Synthetic Biology: Building on Nature's Inspiration: Interdisciplinary Research Team Summaries (2010)

Reconstructing gene circuitry: How can synthetic biology lead us to an understanding of the principles underlying natural genetic circuits and to the discovery of new biology?

Genetic circuits have traditionally been studied using genetics and biochemistry. These studies underpin our current understanding of the regulatory wiring diagrams of organisms. They have also revealed that biological components like regulatory elements in DNA, genes, and proteins are intrinsically modular in nature. However, even when we believe we know the list of circuit components and their interactions, this knowledge often fails to explain/recapitulate the mechanism of the circuit. What is missing from these circuit diagrams? How can we infer those missing components if they have not been revealed by traditional experimentation? How can we test what parts of a given circuit are sufficient for a particular behavior? How different are potential circuit designs, that we imagine, from the actual circuit designs that have evolved to solve biological problems?

Due to an enormous expansion in our knowledge about genetic components and interactions in a number of model systems, we are now in a position to pursue a complementary approach to understanding natural gene circuits, based on reconstruction of genetic circuits. Specifically, we can engineer synthetic genetic circuits out of well-characterized genetic components and analyze their behavior in cells and organisms. These circuits can be based on their natural counterparts or on theories of how natural processes might work. Equally important, they can be engineered to operate as independently as possible from the corresponding endogenous cellular circuits. Circuits can also be created by “rewiring” existing circuits (adding, deleting, or changing regulatory connections). The goals of studying such

reconstructed genetic circuits are to understand how different aspects of circuit architecture contribute to function, to determine what functional tradeoffs are inherent in the design of the circuit, and to establish the sufficiency of particular circuit designs for given biological functions. More generally, they provide a complementary path to identifying both particular circuit interactions and general principles of gene circuit operation.

A reconstructive approach to genetic circuits may allow us to design circuits with unique properties and may provide insight into their underlying mechanisms. With a synthetic approach, it may be possible to construct a replica of a particular natural genetic circuit out of well-understood components and monitor its exact function in living cells. Using a synthetic approach, we could test the sufficiency of an arbitrary circuit made up of well-characterized components for generating a particular function. A major advantage to this approach is that we may be able to study the circuit mechanism without impairing cellular functions or inducing downstream consequences which are often drawbacks of traditional perturbation approaches. Finally, different circuit designs with similar functions can be directly compared to determine the precise properties each design grants a network as well as their relative advantages and disadvantages in particular cellular contexts. Ultimately, these studies may provide us with a deep enough understanding that we can design circuits that perform novel biological functions and we can exploit synthetic circuitry to reveal basic principles about natural circuit design.

Nonetheless, the synthetic approach faces many obstacles. For example, while we often know the components in a circuit, we frequently do not have in vivo information regarding kinetic parameters (affinities, binding and degradation rates, etc.). How can we infer these values if we cannot or have not measured them directly? Additionally, the intracellular environment is intrinsically “noisy,” and small copy numbers of molecular species limit the predictability of biochemical reactions. How can we interpret or predict circuit functions in the face of such noise? Can we devise synthetic circuits that suppress such noise to operate reliably, or take advantage of such noise to enable probabilistic cellular behaviors?

• What are the major advantages and limitations of synthetic circuits as a means of understanding the principles of genetic circuit design?

• How do we determine the basic principles underlying which circuit architectures can generate particular functions in cells and organisms?

• How do we identify missing components from natural circuits if they have not been revealed by traditional experimentation? How can we infer in vivo kinetic values if we cannot or have not measured them directly?

• To what extent can we analyze genetic circuits without comprehensive knowledge of all components and interactions?

• How can we evaluate how a circuit operates in the context of a complete organism?

• What new challenges and opportunities do particular classes of circuits present? In particular, what can synthetic biology do to better understand probabilistic behaviors, developmental circuits, neural circuits, immune circuits, and plant circuits?

• Can we delete natural circuits and replace them with synthetic counterparts within organisms?

• How can we engineer circuits that perform robustly in a noisy environment?

• If synthetic circuits completely fail to work, or work exactly as expected, they may appear to have taught us nothing. How do we develop synthetic projects that are as informative as possible?

