Genomes on the shrink.

B ecause genetic events do not fossilize, we are forced to deduce the evolution of bacterial genomes by comparing the features of contemporary organisms. This comparative approach has exposed many of the alterations endured by bacterial genomes (bouts of expansion and contraction or changes in base compositions), but such reconstructions are often imprecise, and sometimes incorrect, because they are limited by the spectrum and relationships of the sequenced genomes that are available. Determining whether changes in gene repertoires and genome size were gradual or episodic is not feasible when the genomes being considered diverged several hundred million years ago. Fortunately, the results reported by Nilsson et al. (1) in this issue of PNAS indicate how such transformations proceed and help to explain one of the most interesting and pervasive trends in the evolution of bacterial genomes. When analyzed in a molecular phylogenetic perspective, every clade of bacteria with genome sizes of 2 Mb was derived from ancestors with substantially larger genomes (Fig. 1). This pattern dispels the long-held notion that bacteria evolved by the successive doubling of small-genomed progenitors (2, 3) and raises numerous questions about an evolutionary process that seems to affect all bacterial lineages. Among the groups best suited for investigating the progression toward reduced genomes are the -proteobacteria, due principally to the large number of fully sequenced constituents (53 at last count). Within this phylum, which includes the workhorses of bacterial genetics and pathogenesis, Escherichia coli and Salmonella typhimurium, the sizes of already-sequenced genomes vary over an order of magnitude, from 600 kb in Buchnera aphidicola (4) to 7,000 kb in Pseudomonas fluorescens (5). With the near-perfect correlation between genome size and gene number in bacteria (6), reductions in genome size will usually result in the loss of some functional capabilities. Given the observed range of genome sizes, what circumstances might allow elimination of 80% or more of the coding capacity of an organism? Those bacteria with the smallest genomes are intracellular pathogens and symbionts that maintain obligate associations with eukaryotic hosts. In these cases, the hosts provision bacteria with a constant supply of nutrients, thereby rendering unnecessary many genes that were previously needed in less certain environments, such as those encountered by free-living bacteria. Just because a gene is superfluous does not assure its removal from a genome. For example, the human genome maintains hundreds of nonfunctional olfactory receptor genes, including some that date to the origin of tetrapods (7, 8). However, as evident from comparisons of bacterial pseudogenes with their functional counterparts, the mutational process in bacterial genomes is strongly biased toward deletions (6, 9). Although nonfunctional regions can be maintained in a bacterial genome for some time, they gradually erode and are eventually eliminated. The deletional bias observed in bacterial pseudogenes goes a long way toward explaining why bacterial genomes are compact and gene-rich, and why large nonfunctional regions do not accumulate within their genomes. However, the extent to which this process has been responsible for the extreme reduction of symbiont genomes is difficult to evaluate using conventional methods that align and compare homologous sequences. Because the majority of genes present in their large-genome relatives are missing from highly reduced genomes, there is no information about the manner in which these sequences were eliminated: the extreme genome reduction could proceed by a slow and continual erosion of individual genes, or, alternatively, by expansive deletions that jettison numerous genes with each event. The presence of hundreds of pseudogenes scattered around the genomes of some recent pathogens (10–13), as well as the fact