In the 19th century, Darwin based his thinking on the observations of animal breeders and naturalists. Lacking detailed studies of inheritance, he postulated that change arose randomly as 'numerous, successive, slight variations' [2]. Applying the uniformitarian principle he learned from Charles Lyell, his professor of geology [3], Darwin extrapolated that these small changes would accumulate over long periods of time, under the guidance of natural selection, to produce major adaptive characters, such as the eye, and eventually would lead to the branching off of new species. Thus, classical Darwinism was characterized by its gradualist view of change and ascribed the major role in adaptive innovation to the positive action of natural selection in sequentially favouring ever fitter variants.

In the 20th century, evolutionists were confronted by an apparent contradiction between Darwinian gradualism and the abrupt changes in individual traits that were observed to undergo Mendelian segregations in genetic crosses. This contradiction was resolved at mid-century by the neo-Darwinian 'modern synthesis' that integrated Darwinian gradualism with mathematical population genetics [4, 5]. Like Darwin, his neo-Darwinian followers postulated that the mutational process, which generated allelic variants of individual genes, has to be random in nature. In opposition to Lamarckian ideas, any possibility that organismal history could influence hereditary variation was excluded. The primary role in determining evolutionary novelty remained with natural selection.

In the 21st century, we have the legacy of more than five decades of molecular biology. Knowledge of DNA has allowed us to study the mutational process with nucleotide and phosphodiester bond precision [6]. Our DNA-based technology has made it possible to acquire a growing database of genome sequences that permit us to read the history of evolutionary events preserved in the nucleic acid and protein record.

Molecular cell biology has uncovered sophisticated networks in all organisms. They acquire information about external and internal conditions, transmit and process that information inside the cell, compute the appropriate biochemical or biomechanical response, and activate the molecules needed to execute that response. These information-processing networks are central to the systems biology perspective of the new century. Altogether, we have a radically different conceptual perspective on living organisms than our predecessors. As a result, we need to ask how this new perspective affects our 21st century understanding of the evolutionary process. Posing this question and outlining a provisional answer are the goals of this review.

In addition to the discoveries of molecular biology, our 21st century thinking benefits from another major strand of 20th century research - McClintock's cytogenetic studies that led her to recognize the internal capabilities cells possess to repair and restructure their genomes. Starting in the 1930s with X-ray-induced chromosome rearrangements, she analysed how maize cells dealt with broken ends. These studies taught her that maize had the ability to detect broken ends, bring them together and fuse them to generate novel chromosome structures, including deletions, inversions, translocations, and rings [7–11]. She also found that cells in the embryo, but not in the terminally differentiated endosperm, could 'heal' a single broken end by the addition of a telomere. In the course of exploiting these repair capabilities to generate deficiencies of maize chromosome IX, she made the discovery of transposable elements, for which she is best known today [12].

Although the general view is that McClintock's discovery of transposition was most important for revealing a novel mechanism of genomic change, she herself placed the emphasis on another aspect of her work. In conversation, she would often say that she was far more interested in control than she was in transposition. By this, she meant that the ability of her maize plants to regulate expression and restructure their genomes in accordance with their needs was more significant than the mechanics of chromosome rearrangement. She was primarily interested in the sensory and decision-making (that is, cognitive) capacities of cells with damaged genomes. As she expressed it at the conclusion of her Nobel Prize lecture: 'In the future, attention undoubtedly will be centred on the genome, with greater appreciation of its significance as a highly sensitive organ of the cell that monitors genomic activities and corrects common errors, senses unusual and unexpected events, and responds to them, often by restructuring the genome' [1]. In the next section, we will see how prescient and compatible with molecular analysis her vision was.

The pioneering molecular biologists expected to provide a firm physical-chemical basis for the traditional 20th century view that genotype determines phenotype and that genotype changes accidentally during replication [13]. This expectation of one-way cellular information transfer was articulated most succinctly in Crick's Central Dogma of Molecular Biology[14]. Even when Temin and Mitzutani discovered reverse transcriptase [15], Crick insisted that the genome was the source of phenotypic information and that nucleic acids as a class were the basic information molecules of the cell [16].