Bhattacharyya RP, Reményi A, Yeh BJ, and Lim WA. Domains, motifs, and scaffolds: The role of modular interactions in the evolution and wiring of cell signaling circuits. Annual Review of Biochemistry 2006;75:655-680: http://arjournals.annualreviews.org/doi/full/10.1146/annurev.biochem.75.103004.142710?amp;searchHistoryKey=%24%7BsearchHistoryKey%7D. Accessed online 28 July 2009.

Dunlop MJ, Cox RS III, Levine JH, Murray RM, Elowitz MB. Regulatory activity revealed by dynamic correlations in gene expression noise. Nature Genetics 2008;40:1493-1498: http://www.nature.com/ng/journal/v40/n12/abs/ng.281.html. Accessed online 28 July 2009.

Gardner TS, Cantor CR, Collins JJ. Construction of a genetic toggle switch in Escherichia coli. Nature 2000;403:339-42: http://www.nature.com/nature/journal/v403/n6767/full/403339a0.html. Accessed online 28 July 2009.

Gertz J and Cohen BA. Environment-specific combinatorial cis-regulation in synthetic promoters. Molecular Systems Biology 2009;5:244: http://www.nature.com/msb/journal/v5/n1/full/msb20091.html. Accessed online 28 July 2009.

Sprinzak D and Elowitz MB. Reconstruction of genetic circuits. Nature 2005; 438:443-8: http://www.nature.com/nature/journal/v438/n7067/full/nature04335.html. Accessed online 28 July 2009.

Tigges M, Marquez-Lago TT, Stelling J, and Fussenegger M. A tunable synthetic mammalian oscillator. Nature 2009;457:309-312: http://www.nature.com/nature/journal/v457/n7227/abs/nature07616.html. Accessed online 28 July 2009.

Due to the popularity of this topic, two groups explored this subject. Please be sure to review the second write-up, which immediately follows this one.

• Gábor Balázsi, The University of Texas M.D. Anderson Cancer Center

• Meredith Betterton, University of Colorado at Boulder

• Nicolas Buchler, Duke University

• Donald Burke-Agüero, University of Missouri

• Joshua Gilmore, Joint BioEnergy Institute

• Joshua Leonard, Northwestern University

• John March, Cornell University

• Alexander J. Ninfa, University of Michigan Medical School

• Santa J. Ono, Emory University

• Jeffrey Tabor, University of California, San Francisco

• Leor Weinberger, University of California, San Diego

• Lingchong You, Duke University

• Daniel Strain, University of California, Santa Cruz

By Daniel Strain, Graduate Science Writing Student, University of California, Santa Cruz

It’s time for biologists to stop worrying about failure.

The world is a knotty and complex place, and biology in all its criss-crossing parts is the knottiest of all. Synthetic biologists often think they know a genetic circuit, a network of genes that interact to stimulate or repress each other, only to find that in living cells, their lab-built imitations tend to fizzle.

At the National Academies Keck Futures Initiative Conference on Synthetic Biology, an Interdisciplinary Research (IDR) team examined how synthetic biologists might stop this fizzle, designing circuits that generate

the right products at the right time, all the time. But it also explored how scientists can use the unexpected results of a circuit experiment to their advantage. The group asked if a “failed” synthetic circuit could help scientists better understand natural circuits and locate missing genes, proteins or chemical reactions. Can failures help untie the knots?

In many cases, context leads to unpredictability. Two separate cells, for instance, may express even the simplest synthetic circuits in different ways. A circuit that works just fine in muscle cells might shut down when it is expressed in a skin cell. Considerable research in synthetic biology focuses on bypassing the influences of context with modular circuits, circuits that produce dependable, or “robust,” results in many contexts.

While complete modularity may not be realistic, there are techniques that biologists can use to make synthetic circuits more robust. Redundancy is one of them. E. coli and staphylococcus bacteria might like to gobble the same sugar molecule but because they have different promoters, they respond differently to the same stimulus. Many synthetic biologists spend months at the computer or hand-to-pipette designing promoters that respond similarly to the same sugary treat.

But what if scientists want to learn more about context, not bypass it? Context, after all, makes a muscle cell a muscle cell and not a skin cell. The unexpected results of synthetic circuit experiments—which, though unexpected, can certainly not be called failures—can provide important information on the intracellular and extracellular hubbub that makes a muscle cell what it is.

