Queueing Up for Translation

Regulation of gene expression is a central component of the cellular response to changing conditions. Typically, changes in external conditions lead to differential expression of key regulators which, in turn, alters the expression of downstream targets. While direct mechanisms of regulation have been the focus of many studies, current research has highlighted the importance of indirect mechanisms based on crosstalk arising from competition for shared resources.

A representative example is the recent discovery that the expression of competing endogenous RNAs can be used to modulate microRNA-based regulation of key targets such as the tumor suppressor PTEN (1–3). In this case, competition for binding to a shared set of microRNAs leads to crosstalk between RNA molecules with important implications for cancer. These experiments have motivated the development of quantitative biophysical models for interactions between competing endogenous RNAs based on competition for binding to microRNAs (4,5). However, these models are based on chemical rate equations which do not take into account the intrinsic stochasticity of gene expression. The development of analytical approaches that integrate stochastic gene expression with competition for limited resources is thus an important development that can lead to new insights into crosstalk-based regulation of gene expression.

In this issue, Mather et al. (6) analyze stochastic models for crosstalk between two mRNA transcripts induced by competition for a shared set of ribosomes. While the focus is thus on competition for the translational machinery, the analytical approaches developed are applicable to a broader range of problems. In their analysis, the authors use concepts from queueing theory, an approach that is increasingly being used to obtain new results and insights into stochastic models of cellular processes (7–9). Queueing theory traditionally involves the study of systems with waiting lines for customers who arrive to receive service, thus it is of interest to consider the connection to models of gene expression. In this mapping, the creation of cellular macromolecules such as mRNAs/proteins is the analog of the arrival of customers in a queueing network, whereas the departure of customers from the queue upon receiving service corresponds to the degradation of mRNAs/proteins. Besides the application of mathematical techniques originally developed for queueing models, the mapping is useful because it provides a qualitative framework for analyzing the problem. For example, the mapping to queueing models suggests that the system shows qualitatively different behaviors as it transitions from an underloaded regime (wherein ribosomes are in excess of mRNAs) to the overloaded regime (wherein mRNAs are in excess of ribosomes).

The approach taken by Mather et al. (6) is to first consider a simplified version of the model (not including mRNA and ribosome fluctuations) which is analytically tractable. Using this, they obtain exact analytical expressions for key quantities such as the moments and covariance between protein levels. These expressions provide insights into signatures of indirect repression between the two protein species based on observations of means and variances. They proceed to consider more general models (which include slow mRNA and ribosome fluctuations) using both analytical approaches (with approximations) and simulation to demonstrate that the essential features seen in the simplified model are also observed in the general model. Not surprisingly, they find that significant crosstalk between the mRNAs occurs in the overloaded regime. Furthermore, they find evidence for an extremum in the correlation between protein levels (which they term negative correlation resonance) as the system transitions to the overloaded regime. Interestingly, in previous work by the authors, a similar (albeit opposite in sign) correlation resonance was observed in systems with competition for posttranslational processing. It will be of interest in future studies to explore whether such a correlation resonance is a universal feature of stochastic cellular systems transitioning from the underloaded to overloaded regimes.

The approaches developed by Mather et al. (6) will be useful for future studies in diverse systems involving crosstalk between network components. Future extensions to larger networks and the inclusion of feedback effects can provide new insights into how competition for shared resources can be used as a mechanism for global regulation. The results also have significance for the field of synthetic biology given the need for avoidance of crosstalk while engineering genetic circuits. Finally, the connection between models of gene expression and systems considered in queueing theory can potentially lead to biologically motivated developments and novel results in both fields.