Learning is often thought to require a brain. But learning is a broad concept that does not necessarily depend on neurons.1,2
If an organism uses information from past experiences to shape its future decisions, it is also learning.3
Research from my lab, published in the journal PRX Life, shows that even a single bacterium can learn from experience, store memories of the past and use those memories to prepare for the future.4
Keeping Track of Nutrients
Bacteria live in environments that change constantly and on many different timescales.5 In the human gut, for example, nutrient levels go up and down, temperatures shift and antibiotic threats come and go.
To survive, a bacterium has to respond quickly to what is happening right now while still preserving useful information about what it recently experienced. Adapting too quickly leaves the bacterium vulnerable to changing conditions, while forgetting too readily makes it unable to anticipate a recurring threat.
How does a bacterium manage this balancing act? This question interested me as a computational biophysicist who studies how living systems process information and adapt to changing environments.
To investigate whether single-celled organisms such as bacteria can learn from past experience, my colleagues and I used what’s called a microfluidic device to track the behavior of tens of thousands of individual E. coli cells as we switched their nutrient supply on and off at different rates.6
We found that bacteria not only react to current nutrient levels in their environment, they also keep track of their nutrient history to cope with changing conditions.4
If the bacteria were simply reacting to their present environment, they would respond to a sudden pulse of food in exactly the same way, regardless of whether their previous environment was stable or rapidly fluctuating. Instead, when exposed to the same influx of nutrients, bacteria that had just experienced a feast-and-famine environment adapted much faster than bacteria coming from a stable environment.
Because the immediate conditions were identical for both bacteria, we reasoned that the difference in their behavior must originate from a stored internal record of their past rather than a simple reaction to their present.
In other words, the bacteria’s past experiences were shaping their present behavior – they were learning.
Where Bacteria Store Memory
But where were these cells storing this information?
To better understand how bacteria were responding to changes in their nutrient environment, my team and I constructed a mathematical model of the internal molecular network that controls bacterial growth.4 Our model not only reproduced how bacteria behaved under different nutrient environments, but also revealed where their memory likely resides.
Our model pointed to a component of cells called ribosomes as a potential memory storage site. Ribosomes are the molecular factories of cells, building proteins and setting how fast a cell grows.7 Because bacterial memory revealed itself to us as changes in growth rate, our model located the source of this memory within the factories that determine growth. No other component we tested in the model reproduced this behavior.
Some ribosomes responded quickly to nutrient changes, while others changed more slowly. We reasoned that the fast responders track what’s happening in the present, while the slow responders retain traces of the past. Together, they give the cell a memory that spans many timescales, from minutes to hours.
Bacteria and AI Systems
Using our mathematical model, we found that this molecular system follows the same basic computational logic as a gated recurrent neural network: a type of artificial intelligence used to process sequences such as speech and sensor data.8
Key to this process is what computer scientists call a gate – that is, a molecular switch that decides how much of an existing memory to keep and how much to overwrite when new information arrives. Cells hold onto memory at a cost, because maintaining readiness to adapt comes at the expense of growth. If a bacteria is spending its energy adapting to its current environment, it has less resources to spend on growth. Bacteria use molecular gates to actively tune how much to retain and how much to discard.
Modern AI systems rely on gates to remember important details while staying flexible enough to learn something new.9 In the bacterium, the gate is not built from software but from the chemistry within the cell.
This strategy also helps solve a problem familiar to AI researchers: how to learn something new without erasing what was learned before.10 Bacterial cells balance memory and flexibility by storing information across many timescales. This kind of history-dependent computation does not require neurons. Instead, our study showed that a network of ribosome populations is sufficient.
Long before people built artificial intelligence, bacteria had already evolved a remarkably similar strategy for using the past to prepare for the future.
Using Bacterial Memory
Using the broadest definition of learning, our findings show that memory and learning can emerge from chemical networks in a cell, with no brain or neurons of any kind.
Our work also reveals an unexpected connection between biology and artificial intelligence. For biology, our mathematical model provides a precise, quantitative language for describing how cells process information and make decisions. For artificial intelligence, it offers a biological blueprint for building systems that can learn continuously under tight energy constraints.
These findings could also have implications for medicine. Many pathogenic bacteria survive by adapting to changing conditions inside the body.11 If this adaptability depends on cellular memory, then future drugs to treat infection could disrupt the molecular components that allow cells to store and use information about their environment.
Shiladitya Banerjee, Associate Professor of Physics, Georgia Institute of Technology
This article is republished from The Conversation under a Creative Commons license. Read the original article.
- Gershman SJ, et al. Reconsidering the evidence for learning in single cells. eLife 2021;10:e61907.
- Rajan DH, et al. Molecular pathways for learning in the single-cell Stentor coeruleus. Curr Biol. 2026;36(9):2367-2381.e9.
- De Houwer J, et al. What is learning? On the nature and merits of a functional definition of learning. Psychon Bull Rev. 2013;20:631-642.
- Kratz JC, et al. Multi-timescale adaptation and emergent learning in single bacterial cells. PRX Life. 2026;4:023015.
- Nguyen J, et al. Environmental fluctuations and their effects on microbial communities, populations and individuals. FEMS Microbiology Reviews. 2021;45(4):fuaa068.
- Thiermann R, et al. Tools and methods for high-throughput single-cell imaging with the mother machine. eLife. 2024;12:RP88463.
- Scott M, et al. Interdependence of cell growth and gene expression: Origins and consequences. Science 2010;330:1099-1102.
- Chung J, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv. 2014.
- Gers FA, et al. Learning to forget: Continual prediction with LSTM. Neural Comput. 2000;12(10):2451-2471.
- McCloskey M, et al. Catastrophic interference in connectionist networks: The sequential learning problem. Psychol Learn Motiv. 1989;24:109-165.
- Cotten KL, Davis KM. Bacterial heterogeneity and antibiotic persistence: Bacterial mechanisms utilized in the host environment. Microbiol Mol Biol Rev. 2023;87:e00174-22.
