In This Issue
Expansion of Frontiers of Engineering
December 1, 2003 Volume 33 Issue 4

Challenges and Opportunities in Programming Living Cells

Wednesday, December 3, 2008

Author: Ron Weiss

Someday we may be able to program cell behavior as easily as we program computers.

With recent advances in our understanding of cellular processes and DNA synthesis methods, we can now regard cells as programmable matter. Cells naturally process internal and environmental information in complex fashions and interact with neighboring cells to achieve coordinated behavior. Through genetic engineering, we can now equip cells with sophisticated capabilities for gene regulation, information processing, and communication. These new capabilities serve as catalysts for synthetic biology, an emerging engineering discipline to program cell behaviors as easily as we program computers.

Synthetic biology will benefit a wide variety of existing fields and enable us to harness cells for applications that are not feasible today. Applications include tissue engineering, molecular fabrication of biomaterials and nanostructures, synthesis of pharmaceutical products, and biosensing and will surely lead to quantitative insights into the operating principles that govern living organisms. Research so far has focused on building an enabling infrastructure for synthetic biology applications. A particular emphasis has been on constructing prototype synthetic gene networks that perform digital computation, analog computation, signal processing, and communications. In this paper, I will describe the building blocks for these genetic circuits, several intracellular and intercellular prototype genetic circuits that have been implemented recently, some of the challenges in designing such circuits, and the long-term significance of this work.

Synthetic Gene Networks
Genetic circuits are collections of basic elements that interact to produce a particular behavior. These elements include: DNA regions where RNA polymerase molecules bind and initiate transcription of DNA into messenger RNA (mRNA); mRNA sequences where ribosomal RNA molecules bind and initiate translation of mRNA into proteins; proteins that regulate the activity and production of other proteins; DNA regions that terminate transcription; and motifs that determine protein and mRNA stability. Each element serves a particular function that helps accomplish the overall behavior of the genetic circuit.

The basic elements can be grouped into units (or devices) that perform logic operations. For example, Figure 1 (see full pdf version) shows a logic unit that implements the digital NOT operation using the biochemical reactions of transcription, translation, and protein decay. Based on this mechanism, we have constructed and tested synthetic gene networks with units that implement the NOT, AND, and IMPLIES logic functions (Weiss, 2001; Weiss and Knight, 2000). In addition, we also proposed and modeled a biochemical NAND gate that consists of two NOT gates whose outputs are connected with a wire-OR (Weiss et al., 1999). Theoretically, any arbitrary digital logic function can be implemented with genetic circuits using these gates.

By constructing biochemical logic circuits and embedding them in cells, one can extend or modify the behavior of cells. Consider how computer designers program the behavior of computers or robots by fabricating silicon-based logic circuits. To program cells in an analogous way, the genetic circuit designer constructs a DNA sequence that encodes a particular circuit and then embeds this DNA molecule in cells (a process called transformation). Typically, the purpose of this network is to regulate precisely the production of proteins. Because proteins essentially perform all the "work" and information processing in the cell, cell behavior can be programmed by controlling when and under what conditions proteins are produced and degraded in the cell.

To date, several small synthetic gene networks have been built that accomplish specific genetic regulatory functions in vivo: the autorepressor (Figure 2a), in which a repressor regulates its own production to reduce noise in gene expression (Becskei and Serrano, 2000); the toggle-switch (Figure 2b), in which two repressors inhibit each other’s production to achieve a bistable system (Gardner et al., 2000); the repressilator (Figure 2c), in which three repressors are connected in a ring topology to produce repeated oscillation (Elowitz and Leibler, 2000); the genetic clock and toggle switch constructed from transcriptional repressors and activators (Atkinson et al., 2003); and our synthetic gene networks used for engineering digital logic gates and circuits in cells (Weiss and Basu, 2002; Weiss et al., 1999).

Despite their logically simple functions, the difficulties in building these networks revealed that many challenges will be faced in building circuits of increasing complexity. For example, designers of genetic circuits will have to cope with the significant noise inherent in gene expression (Becskei and Serrano, 2000; Elowitz et al., 2002; McAdams and Arkin, 1997) and will have to match carefully the "impedances" of constituent devices to achieve compound circuits that operate properly and reliably (Weiss and Basu, 2002). Because these circuits operate in the context of a living organism that already has an existing "program," understanding the interactions of the exogenous circuit with the endogenous elements will be critical. When designing these circuits, such interactions will have to be minimized to avoid undesired cross talk and interference with normal cellular processes. The exception to this design rule applies when control over endogenous cell behavior is intentional and desired (e.g., controlling cellular metabolism or growth).

