How bacteria avoid using bad words

Bacteria use restriction-modification systems to cleave foreign DNA at specific target sites, while protecting these same sites in their own genomes using methylation marks. Despite the protection afforded by methylation, restriction targets – short DNA ‘words’ recognized by a restriction enzyme – are significantly avoided in bacterial genomes. We hypothesize that such avoidance is the result of selective pressures due to rare events of bacterial DNA cleavage. Our theory demonstrates that spurious DNA cleavage, even at extremely low rates, provides sufficient evolutionary pressure to eliminate target sites. Furthermore, we show that genomes can evolve to avoid words that are similar to the true targets, but which are not directly recognized by restriction enzymes.

HFSP Young Investigator Grant holder Edo Kussell and colleagues
authored on Fri, 08 June 2012

Restriction-modification systems have the potential to exert global selective pressures on bacterial genome sequences.  As such, their presence could significantly alter the constraints on genome evolution.  Since most bacterial strains harbor one or more restriction-modification systems, elucidating the pressures they exert on genomes will provide a key piece in our broader understanding of bacterial evolution.

While the biochemical functioning of restriction-modification systems is well-characterized, their biological functions remain poorly understood.  The restriction enzyme scans DNA for a specific target word, and cleaves it if the target is found. The modification enzyme methylates the same DNA targets rendering the genome resistant to restriction cleavage. While restriction-modification systems can in certain cases protect bacteria against phage infection, they are also thought to function as purely ‘selfish’ DNA elements. Regardless of their evolutionary role, however, their presence can have severe consequences for genome evolution.

For example, consider what could happen if a methylation enzyme occasionally fails to methylate a target site. Bacterial cells would then pay a heavy price for harboring the restriction enzyme, since the bacterial DNA will be cleaved. We simulated populations of DNA sequences in the presence of an imperfect methylation enzyme. The fitness of a single sequence became exponentially smaller with the number of target words it contained. As sequences devoid of targets were amplified in the population, the average target count was reduced to extremely low levels depending on how error-prone, or ‘noisy’, the restriction-modification system was.

Interestingly, the frequency of near-perfect targets – words that have one base-pair mismatch to the target – was also consistently reduced by approximately ten percent of its background level. Instead of being directly selected against, these near-perfect words were avoided because they could easily mutate into perfect targets.  This evolved property of sequences is known as mutational robustness.

We obtained the analytical expression for the population’s equilibrium sequence distribution. Our theory allows the cost of carrying restriction-modification systems to be estimated. We showed that relative frequencies for the perfect and near-perfect targets scale with the ratio between the strength of restriction (or the methylation failure rate) and the mutation rate in a power-law fashion. For systems with very low failure rate – as low as one order of magnitude larger than the mutation rate – avoidance is already detectable. The theory applies more generally, in cases when specific words are associated with fitness costs.  These results enable much further analysis of real genomes to infer the nature of global pressures they have experienced during evolution.


Evolutionary Dynamics of Restriction Site Avoidance. Long Qian and Edo Kussell. Phys. Rev. Lett. (2012) 108(15): 158105.

Pubmed link