Positive feedback in eukaryotic gene networks: cell differentiation by graded to binary response conversion

From Wiki FKKT

Jump to: navigation, search

(Andreja Bratovš)

Becskei, A., Séraphin, B., Serrano L. Positive feedback in eukaryotic gene networks: cell differentiation by graded to binary response conversion. The EMBO Journal 20, 2528-2535. 2001



The paper[1] presented in this seminar was published in 2001 in The EMBO Journal. Becskei, Séraphin and Serrano constructed a synthetic eukaryotic gene switch in S.cerevisiae by using positive feedback. They analysed the consequences of adding a positive feedback loop in a system with well-defined components. The components used were the transactivator from the tetracycline-controlled transcriptional activation, its binding sequence and the reporter gene GFP. These were used in different constructs; the resulting fluorescence was then measured in single cells with fluorescence microscopy. The findings were also consistent with a mathematical model. In the first part of this presentation, I briefly describe the basics needed to understand the results of this paper and their implications. The topics included are a brief description of the difference between eukaryotic and prokaryotic gene regulation, enhancers and two models of their operation, namely the graded and the binary mode, positive feedback and its role in biological systems. Lastly, the tetracycline-induced transcription activation is described since it was used in this research and understanding of its mechanism is required to understand the results. Next, we move on to the second part, where I present the approach and the results of the research.

Eukaryotic gene regulation

Although basic principles of gene regulation are the same in prokaryotes and eukaryotes, the logic of gene regulation is different. The main reason for differences arises from different genome organization, namely eukaryotic DNA is packaged into chromatin templates while prokaryotic DNA is not. Prokaryotic DNA is accessible and RNA polymerase can bind to promoter sequences. The ground state is thus non-restrictive. Strong promoters initiate transcription at maximum rate, while activators increase transcription initiation at inherently weak or repressed promoters by directly interacting with RNA polymerase. Repressors keep the expression rate at a very low level by preventing RNA polymerase to bind to the promoter. In eukaryotes, the ground state is restrictive because of chromatin packaging. Strong promoters are inactive in eukaryotic cells, even with no repressors present. Pol II holoenzyme, which is responsible for gene transcription in eukaryotes, consists of the core RNA polymerase and many different transcription factors and other associated proteins, which are only loosely bound to the complex. For the complex to efficiently bind to the promoter and start the transcription chromatin needs to become accessible and many activators have to bind to their upstream or downstream binding sites. Activators can regulate transcription in two ways; they can bind to Pol II holoenzyme directly or they can remodel chromatin structure and thus indirectly increase recruitment of the transcription machinery to the promoter.[2]

Enhancers and their mode of operation

Activator binding sites are named enhancers and they usually bind transcription factors. They are short sequences, usually 50 to 1500 base pairs long and can be located downstream or upstream of the promoter, or can even be on a different chromosome. There are two different theories explaining how enhancers operate: rheostatic mode and binary mode. In the rheostatic mode theory, activators bound to enhancers cause an increase in the rate of transcription. They do so by directly interacting with the transcription machinery. The presence of an activator therefore causes the same effect in all the cells – increased expression of genes that are affected by the activated enhancer. The level of the increase is positively correlated with the concentration of the activator; the response of a cell population is thus graded. In the binary mode theory, enhancers increase the probability of initiating the transcription rather than the rate of transcription. In other words, they raise the probability of genes activating and staying active, but do not cause a certainty. The consequence of the probabilistic response is that cells in the same environment of transcription factors have different gene expressions – a mosaic pattern is created. The population of cells is divided into two pools; cells with high expression of the gene and cells with low expression of the gene. This is called a bimodal distribution. Different levels of the activator change the proportions of cells in the pools. The response is thus binary. Activators bound to enhancers may operate by remodelling the chromatin and making it accessible, or by recruiting other chromatin remodelling machinery. Stochastic gene activation may be important in cell differentiation, where the expression of some genes leads to commitment of a stem cell to differentiate to a certain type. It is important that a proportion of stem cells remains uncommitted. In the stochastic model, stem cells are scattered between differentiated cells in a tissue, even though the environment is the same throughout the tissue. It is not possible to distinguish between both modes of enhancer operation, unless expression is analysed in single cells by methods such as flow cytometry or fluorescence microscopy.[3][4]

