Genome-wide determination of barriers to horizontal gene transfer

Abstract

Horizontal gene transfer (HGT) is a major contributor to bacterial genome evolution, generating phenotypic diversity, driving the expansion of protein families, and facilitating the evolution of new phenotypes, new metabolic pathways, and new species. Comparative studies of gene gain in bacteria suggest that the frequency with which individual genes successfully undergo HGT varies considerably and may be associated with the number of protein-protein interactions in which the gene participates--its connectivity. Two non-exclusive hypotheses have emerged to explain why transferability should decrease with connectivity: the Complexity Hypothesis (Jain, Rivera, & Lake, 1999) and the Balance Hypothesis (Papp, Pal, & Hurst, 2003). These hypotheses predict that the functional costs of HGT arise from a failure of divergent homologues to make normal protein-protein interactions or from gene mis-expression, respectively. Here we describe genome-wide assessments of these hypotheses in which we used 74 existing prokaryotic whole genome shotgun libraries to estimate rates of horizontal transfer of genes from taxonomically diverse prokaryotic donors into E. coli. We show that transferability declines as connectivity increases, but that this effect is driven primarily by the ribosomal genes. We also show that transferability declines as the divergence (% amino acid difference) between donor and recipient orthologs increases and that this effect of divergence increases with connectivity. We explain how these results, even the stronger effect of connectivity on the transferability of ribosomal compared to non-ribosomal genes, provide strong support for both the Balance and Complexity Hypotheses. Significance StatementComparisons between prokaryotic genomes consistently show that genes with informational functions, e.g. in genome replication, transcription, and translation, have been subject to horizontal gene transfer between species more often than genes with operational functions, e.g. in metabolism and environmental sensing. In this study, we use a genome wide analysis of transferability data obtained from 74 genomes to provide the first experimental evidence that this pattern results from differences between informational and operational genes in the number of other proteins with which they interact, i.e., their connectivity, rather than from their functional differences. The importance of our exceptionally large dataset to the detection of connectivity effects on transferability explains why past experimental studies failed to replicate the consistent finding from comparative genomic studies.