In the late 1990s, a young physicist named Michael Elowitz decided to “program” living cells.

A graduate student at Princeton University, Elowitz was spending a great deal of his free time poring over esoteric papers about circadian clocks, the molecular networks that control an organism’s behavior in roughly 24-hour cycles.

“And as I’m reading all of these papers,” he told Asimov Press, “I noticed that many of them concluded with a cartoon model of the biological circuit deduced from the genetic or biochemical measurements in the paper. What I most remember is just feeling like, are these circuit sketches really sufficient to explain the behavior? Or are they just a summary of observed interactions, possibly omitting many other critical components? It was driving me crazy.”

Richard Feynman’s iconic admonition — “What I cannot create, I do not understand” — resonated with Elowitz, who decided to answer his own questions by building a synthetic molecular clock; one not found anywhere in nature. Inspired by his readings, Elowitz began designing an oscillator that would force living cells to flash on and off in a periodic rhythm. However, even despite initial enthusiasm, doubts quickly mounted.

“When I asked people what they thought of the project,” he said, “I got very different answers. A few well-known biologists would say, ‘No, it’ll never work that way. It just won’t work.’ And I’d ask them, ‘Why won’t it work?’ And they’d say, ‘Biology just doesn’t really work that way. You can’t predict what’s going to happen.’”

By 2000, though, Elowitz had proven those doubters wrong. Alongside physicist Stanislas Leibler, he published his findings in Nature; the duo had successfully created a biological oscillator, endowing living cells with an artificial rhythm.That Nature paper appeared back-to-back with another report of a synthetic gene circuit, called the “toggle switch,” developed by Jim Collins, Timothy Gardner, and Charles Cantor at Boston University. Neither group knew about the other’s work,¹but together they launched the field of synthetic biology.

Technologies to “program biology” have come a long way since the repressilator was first introduced twenty-five years ago. Synthetic biologists have recently designed interacting protein clusters that act as neural networks inside living cells, gene circuits that can switch between OR and AND logic gates based on small molecule triggers, and even programmed a community of cells to execute a hashing function widely used in cryptography.²

Many modern gene circuits, though, are incredibly complicated. They are depicted in research papers as dense tangles of arrows, triangles, and other symbols; similar to the diagrams found in electronic engineering textbooks. The repressilator offers a useful starting point to begin parsing this complexity. After all, the same design approach that Elowitz used to build his oscillator still provides the basics for assembling modern gene circuits. By understanding how the repressilator was made and how it really works, anyone can grasp the basic principles upon which synthetic biology was built.

With that in mind, we created an interactive chart that visualizes the repressilator’s dynamics — and allows you to play around with them.

Building a Circuit

Much like electronic circuits, which use wires and transistors to create or process signals, a synthetic gene circuit is a network of biological parts — DNA, RNA, and proteins — that process inputs and generate outputs. A working gene circuit can detect many different inputs, such as a molecule, flash of light, or a physical force, and then translate those signals into everything from making a fluorescent protein to emitting an odor molecule. For an oscillator, the output is simply a pattern of on-off signaling that repeats over time, akin to a blinking light on a circuit board.

Elowitz’s oscillator design was made of just three genes — lacI, tetR, and cI — linked together in a negative feedback loop, reminiscent of a three-ring oscillator in electronics. Each gene encodes a repressor protein that binds to a specific DNA sequence upstream of a different repressor gene, called a promoter, and blocks RNA polymerase from transcribing that gene.³

{{signup}}

In the repressilator, LacI blocks tetR, TetR shuts down cI, and cI represses lacI. When LacI is abundant, it switches off the tetR gene. But LacI eventually degrades or falls off the DNA, allowing tetR to turn on and make TetR proteins that then repress cI. And so on. This cyclical process is ultimately what drives the cell’s oscillations.

To confirm these oscillations, Elowitz added a separate reporter plasmid encoding green fluorescent protein (GFP) to his cells. This reporter uses a DNA sequence that TetR recognizes, so when TetR levels rise, TetR blocks the GFP gene and cells stay dark. When TetR levels drop, the cells flash a brilliant green.

