DNA supercoiling, the level of over- or underwinding of the DNA molecule around itself, plays an important role in the global regulation of gene expression in bacteria. Indeed, underwound DNA makes the initiation of gene transcription easier, thereby increasing gene expression, while overwound DNA conversely makes the initiation of gene transcription harder and decreases gene expression. As gene transcription in turns affects the local level of DNA supercoiling, deciphering the nature and extent of the interplay between these two processes is integral to a fine understanding of the regulation of gene expression in bacteria. In this thesis, I study this problem using an _in silico_ experimental evolution approach, by introducing a mathematical and computational model tailored to study the mutual interaction between gene transcription and DNA supercoiling (the _Transcription-Supercoiling Coupling_ or TSC), embedded into an evolutionary simulation. Using this model, I first show that environment-specific gene expression levels based on the supercoiling-mediated interaction between neighboring genes can indeed evolve when the relative position of genes on the genome changes through genomic inversions. In particular, I observe the emergence of relaxation-activated genes, as found in many bacterial species. Then, I show that these expression patterns are generated by gene interactions at multiple scales, ranging from local structures, such as pairs of convergent or divergent genes, up to a densely connected genome-wide regulatory network. The description of such a supercoiling-mediated regulatory network thereby strengthens the hypothesis that supraoperonic gene syntenies, observed in several bacterial species, might have been evolutionarily conserved because of a regulatory function. Finally, I show that mutations in the regulation of DNA supercoiling could play an early role in adaptation to new environments, by locally exploring the fitness landscapes that emerge from the transcription-supercoiling coupling.