Fig. 4

The mutational distance from a 100-base random sequence to a promoter. a The number of mutations needed to transform a random sequence to a functional promoter. The experimental results from the evolution library (n = 40) are shown in blue, zero represents random sequences that were active already before any adaptation and ≥2 represent random sequences that require two or more mutation in which we include all the library strains that could not evolve expression via mutations in the random sequences. Computationally generated sequences (n = 30,000) were scanned for their resemblance to a canonical promoter and were mutated in silico until they pass our criterion for a promoter. We used two different criteria: the ability to capture the most important bases in the −10 and −35 promoter motifs TTGnnn and TAnnnT, with a valid spacer (light green), and the ability to score as the median score of E. coli WT constitutive promoters, according to a position-specific weight matrix score (olive). Both criteria yield similar results to those of our experimental library (error bars represent s.d.). b Comparing promoter scores according to a position-specific weight matrix. Upper panel shows a histogram of scores for E. coli constitutive promoters (n = 556). Lower panel shows a histogram of scores for random generated sequences (n = 30,000), before in silico evolution (olive), and after the first mutation selected (orange). The overlap between the scores of the constitutive promoters to those of the random sequences (before evolution) suggests for our experimental observation of the fraction of random sequences that are already active promoters. The overlap between the scores of the constitutive promoters to those of mutated random sequences strengthen our experimental result that a random sequence of ~100 bases is typically one mutation away from functioning as a promoter. Data from these histograms were the base for the data shown in sub- figure A (olive bars)