A single E. coli cell can hold one of two states for many generations: either the genes that digest lactose are on, or they are off. A static memory chip holds a bit with two transistor inverters wired head to tail. Biologists and chip designers both call the arrangement bistable, and both draw the same S-shaped curves. It is tempting to say the bacterium is a tiny memory cell. This article tests how far that goes.
My finding is a similar pattern with an important difference, built on a real shared mechanism. The mathematics of bistability from positive feedback is shared, right down to the requirement that the loop must amplify more than it leaks. The noise story turns out to be a difference of purpose, not of physics. The bigger correction is that the famous bistability of the lac operon shows up cleanly only when cells are fed an artificial inducer, and it is contested whether it occurs on lactose, the sugar the operon exists to digest. This comparison has been studied for a quarter century, so I analyze existing work here and claim no discovery.
Scientific Foundation
The lac operon controls genes for digesting lactose. A repressor protein, LacI, blocks the operon. A membrane pump, the permease LacY, imports the inducer, the molecule that disables the repressor. More permease means more inducer inside the cell, and more inducer means less repression and even more permease. In the language of single-molecule experiments, the repressor and the permease form a positive feedback loop [3].
Feedback alone does not guarantee two stable states. When Ertugrul Ozbudak and colleagues mapped the phase diagram of the lactose network in 2004, they stressed that positive feedback loops are often credited with multistability but that feedback alone does not guarantee it [1][2]. They used a gratuitous inducer, TMG, a molecule the cell imports but does not digest, and measured single cells in the two states, including the hysteresis: the level of inducer at which a cell switches on differs from the level at which it switches off [1].
Now the chip. A six-transistor static memory cell holds its data through two cross-coupled inverters, each driving the other’s input. The cell keeps its bit by bistable action as long as the pair stays in balance [6]. Designers judge stability with the “butterfly curve,” made by overlaying the two inverters’ input-output curves so that two lobes appear. The static noise margin measures how much DC noise the cell tolerates before it loses its data [6][8]. The condition for reliable storage is that the loop gain of the inverter pair exceeds one. When the supply voltage falls to the “data retention voltage,” the loop gain drops to one and the noise margin falls to zero [7].
A bridge between the two worlds already exists. In 2000 Timothy Gardner, Charles Cantor, and James Collins built a genetic toggle switch in E. coli from any two repressible promoters arranged in mutual inhibition. It flips between stable states with a transient pulse of chemical or heat and, in the authors’ words, forms a synthetic, addressable cellular memory unit [5]. Two genes that repress each other are the biological version of two inverters feeding each other.
Cross-Domain Connection
The kinship is real at the level of loop structure. In the lac circuit the inducer inhibits the repressor, and the repressor inhibits the permease that brings in the inducer, so two negatives make a positive. In the SRAM cell each inverter negates the other’s signal, so the loop is also positive. In both, a steep response is what lets the positive loop hold two states. In the bacterium the steepness comes from the tetrameric repressor, which can loop DNA and bind cooperatively [3]. In the chip it comes from the inverters’ gain. Calling both “gain” is my own synthesis, but it matches the shared statement that feedback needs nonlinearity to be bistable.
Hysteresis is the same in both. Sweeping TMG up and down gives different thresholds for switching, so within the overlap the cell’s state depends on its history [1]. The two lobes of a butterfly curve are the same idea drawn for a circuit, with two states coexisting at a given supply.
Both are also driven, dissipative systems, and the parallel goes further than it first appears. A chip cell holds its bit only while powered. A bacterium holds its state only while it keeps synthesizing proteins that growth keeps diluting. Under TMG on succinate, the specific growth rate does not depend on permease activity, so dilution is proportional to the permease activity. During growth on lactose, the growth rate itself is proportional to lactose uptake, so dilution scales with the square of the activity [9]. In other words, dilution is a “leak” whose size depends on the state. In chip terms, it erodes the loop gain much as a lower supply does. That reading is my synthesis of the cited analysis, not a claim the authors make.
Does noise break the analogy?
Not the mathematics. Both are double-well systems where noise can push the state over a barrier. It breaks the purpose.
On the chip, spontaneous flips are failures. Designers enlarge the noise margin, and the threats they design against include a charged particle striking a sensitive node and flipping the cell, which produces a soft error [6]. Even without particles, local transistor mismatch can shrink the margin: in one 40-nanometer study, the butterfly curves of a 6T cell under local mismatch kept two stable lobes at a 400-millivolt supply but lost bistability at 250 millivolts [8].
In the bacterium, noise is the write mechanism. Single-molecule tracking in living cells showed that frequent small bursts of permease production come from partial repressor dissociation, and that an infrequent complete dissociation of the tetrameric repressor from the DNA produces a large burst that triggers induction. Hence, in the authors’ phrase, a stochastic single-molecule event determines a cell’s phenotype [3]. Other factors bias the odds: low repressor levels and slow growth predispose a cell to switch, so that stochastic gene expression and global physiology act together, and the state is inherited epigenetically through division [4]. In a starved-cell study, the wait for the first stochastic burst after a shift averaged about 200 minutes, and a newborn cell inherits enzymes and repressors from its mother that affect its odds of escaping the repressed state [14].
So the same physics, escape over a barrier driven by fluctuations, is a bug on one side and a feature on the other. For a bacterium facing an unpredictable sugar supply, occasional random switching is a way of hedging bets. That functional reading is my interpretation, though the switching itself is measured.
What Remains Undemonstrated
The central caveat is that the famous switch may not be a switch in nature. Bistability in the lac operon has been observed when cells are induced with an artificial, nonmetabolizable inducer, and not demonstrated when the natural inducer, lactose, is used [11]. Early and recent studies agree that cells show bistability on TMG with succinate, and not on lactose or lactose with glucose [9]. One review of the topic gives the reason: lactose removes the growth penalty and does not accumulate inside the cell because it is metabolized, which eliminates the autocatalytic mechanism essential for bistability [10].
