Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

An Electronic Analog of Synthetic Genetic Networks

  • Edward H. Hellen ,

    ehhellen@uncg.edu

    Affiliation Department of Physics and Astronomy, University of North Carolina Greensboro, Greensboro, North Carolina, United States of America

  • Evgenii Volkov,

    Affiliation Department of Theoretical Physics, Lebedev Physical Institute, Moscow, Russia

  • Jurgen Kurths,

    Affiliation Potsdam Institute for Climate Impact Research, Potsdam, Germany

  • Syamal Kumar Dana

    Affiliation Indian Institute of Chemical Biology, Council of Scientific and Industrial Research, Kolkata, India

An Electronic Analog of Synthetic Genetic Networks

  • Edward H. Hellen, 
  • Evgenii Volkov, 
  • Jurgen Kurths, 
  • Syamal Kumar Dana
PLOS
x

Abstract

An electronic analog of a synthetic genetic network known as the repressilator is proposed. The repressilator is a synthetic biological clock consisting of a cyclic inhibitory network of three negative regulatory genes which produces oscillations in the expressed protein concentrations. Compared to previous circuit analogs of the repressilator, the circuit here takes into account more accurately the kinetics of gene expression, inhibition, and protein degradation. A good agreement between circuit measurements and numerical prediction is observed. The circuit allows for easy control of the kinetic parameters thereby aiding investigations of large varieties of potential dynamics.

Introduction

The concept of synthesizing simple gene units to realize a desired function or to reproduce a known function is new [1] in biological systems. After confirmation of the unit's desired functional behavior, a large assembly of such units can be organized to perform complex biological functions [2][4]. This is like engineering small integrated chips to build a computer to derive a targeted function. In efforts towards engineering biological functions, a repressilator was first demonstrated as a synthetic genetic clock expressed in Escherchia coli [5] producing oscillations in expressed protein concentrations. A mathematical model based on standard chemical kinetics was also proposed that predicted the observed oscillations.

The dynamics of coupled synthetic genetic networks (SGNs) was also investigated [6][7] theoretically to understand the generation of synchronous rhythm in an assembly of repressilators via quorum sensing type interaction. Quorum sensing [8] is a form of exchanging information that a bacterial colony uses to develop a common rhythm. This quorum sensing type of indirect coupling is set-up between the SGN cells through diffusion of auto-inducing small molecules in a common medium. When the feedback via auto-inducing agents inside a SGN cell is reinforcing [7], the coupled dynamics show a state of in-phase synchrony, whereas when the feedback is repulsive, then the coupled dynamics show various possible states [9][11]: in-phase and anti-phase synchrony, inhomogeneous limit cycles, inhomogeneous steady states, and homogeneous steady states. Recently, in a biological experiment [12], evidence of in-phase synchronized quorum of genetic clock units was found. However, more complex features, as chaos, antiphase, and multistability in synchronous rhythm of coupled genetic clocks are yet to be observed experimentally.

Mathematical models are always a very useful tool to predict complex behaviors of dynamical systems using numerical simulations. Experimental verification of rich multistability requires an accurate knowledge of the model parameters which is often very challenging in biological experiments. An alternative experimental approach using electronic analogs of the SGN was undertaken [13][15] to confirm the numerical results and to search for possible coupled dynamics. Although it is difficult to simulate the biological experiment exactly in a circuit, the advantage of an electronic SGN is accessibility of system parameters and their controllability that allows a systematic exploration of known and predictable dynamics. Earlier [13][15] electronic circuit analogs of SGN displayed oscillations with 120° phase shifts between the oscillating variables, qualitatively in agreement with the genetic repressilator, however, the multistability of coupled SGNs was missing since access to and control of the system parameters was lacking.

In this paper, an electronic analog of the SGN is specifically designed to derive more accurate kinetic parameters of the repressilator. The goal is to control the parameters and thereby to realize desired sets of various kinetic parameters used in the simulations of the mathematical model. As a result, the circuit shows agreement between the measurements and the numerical predictions. The building block for the electronic repressilator is a circuit model for a single negative regulatory gene. This circuit shall be useful in a variety of other SGN investigations in addition to the repressilator. Designing electronic circuits of SGN also has the purpose of reverse engineering where knowledge of the biological experiments can be utilized for new technology and applications [16].

