PDF, English - main document Download (20MB) | Terms of use

Abstract

Biological signal transduction pathways evolved to reliably transmit information from input signals inducing appropriate cellular responses in the process. Along signaling pathways, information is often successively relayed to several types of transmitter molecules. In some cases, particular transmitter molecules do not only receive one kind of information, but several. To this end, information can for instance refer to the identity or quantity of first messengers. By encoding particular bits of information into specific characteristics of such shared transmitters, the information can be decoded downstream. Here, transmitter characteristics may refer to the absolute level of transmitter molecules, the duration of transmitter activation or, in case of activation pulses, the pulse frequency. In this thesis, I analyzed encoding and decoding in two prime examples of signal transduction: calcium signaling in non-excitable cells and Escherichia coli chemotaxis. For this purpose, I present several methods allowing for a quantitative analysis of information transfer, whereas methods are partly based on measures from the field of information theory.

With regards to calcium signaling, I focused on the frequency-decoding of calcium oscillations by dependent proteins. Particularly, variations in the quantity of input signals can account for modulations of the calcium oscillation frequency. Several proteins like NFAT, NF-κB, CaMKII and calpain were found to be sensitive to such frequency-modulations. To this end, most frequency-decoding proteins exhibit increased activities for fast calcium oscillations and decreased activities for slow oscillations. I refer to this form of frequency-decoding as high-pass activation. In contrast, the transcription factor NFAT was reported to exhibit an optimal frequency for its activation, while slower or faster frequencies only result in a reduced protein activity. In turn, I refer to this form of frequency-decoding as band-pass activation. On the basis of kinetic models, I identified requirements for high-pass and band-pass activation. In more detail, I employed optimization algorithms aiming at a maximization of the high-pass or band-pass activation distinctness. Among other things, I found that antagonistic, oscillatordependent regulation of the decoder was essential for band-pass activation, whereas regulator species had to be differently responsive to upstream calcium oscillations. Further, I defined favorable parameter margins and confirmed reports on the importance of cooperative protein activation for distinct frequency-decoding. Additionally, I employed channel capacity estimates to quantify the discriminability of particular calcium oscillation frequencies in the presence of realistic stochastic fluctuations. For the application of channel capacity estimations and the interpretation of the resultant estimates, I discuss several possible pitfalls.

With regards to Escherichia coli chemotaxis, I focused on the encoding of attractant levels into receptor methylation levels using an established kinetic model. On the basis of results by a collaborateur, encoding was investigated by inferring expected attractant levels from present receptor methylation levels. In addition, I used delayed mutual information estimates to quantify the dynamic processes of memory formation and memory loss. Here, memory formation and memory loss were characterized by targeted transient changes in receptor methylation levels in response to changes in ambient attractant levels. In Escherichia coli chemotaxis, single receptors can be methylated multiple times. By means of the aforementioned methods, I found that, for extreme attractant levels, chemotactic behavior failed due to limitations in the encoding of ambient attractant levels into receptor methylation levels, whereas a reduction of the maximal number of methylations per receptor resulted in severer limitations in the encoding, thus, greater impairments in Escherichia coli chemotaxis.

For both examples of signal transduction, I examined information transmission through molecular communication channels. To this end, the input was the variable to be encoded or decoded and the output was the encoding or decoding variable. Changes in model characteristics, such as the model parameterization or network structure, greatly impacted the number of input signals that could be reliably encoded or decoded. Both example systems distinguished themselves by a pronounced ultrasensitivity of the output variable to changes in the input variable. I found that this ultrasensitivity helped in increasing the discriminability between input signals.