(A) We created an initial library of approximately 106 unique synthetic promoters by cloning random nucleotide sequences, of approximately 100–150 base pairs (bp) in length, upstream of a strong ribosomal binding site followed by an open reading frame for GFP, as used to quantify the expression of native E. coli promoters (Zaslaver et al., 2006), and transformed this library into a population of cells (‘Materials and methods’). We evolved populations of synthetic promoters by performing five rounds of selection and mutation on this library. In each round we used fluorescence activated cell sorting (FACS) to select 2 × 105 cells that lie within a gate comprising the 5% of the population closest in fluorescence to a given target level. Next, plasmids were isolated from the selected cells and PCR mutagenesis was used to introduce new genetic variation into the promoters. We then re-cloned the mutated promoters into fresh plasmids and transformed them into a fresh population of cells. We performed this evolutionary scheme on three replicate populations in which we selected for a target expression level equal to the median expression level (50th percentile) of all native E. coli promoters and three replicate populations in which we selected for a target expression level at the 97.5th percentile of all native promoters (referred to here as medium and high expression levels, respectively). (B) Changes in the fluorescence distribution for one evolutionary run selecting for medium target expression (top) and one evolutionary run selecting for high target expression (bottom). The curves show the population's expression distributions before selection, with the numbers above each curve indicating the selection round. The colored bars at the top indicate the FACS gates that were used to select cells from the populations at each corresponding round. (C) Examples of fluorescence distributions for individual clones obtained after five rounds of evolution. Microscopy pictures of two individual clonal promoter populations are shown as insets. (D) For each native E. coli promoter (blue) and synthetic promoter (red), the mean (x-axis) and variance (y-axis) of log-fluorescence intensities across cells were measured using flow cytometry. Fluorescence values are expressed in units of number of GFP molecules. The green curve shows the theoretically predicted minimal variance as a function of mean expression (Appendix 1). The insets show the log-fluorescence distributions for two example promoters (corresponding to the larger dark blue and light blue dots). (E) Cumulative distributions of excess noise levels of native (blue) and synthetic (red) promoters.