Boosting functionality of synthetic DNA circuits with tailored deactivation

Molecular programming takes advantage of synthetic nucleic acid biochemistry to assemble networks of reactions, in vitro, with the double goal of better understanding cellular regulation and providing information-processing capabilities to man-made chemical systems. The function of molecular circuits is deeply related to their topological structure, but dynamical features (rate laws) also play a critical role. Here we introduce a mechanism to tune the nonlinearities associated with individual nodes of a synthetic network. This mechanism is based on programming deactivation laws using dedicated saturable pathways. We demonstrate this approach through the conversion of a single-node homoeostatic network into a bistable and reversible switch. Furthermore, we prove its generality by adding new functions to the library of reported man-made molecular devices: a system with three addressable bits of memory, and the first DNA-encoded excitable circuit. Specific saturable deactivation pathways thus greatly enrich the functional capability of a given circuit topology.

data. The authors demonstrate that their bistable system can be switched on or off multiple times, study how the autocatalytic reaction order changes in the presence of the drain, and finally test the behavior of multiple switches operating in parallel. -Although the whole project is nicely put together, I would not say that tuning degradation rates to determine dynamic/static behaviors in a network is a majorly new idea. This might be one of the few papers to spell the idea out, but this principle is routinely used in synthetic biology.
-The PEN toolbox is not new, and although a significant amount of labor went into these experiments, the overall contribution seems incremental. A major issue with the PEN toolbox is that it is largely incompatible with other DNA nanotechnology systems, and it is unclear what is its usefulness, beyond the synthesis of molecular reaction networks for proof of principle. -The paper starts off suggesting the following recipe: to obtain bistable behavior it is sufficient to design a system with 3 steady states. This recipe is misleading, because this is not true in general. The presence of 3 steady states guarantees bistability only if the system satisfies other assumptions, such as monotonicity and dissipativity -see, for example, Angeli, Ferrell and Sontag, PNAS 2004. It is unlikely that this system satisfies such assumptions, so the presence of 3 steady states does not automatically imply bistability. (Bistability might just be a lucky coincidence.) -Demonstration that this bistable system can be toggled between states is the most interesting contribution of this paper, as far as I am concerned. Unfortunately, this sections is somewhat poorly written. There is mention of a degradable inhibitor, but there is no explanation of what additional reaction this is. It would be very helpful to mention that reporter systems and normalization procedures are described in the Methods section at the end of the paper.
-Multistable memory units: the way these results are purported is controversial. Essentially, this is a demonstration that one can operate together several individual bistable circuits that are functionally disconnected from each other. I would not call this a multistable system, rather an array of bistable systems. These bistable circuits may be coupled indirectly because they compete for enzymatic resources or by unwanted binding of strands (cross-talk), but it looks like these are not significant problems. Accurate sequence design to avoid crosstalk is routinely done in DNA nanotechnology, with demonstrations of dozens of components operating together.
A better way to spin the importance of these results is that they indicate that it is possible to build a 2 bit or a 3 bit system, which -based on the results at Fig. 5, might be reversibly switched.
-In the discussion section, the authors observe that a decoy strand that simply sequesters \alpha (or input, referring to Fig. 2) without degrading it is not sufficient to obtain bistability, although it (unsurprisingly) introduces delay. Well, this decoy can't really be compared with a degradation rate, so I am not sure how interesting this is.
-What is the contribution of the drain/sink in terms of competition for DNA polymerase? What if in practice bistability was also due to the fact that production decreases due to competition between drain and template? A reduction in free DNA polymerase concentration could be consistent with the transition to a second order rate-like behavior; because there is less DNA polymerase available (some of it is wasted to produce inert copies of \alpha from the drain), then two copies of \alpha are required for a net increase of one copy of \alpha. If my reasoning is correct, then it might be misleading to claim that bistability was obtained by exclusively tuning the degradation rates in the system. Because of the way the system is built, there may be an indirect change in the production rates as well.
-At page 8, I found the approach to derive information on rate order from delay to be interesting. However, I was confused by the reasoning "the order of the reaction is difficult to assess directly from the shape of the time traces, because the fluorescence signal is a composite of many species contributions". Wouldn't the delay also be a quantity that reflects the behavior of whole sample, because it is determined from that same curve that "is a composite of many species contributions"? I can buy the idea that initially the drain reaction dominates before autocatalysis kicks in, so by assessing the time delay it takes before autocatalysis goes off is an indirect readout of the kinetics of the drain/input dynamics. However, I suggest that these paragraphs are rewritten. As for the conclusion of this paragraph, it seems to me rather obvious that by introducing new reactions one does not only change the stationary behavior of a system, but also its kinetics.
