Construction of a genetic toggle switch in Escherichia coli

From Wiki FKKT
Revision as of 15:54, 29 December 2014 by UrbanB (talk | contribs)
Jump to navigationJump to search

In year 2000, a seminal paper (1) by Timothy Gardner and colleagues at Boston University, was published in scientific journal Nature. The paper described the construction of engineered gene-regulatory network in bacterium Escherichia coli that preformed as a toggle switch – analogous to the one we use to turn lights on or off. This work was one of the first examples in then-nascent field of synthetic biology that combined computer simulations based on mathematical models with experiments in the lab to design the complex gene circuit and predict the outcomes. This seminar will discuss both mathematical and molecular biology methods used to engineer and implement a genetic toggle switch in bacteria.

Toggle switches in nature

A classic example of a toggle switch in nature is the regulation of bacteriophage λ. When the phage infects its host such as Escherichia coli, it can follow lytic or lysogenic life cycle. In the lytic pathway, the phage replicates its DNA and new virus particles are produced. This leads to cell lysis and release of newly assembled virions into the environment. On the other hand, in the lysogenic life cycle, the viral DNA integrates itself into the host’s chromosome and replicates with the host DNA (2). The switching is made possible by two proteins that repress each other’s expression – λ repressor, also known as CI, and repressor Cro (controller of repressor and others). In lysogenic cycle, the only bacteriophage protein produced is CI, which represses the expression of all other phage proteins. Importantly, CI dimer binds response elements in the PR promoter (R stands for repressor) and prevents expression of viral gene cro. Bacteriophage λ switches to lytic pathway if the host DNA is damaged. During bacteria’s response to DNA damage, CI is cleaved by bacterial RecA protein and thus cannot bind its response elements. This allows production of Cro repressor, which in turn binds to PRM promoter (RM stands for repressor maintenance), which controls expression of cI gene. Cro represses synthesis of CI, which leads to production of other phage proteins, assembly of virus particles and finally host cell lysis. Presented bacteriophage λ genes act as a genetic toggle switch with two states – it achieves bistability. One stable state represents lysogenic life cycle (CI repressor concentration is high, Cro concentration is low) and the other represents lytic cycle (CI repressor concentration is low, Cro concentration is high) (3–5).

[slika fag lambda]


Switch design

Gardner and colleagues used the above-described nature’s blueprint to design their synthetic genetic toggle switch. Their genetic circuit is composed of two operons that in turn both contain a constitutive promoter and a gene that encodes for a repressor. Each repressor negatively controls the production of the other, just as Cl inhibits transcription of cro gene and Cro protein inhibits transcription of cI gene. The switching is carried out by addition of an inducer (inducer 1) that forces the system into one of the possible steady states. The added inducer 1 prevents binding of the repressor 1 to the opposing promoter 1 and thus induces the synthesis of the repressor 2, which in turn inhibits production of the repressor 1. Even upon removal of the inducer 1, production of the repressor 2 remains high and the concentration of the repressor 1 remains low. This means that the engineered genetic network can store information about its past, in this case the stimulation with the inducer 1 (1). The presented mutual repressor switch resembles the Reset-Set flip-flop or latch (RS flip-flop) in electronics. A RS flip-flop can be constructed with two coupled NOR gates. Only when a NOR gate receives two low inputs (0), the resulting output is high (1). The two inputs a NOR gate receives can be thoughts as the level of transcription (high/low) for the corresponding repressor and presence or absence of an inducer that inhibits the same repressor’s function. In schematic representation of the SR latch, input 1 is analogous to Inducer 1 and Input 2 is analogous to Inducer 2 in designed genetic toggle switch. Similarly, Output 1 corresponds to high Repressor 1 concentration (and low Repressor 2 concentration) and Output 2 to high Repressor 2 concentration (and low Repressor 1 concentration). High Input 1 forces the Output 1 to be low and Output 2 to be high (3).

