A characteristic feature of bacteriophage genomes is that they are architecturally mosaic, with each individual genome representing a unique assemblage of individual exchangeable modules. Plausible mechanisms for generating mosaicism include homologous recombination at shared boundary sequences of module junctions, illegitimate recombination in a non-sequence-directed process, and site-specific recombination. Analysis of the novel mycobacteriophage Giles genome not only extends our current perspective on bacteriophage genetic diversity, with more than 60% of the genes unrelated to other mycobacteriophages, but offers novel insights into how mosaic genomes are created. In one example, the integration/excision cassette is atypically situated within the structural gene operon and could have moved there either by illegitimate recombination or more plausibly via integrase-mediated site-specific recombination. In a second example, a DNA segment has been recently acquired from the host bacterial chromosome by illegitimate recombination, providing further evidence that phage genomic mosaicism is generated by nontargeted recombination processes.The total estimated number of 10 31 bacteriophage particles in the biosphere suggests that phages represent a majority of biological entities (14). The complete genome sequences of approximately 400 double-stranded DNA tailed phages have been determined, which show, together with viral metagenomic studies (1,5,8), that the phage population has very high genetic diversity and is replete with genetic information that either is very distant from other known sequences or has a distinct evolutionary origin. Furthermore, the completely sequenced phage genomes have characteristic mosaic architectures, so that modules, composed of either single genes or several genes, are shared between individual genomes (6, 15, 18, 28); each phage genome can thus be considered a unique assemblage of individual modules. While the number of different modules remains ill-defined, the combinatorial possibilities of these modules far exceeds the estimated total number of phage particles in the biosphere (12).The evolution of mosaic genomes requires the creation of novel junctions at module boundaries. These module junctions often correspond closely with gene boundaries, and two distinct models have been proposed to explain this mosaicism. The first was proposed by Susskind and Botstein (34) and supposes that there are short conserved sequences at gene boundaries that act as recombination targets. Such boundary sequences have been described in some Escherichia coli phages (6), but these appear to be exceptional rather than typical examples. The second model proposes that module junctions are a result of illegitimate recombination events that occur without DNA sequence identity (beyond a few nucleotides in length), accompanied by selection for gene function and genome length of a packageable size (15). While such events are likely to be rare, the large population size, a long evolutionary history, and active p...