Fig 1.
Principle and construction of a Riboverlap.
(a) Standard case: loss-of-function mutations in the synthetic circuits are selected, because they alleviate the cost of the system. (b) The synthetic gene overlaps with an essential gene: loss-of-function mutations are discarded by natural selection because they induce a lethal cost. (c) Rational design of a Riboverlap: a new reading frame is created within the gene to protect (blue). An essential gene (green) is cloned downstream, within this reading frame. The CDS of this essential gene is entirely downstream of the pre-existing gene, however the translation initiation motif and thus the beginning of the reading frame lies within the pre-existing gene. The overlapping part (orange) is fused on the 5’ end of the essential protein. (d) Examples of synonymous changes made by the algorithm.
Fig 2.
Theoretical protection from mutations conferred by overlapping reading frames.
Simulated protection conferred by a gene overlap, depending on the size of the overlap and the fraction of frameshift mutations. The protection is the fraction of loss-of-function mutations in the costly gene that are purged due to the pleiotropic cost induced by the overlap. A true protein overlap (dashed lines) would also protect from non-polar mutations. The average deleteriousness of each amino acid substitutions was Pe = 0.1 (the effect of other values of Pe as well as of alternative models of protein loss-of-function are shown in S2 Fig).
Fig 3.
Large overlapping reading frames can be created in many genes.
(a) Potential Riboverlaps in E. coli: cumulative distribution of the first position at which an overlapping reading frames can be created, for all genes of E. coli MG1655. Different colors represent different levels of stringency: 0 (blue), 1 (green), or 2 (red) amino acid changes allowed. (b) Potential Riboverlaps in other bacterial species: relative size of the largest overlapping frame that can be created. All coding sequences of each organism are averaged, and intermediate level of stringency (1 AA change) was chosen. Error bars represent 95% confidence intervals for all coding sequences of the focal organism. (c) Potential Riboverlaps in all COG (Cluster Of Genes) categories. Pooling the 156,542 coding sequences of 50 bacterial species, average relative size of the largest overlapping frame can be created, for each COG category. Error bars represent 95% confidence intervals for all proteins attributed to the focal functional category.
Fig 4.
Synthesized construct and experimental design.
(a) Synthetic construct: overlapping reading frame between galK, which confers sensitivity to DOG, and kanR, which confers resistance to kanamycin. The expression of both genes is controlled by the pLac promoter: in absence of the IPTG inducer, the operon is repressed by lacIq. (b) galK metabolises DOG into a toxic compound (see S4 Fig for more details). (c) Experimental design: random mutations appear during growth in absence of selection and are subsequently scored by plating on selective medium. This design permits to test whether purifying selection applied on kanR (presence of kanamcyin) protects galK from loss-of-function mutations.
Fig 5.
Experimentally measured protection from mutations and expected resulting increase in circuit lifetime.
(a) The overlap protects galK from loss-of-function mutations. We measure the fraction of galK loss-of-functions (permitting growth on DOG) that did not affect kanR. The lower this fraction, the higher the protection. Each point represents an independent population. The violin plots show the median and kernel density estimation for 12 independent populations. (b) Theoretical median lifetime of the circuit, with the measured protection for the ΔmutS strain, depending on the rate of loss-of-function mutations and on population size. The lifetime is expressed as the number of generations before the first beneficial loss-of-function mutation emerges.