Modular Cell Biology: Retroactivity and Insulation

Modular Cell Biology: Retroactivity and Insulation Domitilla Del Vecchio EECS, University of Michigan at Ann Arbor MechE, MIT Oxford, September 2009 TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAAAAAA

Modularity: A fundamental property Modularity guarantees that the input/output behavior of a component (a module) does not change upon interconnection. Electronics and Control Systems Engineering rely on modularity to predict the behavior of a complex network by the behavior of the composing subsystems. Result: Computers, Videos, cell phones… Internal circuitry of an OPAMP: It is composed of well defined modules Functional modules seem to recur also in biological networks (e.g. Alon (2007)). But… But can they be interconnected and still maintain their behavior unchanged? If not, what mechanism can be used to interconnect modules without altering their behavior? Does nature already employ such mechanisms? The Emergent integrated circuit of the cell [Hanahan & Weinberg (2000)]

Modularity: A grand challenge in synthetic biology X Z Repressilator (Experimental Results) (Elowitz and Leibler, Nature 2000) Courtesy of Elowitz Lab at Caltech WORKING “MODULES” NOT WORKING INTERCONNECTIONS !

LacI-rep NRI-act glnKp glnG lacI IPTG LOAD Courtesy of Ninfa Lab at Umich Modularity is not a natural property of bio-molecular circuits Experimental data (Atkinson et al, Cell 2003) Retroactivity! How do we model these effects? How do we prevent them?

Outline • A modeling framework for systems with retroactivity • Retroactivity in transcriptional networks • A lesson from OPAMPs: Insulation devices • Design of a bio-molecular insulation device based on phosphorylation/dephosphorylation • Fast time-scales as a key mechanism for insulation

y u u y’ A systems theory with retroactivity Basic Idea: The interconnection changes the behavior of the upstream system Familiar Examples:

A systems theory with retroactivity y u r s Retroactivity to the output Retroactivity to the input Def: The I/O model of the isolated system is obtained when s=0 and when r is not an additional output The interconnection of two systems is possible only when the internal state variable sets are disjoint: u2=y1 u2 y1 r2 s1 s1=r2

X Z Gene regulatory circuitry: A network of transcriptional modules A transcriptional component is typically viewed as an input/output module But, is its input/output response unchanged upon interconnection?

Downstream component (isolated) (connected) s Retroactivity in transcriptional networks has dramatic effects on the dynamics

(1D) Connected system (2D) Isolated system s To compare the X dynamics we seek a 1D approximation for the connected system: Measure of retroactivity will be given by Measure of the retroactivity We seek to quantify the difference in the dynamics of the state X between the connected and isolated system

Calculation of s We exploit the time-scale separation between the output X dynamics and the dynamics of the input stage of the downstream component

Isolated system dynamics: Approximate connected system dynamics: percentage difference between the isolated system dynamics and the approximate connected system dynamics The value of the retroactivity measure for the interconnection through transcriptional regulation Meaning and value of Del Vecchio et al., Nature Molecular Systems Biology 2008

Downstream component (isolated) (connected) Effect of R(X) on the dynamics Retroactivity shifts the poles of the transfer function of the linearized system toward low frequency

r≈0 Dealing with retroactivity: Insulation devices In general, we cannot design the downstream system (the load) such that it has low retroactivity. But, we can design an insulation system to be placed between the upstream and downstream systems. u y s • The retroactivity to the input is approx zero: r≈0 • 2. The retroactivity to the output s is attenuated • 3. The output is proportional to the input: y=c u

Non-inverting amplifier: because the input stage of an OPAMP absorbs almost zero current Choose the biochemical parameters of the input stage to allow a small value of For example: Reaching small retroactivity to the input r

r≈ 0 Dealing with retroactivity: Insulation devices In general, we cannot design the downstream system (the load) such that it has low retroactivity. But, we can design an insulation system to be placed between the upstream and downstream systems. u y s • The retroactivity to the input is approx zero: r≈0 • 2. The retroactivity to the output s is attenuated • 3. The output is proportional to the input: y=c u

Non-inverting amplifier: For G large enough: Conceptually: Attenuation of the retroactivity to the output “s”: Large feedback and large amplification

Apply large input amplification G and large output feedback G’ Attenuation of the retroactivity to the output “s” in the transcriptional component Isolated system Connected system approximated dynamics How do we realize a large input amplification and a large negative feedback?