Take a circuit in a hypothetical “grad student” bacterium. In this microbe, morning sunlight activates gene A, which turns on gene B, which turns on gene C, which produces an enzyme that secretes caffeine into the grad student’s environment. The benefits of such a bacterium are obvious. From genetic experimentation, researchers already know a little bit about A and C. Their screens found gene B but didn’t place it in the caffeine circuit. For all the researchers know, the caffeine circuit consists only of genes A and C. To understand this natural circuit better, the researchers make a proxy circuit, tying a synthetic mimic of A to a mimic of C in such a way that these genes effectively produce caffeine. But before they celebrate over espressos, the researchers want to find out if their synthetic circuit is anything like the real deal.

To do so, the group decided, the researchers will need to swap their synthetic circuit for the natural circuit and see what happens. The group developed a technique called genetic photocopying to achieve this swap.

In photocopying, researchers give a synthetic circuit a fitness benefit to bacteria, antibiotic resistance, for instance, so that bacteria will prefer the researchers’ circuit to their own. As the experimental colony grows and evolves, the synthetic and beneficial A and C will slowly replace the redundant but natural A and C in the bacterial genomes. After the experiment, researchers can take a look at these genomes to see which segments of DNA went missing.

This synthetic proxy technique can show scientists what genes, at minimum, are sufficient to complete a particular cellular function and what the DNA sequences of those genes are. It can’t, however, reveal that there’s still a missing component to the circuit, gene B. To learn more, scientists need to run a “parameter screen.”

Every genetic circuit works within a range of environmental conditions. Viruses like the lambda phage have simple genomes that function like circuits and can only infect certain cell types.

If scientists think they understand the circuitry of a virus like lambda, they can create a synthetic lambda and let it run free among a range of bacteria. If the synthetic virus lyses just as successfully as the natural lambda, it’s likely an accurate model. But if it’s less effective, scientists will know there’s still a missing piece to the puzzle. Scientists can also use the synthetic lambda in reverse—as a probe. If they know it only infects cells with certain types of receptors, they can find these receptors by seeing which cells the virus infects.

The idea of the synthetic circuit as probe isn’t limited to viruses and cellular receptors. Scientists can use synthetic circuits as signaling devices for many stimuli across a range of cell types.

The group developed the probe system to look for “noise” in gene expression. Noise describes the random fluctuation in genetic activity that occurs within every cell. It results from the game of chance that governs how often the molecules and proteins that form the machinery of gene expression bump into genes or mRNA strands, setting them in motion. Some cells are noisier than others, and the level of noise can change over time like when a cell undergoes stress or starvation.

Synthetic biologists often consider noise a hassle, making finely constructed synthetic circuits fizzle. But such randomness may be critical to cellular function. The group proposed to test whether the type of random fluctuations of cellular contents changes as an embryonic stem cell differentiates into another type of cell, such as a muscle or skin cell.

The group proposed a noise “decomposer” made of three separate

circuits that biologist could plunk into a variety of locales, such as zebra fish stem cells. Say biologists have a catalogue of circuits whose functions they know—from previous experiments that went kaput or didn’t work according to plan—are either sensitive or insensitive to noise. Scientists could engineer an insensitive circuit to produce a decaying red fluorescent protein in response to chaotic cellular environments, such as unpredictable fluctuations in ribosome abundance. They would then train two increasingly sensitive circuits to produce green and yellow glowing proteins, respectively, in response to medium and low noise levels. Because the fluorescent markers dim over time, scientists can use them to track the randomness in stem cells as they become part of the fish’s tail or slippery skin.

The team also proposed assembling a catalogue of instances in which synthetic circuits produce unwanted or unpredictable results called a deviance library. If a research team designed a synthetic circuit that produced caffeine in E. coli but failed in staphylococcus, that observation could go in the library. The next team that wants to build a caffeine circuit would then know that they should modify the design to get it to work in staph. But the library wouldn’t just record that a circuit had failed but also how it failed. With enough “how” data, biologists could begin to decode why circuits, in general, either move full steam ahead or grind to a halt.

A genetic mutation is a biological failure of sorts. Its effects can be good or bad but they’re always unexpected. Over time, however, mutations can lead to handy new inventions like the opposable thumb or bat claw. This group concluded that biologists should take a tip from nature and use the frequent “failures” of synthetic genetic circuits as a learning tool.

While robust synthetic circuits have many applications, the best circuits for learning about the natural world aren’t entirely predictable. Differences in the way the same synthetic circuit works across contexts, like muscle and skin cells, can help scientists understand what makes those contexts unique. They can help untangle what factors—like noise—make one cell good for running and another suited to a nice tan.