Communication and Signal Processing
Intercellular communication allows individual cells to coordinate their behavior and accomplish sophisticated tasks they simply cannot perform alone. In higher level organisms, such as mammals, eukaryotic cells send and receive signals to perform a wide range of activities, from differentiation and growth during embryogenesis to immune and stress responses during adult life. However, cell-to-cell communication is not exclusive to higher level organisms. Bacterial cells, for example, are known to regulate gene expression based on their own cell density by secreting and then detecting concentrations of unique biochemical signals in a behavior known as quorum sensing (Bassler, 1999). This coordinated behavior gives bacterial cells a competitive advantage both as pathogens and in symbiotic relationships with their hosts.

To explore potential applications that would benefit from coordinated, multicellular behavior, we have begun to integrate communication capabilities with various synthetic genetic regulatory and information-processing networks. Toward this end, we first programmed Escherichia coli cells to communicate with each other by connecting several transcriptional regulatory elements with previously unrelated signaling elements from the marine bacterium Vibrio fischeri (Weiss and Knight, 2000). In the experimental system, we externally induced "sender" cells to synthesize the production of a small molecule (30C6HSL) that then diffused outside the cell membrane and entered the cytoplasm of nearby "receiver" cells. The receiver cells responded to this chemical message by expressing a green fluorescent protein that was visible under a microscope (Figure 3). We are now extending this work to achieve programmed, two-way communications by constructing new circuits that use multiple signal synthesis and response elements from various bacterial sources. Two important challenges in engineering sophisticated and robust, multisignal, cell-to-cell communication networks will be matching response sensitivities and reducing cross talk between the signals.

Cells naturally analyze cell-to-cell communication and various environmental conditions with elaborate signal-processing circuitry that includes both digital and analog components. The ability of cells to detect and subsequently react to environmental and internal signals is a principal characteristic of many biological phenomena. Examples include the movement of bacteria toward higher concentrations of nutrients through chemotaxis, the release of fuel molecules in response to hormones that signal hunger, and cell differentiation during embryogenesis based on chemical gradients. To extend a cell’s ability to respond to internal and environmental stimuli beyond the digital realm, we began to engineer analog signal-processing capabilities using synthetic gene networks. In our laboratory, we have built genetic circuits that enable cells to detect various chemical concentrations (below and above prespecified thresholds or only within certain ranges) and other circuits that allow cells to respond to multiple environmental signals (Basu et al., in press; Weiss et al., 2003).

As a prototype for exploring issues in engineering signal-processing capabilities, we designed a new genetic chemical-source pinpointing circuit. The circuit is able to detect the presence of a particular extracellular molecule and then distinguish between various chemical concentrations of the molecule (Basu et al., in press). Consider a toxin analyte whose location or even presence in the environment is unknown. The analyte is secreted from a particular pathogen and forms a chemical gradient centered on the source. A potential method of determining the location of the pathogen is to spread engineered sentinel cells in the suspected environment that can detect prespecified chemical concentrations. For a given concentration range, these cells will fluoresce in a ring pattern around the source. If the cells are engineered to detect multiple ranges distinguishable by different fluorescent colors, a bullseye pattern will emerge with several concentric rings around the analyte source (Figure 4).

The source-pinpointing circuit consists of several components that first detect the external analyte concentration and then determine whether the concentration falls above a prespecified low threshold and below a high threshold. Figure 5 shows two separate experiments of "sentinel" bacterial cells that have been programmed to respond to either low or high concentration thresholds of a biochemical secreted from nearby sender cells (Basu et al., in press). By combining and inverting the low and high threshold outputs, one can realize a circuit that responds with a high fluorescent output only when the analyte concentration is within the prespecified range. With the aid of simulation tools, we have also been able to fine tune the threshold responses of these circuits by modifying the DNA sequences of chosen genetic elements. An important long-term goal for this type of research is to be able to tune the responses of genetic circuits with the same predictability and reliability as we can when we design electrical devices. By combining this type of analog information processing with digital computation and programmed cell-to-cell communication, we may be able to create a flexible and powerful engineering discipline for programmed cell behaviors.

Key Challenges
The initial modeling and experimental efforts in this field have generated a great deal of excitement but have also revealed many challenges that must be faced. Some of the significant challenges are described below.

Programming complex behavior will require the assembly of a large, well characterized library of intracellular and intercellular components. The library should include elements for regulating transcription and translation, as well as elements for regulating protein-to-protein interactions, such as phosphorylation-based signal-transduction cascades. Most of the library elements must exhibit minimal cross talk and must not affect the host’s behavior. However, an important subset of this component library should be devoted to interfacing with the host to control desired cellular functions, such as the production of specific enzymes during predefined conditions, control over cell replication, and programmed secretion of various chemicals.