Methods

Genetic Network Repressilator

The structure of the repressilator consists of three repressive genes connected in a loop [5], with each gene producing repressor to the subsequent gene. The genes (i = 1,2,3) each produce their own mRNA, which translate the repressor protein. Gene 1′s repressor inhibits transcription of gene 2′s mRNA, 2′s repressor inhibits 3′s mRNA, and 3′s repressor inhibits 1′s mRNA. Taking into account standard chemical kinetics for production, degradation and inhibition, a dynamical system model of 6 first order differential equations was used for the mRNA and the protein concentrations [5].

An electronic analog of the repressilator was also proposed earlier [14][15], where they used three voltages as the variables, thus reducing the above model to a set of three differential equations. However, these circuits did not simulate the kinetic parameters of the repressilator model. A reduced three variable model is also used here but special care is taken to retain the parameters for the chemical kinetics as in the original model [5]. The reduced genetic network (RGN) repressilator is defined as,(1)where i = 1, 2, 3 for three genes, and the loop is closed by the condition x0 = x3. The product of the ith gene is xi, and its production is inhibited by xi-1. The parameter α accounts for the maximum transcription rate in the absence of an inhibitor, β is the decay rate of protein degradation and n is the Hill coefficient for inhibition. The Hill function is commonly used to account for sigmoidal binding kinetics. In this RGN model, there is no distinction between the mRNA and the transcribed repressor protein. We find that by this reduction, the fundamental features of the SGN model are not affected.

Description of the RGN Circuit

The building block for the RGN repressilator circuit is the circuit for a single negative regulatory gene shown in Fig. 1. The desired dynamics are given by (1), where xi corresponds to the output of the circuit and xi-1 is the input. The goal is to use a circuit which accounts for the kinetics of gene inhibition, production of repressor, and degradation of repressor. The gene's product is analogous to the charge coming from the collector of the transistor and the rate of production is the transistor current. The concentration of product xi is proportional to the voltage Vi across the capacitor, and the rate of decay is the current through RC. The kinetics of gene inhibition is determined by the circuitry that couples the input voltage Vi-1 (the inhibitor) to the voltage at the base of the transistor.

thumbnail
Figure 1. Electronic circuit analog of negative regulatory gene.

Vi-1 at the input is the concentration of inhibitory repressor. The circuit output Vi is the concentration of the gene's product. Vth accounts for the binding constant of the repressor to the gene's DNA. As inhibitor concentration Vi-1 increases past Vth, the voltage at the transistor base rises, turning the transistor's collector current off, thereby stopping gene production. Op-amps are LF412, diodes are 1N4148, pnp transistor is 2N3906, and +/−5 V supply.

https://doi.org/10.1371/journal.pone.0023286.g001

Circuit Analysis.

Here the circuit parameters are determined that correspond to particular values of kinetic parameters α, βi, and n in (1). The dynamical equation for voltage Vi is(2)where It(Vi-1) is the transistor's collector current and Vi-1 is the variable input voltage. The equations are expressed in dimensionless form using dimensionless time t/τ where the time-scale is chosen by τ  =  RCC0. The capacitor value is then Ci  =  C0/βi and (2) becomes(3)where the dot denotes time derivative in dimensionless time τ. In order for the circuit to model the gene kinetics it is desired that It(Vi-1) approximates the Hill function,(4)where xi-1 =  Vi-1/Vth and Vth represents an equilibrium constant for binding of repressor xi-1 to the gene's DNA. Imax is the maximum current through the transistor corresponding to gene transcription in the absence of an inhibitor. The production is half-maximal, It  =  Imax/2, when Vi-1  =  Vth. Dividing both sides of (3) by Vth, the α and βi are given by α  =  ImaxRC/Vth and βi  =  C0/Ci. Note that Imax is not due to saturation of the transistor since the voltage ImaxRC is chosen so that the emitter-collector voltage across the transistor does not reach zero. Instead Imax is due to saturation of the op-amp as discussed below.