Feedback through dephosphorylation Amplification through phosphorylation Phospho/Dephospho Reactions: r s Full ODE Model A phosphorylation-based design for a bio-molecular insulation device

s Simplified analysis: Why should it attenuate “s”? Del Vecchio et al., Nature Molecular Systems Biology 2008

Full system: The fast time-scale of the device is a key feature for attenuating “s” Phosphorylation and dephosphorylation reactions are often much faster than protein production and decay:

Interconnection through binding/unbinding (possibly large) Large Fast time scales: A key mechanism for insulation Basic Idea: Claim: if G is large enough, signal x at the QSS is not affected by y.

Fast time scales: A key mechanism for insulation Why would it work? x(t) does not depend on y on the slow manifold Del Vecchio and Jayanthi, CDC 2008

Simulation results for the pho/depho insulation device Slow time-scale Fast time-scale The fast time-scale of the phosphorylation cycle allows to reach insensitivity to very large loads (p=100) Xp for the isolated system Xp for the connected system

We have proposed a systems theory with retroactivity We have provided a measure of retroactivity in transcriptional networks We have introduced the notion of insulation device r=0 We have presented a general (very well known in control systems engineering) mechanism to attenuate retroactivity to the output Futile cycles, which are ubiquitous in natural signal transduction systems are excellent insulation device: they use time-scale separation as an insulation mechanism Conclusions

Thanks to: • Alexander J. Ninfa (University of Michigan Medical School, Prof. of Biological Chemistry) • Eduardo D. Sontag (Rutgers University, Prof. of Mathematics) • Sofia Merajver (University of Michigan Medical School, Cancer Center, Medical Doctor) • Shridhar Jayanthi (EE: Systems, University of Michigan, graduate student) • Hamid Ossareh (EE: Systems, University of Michigan, graduate student) • Prasanna Varadarajan (ME, University of Michigan, graduate student) • Polina Mlynarzh (BME, University of Michigan, graduate student) • Rackham Graduate School at University of Michigan/ CCMB/AFOSR

Downstream component High gains improve signal-to-noise ratio… Bio-molecular processes are intrinsically stochastic How do high gains (required for retroactivity attenuation) impact noise? Courtesy of Elowitz lab (Caltech) calculated by linearizing the system about its equilibrium corresponding to calculated the Fokker-Planck Equation deriving from the Master Equation

Downstream component …but they also increase intrinsic noise at higher frequency Use linearized Langevin approximation Jayanthi and Del Vecchio, CDC 2009

Approx linear input/output relationship: Resulting input/output gain The parts of the insulation device can be designed so as to have small “r” Retroactivity r after a fast transient is small if:

IPTG LacI repressor lac promoter lacY lacA lacZ LacI binding sites (lacOp operators) Ongoing and Future Work Experimental demonstration of genetic retroactivity in living cells (with Shridhar Jayanthi (EECS) and Alexander J Ninfa (Med School) at Umich) Periodic and step injection routines

IPTG Output Lac Repressor kinase (NRII L16R) lac promoter glnK promoter RFP lacZ NRI~P binding sites (enhancers) NRI~P NRI phosphatase (NRII H139N) Amplifier/Insulator Ongoing and Future Work Construction of a phosphorylation-based insulation device (with Shridhar Jayanthi (EECS) and Alexander J Ninfa (Med School) at Umich)

Ongoing and Future Work Computation of a retroactivity measure in signaling pathways and of the “dampening” factor across stages (with Alejandra Ventura and Sofia Merajver in the Cancer Center at Umich) Upstream Effect Downstream Perturbation Does nature uses insulation devices by accident? Can we show that some natural systems would not work they way they work if the phosphorylation/dephosphorylation signaling cascades did not enjoy insulation properties? How general/descriptive is the system modeling with retroactivity? (with Eduardo D. Sontag at Rutgers)

Higher gains may contribute to higher biological noise Faster phosphorylation and dephosphorylation reactions lead to higher amplification and feedback gains (“higher OPAMP amplification”), which lead to higher coefficient of variation.

Downstream component (isolated) (connected) Effect of R(X) on the dynamics Retroactivity shifts the poles of the transfer function of the linearized system toward low frequency

Modular Cell Biology: Retroactivity and Insulation