A review of the past five decades of molecular cell biology, including an analysis of how mobile DNA operates, leads to a dramatically different picture of cellular informatics [17, 18]. We realize that DNA by itself is inert. It must interact with other molecules for transcription, replication, transmission to daughter cells and repair. DNA does not change by itself, even when damaged. Point mutations and DNA rearrangements depend upon the action of protein and RNA molecules, and many organisms undergo genome restructurings as a necessary part of the normal life cycle [19–22]. A great deal of cellular information processing occurs outside the genome - for example, transcript processing, protein processing and decoration and signal transduction - yet the results of this information processing also feed back onto the genome in the form of alterations in nucleoprotein complexes, chromatin configurations, subnucleoid or subnuclear localization, and sequence or structural changes. In other words, the cell is a multilevel information-processing entity, and the genome is only a part of the entire interactive complex.

We can summarize the change from the simple linear view of the Central Dogma to today's complex systems-based picture of cell informatics by writing out molecular information transfer events as sets of Boolean propositions (adapted from [18]):

SUMMARY: DNA + protein + ncRNA + signals + other molecules < == > Genome structure and phenotype

In this review, our focus is on evolution. So it is most appropriate to ask what lessons we have learned from genome sequencing. There are many, and we discuss the ones that are most relevant to the action of mobile DNA.

The rapid natural genetic engineering response to genome doubling reflects a tendency to return to the normal diploid state. This poorly understood process of diploidization involves chromosome loss, deletions and chromosome rearrangements [150]. The chief mechanistic basis for activation of natural genetic engineering in response to hybridization and genome doubling appears to be changes in chromatin organization and in epigenetic modifications of the DNA that normally inhibit activity of mobile elements (Tables 3 and 4) [151–154].

One of the traditional objections to Darwinian gradualism has been that it is too slow and indeterminate a process to account for natural adaptations, even allowing for long periods of random mutation and selection. A successful random walk through the virtually infinite dimensions of possible genome configurations simply has too low a probability of success [155]. Is there a more efficient way for cells to search 'genome space' and increase their probability of hitting upon useful new DNA structures? There is, and the underlying molecular mechanisms utilize the demonstrated capabilities of mobile DNA and other natural genetic engineering systems [156, 157].

Perhaps the most important aspect of evolutionary change by natural genetic engineering is that it employs a combinatorial search process based upon DNA modules that already possess functionality. The evolutionary reuse of functional components has been recognized for many years [158, 159], but it is only with genome sequencing that we have come to appreciate how fundamental and virtually ubiquitous such reuse is. A well-established engineering principle is to build new structures to meet specific requirements by rearranging proven, existing components, as in mechanical structures and electronic circuits. The evolution of proteins by domain accretion and shuffling is one example of an analogous biological process. Mixing functional domains in new combinations is far more likely to produce a protein with novel activities than is the modification of one amino acid at a time. Single amino acid changes are more suitable for modulating existing functional properties (for example, ligand binding and allosteric responses) than for generating capabilities that did not previously exist. In addition to the combinatorial search via shuffling of existing exons, further variability results from the formation of novel exons. We do not yet know a great deal about any biases that may exist in the exonization process. If it is correct to postulate that new functional exons arise by the exaptation of segments of mobile DNA, such as SINE elements, then it will be worthwhile to investigate the coding content of these elements to see if there is any tendency favouring sequences that encode useful folded polypeptide structures.

The second major aspect of evolutionary change by natural genetic engineering is that it generally takes place after an activating event which produces what McClintock called a 'genome shock' [160]. Activating events include loss of food [18], infection and interspecific hybridization (Tables 3 and 4) - just the events that we can infer from the geological and genomic records have happened repeatedly. Episodic activation of natural genetic engineering functions means that alterations to the genome occur in bursts rather than as independent events. Thus, novel adaptations that require changes at multiple locations in the genome can arise within a single generation and can produce progeny expressing all the changes at once. There is no requirement, as in conventional theory, that each individual change be beneficial by itself. The episodic occurrence of natural genetic engineering bursts also makes it very easy to understand the punctuated pattern of the geological record [161]. Moreover, the nature of activating challenges provides a comprehensible link to periodic disruptions in earth history. Geological upheavals that perturb an existing ecology are likely to lead to starvation, alteration of host-parasite relationships and unusual mating events between individuals from depleted populations.