-The paper needs to be read more carefully, to make sure that the necessary information is provided timely and accurately to the reader. I found several omitted definitions, and unclear sentences; for example, at the top of p 8 there is a sentence "results obtained with the other nicking enzyme used in PEN systems (Nt.BstNBI)", that seems completely disconnected from the rest; so far no specific enzyme was mentioned in the main text -prompting the reader to confusion. Fig 2 and the main text should explain what enzymes are being used; what are the system inputs and what are its outputs or readouts. The paragraph in the middle of page 2 (introduction) describing the PEN toolbox is too vague and requires readers to look up earlier papers. The time C_t it takes for the system to spontaneously switch steady state from low to high is used early in the paper (Fig 2), but it is only loosely defined at the end of page 8. Please provide a clear explanation of how this time was computed.
-The supplement should be cleaned up and better connected to the main paper. References in the supplement are not compiled and all appear as question marks.
Reviewer #2 (Remarks to the Author) The authors described a simple and versatile method to change the nonlinearity of the network by regulating the degradation kinetics and created a bistable switch without changing the network topology. In addition, they showed several possible systems by using this bistable switch.
The proposed systems (i.e. memory, excitable molecular network) lacks novelty and usefulness because the systems in and of themselves have been already reported and authors did not show real world applications. However, the concept that the linearity can be changed by tuning the degradation kinetics without changing of the network topology is a significant idea for extending the network functions. Thus, this manuscript seems to be sufficient to publish in Nature Communications if the idea has not been reported yet.
Authors should address the following questions and concerns.
Question 1: Authors describe "see SI Appendix for a mathematical discussion" in page 5, but do not show section number or figure number. Authors should point where the discussion about "additional degradation path" and "small bottleneck" is in "SI appendix for mathematical discussion".
Question 2: Authors draw regression lines in figure 3-B. However, the number of measurement point is too few, for example, the regression line of drain-alpha 4 seems to be calculated by using single measurement value. Authors should measure the 1/Ct at more than three drain concentrations for each line in order to verify that the values of 1/Ct are lined on the straight lines. In addition, authors should add error bars in all plots. Question 3: In figure 5, authors use dye-labeled template to measure the ON and OFF state. However, ON and OFF state should be defined by the amount of produced DNAs rather than template duplex. So, authors should show the relationship between the fluorescent signals and the amount of the product by direct quantification of the product by other method like qPCR.

Question 4:
Author describe that two independent bistable switches can make four stable states in figure 7-B. I expect that "no trigger" and "beta" shows equivalent Cy3.5 signal because the "beta" does not interact with the gamma-template which is labeled with Cy3.5. Similarly, the value of the beta bistable switch in figure 7-E increases in the presence of other triggers. I'm confused because authors describe "independent" but the results seems not to be independent. Authors should mention about this conflict.

Question 5:
Although the delta Gs of the drain templates in table 1 seems to be almost the same, why the trigger concentrations after 200 min are differ in alpha, beta and gamma in figure 7-E? According to authors' claim, I expect that the same binding affinity results in the same kinetics of products, or the same amount of products. Author should explain why the amount of the products differs and experimentally verified the explanation.   -Although the whole project is nicely put together, I would not say that tuning degradation rates to determine dynamic/static behaviors in a network is a majorly new idea. This might be one of the few papers to spell the idea out, but this principle is routinely used in synthetic biology.
-The PEN toolbox is not new, and although a significant amount of labor went into these experiments, the overall contribution seems incremental. A major issue with the PEN toolbox is that it is largely incompatible with other DNA nanotechnology systems, and it is unclear what is its usefulness, beyond the synthesis of molecular reaction networks for proof of principle.
We wich to thank Reviewer#1 for the positive appreciations (and for the many useful suggestions listed below), but we kindly disagree with the last two statements. * First, as explained below, we do not merely tune the degradation kinetic rate. This would never drag the system into a bistable regime, because tuning the rate does not alter the linearity/nonlinearity of the behavior (note that of course, it is trivial to tune degradation rates in our in vitro experiments, by simply adjusting the exonuclease concentration. We do that routinely for the fine-tuning of new circuits).