[slika RS latch z logično tabelo in splošna shema stikala *step-by-step]

Mathematical model

The outcomes of even simplest genetic networks such as toggle switches are difficult to predict (as demonstrated above) solely with human intuition. The behavior of the switch can be captured in deterministic ordinary differential equations (ODEs) to guide the construction of genetic circuit and the rational selection of its parts. The mathematical model describes changes in the concentrations of two repressor proteins over time:

In the simplest model, the production rate of a reppressor is proportional to the fraction of unbound response elements on corresponding promoter. We can use Hill function to describe the rate of production of repressor 1:

<math>f(v) = \frac{\alpha _{1}\cdot K_{d}}{K_{d}v + v^{\beta }}</math>

Where α1 is the maximal production rate of repressor 1 form promoter 2, which include the net effect of RNA polymerase binding, open-complex formation on DNA, transcript elongation, transcript termination, repressor binding, ribosome binding and polypeptide synthesis. Kdv is the dissociation constant of repressor 2 from promoter 2, v is the concentration of repressor 2 and β is exponent representing the degree of cooperativity of repressor 2 binding. <math>\frac{K_{d}}{K_{d}v + v^{\beta }}</math> is thus the fraction of unoccupied promoters 2. By representing the concentration of repressor 2 v in units of Kdv, we can eliminate Kdv and write dimensionless equation: f(v)= α_1/(1+v^β ) To observe the concentrations of repressor 1 and repressor 2 over time, we must include the protein decay due to constant dilution because of cell growth: du/dt= α_1/(1+v^β )-u and dv/dt= α_2/(1+u^γ )-v Where u is the concentration of repressor 1, v is the concentration of repressor 2, α1 and α2 are the relative maximal production rates of repressor 1 and repressor 2, respectively, and the coefficients β and γ represent the cooperativity of binding of repressor 2 and repressor 1 to response elements on promoter 2 and promoter 1, respectively. The first term in equations above describes the cooperative repression of constitutive promoters and the second term represents the decay of repressors due to cell growh. The decay rate is the same for both proteins. To maintain the model dimensionless, the time was rescaled to units of protein half-life on account of dilution (1, 3, 6).

To observe the gene circuit behavior using computer simulation, we must plot these two equations. Because we are interested in steady states of the switch, where du/dt=0 and dv/dt=0 (concentrations of repressors don’t vary over time), we plot two curves: u= α_1/(1+v^β ) and v= α_2/(1+u^γ ) on a graph where x axis represents the concentration of repressor 1 u and y axis corresponds to repressor 2 concentration v.

[graf steady state]

As we can see, the two curves, also called nullclines, intersect at three points. These three points represent three steady states of the toggle switch: low repressor 1 concentration and high repressor 2 concentration, high repressor 1 concentration and low repressor 2 concentration, and one unstable steady state where both repressor concentrations are equal. At this point, even the slightest difference in repressor concentrations will move the system into one of the two stable steady states. It is apparent that we get these three intersections owing to sigmoidal shape of the two curves, which occurs because of cooperative repression of the promoters (β, γ > 1). Binding cooperativity of at least one repressor protein is so the first criteria for bistability of genetic toggle switch we can deduce form derived mathematical model. If we change the values for protein production parameter, we can see that production rates for both repressors must be comparable or else the nullclines cross only once, producing a broken toggle switch with only one state. Thus, the second criteria for a mutual repressor switch to work is balanced maximal expression rate from the two constitutive promoters.

[graf a1 >> a2]

Other valuable information that were extracted from parameter observation are: increasing protein production rates and higher order of cooperativity correlate with increasing robustness of the switch (the collection of parameters that support functioning toggle switch). These findings guided the implementation of engineered mutual repressor switch in Escherichia coli (1).

[graf log(a) in coop.]

DNA constructs and implementation

The toggle switch network was constructed on a low copy number plasmid (7) conferring resistance to ampicillin and containing the ColE1 origin of replication. The two operons both contained constitutive promoter, ribosomal binding site (RBS), repressor genes and a double terminator of transcription. To monitor the state of the switch, green fluorescent protein (GFP) gene GFPmut3 with its own RBS was cloned as the second cistron adjacently to the repressor 1 gene. Consequently, the authors could determine the state of the switch by observing expression of the GFP, which correlated with repressor 1 concentration (State 1: high repressor 1 & GFP concentration, State 2: low repressor 1 concentration).

Two different genetic networks classes were designed. Both used the lac repressor (lacI), whose expression controlled promoter 1, and constitutive Ptrc-2 promoter as Promoter 2. Depending on the second promoter-repressor combination, the plasmids were named pTAK or pIKE class. In the pTAK plasmid class, the temperature-sensitive Cl repressor (cIts) was cloned downstream of the Ptrc-2 promoter and constitutive PLs1con promoter controlled lacI expression, whereas the pIKE family of plasmids contained PLtetO-1 promoter and Tet repressor (tetR) as Promoter 1 and Repressor 1 elements, respectively. When researchers introduced the pTAK plasmids into bacteria, they could therefore switch between stable states using isopropyl β-D-1-thiogalactopyranoside (IPTG) as Inducer 2, lifting repression off the Ptrc-2 promoter, or high-temperature pulse that destabilizes the CI repressor. Conversely, pIKE gene constructs switch from one state to the other with addition of IPTG or anhydrotetracycline (aTc), which causes TetR dissociation from its response elements, to culture medium.

