By introducing TATA box mutations into the upstream (GAL10
) promoter of synthetic gene regulatory cascades and developing a systematic means to control gene expression noise, we sought to achieve several goals. First, we aimed to determine the effect of TATA box mutations on gene expression from the GAL10
promoter itself. Our second and main goal was to determine how these mutations affect several key population-level gene expression characteristics, with specific focus on gene expression noise. In addition to controlling the noise in the expression of a target gene, we have also considered two characteristics of the dose-response: (i) the basal expression and (ii) the dynamic range. Our third and final goal was to examine if the gene expression noise and mean can be decoupled, e.g. if they can be independently controlled using our newly engineered gene expression cascades. To study fitness differences between cell populations due to noise separately from the effects of the gene expression mean, these two gene expression properties must be decoupled, i.e. they should be independently controlled. These goals were motivated by several recent studies suggesting a fitness benefit due to the population-wide phenotypic variation introduced by gene expression noise (25
With these objectives in mind, we constructed three quintuplets of synthetic gene regulatory cascades based on the GAL1 and GAL10 promoters of S. cerevisiae. Each quintuplet consisted of five minimally different (only by one and two base pairs) TATA box variants of the GAL10 promoter controlling TetR repressor expression. The quintuplets differed from each other in the number of tetO2 operator sites inserted into the GAL1 promoter controlling the expression of the reporter yEGFP. In addition, we constructed five control strains expressing yEGFP from the five different GAL10 promoter variants, to measure gene expression directly from the upstream promoter.
In the control strains, we observed decreases in yEGFP
expression levels due to TATA box mutations in the GAL10
promoter, which was expected since these point mutations had been specifically modeled after mutations previously incorporated in the GAL1
TATA box, where similar reductions in gene expression were observed (25
). In general, it has long been known that the TATA box can play a key role in transcription initiation and TATA box mutations can result in a less efficient promoter with reduced expression (62
). While largely consistent with previous studies in yeast (24
), the specific base substitutions we introduced at equivalent positions in the GAL10
TATA box did not produce the same impact on gene expression, illustrating that a given mutation can affect individual promoters differently.
In contrast to the significantly lower mean expression in GAL10 TATA box mutant cells, the GAL10 expression noises remained low and were similar (but distinct) for all strains, likely due to two different reasons. First, growing the cells in 2% galactose-containing medium ensured constitutive, maximal expression from GAL10 in our synthetic constructs, lowering the noise. Higher discrepancies in GAL10 expression noise may be revealed at intermediate induction, as in our previous work. Second, slow degradation of yEGFP (used as a proxy instead of TetR to estimate GAL10 promoter efficiency) filters out fast fluctuations from promoter activation or mRNA production events, thereby reducing the CV closer to the baseline. The differences in CV for GAL10 were small though statistically significant, which we think is due to different contributions from several noise sources. Applying a formula that connects protein noise to the underlying processes (promoter activation/deactivation and mRNA and protein synthesis and degradation) suggests that rare promoter deactivation events coupled with strong protein synthesis are the prime source of noise for the wild-type GAL10 promoter. In contrast, the main source of noise for severe TATA box mutants appears to be fast promoter dynamics coupled with low protein production in these mutants.
Since the main goal of this study was to determine how GAL10 TATA box mutations affect the ‘demographic’ characteristics of yEGFP expression from the T123 promoter, next we discuss in detail the various aspects of these findings. Specifically, we focus on yEGFP basal expression, dynamic range and gene expression noise, all of which are important if the synthetic regulatory cascades are to be used for controlling gene expression across a cell population.
We sought to explain the observed differences in yEGFP
expression through mathematical and computational modeling. This was necessary to infer since there are no direct ways to measure inducer-bound TetR separately from free TetR dimers capable of repressing the T123 promoter. The good agreement between our model and the experimental data indicate that free TetR levels in single cells depend on two key molecular factors: the rate of inducer influx and episodic TetR production controlled by the GAL10
promoter sequence. ATc molecules entering the cell bind and inactivate free TetR dimers, depleting the TetR pool capable of repression. Therefore, once the TetR pool is depleted, new TetR molecules must be synthesized for repression to be possible. The GAL10
TATA box mutations determine the periods of productive and silent promoter states. Importantly, the inducer molecules dissociate from TetR very slowly (66
)—practically never, considering the time scales of other molecular processes. These considerations, and their implications from our model, explain most of yEGFP
’s behavior from the downstream T123 promoter.
Our first functionally important finding is that the severe reductions in TetR repressor levels due to the GAL10
TATA box mutations did not increase T123
basal expression levels significantly. Strong and tight repression is often necessary for achieving stable OFF states in many synthetic gene circuits, inducible switches and devices (1
), and in functional genetics studies of genes encoding toxic proteins (68
). We observed a systematic, but still slight increase in T123 basal expression following mutations of the GAL10
TATA box controlling tetR
gene expression (A). Even for the two most severe GAL10
TATA box mutants (sev1 and sev2) with approximately 10% tetR
production compared to wild type, T123 basal expression stayed at less than 1% of maximal expression. This slight increase in basal expression suggests that TetR dimers dissociating from the DNA take longer to replace in TATA box mutant cells. Nevertheless, TetR is an extremely strong and efficient repressor that, even in low concentrations, can reliably shut down the expression of target genes. Therefore, in the future, synthetic biologists can adjust gene expression noise by reducing the strength of the promoter encoding the repressor TetR, without being concerned about changing the basal level of downstream gene expression.
