Learning without neurons - how a single cell processes information

Cells show a range of complex behaviors which might normally associated with cognition, including basic forms of learning such as habituation. The giant single-celled ciliate Stentor coeruleus is a pond-dwelling organism that attaches to substrates like pond plants and filter feeds by means of a ring of cilia at one end of the cell. Stentor cells can contract in response to mechanical stimulation, an effective escape mechanism but energetically costly. To avoid contracting due to random collision with non-threatening things like algae or pond plants, but still allow contraction when a large predator attacks, Stentor cells use a simple form of learning: when a cell is mechanically stimulated over and over again, it will gradually stop responding, representing a case of habituation. Our goal is to understand how a single cell learns, without a nervous system. We have developed automated devices for applying series of mechanical stimuli under computer control, which we are now using to test predictions of computational models for learning based on known or plausible biochemical mechanisms. Initial experiments showed that learning occurs in a stepwise manner, which suggested a simple two-state Markov model, however, subsequent experiments rule out this model. We next formulated a simplified biochemical model invoking known or suspected processes, which can account for all prior observations as well as new experiments designed to test the model. Based on this framework, we have investigated the molecular pathways involved in learning in Stentor, which suggests a role for CamKII kinase signaling in the learning process.

Speaker: Wallace Marshall, UC San Francisco