Instead, in this report, we explain that tuning the shape of the degradation curve, (i.e. the kinetic law, not the kinetic rate) is much more powerful, and we precisely demonstrate a way to do that in the context of the PEN toolbox. In this regard, as rightly noted below by the reviewer, Fig. 6 is a very important part of the paper, as it demonstrates that crafting the degradation kinetic law allows to emulate second order production dynamics (something that could obviously not be achieved by playing only with the degradation rate).
Indeed, synthetic biology sometimes makes use of degradation tags (such as ssrA) to accelerate turnover of protein components by directing them to efficient endogenous or exogenous proteases (and in some cases this may have unexpected nonlinear consequences as detailed in my PRL 2012 paper 1 ), but this is typically seen merely as a way to balance production and degradation rates. On the contrary, our approach to compensate for featureless production dynamics by constructing ad hoc "bumpy" and species-specific degradation curves is a rational, new and general strategy to achieve non-trivial dynamics, e.g. multistability or excitability as demonstrated in the paper. * Regarding the compatibility and usefulness of the PEN toolbox, we think that the conclusion of the reviewer is a bit hurried, as explained below. In any case, we think that this personal opinion should not directly cause the rejection of this manuscript, which, in addition to introducing a new efficient tool for PEN systems, has the more universal merit of bringing the attention to the degradation pathway as a powerful approach to craft interesting circuits and manage nonlinearities.
One has to note that the PEN toolbox is still a relative newcomer to the field of Molecular Programming (with a first paper in 2011), whereas competing approaches have a much longer history (e.g. toehold mediated strand displacement originated in 2000, and genelets in 2004).
True, we have spent the last five years solving many "proof of principle" challenges in the design of synthetic dynamics. But, together with other groups, we are now very much involved in taking advantage of the robustness/versatility of this system to design practical applications. [redacted] The remark on compatibility is also undeserved in our opinion. Compatibility is tricky for any molecular well-mixed system, but DNA-based approaches have a clear strength in this respect, because DNA can connect to a huge range of other chemistries or physical processes. And indeed, the PEN toolbox is doing quite well in this respect. [redacted] For example Team Sendai at BIOMOD2014 had a functional system (now an oral presentation at DNA22, see http://www.dna-node.com/dna22/accepted-papers-and-abstracts/index.html) that uses a combination of PEN DNA and toehold-mediated strand displacement.
We hope these incomplete arguments will convince the reviewer#1 to reserve its conclusion on PEN for a few more years, as we believe that many exciting applications are coming (in fact, most of the ones we are aware about actually use the new "drain" strategy reported in the present manuscript) Specific comments: Of course we did not aim to imply sufficiency in the general case, and we thank the reviewer for pointing that the manuscript was unclear in this respect. Our discussion is strictly restricted to the 1D case (single variable ODE) where it is easy to demonstrate that steady states need to alternate their stability (briefly, the demonstration goes as: let f correspond to the righthand side We have changed the text (including the abstract) to clarify this and also used the opportunity to emphasize the difference between tuning rates (Fig 1A,  -We added: "Species X is produced by an autocatalytic mechanism (green) and subject to degradation (red). " in Fig caption 1 -We also changed: "We started this analysis by considering a theoretical network containing only a positive production feedback loop and a degradation pathway. In the absence of specific nonlinearities, this simple system provides at most a single stable steady state (Fig. 1). To obtain bistability, some curve twisting (i.e. change in the kinetic law) is required: one may either tweak the shape of the production curve -to make it slower at low concentrations (Fig. 1D)-or adjust that of the degradation curve -to make it faster at low concentrations (Fig. 1E). " To: "We started this analysis by considering a theoretical one-species network containing only a positive production feedback loop and a degradation pathway. In the absence of specific nonlinearities (i.e. linear or Michaelis-Menten kinetics), this simple system provides at most a single stable steady state, whatever the respective rates, Fig.   1A,B,C. To obtain bistability, some curve twisting, that is, a change in the kinetic laws, is required: one may either tweak the shape of the production curve -to make it slower at low concentrations (Fig. 1D)-or adjust that of the degradation curve -to make it faster at low concentrations (Fig. 1E). " -Demonstration that this bistable system can be toggled between states is the most interesting contribution of this paper, as far as I am concerned. Unfortunately, this sections is somewhat poorly written. There is mention of a degradable inhibitor, but there is no explanation of what additional reaction this is. It would be very helpful to mention that reporter systems and normalization procedures are described in the Methods section at the end of the paper.