Maximal T123 expression when fully induced was not significantly affected by the strength of the tetR
promoter (B), indicating that the repressor activity of TetR molecules was completely abolished by ATc. Therefore, there was practically no reduction in dynamic range due to the GAL10
TATA box mutations. This result becomes important when compared to prior studies, where GAL1
(downstream) TATA box mutant noise reductions were tied to large reductions of maximum expression (25
). In contrast, in this study, we were able to maintain as broad a dynamic range as possible from the promoter of interest (T123) by transferring the TATA box mutations upstream to the GAL10
promoter. This maintenance of maximal expression levels and overall dynamic range can be important for synthetic gene expression systems, especially in functional genetics studies or within engineered microbes that need to produce industrially important protein products, enzymes and chemicals (14
) at user-specified rates.
In view of future applications for population-wide gene expression control, one of the more important effects of upstream TATA box mutations is the noise reduction observed downstream, at the level of the T123 promoter. This is relevant because high levels of gene expression noise can have negative effects on the fidelity of information flow in synthetic or natural gene networks (2
). Accordingly, several techniques have been employed to reduce noise in gene expression, including negative feedback loops (29
), alterations of cell volume (29
), temperature (8
), coexpression (72
) and direct TATA box mutations (25
). While in our case the noise peaked at some intermediary inducer concentration for all TATA box mutant strains, we observed a systematic decrease and leftward shift of the noise peak for increasingly severe TATA box mutants. These results suggest that a substantial part of the yEGFP
noise originates from fluctuations in free TetR dimer concentrations, and are consistent with the expectation that upstream noise is amplified more in the region where the dose–response curve has the greatest slope (21
). While most of this noise is masked in the absence of inducer, free TetR fluctuations are unmasked as ATc influx into the cells increases and depletes the free TetR dimer pool. At this point where ATc depletes TetR, the yEGFP
noise peak decreases in increasingly severe TATA box mutants, which is most likely due to the lower TetR fluctuations (variance) of TATA box mutants (see ), as explained in the Results. Overall, we show that noise in free (active) TetR strongly affects the expression of the regulated gene, while the noise of total TetR seems to have a negligible effect on yEGFP noise. Still, in addition to noise from active TetR molecules, we previously showed substantial intrinsic noise contributions from the downstream promoter itself (25
). Slow T123 promoter fluctuations (due to the presence of a consensus TATA box sequence) may be most prone to respond to the slow fluctuations of the upstream gene expression, typical to the wild-type GAL10
promoter. In contrast, faster upstream fluctuations of TATA box mutant strains may be filtered out by the slow T123 promoter. This may give rise to stochastic entrainment of the downstream promoter, which could be tested in the future by mutating both the upstream and downstream TATA boxes in the regulatory cascade.
Our final goal was to examine whether GAL10
TATA box variants could be used to decouple the control of yEGFP
expression noise and mean. To achieve this, we focused on the noticeable reduction in the height of the gene expression noise peak, in addition to the shift in the mean yEGFP
expression level where peak noise occurred. Both effects were due to different TetR repressible promoters, indicating a strain-dependent relationship between the noise and the mean. These shifts and reductions of maximum noise to higher mean expression levels were even more dramatic in our S1 single tetO2
promoter set, as seen in Supplementary Materials
. Thus, the nonoverlapping mean–noise characteristics can be used to decouple the mean and noise of gene expression by preparing cell populations to have identical means, but different noises of gene expression, as illustrated in C. Such decoupling (25
) is crucial for functional genetics studies investigating the effect of gene expression noise on cell population fitness. Accordingly, the gene circuits described here can be used as robust and tunable synthetic noise generators that control a target protein with prescribed levels of noise and mean in natural or other synthetic gene networks, resulting in cell populations with engineered fitness or differentiation capabilities. For example, stem cell populations could be engineered to differentiate at desired rates, depending on the noise levels of a differentiation factor, while maintaining a stable stem cell population by keeping the average expression levels constant.
In addition to our steady-state analysis of this synthetic gene circuit, our time-course measurements also provide useful principles for synthetic design, especially pertinent for natural cellular environments, where input signals often fluctuate (73
) or occur transiently. Our results demonstrate an important design principle involving the robustness of gene repression against transient input signals and gene expression variability at steady state. High expression of TetR repressor protein from the wild-type GAL10
promoter resulted in robust maintenance of T123 repression in the presence of a strong, but transient inducer signal input (A). At steady state, this promoter demonstrated the highest overall noise levels (, in black). In contrast, the reduced mean and variance in TetR expression from the mutant TATA promoter int2
lead to a weaker repression of T123, which responded strongly to the short pulses of inducer signal (C). At steady state, though, this promoter exhibited significantly reduced gene expression noise (, in teal). These results demonstrate an important cost-benefit relationship between robustness of repression during short transient pulses and steady-state variability. The future designs of synthetic gene circuits, especially when dealing with fluctuating input signals, will need to balance these factors in order to achieve optimal performance.
In conclusion, the use of TATA box mutations in synthetic gene regulation offers a means of improving the controllability of gene expression across cell populations. Our findings demonstrate the possibility of improving noise control with a clear benefit from TATA box mutations, reducing the expression level of repressor proteins that inhibit expression of a downstream promoter. High repressor protein expression (in our case, TetR) causes no significant change in the dynamic range, while strongly elevating variability in the gene expression of interest. The introduction of TATA box mutations in the regulatory promoters of inducible gene expression systems thus offers an attractive solution for adjusting the noise of gene expression independently of the mean. Our method involves minimal alteration to synthetic gene circuits, maintains low basal expression of repressed promoters, and retains the maximum gene expression output from the regulated promoter of interest. Admittedly, the gains in noise control and the maintenance of the dynamic range come at the cost of increasing inducer sensitivity (the dose–response curves become steeper for the mutants in ). Due to the ease of implementing such mutations for controlling gene expression noise, together with the clear benefits, this strategy may prove worthwhile to include in the future design of synthetic gene networks as well as repressor protein-based inducible gene expression systems.