A stochastic single-molecule event triggers phenotype switching of a bacterial cell

Abstract

By monitoring fluorescently labeled lactose permease with single-molecule sensitivity, we investigated the molecular mechanism of how an Escherichia coli cell with the lac operon switches from one phenotype to another. At intermediate inducer concentrations, a population of genetically identical cells exhibits two phenotypes: induced cells with highly fluorescent membranes and uninduced cells with a small number of membrane-bound permeases. We found that this basal-level expression results from partial dissociation of the tetrameric lactose repressor from one of its operators on looped DNA. In contrast, infrequent events of complete dissociation of the repressor from DNA result in large bursts of permease expression that trigger induction of the lac operon. Hence, a stochastic single-molecule event determines a cell's phenotype.

Figure 1

The expression of lactose permease in E. coli. (A) The repressor LacI and permease LacY form a positive feedback loop. Expression of permease increases the intracellular concentration of the inducer, TMG, which causes dissociation of LacI from the promoter, leading to even more expression of permeases. Cells with a sufficient number of permeases will quickly reach a state of full induction, while cells with too few permeases will stay uninduced. (B) After 24 hours in M9 media containing 30 µM TMG, strain SX700 expressing a LacY-YFP fusion exhibits all-or-none fluorescence in a fluorescence-phase contrast overlay. Fluorescence imaging with high sensitivity reveals single molecules of permease in the uninduced cells zoomed in on the red box from B. (C) After one day of continuous growth in media containing 0 to 50 µM TMG, the resulting bimodal fluorescence distributions show that a fraction of the population exists either in an uninduced or induced state, with the relative fractions depending on the TMG concentration. (D) The distributions of LacY-YFP molecules in the uninduced fraction of the bimodal population at different TMG concentrations, measured with single-molecule sensitivity, indicate that one permease molecule is not enough to induce the lac operon, as previously hypothesized (12). Over 100 cells were analyzed at each concentration. Error bars are standard errors determined by bootstrapping.

Figure 2

Measurement of the threshold of permease molecules for induction. (A) Single cell time traces of fluorescence intensity, normalized by cell size, starting from different initial permease numbers. The initial LacY-YFP numbers are prepared through dilution by cell division of fully induced cells after removal of inducer. Upon adding 40 µM TMG at time zero, those cells with low initial permease numbers lose fluorescence with time as a result of dilution by cell division and photobleaching, while those cells with high initial permease numbers exhibit an increase in fluorescence as a result of reinduction. Permease molecule numbers are estimated from cell fluorescence (28). (B) The probability of induction of a cell within three hours as a function of initial permease number was determined using traces from 90 cells. The probability of induction, p, is fit with a Hill equation p = y^4.5 / (y^4.5 + 375^4.5) for initial permease number, y. The threshold of permease numbers for induction is thus determined to be 375 molecules. Error bars are the inverse square root of sample size at each point. (C) To prove that complete dissociation of tetrameric repressor from two operators triggers induction, we constructed strain SX702 with auxiliary operators removed (no DNA looping). The figure shows single-cell traces of permease numbers in single cells grown in 40 µM TMG as a function of time. Unlike the looping strain SX700, the rapid induction of SX702 is no longer dependent on the initial number of permease molecules. This proves that phenotype switching is the result of a complete dissociation of the tetrameric repressor as shown in B. (D) In the absence of DNA looping, the entire population of strain SX702 rapidly induces in a coordinated manner from far below the threshold for a concentration as low as 20 µM TMG. DNA looping is necessary for bistability of the lac operon under these conditions.

Figure 3

Small and large bursts in the absence of positive feedback. In order to eliminate positive feedback from permease transport, we constructed strain SX701, replacing the lactose permease with the membrane protein fusion, Tsr-YFP. (A) Real-time traces of protein production in SX701 in 200 µM TMG. The FPs are photobleached immediately after detection to ensure that only newly produced proteins are measured after each 4 minute interval. Representative traces of single cells show frequent small bursts, associated with partial dissociation events, and a rare large burst, associated with a complete dissociation event, of protein production. (B) Distribution of burst sizes determined for 208 bursts from the real-time experiment depicted in B. Although the majority of bursts are small, a number of unusually large bursts are observed. The former are attributed to partial dissociations, and the latter to complete dissociations of the tetrameric repressor. Inset: The occurrence of burst sizes smaller than 10 molecules is well-fit by an exponential distribution (red line). (C) Analysis of small bursts from partial dissociations of the tetrameric repressor shows that the distribution of YFP molecules in the range of zero to ten molecules does not change in the range of 0 to 200 µM TMG. A small percentage of cells have much more than ten molecules and do not appear on the axis of this plot (see Fig. S7C). Over 100 cells were analyzed for each concentration. Inset: The frequency of bursts per cell cycle, (purple circles) increases slightly while the average number of proteins per burst, (gray triangles), remains approximately constant. These kinetic parameters are determined from the steady-state distribution of YFP in those cells with less than ten molecules (see text). Error bars are standard errors determined by bootstrapping. Gamma distributions using the determined parameters are overlayed as dashed lines. The gamma distribution for 200 µM is normalized for the subpopulation of cells with less than 10 molecules. (D) In addition to having the permease replaced with Tsr-YFP, strain SX703 has the auxiliary operators removed, eliminating DNA looping so that every dissociation event is a complete dissociation. The protein number distributions should reflect the bursts of protein expression from complete dissociations alone. Inset: the noise parameters µ²/σ² (purple circles ) and σ²/µ (gray triangles) suggest that the frequency of large bursts is independent of the TMG concentration but that the burst size is not.

Figure 4

Complete dissociation of the tetrameric repressor triggers induction. (A) A high concentration of intracellular inducer can force dissociation of the repressor from its operators, as described by Jacob and Monod (6). (B) At low or intermediate concentrations of intracellular inducer, partial dissociation from one operator by the tetrameric LacI repressor is followed by a fast rebinding. Consequently, no more than one transcript is generated during such a brief dissociation event. However, the tetrameric repressor can dissociate from both operators spontaneously and stochastically, then sequestered by inducer such that it cannot rebind, leading to a large burst of expression. (C) A time-lapse sequence captures a phenotype switching event. In the presence of 50 µM TMG, one daughter cell of a dividing cell switches phenotype to express many LacY-YFP molecules (yellow fluorescence overlay) while the other daughter cell does not (see Movie S1).