One of the major challenges that synthetic biology must face is the intrinsic variability and genetic instability of living organisms . As the complexity of synthetic systems increase, the emergence and selection of new features will become a significant impediment in achieving robust and stable performance. Improving the genetic stability of the host organism, or synthetic biological chassis is therefore a validated goal.
Previously, we have demonstrated that genome stabilization by elimination of mobile genetic elements has advantages in certain cloning applications [18, 20]. To achieve additional genetic stabilization of the host, we targeted and eliminated error-prone DNA polymerases (Pol II, Pol IV, Pol V), major sources of frameshift and point mutations. Possible alternative approaches to lower the mutation rate include the introduction of a so-called antimutator dnaE allele or upstream inactivation of the SOS response by introducing a recA or lexA mutation. Previous studies on the effect of antimutators found a 5 to 30 fold reduction in mutation rate . It was later shown that the mode of action of these DnaE antimutators was a more effective ability to exclude error-prone DNA polymerases at sites of DNA synthesis during DSB-repair associated stress-induced mutagenesis , suggesting that the elimination of these enzymes would reproduce the antimutator effect. Testing the other alternative approaches, MDS42recA, compared to MDS42pdu, showed a much less pronounced reduction in spontaneous mutation rate. Furthermore, owing to the central role of the RecA enzyme in cell physiology , unwanted pleiotropic effects might arise within the cell, manifested, among others, in sensitivity to mitomycin-C, presumably due to insufficient repair activity. MDS42lexA(S119A), carrying a non-cleavable form of the LexA repressor, did not significantly lower the mutation rate under the conditions applied. The small effect of the recA and lexA mutations can be explained by the relative SOS-independence of the error-prone polymerases: the enzymes are present even in unstressed cells, and can be up-regulated by a number of (not just SOS) stress responses (Figure 1) .
Several studies have been made on the effects of error-prone polymerases on mutation rates using various strains of E. coli and various methods of measurement. Supporting our findings is the observation that deletion of dinB significantly decreases the mutation rates for both frameshift and base substitution mutations in a Lac+ reversion system, as well as in a rifampicin resistance assay . In another study, the lack of Pol V caused a decrease in the number of Arg+ growth dependent revertants . Later studies also showed that post-exposure mutation rates in the presence of ciprofloxacin were markedly reduced when all three inducible polymerases were separately eliminated .
Here, using a D-cycloserine resistance-based fluctuation analysis , confirmed in some cases by a rifampicin resistance assay as well, we carefully quantified the effect of individual and combined error-prone polymerase deletions on the mutation rate, under either unstressed or stressed conditions. We determined that, under unstressed conditions, elimination of each error-prone polymerase by itself significantly decreased the spontaneous mutation rate. The effect of combining deletions ΔpolB and ΔdinB was additive, indicating an independent mode of action for these polymerases. However, ΔumuDC generated no additional decrease of the mutation rate when any of the other two error prone polymerases was missing, possibly marking an interaction among the genes or their products. This phenomenon has been described previously regarding ΔdinB and ΔumuDC, the nature of their putative interaction, however is not yet known .
As expected, the most dramatic differences in mutation rate between MDS42 and MDS42pdu were observed under various stress conditions. A sub-inhibitory concentration of the SOS response-activating mitomycin-C, overproduction of either the non-toxic GFP protein or of the highly toxic ORF238 hydrophobic protein all significantly increased the mutation rate of MDS42. The values for MDS42pdu remained stable under the same conditions. It is also noteworthy, that among the strains tested, the commonly used production strain BL21(DE3) showed not just the highest spontaneous mutation rate, but also the highest increase in mutation rates in response to the various stresses. A difference of almost two orders of magnitude was observed between the mutation rate of BL21(DE3) and MDS42pdu when overproducing the toxic ORF238 protein. This elevated rate of mutation in BL21(DE3) can be mostly attributed to an increased rate of IS insertions.
A clear practical advantage of working with MDS42pdu was demonstrated in a protein production experiment, where the SinI methyltransferase was expressed from an inducible plasmid construct. SinI, producing 5-methylcytosines is toxic to cells that carry the McrBC endonuclease. Even in cells lacking McrBC, we observed a negative effect on cell fitness. When SinI was produced, we found that the sinI gene, carried on a plasmid, acquired loss-of-function mutations approximately three times less frequently in MDS42pdu than in MDS42, and over five times less frequently than in BL21(DE3)mcrBC. Remarkably, after only 16 hours of production in BL21(DE3)mcrBC, almost half of all sinI genes encoded on the plasmids had suffered a disabling mutation.
Clearly, the unexpectedly high ratio of mutated clones in the SinI-expressing culture cannot be explained solely by the stress-induced mutagenesis, the overall mutation rate of which being too low in absolute values (in the order of 10-6 mutations/gene/generation) to cause such a dramatic effect. Rather, the phenomenon is in large part due to the growth inhibitory effect of the plasmid carrying the toxic gene. The chain of events is the following: Upon induction of expression of the toxic gene, growth rate of the cell is reduced. At the same time, mutation rate is increased by the stress. Once a mutant, not producing the toxic protein, arises in the plasmid population, the cell harboring it can resume normal growth and become dominant in the culture. In low-mutation-rate MDS42pdu, appearance of such mutants is delayed, and the cells can produce the functional toxic protein for an extended period of time.
Calculating the precise advantage of MDS42pdu over the parental MDS42 or the commonly used production strain BL21(DE3) can be challenging, due to the stochastic nature of mutagenesis, as well as the lack of exact data on the fitness cost of overproduction. Nevertheless, it is clear that the more severe the stress of overproducing a product is (resulting in an elevated mutation rate and growth inhibition), the greater the advantage of the stabilized host.