Elowitz inserted his assembled DNA into E. coli and then used a fluorescence microscope to observe the cells in real time. Unfortunately, his microscope lacked autofocus, causing cells to drift in and out of focus. He had to remain at the microscope day and night to adjust it manually before taking each photograph. In one especially grueling stretch, he slept next to the microscope with an alarm clock, waking every hour to refocus the image.

Fortunately, his work paid off. Elowitz’s repressilator caused cells to flash green about every 150 minutes, albeit with variations between cells. The results convinced skeptics that researchers could engineer predictable circuits into living cells.

Modeling a Circuit

Before building the repressilator, Elowitz needed to understand the quantitative values of various parameters that would ensure its oscillations in the cell. He modeled the repressilator using six differential equations, a standard mathematical approach for describing how molecular concentrations change over time. Each of the three genes — lacI, tetR, and cI — makes both mRNA and a corresponding protein, so Elowitz wrote separate equations for both molecules and for each gene.

In real cells, each gene and protein differs in subtle ways. Some repressors bind DNA more tightly than others, for example, or decay more rapidly. Elowitz simplified his mathematical model by assuming that all three genes behave identically and assigning them the same parameter values. This choice allowed him to understand what parameter regimes generally favored oscillation, while neglecting less significant effects due to differences among the three genes.⁴ In total, he used six parameters:

1. Leakiness (μ) measures how much gene expression slips through when a promoter is bound by its repressor. Even when LacI binds to the tetR promoter, for example, a small amount of TetR still gets made (LacI sometimes falls off the DNA, leaving a small gap of time for RNA polymerase to bind and transcribe the gene.) Too much leakiness can ruin an oscillator by preventing it from cycling.
2. Promoter strength (α) indicates how quickly RNA polymerase transcribes mRNA when the repressor is not bound. A “strong” promoter makes more mRNA, which typically leads to more repressor protein. If a promoter is too weak, for example, not enough of a repressor will get made to shut down the next gene in the loop, and so the rhythms stop.
3. Decay Rate Ratio (β) compares how fast proteins degrade relative to their mRNA. Bacterial mRNA usually breaks down in a few minutes, whereas proteins stick around the cell for longer. Researchers can tweak this ratio by engineering proteins to degrade more quickly, such as by fusing them to little peptide tags — called degrons — that signal cellular proteases to break them down. Higher protein turnover rates usually speed up oscillations.
4. The Hill Coefficient (n) measures how sharply transcription flips between off and on once a repressor binds. A value near 1 causes gradual shifts that might not allow enough “overshoot” of protein concentrations to sustain oscillations, while higher values (around 2 or 3) create abrupt transitions that favor such overshooting, leading to more stable cycles.
5. m and ρ refer to the concentrations of mRNA and protein, respectively, for each gene.

Try sliding these parameters around in the interactive diagram to see how each parameter affects the repressilator’s dynamics. Notice how reducing β makes proteins linger longer, thus stretching out each cycle. Boosting α (promoter strength) increases the amount of repressors made, tweaking the amplitude of oscillations. Increasing μ (leakiness) past a certain point deteriorates the oscillations, or, at high enough levels, prevents it from beginning altogether.

Remember that these equations are approximations. This model aims to guide the design of the circuit, rather than to represent all of its molecular interactions in full detail. In reality, repressor proteins bind to DNA in a discrete and probabilistic way, but these equations approximate their behaviors as if the number of molecules in the cell were continuous. Also, RNA polymerase moves in bursts along DNA, rather than at a steady pace. Despite these limitations, these six equations provided the key insights needed to help Elowitz design an actual repressilator by stitching together genes and inserting them into microbes.

Tuning a Circuit

Early versions of the repressilator didn’t create a perfect rhythm in every cell. Elowitz observed oscillations of about 150 minutes, but the period varied from one cell to the next, and only around forty percent of the cells oscillated during any given movie. This variation underscored the immense challenge of coaxing thousands of individual bacterial cells to “tick” in near-unison. It also provoked the question of why a well-defined circuit, designed down to the nucleotide by a human being, behaved so differently in genetically-identical cells.