• Aseem Ansari, University of Wisconsin-Madison

• David Arnosti, Michigan State University

• Richard Bonneau, New York University

• Julie Dickerson, Iowa State University

• Charles Gersbach, Duke University

• Jane Kondev, Brandeis University

• Kathleen Matthews, Rice University

• Alexander Mitrophanov, Biotechnology High-Performance Computing Software Applications Institute

• Russell Monds, Stanford University

• Christopher Rao, University of Illinois at Urbana-Champaign

• Christopher Voigt, University of California, San Francisco

• Joshua Weitz, Georgia Institute of Technology

• Tia Ghose, University of California, Santa Cruz

By Tia Ghose, Graduate Science Writing Student, University of California, Santa Cruz

For many biological problems, the instruments used to uncover new connections are scattershot, and sometimes very blunt. To tease out the role of a certain gene, scientists often knock out its function completely, or send it into overdrive so that it produces far more protein than it would in nature. Others blast a cell with a huge amount of a chemical and then measure everything they can to see what turns up. At the 2009 National Academies Keck Futures Initiative Conference on Synthetic Biology, an Interdisciplinary Research Team (IDR) charged with “reconstructing gene circuitry” asked: what if synthetic biology could uncover what’s really going on in biological systems, in a way that is more precise, informative, and systematic than anything we can do now? The team agreed that a toolbox of controls that precisely dial different parts of a biological system up or down would provide synthetic biologists with significant opportunities to discover new information about living systems.

The group consisted of a diverse set of physicists, engineers, developmental biologists, computational biologists, biochemists, and chemists. Its stated challenge was to determine how synthetic biology can lead to an understanding of the principles underlying natural genetic circuits and to the discovery of new biology. For instance, scientists believe they have a complete list of circuit components and their interactions, yet this knowledge often fails to capture exactly how the circuit works. The team was charged with using synthetic biology to determine what is missing from these circuit diagrams, and how to infer these missing components when traditional

experimentation fails. The team was also tasked with devising a method to test what parts of a given circuit are sufficient for a particular behavior, and to imagine how circuit designs we construct differ from the actual circuit designs that have evolved to solve biological problems. To begin with, team members suggested several different ways that synthetic biology could answer fundamental biological questions. Can synthetic biology be used to determine the “minimal circuit,” or the smallest number of elements and connections that can mimic the behavior of a real network in nature?

Ultimately the team rejected pursuing that path, because nature often creates a tangle of redundant or even dead-end connections between genes, proteins or transcription factors. This redundancy helps keep the network stable in response to changing conditions. So, the simplest network may reveal very little about natural systems. For instance, the fruit fly embryo develops in the presence of a protein called bicoid. A fly can develop normally, even if in the spatiotemporal distribution of bicoid in the embryo changes wildly, demonstrating that it is insufficient to know only that bicoid is important. Obviously there is more to it than that.

The team also raised the idea that synthetic biology could help determine underlying principles that govern cellular behavior. For instance, if all bacteria that use a gradient of chemicals to sense and move toward food rely on a certain fundamental set of genes, proteins, or chemical signals, synthetic biology might confirm that. And if the bacterial genes differ, perhaps the underlying type of network stays the same. Synthetic biology could provide tools to uncover these similarities.

Team members agreed that different scales are likely to play a role in the way network problems are studied. For instance, the approaches used for uncovering how the spinal cord assembles may be completely different from those needed to probe how individual E. coli in a biofilm talk to each other. Those will differ from techniques used to examine how proteins bind to each other, inactivate DNA, or change shape. This observation about scale helped the team focus its discussion.

Despite their divergent interests, each of the scientists longed to improve upon some of the sledgehammer approaches of some now traditional genetic technologies. They wanted instead to gently nudge biological systems from their ordinary states and then measure and analyze how these systems respond in real-time. To do that, they converged on the idea of using the electrical circuit as a metaphor for biological connections. Just as electrical engineers send a pulse or an oscillating wave into circuits and then use an oscilloscope to measure the output, biologists need a set of

tools to precisely perturb natural systems and then observe and understand how they react. These artificial networks could be plugged in to the natural networks to send in different inputs, they suggested. Synthetic tools may also help scientists detect changes.