Positive feedback in eukaryotic systems

A system has a feedback loop if a certain component influences its own levels or activity. Feedback loops can be positive or negative. In a negative feedback loop, an increase in levels/activity of the component is counterbalanced and the change is minimised. In a positive feedback loop, the system acts by potentiating the change, for instance, an increase in levels/activity leads to an even bigger increase. Negative feedback therefore stabilises the system while positive feedback amplifies changes and destabilises the system. For example, in genetic networks, a positive feedback loop consists of a transcription factor that positively regulates its own transcription. When its levels increase, the binding of the transcription factor to its operator amplifies the change and further increases the transcription. A negative feedback loop in a genetic network involves a repressor that lowers its own activity. Both positive and negative feedback loops can be simple, composed of only a single component, or very complex with many interactions; they can also be direct or indirect. Feedback loops occur on the level of transcription regulation as well as in protein-protein interactions. A common feature of a system with positive feedback is bistability. Bistable systems can exist in one of two stable steady states. One of them is usually a high activity state and the other a low activity state. A variable, which in biochemical systems is often the concentration of a molecular species, determines the state of the system. Bistability is closely associated with bimodal distribution of a cell population. However, positive feedback alone is not enough to cause bistability in a system. Usually, it has to be paired with a sigmoidal response curve within the feedback loop. An example of a sigmoidal response is cooperative binding of a regulator to DNA. Positive feedback in nature often takes place when cells need to commit to an irreversible development. Such cases arise, for example, in the cell cycle, oocyte maturation and cell differentiation.[5][6][7]

Tetracycline-controlled transcriptional activation

Tetracycline inducible systems are used for controlling protein expression in prokaryotic and eukaryotic gene networks. The systems used in research are derived from E.coli tetracycline resistance operon. The two systems most used are Tet-Off and Tet-On. Tet-Off: This system uses tetracycline transactivator (tTA). This protein consists of two parts: tetracycline receptor (TetR) from E.coli and an activation domain VP16 from Herpes Simplex Virus. tTA binds to specific tetO sequences, which are usually placed in front of a minimal promoter, and activates the transcription of the target gene. When tetracycline (or its analog, e.g. doxycycline) is added to the system, it binds to tTA. tTA with bound inducer can no longer bind to tetO and the expression of the target gene is repressed. Tet-On: The transactivator of this system is rtTA. rtTA is similar to tTA, but has a few point mutations, which make its response to tetracycline (or analogs) reverse. When there is no inducer present, rtTA cannot bind to tetO and thus the target gene is not expressed or has a very low expression. When the inducer is added, it binds rtTA and the complex can then bind to TetO, resulting in the expression of the target gene. A schematic representation of both systems can be found here. In the research described in this seminar, they used the Tet-On system to characterize the effects of an addition of positive feedback to the gene network.

Converting graded response to binary response with positive feedback


To test the ability of a positive feedback to convert a graded response to a binary response, they performed two series of experiments. The design of the constructs used is shown here. They used yeast centromeric plasmids, which are low copy vectors. Expression systems incorporate the following components:

  • rtTA gene, which is a fusion of TetR and VP16
  • constitutive promoters CMV and CYC1
  • tetreg regulatory sequence, composed of the promoter CYC1 and two tetO sequences
  • GFP gene.

When constructs were chromosomally integrated, different copy numbers of the reporter gene (n) were used. In the first series, they compared rtTA expression in constitutive and autocatalytic systems. To build the constitutive system they inserted the rtTA gene into plasmids and linked it to two different constitutive promoters: CMV and CYC1. CMV is a promoter from human cytomegalovirus, while CYC1 is a yeast promoter. They built the autocatalytic system by adding an rtTA binding sequence to the CYC1 promoter, thus creating tetreg – a regulatory sequence for tetracycline-induced transcriptional activation. They then transformed yeast strains containing chromosomally integrated green fluorescent protein (GFP) with an rtTA binding sequence. Thus, when rtTA on the plasmids was being expressed it bound to its chromosomal binding site and caused the expression of GFP, which was then quantified by measuring fluorescence intensity in single cells. In the second series of experiments, they detected the expression of rtTA directly in autocatalytic systems. rtTA was fused to GFP and linked to tetreg promoter sequence. This construct was either inserted in a plasmid or integrated into the chromosome and the expression of rtTA was, as above, determined by measuring fluorescence intensity.[1]