Next we determine how the Hill coefficient n relates to the circuit parameters. The Hill function behavior consists of a transition of Vi from one value to another one as Vi-1 increases, with the transition occurring in the region around Vi-1  =  Vth as shown in Fig. 2. In addition the slope in the transition region is not constant. The circuit using the two op-amps U1 and U2 approximates this behavior by saturating the op-amp output and by using different gains in the transition region, a larger gain for Vi-1 < Vth and smaller for Vi-1 > Vth. Op-amp U1 is configured as a subtraction amplifier with output G1(Vi-1Vth)  =  G1ΔV where G1 is negative. Op-amp U2 has different gains G+2 and G-2 depending on the sign of ΔV. Taking saturation of the outputs into account gives the voltage at the output of U2,

thumbnail
Figure 2. Hill function inhibitory response.

Hill function inhibition of gene expression is approximated by the transistor current's dependence on input voltage Vi-1 for the circuit in Fig. 1. Hill function (red dashed line), predicted current (green solid line), and measured current (blue dots). Maximum transistor current of 3 mA corresponds to maximum transcription rate α = 60. The Hill coefficient is n = 3.75.

https://doi.org/10.1371/journal.pone.0023286.g002

(5)where V±sat are the saturation levels.

The next step is to consider how GΔV controls the transistor current. The voltage drop across Rb1 is(6)where the forward bias voltage drop across the diode is 0.6 V and VCC is the supply voltage, +5 V. The diode in series with Rb1 compensates for the transistor's emitter-base voltage drop, so that the voltage across Rb1 is approximately the same as the voltage across RE. The current It(Vi-1) through RE and the transistor is therefore (6) divided by RE. The maximum current Imax occurs when the output of U2 is saturated at GΔV  =  V-sat, giving(7)

For comparison with the Hill function it is useful to express It in terms of Imax and the normalized input voltage xi-1,(8)where Δxi-1  =  (xi-1−1).

In order to approximate the Hill function the overall gain G1G-2 for Vi-1<Vth is chosen such that slope dIt/dxi-1 of the transistor current matches the slope of the Hill function at xi-1 =  1 (Vi-1 = Vth). At xi-1 = 1, the output GΔV is not saturated, so G  =  G1G-2. Equating the slopes gives the condition relating Hill coefficient n to the overall gain G1G-2,(9)

Choosing the gain G+2 (for Vi-1>Vth) to be less than G-2 improves the transistor current's approximation to the Hill function. We find that choosing G+2 ≅0.3G-2 works well for a range of parameter values α, βi, and n. Fig. 2 shows the Hill function (red dashed) and the predicted (green solid) and measured (blue dots) transistor current for n = 3.75.

Model Parameters, Circuit Parameters, and Design Considerations.

Given a circuit it is useful to be able to easily determine the corresponding model parameters. From the previous section α, βi, and n, are expressed in terms of circuit parameters by(10)(11)(12)

As an example, in Fig. 1 the gain for op-amp U1 is G1 = −6.8 and gain for op-amp U2 is G-2 =  −22 for non-saturated overall gain G1G-2 = 150. For Vi-1 > Vth the gain G+2 is approximately −6.9. For Vth = 50 mV, C0 = 1 µf, Imax = 3 mA, supply VCC = 5 V, and LF412 op-amp saturation V−sat = −3.5 V, the resulting model parameters are α = 60, βi =  1, and n = 3.8.