Thank you for the positive comment. We have improved this part using these recommendations.
The Methods section is now mentioned (and has been reorganized). The new paragraph is as follows: We also confirmed that it was possible to switch back from the high stable state to the low stable state by injecting a degradable inhibitor -a DNA strand that acts as a drain, but is not protected against the exonuclease and thus has a short lifetime in the solution. In Fig. 5 A better way to spin the importance of these results is that they indicate that it is possible to build a 2 bit or a 3 bit system, which -based on the results at Fig. 5, might be reversibly switched.
We have followed this suggestion. We changed our terminology to "n-bit system", and mention the overall number of stable states. Concerning the competition for enzymatic resources, it is indeed a very general feature of enzyme-driven networks, but as the reviewer notes, it does not forbid the building of such multinode systems, given that appropriate measures are taken (as we mention in the text, we simply kept: "the total concentration of autocatalytic templates constant to mitigate enzyme load 2 ", and this strategy worked).
-In the discussion section, the authors observe that a decoy strand that simply sequesters \alpha (or input, referring to Fig. 2)  And: "Similarly, if the deactivation process becomes stoechiometric (no exonuclease), the transition to bistability is also lost (Supplementary Fig. S11)." -What is the contribution of the drain/sink in terms of competition for DNA polymerase? What if in practice bistability was also due to the fact that production decreases due to competition between drain and template?
A reduction in free DNA polymerase concentration could be consistent with the transition to a second order rate-like behavior; because there is less DNA polymerase available (some of it is wasted to produce inert copies of \alpha from the drain), then two copies of \alpha are required for a net increase of one copy of \alpha. If my reasoning is correct, then it might be misleading to claim that bistability was obtained by exclusively tuning the degradation rates in the system.
Because of the way the system is built, there may be an indirect change in the production rates as well.
As can be seen in Fig. 2, the basin of attraction of the null state is typically very small, which implies that departure from this state is a phenomenon that happens at low trigger concentration.
Hence it is controlled by the linear regime of the enzymes, and saturation/competition is not critical here.
Note also that the transition to bistability is obtained not by changing the rate, but by changing the shape of production or degradation curves. In the case of production one would need to lower the production rate ONLY at low concentrations, and it is not clear how this could be obtained by polymerase competition only (if the competitive substrate for the polymease has no relation to the autocatalytic species).
The new experiments in Fig. S11 confirm these arguments, as it shows that inclusion of a competitive suicide inhibitor (which also loads the polymerase) does not lead to bistability.
As a side note, once again, we never claimed that "bistability was obtained by exclusively tuning the degradation rates in the system". On the contrary, we emphasize the need to adjust the kinetic laws of the degradation pathway.
To clarify this point, we added "Finally note that since the basin of attraction of the null state is typically small (see Fig. 2 in the main text), departure from this state is a phenomenon that involves low trigger concentration. Hence it is controlled by the linear regime of the enzymes, and we do not expect polymerase competition (between the template and the drain template) to play a significant role in the transition to bistability." in SI note 1.1.
-At page 8, I found the approach to derive information on rate order from delay to be interesting.
However, I was confused by the reasoning "the order of the reaction is difficult to assess directly from the shape of the time traces, because the fluorescence signal is a composite of many species contributions". Wouldn't the delay also be a quantity that reflects the behavior of whole sample, because it is determined from that same curve that "is a composite of many species contributions"? I can buy the idea that initially the drain reaction dominates before autocatalysis kicks in, so by assessing the time delay it takes before autocatalysis goes off is an indirect readout of the kinetics of the drain/input dynamics. However, I suggest that these paragraphs are rewritten.
The delay gives an indication of when something happens for the autocatalytic species, whatever it is. On the contrary, the shape of the curve is a composite of fluorescent signals generated by many species (precisely 9), with unknown weights. The paragraph has been modified, see the new version below: "Mathematically, it can be shown that, just above the critical concentration, the template/drain system should behave as a second-order autocatalytic system (Supplementary Note 3.1).