Graded response to rtTA in constitutive systems

In this simple system, rtTA is under the control of the CMV promoter and is expressed constitutively. In the presence of doxycycline, rtTA activates the expression of GFP. After activation, all the cells in the population began expressing GFP (Figure 3 A and B) and the distribution of the population was approximately Gaussian. The mean values of this distribution increased with the concentration of the inductor (Figure 2 A). These properties are typical for a graded response. When the CMV promoter was substituted with the CYC1 promoter, the mean values were lower and the distribution was less regular (Figure 2 B, lanes labelled constit). However, the population did not split into two sub-population and the response was still graded, as opposed to binary.[1]

Converting graded response to binary

Adding a regulatory sequence to the CYC1 promoter converted the system from constitutive to autocatalytic. The expression of rtTA is induced by adding doxycycline. In the autocatalytic system, rtTA activates not only the expression of GFP but also induces its own transcription beyond the basal rate. In contrast to the constitutive system, inducing the expression of rtTA caused the division of cells into two distinct pools: fluorescent (‘on’) cells and non-fluorescent (‘off’) cells (Figure 3 C and D). At higher inducer concentrations, more cells were in the ‘on’ state and less in the ‘off’ state, but the position of the peaks did not change (Figure 2 B – lanes labelled autocat – and Figure 4 A). Therefore, the response in this autocatalytic system was binary. Increasing the copy number of the GFP gene increased the fluorescence in on-cells, but did not change the proportion of on- to off-cells. Once this bimodal distribution was established, the percentage of bright and dark cells did not change until the inducer was removed. GFP was fused to the activator rtTA to allow direct detection. The construct was on a plasmid or chromosomally integrated. In both cases, the distribution of the cells remained bimodal, however, the number of bright cells was higher and the difference in fluorescence between both sub-populations was smaller. (Figure 4) The proportion of on-cells was positively correlated with the gene copy number in contrast to the system with chromosomal reporter. The differences between the systems with direct and indirect detection can occur because of two possible reasons. The first one is different noise sensitivity: rtTA-GFP fusion protein has slightly different properties than rtTA. Secondly, the chromosomal reporter system in addition to being bistable also has a sigmoidal response of the reporter system to the activator (i.e. the expression of GFP because of rtTA binding). This combination results in a better separation of the two groups of cells. They also compared this autocatalytic system with a system with indirect positive feedback. For that purpose, they tested the expression of GFP under the control of GAL1 promoter. This promoter controls the expression of Gal4 in S. cerevisiae, but Gal4 has no binding site on the promoter, therefore, there is no direct positive feedback. Gal4 increases the expression of galactose transporters. This leads to higher levels of galactose in the cell, which increases Gal4 activity. In this case, the positive feedback is indirect. As you can see in Figure 4B in the black lane, the indirect positive feedback also results in a bimodal distribution of cell fluorescence.[1]

Autocatalytic switch

The autocatalytic system is a switch, in which single off-cells can become on-cells in a stochastic way. Switching in the other direction (from ‘on’ to ‘off’ state) is theoretically possible but was not observed. Because the switching is continuous, this switch is noise-based, in contrast to the toggle switch. The two switches are both bistable, but the toggle switch remains stable and has no random transition.[1]

Theoretical background of the switch

To explain the results of the positive feedback, they performed a mathematical analysis using probabilistic methods, which allowed for noise and fluctuations. Theoretical models show that a system with a constitutively expressed activator results in a graded response with a sigmoidal shape. The distribution is unimodal, which means that there are no subpopulations. Introducing a positive feedback loop to this model and thus converting the system to autocatalytical results in bistability. Bistability is not an obligatory consequence of the positive feedback, since other parameters of the gene circuit also influence it. High degree of cooperativity, which is characteristic for eukaryotic transcription activation and is a consequence of nucleosomal rearrangement and protein-protein interactions, in concert with positive feedback gives rise to a robust bistability. If the two steady states are sufficiently separated, the distribution of the population is bimodal. To explain the properties of the autocatalytic system, they used the following model:

f(x) = \frac {dx}{dt} = \frac {s \cdot x^n}{d + x^n} + r - k \cdot x

Where x is the concentration of the activator, s is the maximal rate and r is the basal rate of the synthesis of the activator, d is the dissociation constant of the activator from DNA, n is the cooperativity of the activation and k is the degradation rate of the activator. Parameter s was experimentally tuned by doxycycline and gene dosage, as it includes the copy number of the gene circuit, the inducer concentration and the proportionality constant between transcription and translation. The potential of this equation, which is negatively correlated with the probability distribution of the concentration of transcriptional activator, was calculated. The probability distribution (or negative potential) for the concentration of transcriptional activator is shown in Figure 6. To achieve different levels of activation, the parameter s (rate of synthesis) was changed. The lower steady state reflects the basal expression rate and the upper steady state corresponds with the maximal rate of the expression. For smaller rates of synthesis, which correspond to lower activation levels, the probability for the system to adopt the lower steady state is higher. The situation becomes reversed when the degree of activation increases. This theoretical model is in agreement with the results of the experiments. The lower steady state corresponds to off-cells and the upper one to on-cells. The probability distribution reflects the fluorescence distributions. The lower steady state overlaps with the unimodal distribution of the constitutive system, while the second peak – the upper steady state – only appears in autocatalytic system and corresponds to maximally activated expression rates. The degree of activation, which is analog information, is converted to binary information; '0' and '1' correspond to lower and upper steady states.[1]

Autocatalytic expression in cell differentiation

To follow the fates of single cells during population growth, their growth was monitored on microscope slides (Figure 5). In Figure 5 A and B are non-fluorescent (off-) and fluorescent (on-) cells at the beginning of the experiment. The colony in Figure 5 C grew from off-cells and contained both on- and off-cells. However, the colony in Figure 5 D, which grew from on- cells, only contained fluorescent cells. This shows that cells can switch randomly from ‘off’ to ‘on’ state. It was also observed that the cell-doubling time is shorter than the average time required for a cell to switch to the ‘on’ state. The slow switching time may be the consequence of the robust bistability of eukaryotic gene expression. The off-cells also replicate faster than the on-cells. The above-mentioned properties maintain a pool of off-cells even at high concentrations of the inducer. Even though the switching-time is random, the inducer controls the percentage of on- and off-cells. Processes that are similar to this autocatalytic expression might be involved in blood stem cell differentiation. A stochastic decision activates genetic programs in stem cells and differentiates them to distinct blood cell lineages. Similarly, a pool of undifferentiated stem cells is maintained and the percentages of the cell types are regulated.[1]


To summarize, positive feedback is an important mechanism of gene regulation. In the presented research paper, they constructed a positive feedback loop with well-characterized components – the tetracycline-induced regulator and the reporter protein GFP. By implementing the positive feedback loop, they managed to convert the response of the cell population from graded to binary. Thus, they constructed a eukaryotic autocatalytic switch, which is similar to the toggle switch in its bistability, but is less stable since random transition occurs. Since properties, similar to this system, are found in cell differentiation, positive feedback might play an important role in regulating the processes involved. The findings of this paper might also have implications in explaining the way enhancers operate, namely those where a binary response has been described.


  1. 1.0 1.1 1.2 1.3 1.4 1.5 1.6 Positive feedback in eukaryotic gene networks: cell differentiation by graded to binary response conversion. Becskei et al., EMBO J., 2001
  2. Fundamentally different logic of gene regulation in eukaryotes and prokaryotes. Struhl, Cell, 1999
  3. To be or not to be active: the stochastic nature of enhancer action. Fiering et al., BioEssays, 2000
  4. Probability in transcriptional regulation and its implications for leukocyte differentiation and inducible gene expression. Hume, Blood, 2000
  5. Positive feedback in cellular control systems. Mitrophanov and Groisman, BioEssays, 2008
  6. Transcriptional autoregulation in development. Crews and Pearson, Curr. Biol., 2009
  7. Simple, realistic models of complex biological processes: Positive feedback and bistability in a cell fate switch and a cell cycle oscillator. Ferrel Jr. et al., FEBS Lett., 2009
Personal tools