It is also useful to be able to determine circuit parameters that achieve a desired set of model parameters. Starting with (10), ImaxRC must be far enough below the supply VCC so that the emitter-collector voltage of the transistor never reaches zero. For VCC = 5 V, ImaxRC is chosen to be around 3 volts and the emitter-collector voltage never gets less than about 1volt so that the transistor never goes into saturation. The choice of Imax has some freedom. For Imax =  3 mA, this determines RC = 1 kΩ and RE = 330 Ω in order to get voltage drops of 1, 1, and 3 volts across RE, the transistor, and RC, respectively, when It  =  Imax. The remaining free circuit parameter to adjust for a desired value of α is Vth. Rearranging (10) gives(13)

For example, for ImaxRC = 3 V, a value of α  = 100 is obtained using Vth ≅30 mV. The capacitor value Ci to use for a desired βi is easily given by (11) as(14)

The third circuit parameter is the overall gain G1G-2 which is determined by n and α. Rearranging (12) gives(15)

For VCC = 5 V and V−sat = −3.5 V, then VCC – 0.6− V−sat = 7.9 V, and with ImaxRC = 3 V, then G1G-2≈2/3. For example, if the desired parameter values are α  = 100 and n = 4, then the required overall gain is G1G-2≈267 which can be split as desired between G1 and G-2. Thus, the circuit parameters that are adjusted to obtain a desired set of model parameters are Vth, Ci, and G1G-2.

A careful selection of the op-amp is important for good approximation of the Hill function by the transistor current. The op-amp must be able to recover satisfactorily from saturation of its output. The circuit in Fig. 1 is tested with Vth set to zero and with Ci = 0. Results for the LF412 are shown in Fig. 3. The input voltage Vi-1 (blue line) is a triangle wave from a signal generator and the measured outputs are GΔV (red line) from the output of U2 and the final output voltage Vi (green line). The red curve shows that the LF412 saturates at V+sat  =  +4.5 V and at V-sat  =  −3.5 V when using a ±5 V supply, and that the circuit makes the transition from high gain to low gain when Vi-1 goes from negative to positive corresponding to xi-1 surpassing one. The green line shows the expected inhibitory response with respect to input Vi-1, and the maximum value of ImaxRC = 3 V.

thumbnail
Figure 3. Measured time response of single gene circuit.

Time response for the single gene circuit in Fig. 1 with Vth = 0 and Ci = 0. Input Vi-1 (blue), op-amp output V (red), final output Vi (green). The red curve shows saturation of the op-amp output at +4.5 and –3.5 V when using a +/–5 V supply. The change in slope when Vi-1  =  Vth is also apparent.

https://doi.org/10.1371/journal.pone.0023286.g003

Repressilator Circuit.

The electronic repressilator circuit consists of three negative regulatory gene circuits (Fig. 1) connected in an inhibitory loop as shown in Fig. 4. The triangle symbol contains the 2 op-amps, transistor, and circuitry which determine parameters α and n.

thumbnail
Figure 4. Electronic repressilator.

Repressilator circuit constructed from a loop of three of the negative regulatory gene circuits of Fig. 1.

https://doi.org/10.1371/journal.pone.0023286.g004

Results and Discussion

The model parameters α, βi, and n are now ably determined by circuit parameters. Our results of circuit measurements and numerical predictions are shown for three cases: (1) identical genes, β-ratio  = 1∶1∶1; (2) gene i = 1 with faster decay, β-ratio  = 3∶1∶1; (3) gene i  = 1 with slower decay, β-ratio  = 0.3∶1∶1. Gene products are the normalized voltages x1 (blue), x2 (red), and x3 (green). Circuit measurements are solid lines, numerical predictions are dashed lines. The circuit parameters Vth and overall gain G1G-2 were varied using (13) and (15) in order to set α and n. Vth varied from 30 mV to 120 mV, and G1G-2 from 73 to 220 corresponding to α = 25 to 100 and n = 3.0 to 6.6. The figures show results for (α, n)  =  (50, 6.6) and (100, 3.3). Other sets of parameters produced results with similar agreement of measurements and numerical predictions. The time constant was τ  =  RCC0  =  (1 kΩ)(1 μf)  = 1 ms. It follows that Ci =  (1 µf)/βi.

Fig. 5 shows the repressilator dynamics for β-ratio = 1∶1∶1 where Ci = 1 µf for each capacitor. The three state variables have the same shape, but with 120° phase shift. Numerical predictions are in close agreement with the measurements.

thumbnail
Figure 5. Electronic repressilator dynamics.