Experimentally, the order of the reaction is difficult to assess directly from the shape of the time traces, because the fluorescence signal is a composite of many species' contributions. To extract the reaction order, it is more convenient to trigger the system with various initial concentrations [x i ] of its input and observe the time the system takes before it crosses a given small fluorescence threshold (Ct). Ct will relate logarithmically with [x i ] for a first order amplification whereas an inverse law will reveal a second order process. Experimentally, in the absence of drain, we observed regular intervals between the amplification curves for a logarithmic range of initial trigger concentrations, therefore a first-order autocatalytic process.
For a system just above the critical drain concentration, this pattern is disrupted and the Ct rather follows an inverse law, symptomatic of a second-order process (Fig. 6). Just below the critical drain concentration we see an intermediate case, which is indicative of an order between 1 and 2. Overall, this confirms that the drain approach, while based on the decay pathway, can be used to change apparent kinetic laws (not rates) of a self-activating positive feedback loop." As for the conclusion of this paragraph, it seems to me rather obvious that by introducing new reactions one does not only change the stationary behavior of a system, but also its kinetics.
Again, we wish to emphasize the difference between kinetic rates and kinetic laws. Changing the first is trivial, crafting the second is the art of molecular programming with dynamic circuits.
We also wish to emphasize that a much more detailed discussion on this point is provided in the SI Notes 3.1 to 3.3, now better connected to the text as suggested.
-The paper needs to be read more carefully, to make sure that the necessary information is provided timely and accurately to the reader. I found several omitted definitions, and unclear sentences; We have carefully proofread the paper and made many minor improvements (in red).  Fig. S2-4). " We have now modified the Fig. 3 caption to include this information: "Experimental implementation. Bst (polymerase), Nb.BsmI (nickase) and ttRecJ (exonuclease) are used to drive an autocatalytic template with or without drains. Fluorescence recordings from an intercalating dye are used to follow in real time the amplification process. (A) the autocatalytic template αtoα2 was incubated without trigger…"

The paragraph in the middle of page 2 (introduction) describing the PEN toolbox is too vague and requires readers to look up earlier papers.
Due to constraints in the length of the manuscript, and as the details of the PEN DNA toolbox, without drains, have been described in at least 5 publications, the present description of the PEN toolbox was indeed made a bit short. Following this remark, we have added "Polymerizing/nicking cycles allow the input strand, acting as a trigger, to activate the generation of the signal strand encoded by the output side of the template." in order to make the introduction easier to understand.
The time C_t it takes for the system to spontaneously switch steady state from low to high is used early in the paper (Fig 2), but it is only loosely defined at the end of page 8. Please provide a clear explanation of how this time was computed.
We have now defined Ct in Fig. 3  -The supplement should be cleaned up and better connected to the main paper.
We have done some reorganizing and it is now much better referenced in the main text. All SI Notes are called at least once in the main text.

Reviewer #2 (Remarks to the Author):
Question 1: Authors describe "see SI Appendix for a mathematical discussion" in page 5, but do not show section number or figure number.
Authors should point where the discussion about "additional degradation path" and "small bottleneck" is in "SI appendix for mathematical discussion".
This particular point has been modified in the main text to send the reader to the right section of the supplementary information: "In other words, the additional degradation path should be fast but have a small bottleneck (see Supplementary Note 3.1)." Additionally, the corresponding part of the SI has been clarified.

Question 2:
Authors draw regression lines in figure 3

In addition, authors should add error bars in all plots.
This point raised by the reviewer is relevant for some, but not all, plots as detailed below. As the drainα4 is very efficient, the bistability is reached at low concentrations of drain, which explains why there are only 2 data points -not one(!)-that give non-zero 1/Ct value for this drain.
Although it is indeed not ideal to perform a linear regression with just 2 points, here we are just interested in obtaining an estimation of the critical drain concentration for each drain, and the best estimation possible given the data is the linear extrapolation presented in the plot. In this case, we can easily give a best estimation of the uncertainty of the value, which is simply the interval between the last concentration of drain for which the amplification is observed and the first bistable point. These error bars are now plotted in the inset, where the linear extrapolations are used to give an estimation of the critical drain concentration within this interval.

12
This procedure is now clearly explained in the caption: according to the Fig. 7E). This may explain why the Cy3.5 signal is slightly lower once the β node is "ON" than for the negative control (with no trigger), and reciprocally, for the BMN3 signal slightly shifted once the γ node is active.
