Links to Other ASM Pages:



• Table of Contents
• ASM News Issues

Strengthening Regulon Response Termed "Learning" in E. coli

Although not the brightest bulb in the class, Escherichia coli can learn, and this studious form of microbial behavior might even resemble a piece of a neural net analog, according to Jan Tommassen and Sally Hoffer of the University of Utrecht in the Netherlands and their collaborators.

"The demonstration of this `learning' behavior is the first evidence for a neural network characteristic in a prokaryotic cell," Tommassen says, referring to findings his group published in the Journal of Bacteriology (83:4914-4917). "Many bacteria have a remarkable capacity to adapt to fluctuating environmental conditions. This adaptation often proceeds via two-component regulatory systems, which usually consist of a sensor in the cytoplasmic membrane and a response regulator in the cytoplasm." When the sensor detects an appropriate signal, it activates the response regulator, which triggers transcription of the relevant genes.

In the Dutch group's experiments, cells of E. coli were exposed to a medium in which phosphate levels were very low. PhoR, a sensor in the cytoplasmic membrane of the cells, registers low phosphate. It signals the response regulator, PhoB, to induce the Pho regulon, which encompasses genes that encode for proteins including "transporters and enzymes that function to scavenge traces of inorganic phosphate and phosphorylated compounds from the extracellular medium," Tommassen says.

Expression of both sensor and response regulator genes rises during phosphate starvation, and in contrast to other regulatory proteins, the proteins that they encode both appear to be stable, which is, the investigators think, key to the "learning" that the cells do. Thus, in subsequent cycles, increased expression of the relevant proteins results in faster induction of the Pho regulon the next time the cells are starved for phosphate, Tommassen explains. This strengthened response represents what Tommassen characterizes as learning in E. coli.

E. coli and other comparable bacteria have many more two-component regulatory systems. "The genome sequence has revealed that E. coli contains about 30 two-component systems, which display considerable mutual sequence homology among the sensors and among the response regulators," says Tommassen. These sensory response systems probably work together to coordinate responses to a multitude of environmental conditions through cross-regulation among the systems, and this crosstalk is what would constitute the functional equivalent of a neural net, according to Tommassen. "The response regulators might act as logical operators at the crosstalk points, integrating different input signals into an appropriate output."

Several studies suggest that cross-regulation occurs. "For example, the Pho regulon can be expressed in the absence of the sensor PhoR by cross-regulation via CreC," Tommassen says, referring to another regulon whose sensor ordinarily responds to carbon instead of phosphate. Additionally, Masahiro Matsubara of Nagoya University in Nagoya, Japan, demonstrated that porin expression, which is regulated by the EnvZ-OmpR two-component system and depends on the osmolarity of the growth medium, is tuned under anaerobic growth by cross-phosphorelay through the ArcB anaerosensor.

This is not the first suggestion that bacteria can perform feats which require learning or memory. Daniel Koshland of the University of California, Berkeley, and others showed that E. coli can sense a gradient of a chemical attractant by sensing a change in the concentration of the attractant over time, a change which they respond to by directing their formerly random swimming in the direction of greater concentration. That requires a memory, brief as it may be. And Gerald Hazelbauer of the University of Missouri-Columbia showed how that memory may work, also in E. coli. He found that modification of receptor sites for galactose provides information about galactose levels in the medium from a few seconds in the past.

Whether a neural net analog really exists in E. coli depends in large part on whether other two-component systems exhibit similar "learning" responses to environmental signals, according to Donald Pettigrew of Texas A&M University in College Station. The big question, he says, is whether "this is the first case of a general phenomenon previously unknown, or some unique system that has some cool properties that we need to try to understand." To Pettigrew, the most surprising result is the stability of the proteins encoded by PhoR and PhoB. "I'm of the school of thought that these things don't last very long in these regulatory systems."

David Holzman
David Holzman writes from Lexington, Mass.