We set out to develop a generic technology for evolving the chemical constitution of microbial populations by using the simplest possible algorithm. Extant living cells polymerize a restricted set of nucleic acid precursors, namely, four nucleoside triphosphates (UTP, CTP, ATP, GTP) and four deoxynucleoside triphosphates (dTTP, dCTP, dATP, dGTP). Synthetic analogues, such as 5-halogenopyrimidines, 7-deazapurines, and 8-azapurines, are known to partially replace canonical bases in cellular RNA and DNA, yet were never demonstrated to sustain unlimited self-reproduction of an organism through complete genome or transcriptome substitution. A hamster cell line serially adapted to grow in the presence of bromodeoxyuridine, while dTMP synthesis was inhibited with aminopterin, has been reported to harbor DNA highly enriched in bromouracil over thymine. However, the significance of these findings could not be ascertained owing to the absence of a direct physical measurement of the base composition of the DNA and the absence of an assay of thymidylate biosynthesis, as well as the likely presence of metabolic components, such as nucleotides in the complex growth medium of the cells. Only certain DNA viruses are known to have undergone full transliteration of a canonical base through the biosynthesis of a noncanonical nucleoside triphosphate, for example, hydroxymethylcytosine in the T4 bacteriophage, presumably to counteract the restriction enzymes of their bacterial hosts. When Weiss and coworkers attempted to substitute thymine in the DNA of Escherichia coli with uracil, over 90% replacement was reached, but further growth was prevented. Genome-scale transliteration has apparently not evolved in any known living cell, possibly owing to a chemical barrier that natural biodiversity cannot overcome. Our experimental plan consisted of the combination of tight metabolic selection with the long-term automated cultivation of fast-growing asexual bacterial populations to change a canonical DNA base for a chemical ersatz. The cultivation setup was elaborated from the GM3 fluidic format (Figure 1), which features the cyclic transfer of the culture between twin growth chambers that alternately undergo sterilization. This cycle ensures that no internal surface of the device is spared from transient periodic cleansing with a sterilizing agent (5m sodium hydroxide), and therefore that no cultivated variant can escape dilution and selection for faster growth through the formation of biofilms. The active elimination of biofilms (wall growth) has proved critical for reprogramming and improving the metabolism of microbial populations. The GM3 cultivation device was connected to two nutrient reservoirs of different composition: a relaxing medium R that contains the canonical nutrient and a stressing medium S that contains the ersatz nutrient. Liquid pulses of defined volume are sent at regular intervals of time from these reservoirs to the culture, which is kept at a constant volume. Depending upon the state of the adapting cells, as measured by turbidity recording of the population density, the culture periodically receives a pulse of fixed volume of either medium R (if the population density falls below a fixed threshold) or medium S (if the density is higher than or equal to the threshold). Successive pulses thus renew the culture at a fixed dilution rate with a nutrient-medium flow whose composition varies with respect to the growth response of the population in such a way that the lowest tolerable concentration of canonical nutrient is automatically maintained over passing generations. We designate this mode of operation as the conditional pulse-feed regime. It qualifies as a simplified and generalized version of a method pioneered by Oliver. Mutations that confer a lower requirement for the canonical nutrient or a higher survival rate under starvation are expected to accumulate in the genome of the adapting population. No attempt was made to implement a finer regulation of differential nutrient supply than the coarse-grained control by medium-switch pulse feed described above. We thus relied on the robustness of biochemical machineries and their evolution to dampen oscillations. [*] Dr. P. Marli re Heurisko USA Inc., Delaware (USA)