The direct quantification performed in Figure 7E provides additional insight. In this case, the increase in concentration of the beta strand when the other switches are ON might be explained by competition mechanisms between the 3 nodes, as underlined by the reviewer 1. For instance, when only beta is ON, the exonuclease is less loaded than when other switches are active, and should degrade beta more efficiently, leading to a lower concentration. However, this phenomenon can be counterbalanced by competition for other resources such as the polymerase or the nickase, making precise predictions difficult. In any case, we do not expect "perfect" positions for the fluorescence shifts or the measured concentrations.
In conclusion to this point, we are aware that competition and fluorescence crosstalk exist in such complex molecular mixture, sharing the same enzymatic processor, and we modified the term "independent" in the Figure 7 caption, which was primarily chosen to characterize the network itself but is not appropriate regarding the underlying chemistry. Changes: "A 2-bit system can be constructed by combining two bistable networks with orthogonal sequences, as schematized, and monitored with two fluorescent dyes attached to the templates 4 . and: "We thus attempted the construction of a system with four stable states, built from a mixture of 2 orthogonal 1-bit molecular circuits of memory (Fig. 7A)." and: "When the four oligonucleotides were combined in a tube, the fluorescence signals indeed suggested four stable states" To make this clearer, we changed the text to: For each initial condition, we measured steady state concentrations in the range 100-400 nM for the bits that had been switched ON, while un-triggered bits yielded only traces values (<1nM).
The differences in the levels of ON switches, from bit to bit and state to state, can be explained by the differences in amplification/degradation efficiencies of the various sequences, as well as competition effects between switches. Overall, these results are consistent with the presence of eight distinct stable chemical states, hence 3 independent memory bits accessible from the adhoc initial conditions (Fig. 7E).

Question 6:
Authors should add an illustration of the network to figure 8 like figure 3-A or figure 2- According to the reviewer's suggestion we modified the Figure 8 with an illustration of the network. We also included a "negative" control containing only the autocatytic template without the drain. The reviewer's suggestion on that point is particularly useful since this control highlights the necessity of the drain template to build the excitable network (without it, one obtains an oscillatory network). The caption has been modified accordingly: The main text has also been modified to stress this important control: "The drain template is necessary to trap the system in the 0 state, and forbid the emergence of oscillatory cycles" Finaly, we want to clarify the difference between our work and Subsoontorn et al. paper mentionned in ref 15: these authors indeed present a single node bistable switch, and they use a production-decay mechanism to implement this molecular memory. In their case the system is based on RNA oligos production and degradation (using an RNA polymerase and a mix of RNAses) whereas in our case, the system is based on DNA oligos production and degradation.
However, the similitude with our work does not extend much further.
The most important difference is that the nonlinearity necessary to obtain the bistable behaviour comes in their case from the very non-linear production curve which, by design, present an inherent sharp "activation threshold" (see plots in Fig2). Quoting the authors: "whether the switch is bistable or monostable depends […] on the output amplitude and the activation threshold". This is to say that bistability depends on features of the production pathway, not the degradation pathway. This can be clearly seen by comparing their plot 6A right, which neatly correspond to our plot in Fig1D, i.e. it is the strategy that we do not take (we take the one of our fig1E, where production is featureless).
It is true that Subsoontorn et al. use a combination of RNAses, which results in a twisty global degradation curve, but this is not essential to their design, and is more an attempt to avoid uncontrolled accumulation of RNA. A perfectly linear degradation curve, if they could obtain it, would do the job as well, because it could still intersect three times with the sigmoidal production curve. In fact, they show that bistability can also be obtained with only RNAse H, in which case the high state is not bounded.
Another important difference is that the degradation mechanism of Subsoontorn et al. is nonspecific (all RNA molecules will be processed by the RNAses) whereas in our case, the drain template are specific to a particular node and its concentration directly control the non-linear behaviour. Therefore drain templates can be used to tune individualy and specifically the nonlinearities of each node in a network. This is an essential asset of our method for the building of large scale networks.
In conclusion I think that it is still fair to say that our work is new in its suggestion to use speciesspecific degradation/deactivation as a way to craft the non-linear dynamics of molecular programs, as opposed to the more traditional focus on non-linear production pathways.