No support for the adaptive hypothesis of lagging-strand encoding in bacterial genomes

Genes are preferentially encoded on the leading instead of the lagging strand of DNA replication in most bacterial genomes1. This bias likely results from selection against lagging-strand encoding, which can cause head-on collisions between DNA polymerases and RNA polymerases that induce transcriptional abortion, replication delay, and possibly mutagenesis1. But there are still genes encoded on the lagging strand, an observation that has been explained by a balance between deleterious mutations bringing genes from the leading to the lagging strand and purifying selection purging such mutations2. This mutation-selection balance hypothesis predicts that the probability that a gene is encoded on the lagging strand decreases with the detriment of its lagging-strand encoding relative to leading-strand encoding, explaining why highly expressed genes and essential genes are underrepresented on the lagging strand3,4. In a recent study, Merrikh and Merrikh proposed that the observed lagging-strand encoding is adaptive instead of detrimental, due to beneficial mutations brought by the potentially increased mutagenesis resulting from head-on collisions5. They reported empirical observations from comparative genomics that were purported to support their hypothesis5. Here we point out methodological flaws and errors in their analyses and logical problems of their interpretation. Our reanalysis of their data finds no evidence for the adaptive hypothesis.