The Value of Tools in Biology
Smolke Lab talk 11-1-06
Framework
• Thesis: our ability to understand and manipulate biology is limited by the quality and scope of our tools
– cellular understanding - what determines the cell's behavior?
– cellular manipulation - how can we control the cell's behavior?
Quantizing Biology
• cellular behavior is determined by physical properties and their variation in time: – Structures – Locations – Energies – Numbers
• Cellular processes often manipulate these quantities in tandem
Natural Systems
• For example, transcriptional processes separate mechanisms for controlling protein (Number) vs (Structure):– Structure then determines the protein’s Location and
Energies, and thereby its function
Independence of Tools• If we could manipulate cellular quantities independently, then more states
would be reachable.– Analogy: like building a house with (nails, a hammer, and a saw) vs with a (nails-
hammer-saw)• We can reappropriate natural systems for our own purposes, but their
independent use is limited.– Example: PCR borrows from the transcriptional network. Some sequences of
DNA are difficult to amplify.• Complete independence is not always possible
– Example: the necessary connection between protein Structure and Energy, which limits functions.
A closer look at Number
• Control over protein number is affected by cellular noise sources– Extrinsic noise: variation in environmental
conditions. (temperature, nutrients, signals)– Intrinsic noise: follows from the stochastic
nature of protein formation
• Laboratory experiments often focus on reducing extrinsic noise– Repeated trials reduces measurement
variance
A Simple model
• Protein produced at an average rate of λ proteins/sec– No RNA, no protein decay – Instrinsic noise is the single cell probability distribution– Extrinsic noise is the sum of many cellular distributions
Adding the effects of Translation
• Translation efficiency is a major source of noise– variance of many small
steps is less than that of fewer large steps
– Translation amplifies transcriptional variation in addition to adding noise
Ozbodak PMID: 11967532
Qualities of protein Number
• Mean and the Variance are both important for cellular behavior
• Example: robustness – Mean influences most probable action
• Cellular robustness through error control averaging
– Variance influences probability of alternative actions• population robustness through diversity
Independent control of protein Number
• Goal: control over the mean and variance of cellular protein– Mean controlled by protein production rates– Variance controlled by feedback on rates
• negative feedback on protein production reduces variance– More protein lower rate less production less protein
– Less protein higher rate more production more protein
Protein Auto-regulation
• Transcriptional feedback: production of a repressor that inhibits transcription
• Becskei PMID: 10850721
• Translational feedback: production of a protein that decreases RNA stability– More efficient at reducing
relative variance– Higher metabolic cost
• Swain PMID: 15544806
A Physical Feedback Mechanism
• Translational regulation via modulation of RNA decay rate– RNA degraded though endogeneous endo/exo-
nuclease pathways in E. Coli– 5’ and 3’ hairpins increase the stability of RNA
RNA modulation
• Removal of protective hairpins decreases stability of RNA transcript less protein produced– Yeast Rnt1p cleaves RNA
hairpins with high sequence specificity
• Express Rnt1p from the protected RNA transcript, closing the feedback loop– Possibility of an orthogonal,
modular feedback system
RNA hairpin substrate specificity
• Rnt1p recognizes sequence dependent domains
• E. Coli RNaseIII also cleaves dsRNA with some sequence dependence
• Goal: high Rnt1p activity, low E. Coli RNaseIII activity– Orthogonal system
Lamontagne PMID: 14581474
System Modularity
• Independence of functional parts:– 5’ and 3’ protective hairpin sequences
determine lifetime control of protein number mean
– Rnt1p hairpin sequence determines level of feedback control of protein number variance
– Hairpin libraries tuning of variance and mean
Correlated Expression of YFGOI
• Polycistronic coding regions have correlated expression levels– Express any other
protein on the same transcript
– Use GFPuv for testing purposes
– Additional correlation if using same RBS
Controls• Open loop system: Rnt1p on
separate plasmid no feedback1. Test for Rnt1p substrate
cleavage and RNA destabilization after the expression of Rnt1p
2. Test for no destabilization with non-active Rnt1p hairpins
3. Test for no destabilization without Rnt1p hairpins
4. Test for no destabilization without protectice 5’ and 3’ hairpins
– With additional combinations for individual 5’ vs 3’ testing if necessary
Applications of controlled variance
• Any decision can be modelled as maximizing over some Utility function
• Cells make decisions to express or not express a specific protein with a certain probability– Rewarded if choice is correct– Penalized if choice is incorrect
• Engineering systems have their own Utility functions
Low Number protein expression
• Proteins toxic in large numbers
• Low number expression is difficult, due to relatively high variance at small N
• Variance control through feedback provides higher net population fitness
Signal Rectification
• Electronic Digital circuits scale well due to voltage rectification after every computation
• In contrast, in electronic Analog circuits, errors can propogate and amplify uncontrollably
• Chemical rectification may be a useful method for reducing error propogation between separate circuit elements– Allowing for larger, more
complicated synthetic circuits and computations
Measurement Probe
• Remember that every measurement is actually the result of many individual measurements of individual cells– Reducing intrinsic
cellular noise increases the accuracy of measurements
Conclusions
• Tools for Independent manipulation of cellular quantities are intrinsically useful
• Negative Feedback as a method for control of number variance
• Modular Rnt1p system for orthogonal control of protein variance in E. Coli
• Circuit designs using low variance systems
Future plans
• Cloning
• Cloning
• Cloning
• Cloning
• More cloning…