A cognitive framework for understanding cellular behavior
Predictive behavior within microbial genetic networks
Microbial ecologies are highly complex and dynamic with fluctuations along many physical, chemical, and biological dimensions. From a homeostatic point of view, we expect microbes to rapidly adapt to individual perturbations by activating processes that bring the intracellular state of the cell towards some optimum. However, such environmental fluctuations may be important not only because of their immediate and direct fitness consequence, but also in the information that they might convey about the overall nature of the environment and its future trajectory. This information can help the organism reduce its uncertainty about the nature of its environment and the direction in which it may change, allowing it, in a statistical sense, to ‘predict’ the future. This requires genetic networks to embody an ‘internal representation’ of the dynamic multi-dimensional structure of their habitat, and to use this model and the incoming stream of sensory information to anticipate the future and carry out appropriate behaviors. As such, microbial physiology may be viewed from this ‘cognitive’ perspective, much as we do for understanding animal behavior.
We have explored the relevance of this cognitive framework through oxygen and temperature perturbations that reflect the correlation of these variables as the bacterium E. coli transitions back and forth between the outside world and the inside of the mammalian gastrointestinal tract—where the environment is anaerobic and temperature is clamped at 37°C. What we have uncovered is a fascinating phenomenology of correlated behaviors that precisely reflects these ecological transitions (Tagkopoulos et al. Science 2008, 320:1313). High temporal resolution microarray experiments reveal strong genome-wide correlations between a temperature up-shift (25°C to 37°C) and an oxygen down-shift (20% O 2 to 0% O 2 ), reflecting E. coli's transition into the host. More than a thousand genes respond to these perturbations, with some pathways showing paradoxically mal-adaptive behaviors. Strikingly, in the presence of maximal O 2 levels, E. coli shuts down the entire machinery of aerobic respiration when temperature is lowered from 25°C to 37°C. This seemingly mal-adaptive transient response, and many others that we have documented, only makes sense when viewed in the context of the predictive ecological transition model above.
Associative learning and rapid rewiring in genetic networks
Correlated genome-wide behaviors can be thought of as the byproduct of an ‘associative learning’ process which, over geological timescales, captures the essential correlation-structure of the environment. Can these correlations be rewired in the same way that new behaviors are learned in neural systems? We have begun to explore this question through laboratory evolution experiments of E. coli populations exposed to a variety of different correlations between temperature and oxygen. We find that, indeed, there is tremendous plasticity in regulatory networks, and that these correlations can be partially uncoupled under appropriate selections where the transitions in parameters are opposite to the ecologically native structure (Tagkopoulos et al. Science 2008, 320:1313). We aim to characterize the molecular basis of this regulatory rewiring through the precise identification, isolation, and characterization of the underlying genetic changes. In particular, we are interested in the nature of mutations that lead to the observed regulatory network re-wiring? Are they mediated through subtle changes in protein sequences, or do they result from null mutations in regulatory genes? Mutations in cis-regulatory sequences may also be an important contributor to the network plasticity we observe. These may change the interaction-affinity of the site for a transcription factor, create novel sites, or perturb combinatorial interactions of multiple transcription factors. We are utilizing a multi-faceted strategy for both identifying the causal mutations, and characterizing how they perturb network connectivity.
Related publications
Predictive behavior within microbial genetic networks
Science (2008) 320:1313-1317, Epub 2008 May 8
PDF
Tagkopoulos I, Liu Y, Tavazoie S