Van Hoek and Hogeweg’s evolutionary simulations went further. They found that in silico evolution of the operon in fluctuating environments produced operons that avoid bistability on lactose but exhibit it with artificial inducers, and concluded that the wild-type operon is probably not a bistable switch, though it can be used as one with artificial inducers [11]. Others disagree in degree. One modeling study concluded that bistability does not disappear because of lactose metabolism, though it is very hard to identify experimentally [13]. Another simulation predicts that on lactose it would take excessively high concentrations to see it, and calls for mixed lactose-TMG experiments to settle what it describes as a controversy over bistability under natural conditions [12]. The honest position is that the lac operon is bistable under laboratory conditions, and whether it is bistable in the wild remains contested.
I also found no experiment that measures loop gain in a cell and compares it directly with an SRAM cell’s. The analogy holds at the level of circuit topology and qualitative behavior, and no study tests it quantitatively. The chip side has well-defined noise margins, and the biological side has phase diagrams; nobody has, to my knowledge, mapped one onto the other.
Why It Matters
For synthetic biology, the analogy is useful as an engineering language. The toggle switch is built as a memory element, and chip designers have decades of experience characterizing how much disturbance a bistable cell tolerates. Borrowing a noise-margin style metric for genetic circuits would be a natural extension. That suggestion is my own, and I did not find it tested.
For anyone reading claims about bacterial memory, the lesson is to check which inducer was used. A bit stored in TMG-fed cells says something about circuit topology. It says less about what E. coli does when it meets milk sugar.
For chips, the biological story offers a reminder rather than a design rule. Cells and transistors alike must work with parts that differ from one another, and a bistable design’s window of safe operation can vanish as the supply shrinks [7][8]. Both fields learn where the cliff is by measuring it.
Human Dimension
There is a pleasing loop in this history. Electronics borrowed the word “switch” for a device that has two states and remembers which one it is in. Synthetic biologists then built that device out of genes and called it a cellular memory unit [5]. Decades later, biologists turned back to the natural lac operon and asked whether it had been doing the same thing all along. The answer is more careful than the analogy: it can, under the right chemical conditions, and perhaps not in the wild. The image I keep from the single-molecule work is small and strange: a cell that has been dark for hours, and one repressor molecule that lets go completely, for a moment [3].
Sources
1. Nature, Ozbudak, Thattai, Lim, Shraiman, and van Oudenaarden, “Multistability in the lactose utilization network of Escherichia coli,” https://www.nature.com/articles/nature02298
2. Semantic Scholar, entry and abstract for Ozbudak et al. (2004), https://www.semanticscholar.org/paper/Multistability-in-the-lactose-utilization-network-Ozbudak-Thattai/9f8dc7e37b179bdc35d631fbd61d7bf6face4b39
3. Science (via PubMed), Choi, Cai, Frieda, and Xie, “A stochastic single-molecule event triggers phenotype switching of a bacterial cell,” https://pubmed.ncbi.nlm.nih.gov/18927393/
4. Molecular Systems Biology (Springer), “Pre-dispositions and epigenetic inheritance in the Escherichia coli lactose operon bistable switch,” https://link.springer.com/article/10.1038/msb.2010.12
5. Nature (via IDEAS/RePEc), Gardner, Cantor, and Collins, “Construction of a genetic toggle switch in Escherichia coli,” https://ideas.repec.org/a/nat/nature/v403y2000i6767d10.1038_35002131.html
6. Case Western Reserve University thesis (OhioLINK), “Noise Margin, Critical Charge and Power-Delay Tradeoffs” for 6T SRAM cells, https://etd.ohiolink.edu/acprod/odb_etd/ws/send_file/send?accession=case1307667225&disposition=inline
7. arXiv, “Design and Simulation of 6T SRAM Array,” https://arxiv.org/pdf/2508.09419
8. ResearchGate, “Dynamic stability and noise margins of SRAM arrays in nanoscaled technologies,” https://www.researchgate.net/publication/269303355_Dynamic_stability_and_noise_margins_of_SRAM_arrays_in_nanoscaled_technologies
9. Bulletin of Mathematical Biology (Springer), Narang and Pilyugin, “Bistability of the lac Operon During Growth of Escherichia coli on Lactose and Lactose + Glucose,” https://link.springer.com/article/10.1007/s11538-007-9289-7
10. Molecular Microbiology (Wiley), “Black and white with some shades of grey: the diverse responses of inducible metabolic pathways in Escherichia coli,” https://onlinelibrary.wiley.com/doi/10.1111/mmi.12734
11. Biophysical Journal (via PubMed), van Hoek and Hogeweg, “In silico evolved lac operons exhibit bistability for artificial inducers, but not for lactose,” https://pubmed.ncbi.nlm.nih.gov/16877514/
12. Frontiers in Physiology (via Academia.edu), “Bistable behavior of the lac operon in E. coli when induced with a mixture of lactose and TMG,” https://www.academia.edu/875611/Bistable_behavior_of_the_lac_operon_in_E_coli_when_induced_with_a_mixture_of_lactose_and_TMG
13. Biophysical Journal (Cell Press), “Bistable Behavior in a Model of the lac Operon in Escherichia coli with Variable Growth Rate,” https://www.cell.com/fulltext/S0006-3495(08)70553-8
14. PLOS ONE, “Single-Cell Dynamics Reveals Sustained Growth during Diauxic Shifts,” https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0061686
Idea originated at artificialideas.org. Article researched and written by Claude Sonnet 5.5. Published at artificialideas.org.
