It is well known that basing phylogenetic reconstructions on uncorrected genetic distances can lead to errors in their reconstruction. Nevertheless, it is often common practice to report simply the most similar BLAST (Altschul et al. 1997) hit in genomic reports that discuss many genes (Ruepp et al. 2000; Freiberg et al. 1997). This is because BLAST hits can provide a rapid, efficient, and concise analysis of many genes at once. These hits are often interpreted to imply that the gene is most closely related to the gene or protein in the databases that returned the closest BLAST hit. Though these two may coincide, for many genes, particularly genes with few homologs, they may not be the same. There are a number of circumstances that can account for such limitations in accuracy (Eisen 2000). We stress here that genes appearing to be the most similar based on BLAST hits are often not each others closest relative phylogenetically. The extent to which this occurs depends on the availability of close relatives present in the databases. As an example we have chosen the analysis of the genomes of a crenarcheaota species Aeropyrum pernix, an organism with few close relatives fully sequenced, and Escherichia coli, an organism whose closest relative, Salmonella typhimurium, is completely sequenced.

Large-scale genome arrangement plays an important role in bacterial genome evolution. A substantial number of genes can be inserted into, deleted from, or rearranged within genomes during evolution. Detecting or inferring gene insertions/deletions is of interest because such information provides insights into bacterial genome evolution and speciation. However, efficient inference of genome events is difficult because genome comparisons alone do not generally supply enough information to distinguish insertions, deletions, and other rearrangements. In this study, homologous genes from the complete genomes of 13 closely related bacteria were examined. The presence or absence of genes from each genome was cataloged, and a maximum likelihood method was used to infer insertion/deletion rates according to the phylogenetic history of the taxa. It was found that whole gene insertions/deletions in genomes occur at rates comparable to or greater than the rate of nucleotide substitution and that higher insertion/deletion rates are often inferred to be present at the tips of the phylogeny with lower rates on more ancient interior branches. Recently transferred genes are under faster and relaxed evolution compared with more ancient genes. Together, this implies that many of the lineage-specific insertions are lost quickly during evolution and that perhaps a few of the genes inserted by lateral transfer are niche specific.

As a means of investigating gene function, we developed a robust transcription fusion reporter vector to measure gene expression in bacteria. The vector, pTH1522, was used to construct a random insert library for the Sinorhizobium meliloti genome. pTH1522 replicates in Escherichia coli and can be transferred to, but cannot replicate in, S. meliloti. Homologous recombination of the DNA fragments cloned in pTH1522 into the S. meliloti genome generates transcriptional fusions to either the reporter genes gfp ؉ and lacZ or gusA and rfp, depending on the orientation of the cloned fragment. Over 12,000 fusion junctions in 6,298 clones were identified by DNA sequence analysis, and the plasmid clones were recombined into S. meliloti. Reporter enzyme activities following growth of these recombinants in complex medium (LBmc) and in minimal medium with glucose or succinate as the sole carbon source allowed the identification of genes highly expressed under one or more growth condition and those expressed at very low to background levels. In addition to generating reporter gene fusions, the vector allows Flp recombinase-directed deletion formation and gene disruption, depending on the nature of the cloned fragment. We report the identification of genes essential for growth on complex medium as deduced from an inability to recover recombinants from pTH1522 clones that carried fragments internal to gene or operon transcripts. A database containing all the gene expression activities together with a web interface showing the precise locations of reporter fusion junctions has been constructed (www.sinorhizobium.org).

A simple sequence is abundant in the proteins that have been sequenced to date. But unusual protein features, such as a simple sequence, are not present in the same high frequency within structural databases. A subset of these simple sequences, a group with a highly repetitive nature has been shown to be abundant in eukaryotes but not in prokaryotes. In this study, an examination of the eukaryotic proteins in the Protein Data Bank (PDB) has revealed a large deficiency of low complexity, highly repetitive protein repeats. Through simulated databases of similar samples of eukaryotic proteins taken from the National Center for Biotechnology Information (NCBI) database, it is shown that the PDB contains a significantly less highly repetitive, simple sequence than artificial databases of similar composition randomly derived from NCBI. When the structural data for those few PDB sequences that did contain a highly repetitive simple sequence is examined in detail, it is found that in most cases the tertiary structure is unknown for the regions consisting of a simple sequence. This lack of a simple sequence both in the PDB database and in the structural information suggests that this type of simple sequence may produce disordered structures that make structural characterization difficult.

For decades proteins were thought to interact in a "lock and key" system, which led to the definition of a paradigm linking stable three-dimensional structure to biological function. As a consequence, any non-structured peptide was considered to be nonfunctional and to evolve neutrally. Surprisingly, the most commonly shared peptides between eukaryotic proteomes are low-complexity sequences that in most conditions do not present a stable three-dimensional structure. However, because these sequences evolve rapidly and because the size variation of a few of them can have deleterious effects, low-complexity sequences have been suggested to be the target of selection. Here we review evidence that supports the idea that these simple sequences should not be considered just "junk" peptides and that selection drives the evolution of many of them.