Under conditions of bistable induction of Escherichia coli lac operon, epigenetic patterns of sublineages of ‘on' and ‘off' cells originate from distinguishable ancestors up to two generations before induction.We found two switching pre-disposing factors, namely low repressor levels and slow growth, demonstrating that stochasticity in gene expression and global physiology synergistically determine the single-cell responses.A quantitative model where growth rate acts through simple dilution of intracellular content and repressor level controls the basal activity of the operon demonstrates that both growth rate and repressor concentration influence the cell switching ability.
The bacterium Escherichia coli, like many other microorganisms can use different sugars as a carbon source and uses some of these sugars in preference to others. For example, when grown in the presence of both lactose and glucose, the bacteria first consume glucose and use lactose only when glucose is exhausted. To this end, the enzymes necessary for lactose uptake and metabolism, grouped in one transcriptional unit called the lac operon (lacZ, Y, A encoding for the lactose degrading enzyme (β-galactosidase), permease and transacetylase, respectively) are produced only in the absence of glucose and in the presence of lactose or its analogs, such as the non-metabolizable analog thiomethyl-β-galactoside (TMG). In the absence of such inducers, the transcription of the operon is inhibited by the repressor molecule LacI. This inhibition is relieved by the inducers, which bind and inactivate LacI, initiating an amplifying feedback loop through the expression of the permease that ensures a high influx of inducer to maintain the operon's expression in the ‘on' state. This phenomenon of adaptive enzyme production has been widely studied since its discovery by Jacques Monod and François Jacob and is one of the most famous and best characterized examples of transcriptional gene regulation. E. coli lactose operon is also a paradigm of cellular differentiation. Indeed, in the presence of an intermediate concentration of TMG, an isogenic bacterial population is divided in two subpopulations of cells with the operon's genes either turned on or remaining off. The differentiation step is generally hypothesized to depend on fluctuations in expression of the operon's proteins. Nevertheless, it is still poorly characterized. On the basis of experimental and theoretical approaches, we explored the determinants of cell fate in this system.
We designed a microfluidic device allowing the observation of single cells growing within a microcolony under conditions that can be changed at will. We used this setup to study phenotypic variability in the lactose operon induction under conditions leading to a transient bimodality of lac expression in the population. We used an E. coli strain modified to express the yellow fluorescent protein (YFP) and the cyan fluorescent protein (CFP), both under the control of a promoter regulated by LacI (PLlacO1). Therefore, yellow and cyan fluorescence intensities both represent the concentration of active repressor molecules and indirectly, the expression state of the lactose operon. Microcolonies originating from a single cell were grown in the microfluidic device and followed by time-lapse microscopy. During the first generations of growth, cells were grown in the absence (or with a very low concentration) of inducer and after several generations, TMG was introduced at intermediate concenteration into the medium and maintained thereafter. In the absence of TMG, cells exhibit an overall weak fluorescence yet with significant variations between cells that were shown to correspond in part to the variability in the intracellular concentration of active LacI molecules. Upon induction, transient bimodality is observed, as the cells are divided between two subpopulations of bright and dim fluorescence.
We found a strong clustering of induced cells within their genealogical trees, indicating a substantial epigenetic inheritance. This epigenetic inheritance can be traced back up to two generations prior induction, suggesting that some molecular determinants of cell fate are epigenetically inherited with a short-range memory lasting around two divisions.
The promoter used to control fluorescence proteins expression is sensitive to small variations in active LacI concentration. Thus, in the absence of inducer, these variations result into correlated variations of YFP and CFP levels. We used the arithmetical mean of yellow and cyan fluorescence intensities to estimate the concentration of active LacI in the cells. We found that the cells exhibiting a low LacI concentration before induction are more likely to be induced upon TMG introduction. Likewise, the slowly growing cells were found to have a higher switching probability than the fast-growing ones. We used a multivariate analysis based on a generalized linear model to estimate the correlations of pre-induction LacI concentration and growth rate with the switching probability (Figure 5C). This analysis confirms that both LacI concentration and growth rate are correlated with the switching probability and demonstrates that even though LacI concentration and growth rate can be linked, their correlations with the switching probability represent independent effects. Together, these effects can account for 90% of the observed switching events.
To gain a better understanding of the possible influence of LacI expression fluctuations and growth rate on the switching probability of a cell, we used a model consisting in a system of differential equations and describing the dynamics of the lactose utilization network. In this model, LacI concentration controls the basal level of expression of the operon and the growth rate acts through the dilution of intracellular components. According to this model, depending on both LacI concentration and growth rate, a cell can be in a monostable or bistable regime. Therefore, monostable and bistable cells can coexist in the population due to parameters' variability. In addition, for cells in the bistable regime, the size of the minimal LacY burst necessary to trigger induction increases with LacI concentration and growth rate. Thus, in agreement with our experimental results, these two variables control the sensitivity of the cell to permease bursts and therefore influence its switching probability.
We thus found pre-disposing factors governing the lactose operon switching in a regime of transient bimodality. Some factors, such as LacI and LacY concentrations result from stochasticity at the local level of the network. On the contrary, growth rate variability represents variations in the cell global physiology. Therefore, the effects of local stochasticity are coupled with the influence of the global physiology, demonstrating the importance of considering the embedding of a particular genetic network in the whole cellular physiology to understand fully its dynamics.
The lactose operon regulation in Escherichia coli is a primary model of phenotypic switching, reminiscent of cell fate determination in higher organisms. Under conditions of bistability, an isogenic cell population partitions into two subpopulations, with the operon's genes turned on or remaining off. It is generally hypothesized that the final state of a cell depends solely on stochastic fluctuations of the network's protein concentrations, particularly on bursts of lactose permease expression. Nevertheless, the mechanisms underlying the cell switching decision are not fully understood. We designed a microfluidic system to follow the formation of a transiently bimodal population within growing microcolonies. The analysis of genealogy and cell history revealed the existence of pre-disposing factors for switching that are epigenetically inherited. Both the pre-induction expression stochasticity of the lactose operon repressor LacI and the cellular growth rate are predictive factors of the cell's response upon induction, with low LacI concentration and slow growth correlating with higher switching probability. Thus, stochasticity at the local level of the network and global physiology are synergistically involved in cell response determination.