At the level of small molecules, biologists need a way to tune precisely the amount of each chemical in specific sites in the cell (such as the nucleus). They may also want to target molecules to a very specific place within the cell. Being able to control how fast chemicals respond, break down, or change state would also be useful. For instance, green fluorescent protein can take a while to fold into its functional shape, while light can activate chemicals in a flash. Researchers need a way to modulate the time-scale of these reactions, depending on what they are measuring. They need ways to easily modify how fast certain genes are transcribed and translated or how, in real time, promoter regions in genes respond to different stimuli.

On the level of individual circuits, researchers want to peer in as proteins assemble, bind to DNA and regulate its transcription, or when chemical modifications like phosphorylation occur.

For large networks, altering the input into a system without actually changing the way the network is configured is a key goal. Currently, the standard tools for changing gene networks are knockout or over-expression experiments. But these are more like all-or-nothing changes that can’t be precisely controlled. Making a gene produce, say, a third as many copies of a protein for twenty minutes, then twice as many for five minutes, is not feasible right now. Because of current technological limitations, the team decided to focus on this smaller scale.

Sometimes scientists cannot see what’s going on inside a living organism in real-time. If the new tools the team proposed could create subtle and dynamic changes on a fine scale, then scientists may also need better techniques to sense the system’s response.

One step would be to remove what the team called the deconvolution problem. For instance, to find out if a certain protein is being made, biologists often add a snippet of DNA that encodes a green fluorescent protein linked to another protein to see if the target protein lights up. But it can

take many minutes for the protein to be transcribed, translated, fold up, fluoresce, and be detected. Though scientists have rough ways of determining when transcription occurred, the team envisioned using synthetic techniques to know exactly when DNA was transcribed.

It’s also important to check that proteins are actually working, not just that all the raw materials are present. For instance, when bacteria build their whip-like tails, it would be useful to know that the proteins are linked together and form functional structures, rather than simply pinpointing them to the same general region. The team envisioned creating a sensor or time-stamp that detected when proteins are assembled. The group’s ultimate goal was a “synthetic oscilloscope”—a grab-bag of different techniques that can detect when and how changes occur inside the cell more easily and quickly.

With such precise control of what is sent into the system, the team wondered whether the existing tools for analyzing data and formulating experiments might be insufficient. Even though a scientist can test the system with a hundred different inputs, that process is often cumbersome and expensive. Given the explosion of new technologies that a synthetic tool box would provide, new analytical techniques may help researchers focus their efforts and design experiments more efficiently. When the inputs are more finely controlled, it may also be necessary to determine which of the numerous variables should be measured, and with what resolution. Once one experimental system is characterized, those results must be analyzed to help scientists plan future experiments.

Detecting a cell or a network’s response may require modelers to develop new analyses of biophysical principles—or even uncover new principles. Another frequently raised issue was the “inverse problem.” Many gene networks may make a fish blue, for instance. But once the scientist knows the fish is blue, it can be tricky to figure out which genes led to its coloring. Untangling the results from these much more complicated experiments may require mining existing analytical tools or even developing new ones.

One question that came up repeatedly was whether synthetic biology is truly better than, or simply complementary to, existing technologies.

Though existing tools are clearly insufficient for answering many biological questions, the team wondered whether synthetic biology was actually the best way to produce better tools. For instance, researchers could envision the usefulness of creating precise inputs using synthetic biology, but it was less clear that synthetic biology could produce better ways of detecting outputs from natural systems.

The team developed a rough model of how synthetic networks could be linked into biological systems. Their “synthetic Swiss army knife” would be genetically encoded into a cell, complete with simple start and stop buttons that work reliably. These would attach to an oscillator or wave generator whose frequency could be tuned. The team also envisioned adding a noise filter which could make the signal sent into the cell more random. Scientists could link this tool to a real system at various points in the natural network.

By modulating the input functions, a researcher could very precisely control how much messenger RNA is made, how many changes like methylation or phosphorylation are added to a completed protein, or the concentration of proteins or ions in a cell. Many of these different components could be altered at once, or each change could be done sequentially. Using this system, the team could explore a larger range of behaviors in the natural networks, perhaps uncovering new principles along the way.

While synthetic biology is traditionally touted as a way to create tailor-made, artificial biology, its potential for understanding the natural world has not yet been realized. Though a multi-purpose, synthetic biology-based tool as envisioned by the team is still a long ways away, it could ultimately provide a deeper understanding of natural biological systems.