MIT's bacterial transistors form printed living circuit boards

A colony of bacteria the size of a pinprick can now do something a silicon chip does: decide whether a signal should pass. Researchers at the Massachusetts Institute of Technology have engineered living cells to behave like transistors and then printed them onto a slab of nutrient gel, where they form what the team calls "living circuit boards." The work, published in Nature Chemical Biology on 17 August 2026, moves computation out of a single engineered cell and into the geometry of a bacterial population.

The idea starts with a familiar component. An electronic transistor is a valve: one input arrives, another determines whether it gets through, and an output leaves. Christopher Voigt, head of MIT's Department of Biological Engineering and the study's senior author, and lead author Hamid Doosthosseini, an MIT postdoc, rebuilt that valve using chemistry instead of current.

Their chassis is Pantoea agglomerans, a Gram-negative bacterium that naturally lives on plant leaves and roots. From it they built two "transistor" strains and three "relay" strains — five in total. Both transistor strains read a control molecule called OC6, but they respond in opposite ways: one switches on when OC6 is present, the other switches off. Each also senses a second molecule, OC12, which carries the actual data being processed. When the right combination arrives, the colony releases a third molecule, OHC14, as its output.

Left alone, OHC14 would simply diffuse in every direction and blur the computation. That is where the three relay strains earn their name. Each picks up OHC14 from one colony and converts it into a different chemical that the next transistor can read. The conversion gives the signal a defined direction, much like a wire routed between two components on a printed board.

To assemble the boards, the researchers used an acoustic liquid handler — a machine that ejects tiny droplets using sound energy rather than a physical pipette tip — to print colonies of the five strains onto agar. Colonies that need to talk sit about five millimetres apart, close enough for a signal to reach the neighbour but far enough that it does not spread indiscriminately. Change the printed pattern and you change the computation; the same five strains are reused, no genetic redesign required. The arrangement mirrors pass-transistor logic, where switches conditionally pass a signal along a chain.

The toolkit is small but versatile. The team demonstrated an OR gate, an IMPLY gate, multi-input operations, a demultiplexer that routes one signal to a chosen destination, and adders that combine two or three inputs. Their largest working circuit was a two-input adder built from 24 bacterial colonies wired together through chemical messages.

Why print computation across many colonies instead of packing it into one cell? That is the study's central argument. Conventional synthetic biology crams every sensor, gate, and output into a single engineered cell, but the available pool of transcription factors is finite, and each added circuit risks crosstalk with the others while draining the cell's protein-making machinery. By splitting the job into specialised colonies, the team shifts the bottleneck from genetic programming to physical layout. Complexity grows by adding dots on a plate, not by rewriting DNA.

The obvious cost is speed. One calculation takes roughly eight hours, because populations must grow, express proteins, emit signals, and wait for them to diffuse to neighbours. Voigt frames that as acceptable on a biological clock. "We're not trying to replace computers, but rather put computational control into biology," he told MIT News, adding that a simple overnight calculation is fast enough "relative to a growth season." He also argued that, in principle, "there's nothing that your iPhone can do that these circuits couldn't do" — a claim about expressive capability, not about competing with a phone on speed or scale.

The intended arena is agriculture. Coating a plant's roots or leaves with these circuits could give the organism its own sensing-and-response layer: for example, a colony that produces a fungicide only when it detects both a pathogen marker and a stress signal from the plant. Because the host bacterium already lives on plants, the platform fits that use case better than a laboratory curiosity.

The honest limits are real. The system is slow, demands carefully controlled growth conditions, and is bounded by the stability and signalling range of living cells. Nothing has yet been shown to work outside a dish. Doosthosseini notes the five strains are enough, in principle, to build any operation, but "in principle" and "in a Petri dish" are not the same as "in a field."

Analysis

The real trade here is not silicon versus biology; it is speed versus actuation. An electronic processor can calculate in nanoseconds but it cannot, by itself, secrete a fungicide or sense a molecule on a leaf. These bacterial circuits are sluggish precisely because they are made of the same stuff they act on: they grow, signal, and respond using the chemistry of life. That slowness is the price of being physically embedded in the environment they compute about. The valuable question is not whether they beat a chip at arithmetic, but where a computation that is itself alive can do something a chip wired to a sensor never could. The answer seems to be wherever the output needs to be a biological action in a biological place.