[plazmidna karta]

Plasmid features

  • ColE1 ori: origin of replication from ColE1 plasmid. It confers 50-70 plasmid copies per cell (7).
  • bla: β-lactamase ampicillin resistance gene.
  • RBS: ribosomal binding site. Protein production rates (α in mathematical model) were controlled by using different RBS sequences. The RBS used downstream of Ptrc-2 promoter (Promoter 2) was rbs E, while different RBS sequences were used to modify the synthesis rates of LacI (Repressor 2). The exact RBS sequences are listed in the Supplementary Information (http://www.nature.com/nature/journal/v403/n6767/extref/403339ai1.pdf ) (1).
  • PLs1con: shortened version of the wild-type bacteriophage λ PL constitutive promoter with point mutations in -10 sequence region to weaken the otherwise very strong wild-type promoter (8).
  • PLtetO-1: constitutive promoter containing two copies of O2 operator from Tn10 tetracycline resistance operon and -35 and -10 sequences from PL promoter (8).
  • Ptrc-2: constitutive promoter, generated by fusing Ptrp and Plac promoter (8).
  • cIts: gene encoding temperature-sensitive mutant of the λ repressor CI. At 42 °C, the CI repressor denatures and cannot bind to PLs1con promoter (9). CI binds to its operator as a dimer moiety. All plasmids containing CL and its responsive elements were grown at 32 °C instead of 37 °C. Thermal induction was performed at 42 °C (1).
  • lacI: lac repressor (LacI) gene of the E. coli lac operon binds the operator as a tetramer. The binding is prevented with inducer molecule e.g. isopropyl β-D-1-thiogalactopyranoside (IPTG) (5).
  • tetR: tetracycline repressor (TetR) gene of the E. coli Tn10 transposon that confers tetracycline resistance to the host bacteria. TetR binds its response elements as a dimer. Anhydrotetracycline (aTc) causes dissociation of the TetR homodimer from its corresponding operator sites (10).
  • GFPmut3: gene encoding mutant green fluorescent protein which is 20 times more fluorescent than wild type (11).
  • T1T2: transcription terminator of E. coli ribosomal RNA operon.

The E. coli strain used was JM2.300 (lacI22, λ-, e14-, rpsL135(strR), malT1(LamR), xyl-7, mtl-1, thi-1) (12). Importantly, this strain doesn’t have the cI λ gene incorporated in its genome and has a mutated non-functional LacI repressor, so it didn’t interfere with any of the switch components (1).

Results

All in all, six different toggle switches were constructed, where four had CI – LacI repressor combination (pTAK plasmids) and two had TetR – LacI repressor pair (pIKE plasmids). The only difference among the genetic switches of the same type was the ribosomal binding site (RBS) downstream of Promoter 1 that defined the LacI synthesis rate (α2 in the mathematical model).

Promoter strengths determination

Gardner et al. determined the strengths of different promoter-RBS combinations used by cloning GFPmut3 or GFPuv reporter gene downstream of the studied promoter-RBS pair (Type I plasmid type in Supplementary Information) and GFP expression was assayed by flow cytometry. The determined promoter efficiency series was: PLs1con > Ptrc-2 > PLtetO-1. The Type I plasmid map and GFP expression measurements can be found listed in the Supplementary Information (http://www.nature.com/nature/journal/v403/n6767/extref/403339ai1.pdf).

Efficacy of repression

The efficacy of repression was determined in similar experiment as the relative promoter strengths. The DNA constructs used to assay repression were modified pTAK and pIKE plasmids described above. The Type II plasmids were used to measure LacI repression. The lac repressor prevented the expression of GFPmut3 from Ptrc-2 promoter. The λ and Tet repressor were expressed from Ptrc-2-rbs E promoter-RBS pair and prevented GFPmut3 transcription downstream of PLs1con or PLtetO-1 promoter on Type III plasmids. Again, the leaky GFP expression (i.e. expression from repressed promoters) data were collected using flow cytometer. The fold repression was estimated as ratio between maximal GFP expression values measured using Type I plasmids and GFP expression under repressed promoters using Type II and Type III plasmids. The CI repressor was the most effective, achieving almost 6000-fold repression of the PLs1con promoter. Furthermore, when TetR or LacI downregulated the GFP synthesis, the measured values were 100- or 20-fold smaller compared to highest recorded values for unrepressed expression. The plasmid maps and leaky GFP expression values are listed in the Supplementary Information (http://www.nature.com/nature/journal/v403/n6767/extref/403339ai1.pdf).