A particular instance of the potential for stress-activated natural genetic engineering to produce complex novelties is the exaptation of an existing functional network following its duplication by WGD. Domains may be added to various proteins in the network to allow them to interact with a novel set of input and output molecules. In addition, insertions of connected regulatory signals at the cognate coding regions can generate a new transcriptional control circuit that may allow the modified network to operate under different conditions from its progenitor.

The idea that genomic restructuring events may be integrated functionally in order to operate coordinately at a number of distinct loci encoding components of a regulatory network may seem extremely unlikely. However, the basic requirement for such integration is the ability to target DNA changes to co-regulated regions of the genome. Precisely this kind of targeting has been demonstrated for mobile elements in yeast, where retrotransposon integration activities interact with transcription [162] or chromatin [163] factors, and in Drosophila, where P elements can be engineered to home in on loci regulated by particular regulatory proteins [164]. In addition, we know that mobile element insertion can be coupled with replication [165] and DNA restructuring with transcription [166]. Of course, the feasibility of such multi-locus functional integration of genome changes remains to be demonstrated in the laboratory. Fortunately, the experiments are straightforward; we can use appropriately engineered transposons and retrotransposons to search for coordinated multilocus mutations after activation. Clearly, the subject of functionally targeted changes to the genome belongs on the 21st century mobile DNA research agenda.

Our ability to think fruitfully about the evolutionary process has greatly expanded, thanks to studies of mobile DNA. Laboratory studies of plasmids, transposons, retrotransposons, NHEJ systems, reverse transcription, antigenic variation in prokaryotic and eukaryotic pathogens, lymphocyte rearrangements and genome reorganization in ciliated protozoa have all made it possible to provide mechanistic explanations for events documented in the historical DNA record [6]. We know that processes similar to those we document in our experiments have been major contributors to genome change in evolution. Using our knowledge of genome restructuring mechanisms, we can generate precise models to account for many duplications, amplifications, dispersals and rearrangements observed at both the genomic and proteomic levels.

The genome DNA record also bears witness to sudden changes that affect multiple characters at once: horizontal transfer of large DNA segments, cell fusions and WGDs. These data are not readily compatible with earlier gradualist views on the nature of evolutionary variation. However, we are now able to apply the results of findings on the regulation of natural genetic engineering functions in the laboratory and in the field to make sense of the DNA record. Cell fusions and WGDs are events we know to activate DNA restructuring functions (Tables 3 and 4). Thus, it is not surprising that bursts of intracellular horizontal transfer, genome reduction and genome rearrangement follow these initial abrupt changes in the cell's DNA. How a newly symbiotic cell or one with a newly doubled genome manages the transition to a stable genome structure that replicates and transfers reliably at cell division is another important subject for future research. The lessons we learn about silencing mobile DNA by internal deletion [12] and RNA-directed chromatin modification [167] are likely to prove helpful starting points.

This 21st century scenario assumes a major role for the kind of cellular sensitivities and genomic responses emphasized by McClintock in her 1984 Nobel Prize address [1]. Such a cognitive component is absent from conventional evolutionary theory because 19th and 20th century evolutionists were not sufficiently knowledgeable about cellular response and control networks. This 21st century view of evolution establishes a reasonable connection between ecological changes, cell and organism responses, widespread genome restructuring, and the rapid emergence of adaptive inventions. It also answers the objections to conventional theory raised by intelligent design advocates, because evolution by natural genetic engineering has the capacity to generate complex novelties. In other words, our best defense against anti-science obscurantism comes from the study of mobile DNA because that is the subject that has most significantly transformed evolution from natural history into a vibrant empirical science.