Supercoiling-mediated feedback rapidly couples and tunes transcription​ in mammalian cells*

Cells coordinate complex behaviors through precise spatiotemporal control of gene expression and a myriad of interwoven control circuits. Gene- and cell-based therapies require synthetic circuits to similarly coordinate dynamic patterns of gene expression across a large population of cells. However, stable, robust expression of synthetic circuits in mammalian cells is challenging due to significant extrinsic and intrinsic noise. In particular, the stochastic nature of transcription makes coordinating expression across multiple genetic elements challenging. We hypothesized that mechanical sources of gene regulation may be a powerful force driving expression changes at the kilobase scale that we could harness through design. Positive and negative supercoiling induced by the motion of RNA polymerases provides a major mechanical force in the genome.

We developed a model of the underlying biophysical interactions that drive RNA polymerase activity—including the motion of polymerases acting under the influence of supercoiling and the supercoiling-dependent initiation process. Applying the resulting stochastic-differential equation model to the behavior of simple genetic circuits, we identify that supercoiling-mediated feedback drives differences in gene expression. We find that syntax—the relative order and orientation of genes—defines the coupling of transcription rates between the two genes. Moreover, the surrounding boundary conditions, intergene spacing, and promoter strength of genes affect the expression profiles.

In order to validate our predictions, we analyzed the behavior of colocalized, integrated reporter circuits in both HEK293T cells and in primary mouse embryonic fibroblasts. We observe syntax-specific profiles of expression that match predictions. Using these results, we designed an improved "all-in-one" inducible construct. By harnessing positive biophysical feedback, this all-in-one system substantially outperformed alternative designs in primary cells. In sum, our model offers design rules for constructing synthetic gene circuits and provides explanatory power for the physical arrangement of genes within native circuits. With these tools, we aim to design robust sensors and actuators of cell state.