Demonstration of bistability

When different genetic switches were tested for bistability, all four pTAK (pTAK117, pTAK130, pTAK131 and pTAK132) and one pIKE plasmid (pIKE107) successfully preformed switching between two stable steady states in E. coli. The only difference among vectors of the same group was the RBS sequence upstream of lacI repressor gene. Control plasmids used were Type II plasmids. These plasmids don’t contain functional genetic toggle switches because they contain only one of the two repressor genes. pTAK102 was used to determine LacI repression efficiency and can support GFP expression only when IPTG is present. This plasmid served as a positive control for IPTG transient induction and as a negative control for conditions, where IPTG is absent – including induction with heath or aTc (pTAK102 doesn’t contain genes for CI and TetR). On the other hand, pTAK106 and pTAK108 control plasmids were used as a negative control for non-induced setting and IPTG induction since they don’t express LacI repressor, and as a positive control for heat or aTc induction, respectively. The plasmid maps for control plasmids are presented in the Supplementary Information (http://www.nature.com/nature/journal/v403/n6767/extref/403339ai1.pdf). The experiment lasted for 23.5 hours from plasmid transformation into E. coli JM2.300 to final sample measurement. During the experiment, samples were washed and diluted at three time points into fresh growth medium with (or without) inducers to preform transient switching between the states. Toggle switch bearing cells were first grown for 6 h in growth media supplemented with IPTG, which set all switches in high GFP expression state. Cells were then diluted in medium without any inducer to ensure high GFP state remained stable over the course of next 5 h. As expected, the pTAK102 control plasmid didn’t support stability and it retuned to ground state. At 11 h post-transformation, the high GFP/Repressor 1 state was perturbed by heat induction at 42 °C or dilution in growth medium with aTc. After 7 h (18 h time point), the cultures were assayed for GFP fluorescence and diluted into fresh medium at 37 °C (pIKE switches) or 32 °C (pTAK switches). The induction caused switching of the toggle switches by inhibiting the repression of LacI synthesis by CI λ repressor (pTAK plasmids – induction by heath) or Tet repressor (pIKE plasmid – induction with aTc). Consequently, the GFP production was downregulated by newly formed LacI repressor protein, except for control constructs pTAK106 and pTAK108 where GFPmut3 was expressed instead of lac repressor. Finally, 23.5 h after transformation, the functioning switches retained the low GFP expression state and the controls returned to uninduced state.

This experiment provided information on switching ability of the cloned genetic circuits. All four of Cl-LacI repressor pairs (i.e. pTAK DNA constructs) conferred bistability, while only one of the two TetR-LacI mutual repressor switches (i.e. pIKE DNA constructs) supported bistablility. The authors hypothesized that pIKE105 didn’t exhibit bistability due to insufficiently balanced maximal expression rates (α in mathematical model) from Ptrc-2 – rbs B and PLtetO-1 – rbs A promoter-RBS combinations in relation to higher CI repressor efficiency compared to Tet repressor. As accurately predicted form the model, less efficient synthesis of CI repressor (lower α1) due to modified RBS sequence successfully shifted the pIKE configuration from monostability (pIKE105) to bistability (pIKE107).

Stability of the switch

To test if genetic toggle switch can store its information (hold its state) for longer periods of time, the CI-LacI mutual repressor switch (pTAK117) was transformed into E. coli cells and left uninduced for 28 h. In addition, one group of transformed cells was induced with IPTG in the first 6 h of the experiment. Afterward, the culture was diluted in fresh unsupplemented medium and left without any inducers for 22 h. The cultures were sampled and diluted every 6-8.5 h. Both experiment groups successfully retained their state after 6 h time point, proving that mutual repressor switch design supports long-term stability of both predicted stable states even after several cell generations (E. coli doubling time in optimal conditions is approx. 20 min).

