The problem. The toggle switch showed a cell could hold a state. Elowitz and Leibler asked the harder dynamical question: could you make a cell cycle through states on a schedule you designed, with no natural clock to borrow from?

The idea. Three repressors in a ring — LacI represses TetR, TetR represses λ cI, cI represses LacI — a cyclic negative-feedback loop with an odd number of inversions, so it can never settle. On a plasmid in E. coli, with GFP reporting one node, the network oscillates with a period of roughly 150 minutes, longer than the cell-division time. Again the design came first, from a continuous model that said sustained oscillation needs strong promoters, tight repression, and comparable protein/mRNA lifetimes.

Why it matters. Two things. First, it completes the toggle switch’s argument: cells are programmable not just in memory but in time. Second — and more interesting in hindsight — the oscillations are noisy. Period and amplitude vary between sibling cells and drift over generations. That failure to be a clean clock is arguably the paper’s most durable contribution: it made stochasticity in gene expression a first-class engineering problem, the thread Elowitz pulled for the next decade. Anyone who has watched a single-cell dataset fan out into unexpected variance recognizes the lesson.

Verdict. A founding paper that also, quietly, opened a second field (noise biology). The engineered dynamics are real but fragile — an argument for feedback control, later delivered by more robust synthetic oscillators. Post it back-to-back with Gardner as “the two papers that turned cells into circuits.”