Normalized voltages xi measured from RGN repressilator circuit (solid) and numerical predictions (dashed). β-ratio  = 1∶1∶1. Panel A (α = 50, n = 6.6); Panel B (α = 100, n = 3.3).

https://doi.org/10.1371/journal.pone.0023286.g005

For the dynamics in Fig. 6 gene i = 1 has its capacitor reduced to 0.33 µf, so it has β1 = 3. Thus the β -ratio for the genes in the repressilator circuit is 3∶1∶1. Increasing the gene product's decay rate causes larger oscillations for the product, reduced oscillations for the gene's inhibitor, and an increased oscillation frequency for the repressilator.

thumbnail
Figure 6. Electronic repressilator dynamics with one fast decay rate.

Normalized voltages xi measured from RGN repressilator circuit (solid) and numerical predictions (dashed). β-ratio  = 3∶1∶1. Panel A (α = 50, n = 6.6); Panel B (α = 100, n = 3.3).

https://doi.org/10.1371/journal.pone.0023286.g006

For the dynamics in Fig. 7 electronic gene i = 1 has its capacitor increased to 3.3 μf, so it has β1 = 0.3. Thus the β-ratio for the genes in the repressilator circuit is 0.3∶1∶1. Gene i = 1 now has reduced oscillations, its inhibitor gene i = 3 has increased oscillations, and the repressilator frequency has decreased.

thumbnail
Figure 7. Electronic repressilator dynamics with one slow decay rate.

Normalized voltages xi measured from RGN repressilator circuit (solid) and numerical predictions (dashed). β-ratio  = 0.3∶1∶1. Panel A (α = 50, n = 6.6); Panel B (α = 100, n = 3.3).

https://doi.org/10.1371/journal.pone.0023286.g007

The circuit presented here as an electronic analog of a synthetic genetic network known as the repressilator shows good agreement between experimental measurements and numerical predictions. The circuit includes control of parameters for the Hill function which is used to model the kinetics of gene expression and inhibition in the cyclic 3-gene network. Previous electronic analogs of the repressilator [14][15] did not concentrate on the kinetics and control of the parameters, and thereby did not capture many complex dynamical features. With the ability to control the model parameters, this circuit will be useful for investigations of multistability of coupled repressilators as well as for other SGN dynamics.

Acknowledgments

The authors thank Siddhartha Roy, Indian Institute of Chemical Biology, for valuable comments and suggestions.

Author Contributions

Conceived and designed the experiments: EHH EV JK SKD. Performed the experiments: EHH. Analyzed the data: EHH SKD. Wrote the paper: EHH SKD.