Toggle switch induction threshold

According to the mathematical toggle switch model, there exist three possible steady states. Two of them are stable and one occurs between these two and is highly unstable in nature. This state also represents the threshold if induction that is characteristic for each switch. At this tipping point, the cell population divides itself (or bifurcates) into two branches: one assumes high and the other low stable state of the switch. To investigate this phenomenon, the pTAK117 switch and pTAK102 control were grown in several increasing concentrations of IPTG inducer. The control exhibited usual dose-response sigmodial curve for induction of GFP expression due to Ptrc-2 promoter derepression. On the contrary, the genetic switch jumped from its ground, or low, state to high state in almost digital fashion at 40 μM IPTG concentration in growth medium. This result is expected, as there are only two predicted stable steady states the switch can adopt. The bifurcation event wasn’t perfect (there was still gradual increase in fluorescence detected) owing to random (stochastic) nature of gene expression within individual cells, which can differ from the average transcription induction of the whole bacterial population. This variability causes slight differences in induction threshold for individual cells and can be observed as wide population distributions in flow cytometry histograms as well as divided cell populations near the induction threshold (bifurcation event). This is understandable because deterministic ODE model can predict only average population behavior and doesn’t take into account stochastic events on cellular level (1, 6).

Switching time

An important characteristic of a toggle switch is its switching time. For a genetic switch to be useful as a form of cellular memory, it has to have a swift switching kinetics. The well-functioning pTAK117 switch was assayed for its switching time between high and low state and vice-versa. The cells in set low state were induced with IPTG and grown for different periods of time before being diluted in fresh medium and studied with the flow cytometer at 10.25 h time point. Similarly, the cells in induced high state were first diluted in fresh medium and grown at 42 °C for 35 min to 6 h. Afterwards, the cells were diluted again and grown at standard temperature (32 °C) until the end of experiment. The results show that pTAK117 toggle switch starts switching from low to high state after 3 h of induction and that all cells complete the switching event 6 h after addition of inducer. However, switching from high to low state is significantly more rapid as the whole cell population returned to ground state after 35 min of heath induction. Observed differences in switching times are consequences of different elimination mechanisms for λ repressor and LacI. While λ repressor is quickly denatured by heath, the IPTG-bound lac repressor is gradually diluted by subsequent cell divisions. In addition, the protein production rate for new LacI from PLs1con is significantly higher than CI production, making the ground state more stable in comparison to the high, CI-dominated state. Thus, it takes more time for CI to overcome the LacI-mediated repression.

Comment

To conclude, the discussed work of Gardner et al. presented the first successful introduction of a genetic toggle switch network in bacteria. Their approach combined mathematical modeling with genetic engineering to construct a functioning complex genetic circuit that paved way for extensive research of artificial genetic network in the last decade. The presented genetic toggle switch functions as a device that can store epigenetic memory of transient events in cells’ past across multiple generations which could be useful in advanced cell or gene therapy, biotechnology and biocomputing (1). An identical mutual repressor switch architecture was also used in design and implementation of a bistable switch in mammalian cells by Kramer et al. in 2004 (13).

References

1. T. Gardner, C. Cantor, J. Collins, Nature , 339–342 (2000). 2. J. Hasty, D. McMillen, F. Isaacs, J. Collins, Nat. Rev. Genet. 2 (2001). 3. J. L. Cherry, F. R. Adler, J. Theor. Biol. 203, 117–33 (2000). 4. J. Hasty, D. McMillen, J. Collins, Nature 420 (2002). 5. J. M. Berg, J. L. Tymoczko, L. Stryer, G. J. Gatto, Biochemistry (W. H. Freeman and Company, 7th ed., 2012). 6. M. Kaern, W. J. Blake, J. J. Collins, Annu. Rev. Biomed. Eng. 5, 179–206 (2003). 7. Copy number of plasmids with ColE1 origin of - Bacteria Escherichia coli - BNID 103857 (available at http://bionumbers.hms.harvard.edu/bionumber.aspx?&id=103857&ver=4). 8. T. Gardner, C. Cantor, J. Collins, Nature , SI (2000). 9. Part:BBa K200016 - parts.igem.org (available at http://parts.igem.org/Part:BBa_K200016). 10. A. Welman, J. Barraclough, Oncogenomics , 17–33 (2007). 11. J. B. Andersen et al., Appl. Environ. Microbiol. 64, 2240–6 (1998). 12. E. coli genotypes - OpenWetWare (available at http://openwetware.org/wiki/E._coli_genotypes#JM2.300). 13. B. P. Kramer et al., Nat. Biotechnol. 22, 867–70 (2004).