Over time, synthetic biologists dissected Elowitz’s results and figured out that small differences in repressor strength, protein decay rates, and cellular resources can throw the oscillator off. The cells’ environment matters, too.

Elowitz grew his cells on glucose gel pads. But as the bacteria multiplied, they eventually ate up all the sugar and began swimming in their own waste, thus disrupting the circuit’s rhythm. In 2010, a group of Harvard scientists invented a device called a “mother machine” to solve this problem. Their device traps individual cells in tiny wells, bathing them in fresh nutrients while washing away waste, stabilizing the repressilator’s cycles.

Scientists have also tweaked the repressilator at the genetic level to improve reliability. In 2016, Laurent Potvin-Trottier and colleagues at Harvard University consolidated the three repressor genes and fluorescent reporter onto a single plasmid so that each gene would be produced at comparable levels. Noticing that TetR bound its DNA target more strongly than LacI or cI, they also created a “DNA sponge” — extra binding sites that soak up excess TetR — to bring its binding strength down to a level that matched the others more closely. With these tweaks, Potvin-Trottier shrank the standard deviation of periods between cells to just 14 percent.

After the Potvin-Trottier study was published in Nature, Elowitz and his student, Xiaojing Gao, fired off a response; their letter strikes at the heart of the idea that one can achieve precision in biology using simple gene circuits (bolding our own):

"...in the most precise of Potvin-Trottier and colleagues' circuits, the standard deviation in period length was reduced from 35% of the mean to around 14%, with strikingly uniform pulse shapes and amplitudes observed. This repressilator generates a pulse of fluorescent-protein expression just once every 14 generations. Assuming a cell-cycle time of 1 hour, it would take around 7.5 days, or 180 cell cycles, for a colony of cells to accumulate a standard deviation of half a period of drift. This extraordinary precision means that even a large population of cells can remain in sync. In fact, the authors were able to visualize oscillation dynamics in a test-tube culture … Evidently, precision does not necessarily demand circuit complexity and, in this case, even seems to benefit from minimalism.”

The repressilator remains a foundational example of synthetic biology’s ability to blend mathematics with predictable outcomes. Elowitz showed that one can design gene networks that exist nowhere in nature, predict their behavior using a bit of calculus, and then actually use them to get cells to carry out new behaviors.

Today, the repressilator’s impact extends far beyond making cells pulse rhythmically. Its core ideas — negative feedback, parameter tuning, cooperativity, and molecular noise — continue to guide the design of everything from biological logic gates and memory modules to biosensors and entire metabolic pathways.

And it all started with some mathematical equations and an idea.

{{divider}}

Thanks to Michael Elowitz for reading a draft of this.

Nehal Udyavar is a design engineer who creates explorable explanations of biological systems at Newt Interactive. He shares his learning journey on his Substack.

Niko McCarty is a founding editor of Asimov Press.

Cite: Udyavar N. & McCarty N. “The Making of a Gene Circuit.” Asimov Press (2025). DOI: 10.62211/23ey-67yy

Lead image by Ella Watkins-Dulaney.

Footnotes

1. There were actually four papers that all came out around the same time, but the Elowitz and Collins papers received the most attention by far. William Farmer and James Liao designed and assembled “a regulatory circuit to control gene expression in response to intracellular metabolic states,” published in Nature Biotechnology in May 2000, whereas Attila Becskei and Luis Serrano built an autoregulation gene circuit, published in Nature in June.
2. Specifically, the MD5 hashing function. It was built by distributing 110 logic gates across 65 unique E. coli strains.
3. “Transcribing” a gene simply means that an enzyme called RNA polymerase turns DNA into RNA.
4. A later study by Johan Paulsson’s group showed that a specific aspect of the TetR system — namely, its very tight binding to DNA — made the circuit more “noisy” than it otherwise would have been. So experimentally, differences between the repressors are extremely important.

This article was published on 16 February 2025.