References

  1. 1. Elowitz M, Lim WA (2010) Build life to understand it. Nature 468: 889–890.M. ElowitzWA Lim2010Build life to understand it.Nature468889890
  2. 2. Gardner TS, Cantor CR, Collins JJ (2000) Construction of a genetic toggle switch in Escherichia Coli. Nature (London) 403: 339–342.TS GardnerCR CantorJJ Collins2000Construction of a genetic toggle switch in Escherichia Coli.Nature (London)403339342
  3. 3. You L, Cox RS, Weiss R, Arnold FH (2004) Programmed population control by cell-cell communication and regulated killing. Nature 428: 868–871.L. YouRS CoxR. WeissFH Arnold2004Programmed population control by cell-cell communication and regulated killing.Nature428868871
  4. 4. Bell-Pederson D, Cassone VM, Earnest DJ, Golden SS, Hardin PE, et al. (2005) Circadian rhythms from multiple oscillators: lessons from diverse organisms. Nat Rev Genet 6: 544–556.D. Bell-PedersonVM CassoneDJ EarnestSS GoldenPE Hardin2005Circadian rhythms from multiple oscillators: lessons from diverse organisms.Nat Rev Genet6544556
  5. 5. Elowitz MB, Leibler S (2000) A synthetic oscillatory network of transcriptional regulators. Nature 403: 335–338.MB ElowitzS. Leibler2000A synthetic oscillatory network of transcriptional regulators.Nature403335338
  6. 6. Yamaguchi S, Isejima H, Matsuo T, Okura R, Yagita K, et al. (2003) Synchronization of cellular clocks in the suprachiasmatic nucleus. Science 302: 1408–1412.S. YamaguchiH. IsejimaT. MatsuoR. OkuraK. Yagita2003Synchronization of cellular clocks in the suprachiasmatic nucleus.Science30214081412
  7. 7. García-Ojalvo J, Elowitz M, Strogatz SH (2004) Modelling a synthetic multicellular clock: repressilators coupled by quorum sensing. Proc Nat Acad Sci 101: 10955–10960.J. García-OjalvoM. ElowitzSH Strogatz2004Modelling a synthetic multicellular clock: repressilators coupled by quorum sensing.Proc Nat Acad Sci1011095510960
  8. 8. Lerat E, Moran NA (2004) Evolutionary history of quorum-sensing systems in bacteria. Molecular Biology and Evolution 21: 903–913.E. LeratNA Moran2004Evolutionary history of quorum-sensing systems in bacteria.Molecular Biology and Evolution21903913
  9. 9. Ullner E, Zaikin A, Volkov EI, García-Ojalvo J (2007) Multistability and clustering in a population of synthetic genetic oscillators via phase-repulsive cell-to-cell communication. Phy Rev Lett 99: 148103.E. UllnerA. ZaikinEI VolkovJ. García-Ojalvo2007Multistability and clustering in a population of synthetic genetic oscillators via phase-repulsive cell-to-cell communication.Phy Rev Lett99148103
  10. 10. Ullner E, Koseska A, Kurths J, Volkov EI, Kantz H, et al. (2008) Multistability of synthetic genetic networks with repressive cell-to-cell communication. Phy Rev E 78: 031904.E. UllnerA. KoseskaJ. KurthsEI VolkovH. Kantz2008Multistability of synthetic genetic networks with repressive cell-to-cell communication.Phy Rev E78031904
  11. 11. Potapov I, Volkov E, Kuznetsov A (2011) Dynamics of coupled repressilators: The role of mRNA kinetics and Transcription cooperativity. Phy Rev E 83: 031901.I. PotapovE. VolkovA. Kuznetsov2011Dynamics of coupled repressilators: The role of mRNA kinetics and Transcription cooperativity.Phy Rev E83031901
  12. 12. Danino T, Mondragón-Palomino O, Tsimring L, Hasty J (2010) A Synchronized quorum of genetic clocks. Nature 463: 326–330.T. DaninoO. Mondragón-PalominoL. TsimringJ. Hasty2010A Synchronized quorum of genetic clocks.Nature463326330
  13. 13. Mason J, Linsay PS, Collins JJ, Glass L (2004) Evolving complex dynamics in electronic models of genetic networks. Chaos 14: 707–715.J. MasonPS LinsayJJ CollinsL. Glass2004Evolving complex dynamics in electronic models of genetic networks.Chaos14707715
  14. 14. Wagemakers A, Buldú JM, García-Ojalvo J, Sanjuán MAF (2006) Synchronization of electronic genetic networks. Chaos 16: 013127.A. WagemakersJM BuldúJ. García-OjalvoMAF Sanjuán2006Synchronization of electronic genetic networks.Chaos16013127
  15. 15. Buldú JM, García-Ojalvo J, Wagemakers A, Sanjuán MAF (2007) Electronic design of synthetic genetic networks. Int J Bifur Chaos 17: 3507–3511.JM BuldúJ. García-OjalvoA. WagemakersMAF Sanjuán2007Electronic design of synthetic genetic networks.Int J Bifur Chaos1735073511
  16. 16. Lucks JB, Arkin AP (2011) The hunt for the biological transistor: How genetic circuits will unlock the true potential of bioengineering. IEEE Spectrum 48: 34–39.JB LucksAP Arkin2011The hunt for the biological transistor: How genetic circuits will unlock the true potential of bioengineering.IEEE Spectrum483439