Designing operational and efficient genetic circuits will require models and simulation tools that can provide accurate quantitative predictions of circuit behavior. Engineered genetic circuits operate within a highly complex environment that is not well characterized and whose effects on the operation of the circuit have not been quantified. Models are still unable to predict the precise concentration averages and population statistics of the molecular components of even relatively simple systems. Solutions to this challenge will require more precise kinetic rates for describing the relevant biochemical reactions, models that incorporate additional cellular states (e.g., accurate RNA polymerase and ribosomal RNA levels), accurate predictions of noise in the biochemical reactions, a physical model of the cells and their surroundings, and potentially completely different modeling techniques. Furthermore, as the complexity of engineered circuits increases, they begin placing a significant metabolic burden on the host. Molecular interactions between the exogenous components and the endogenous cellular circuitry can also affect host behavior. Effects on the host’s environment must be understood and modeled to design complex synthetic circuits.

Genetic circuits must integrate specific components or network motifs that make them robust to fluctuations in the kinetics of biochemical reactions. Gene expression tends to be noisy because of the stochastic nature of the constituent biochemical reactions (Elowitz et al., 2002; McAdams and Arkin, 1997). In addition, fluctuations in environmental conditions, such as temperature and nutrient levels, affect cellular metabolism and consequently the operation of genetic circuits. Circuits that achieve reproducible, reliable behavior must do so despite components whose behavior fluctuates considerably. Mitigating the effects of gene expression noise will probably require a solution that incorporates positive and negative feedback loops.

A major bottleneck in genetic circuit engineering is the difficulty of synthesizing DNA constructs. Anecdotal evidence suggests that the construction of new strands of DNA consumes a significant portion, if not most, of the time spent in circuit engineering. Currently, several ongoing efforts are trying to remove this bottleneck by using various de novo DNA-synthesis methods.

Genetic-circuit design will require novel approaches that may be fundamentally different from existing computer-circuit design methodologies. In constructing our circuits, we have used "rational" design in which detailed models are used to guide the circuit engineering, helping to select appropriate components and suggesting how to mutate them, when necessary, to achieve correct circuit behavior (Weiss and Basu, 2002). We have also used another technique called directed evolution to optimize circuit behavior (Yokobayashi et al., 2002). Building on nature’s fundamental principle of evolution, this unique process directs cells to mutate their own DNA until they find gene network configurations that exhibit the desired system characteristics. Because our understanding of cellular processes is incomplete, efficient circuit design will most likely require a combination of rational design and directed evolution, or perhaps some other completely different approach.

Long-Term Significance
The field of synthetic gene networks is still in its infancy. Researchers are currently trying to build small genetic-regulatory systems that exhibit a particular behavior reliably and predictably. Solving the existing challenges in building complex genetic circuits will take considerable effort. However, once the major challenges are solved, the construction of synthetic gene networks will enable us to direct cells to perform sophisticated digital and analog functions, both as individual entities and as part of larger cell communities. An engineering discipline to program cell behaviors and its associated tools will advance the capabilities of genetic engineering and allow us to harness cells for a myriad of new applications. As a result, someday we will be able to modify and extend the behavior of cells in almost arbitrary ways. Because this technology will allow us to modify the instructions that govern the capabilities of organisms that live around us, as well as our own bodies’ instructions, it has tremendous potential for increasing our control over the environment and our surroundings and improving our quality of life.

References
Atkinson, M.R., M.A. Savageau, J.T. Myers, and A.J. Ninfa. 2003. Development of genetic circuitry exhibiting toggle switch or oscillatory behavior in Escherichia coli. Cell 113(5): 597-607.

Bassler, B.L. 1999. How bacteria talk to each other: regulation of gene expression by quorum sensing. Current Opinion in Microbiology 2(6): 582-587.

Basu, S., D. Karig, and R. Weiss. In press. Engineered Signal Processing of Cells: Towards Molecular Concentration Band Detection. In Eighth International Meeting on DNA-Based Computers, Sapporo, Japan, June 2002. Natural Computing, an International Journal, special issue.

Becskei, A., and L. Serrano. 2000. Engineering stability in gene networks by autoregulation. Nature 405(6786): 590-593.

Elowitz, M.B., and S. Leibler. 2000. A synthetic oscillatory network of transcriptional regulators. Nature 403(6767): 335-338.

Elowitz, M.B., A.J. Levine, E.D. Siggia, and P. Swain. 2002. Stochastic gene expression in a single cell. Science 297(5584): 1183-1186.

Gardner, T.S., C.R. Cantor, and J.J. Collins. 2000. Construction of a genetic toggle switch in Escherichia coli. Nature 403(6767): 339-342.

McAdams, H.H., and A. Arkin. 1997. Stochastic mechanisms in gene expression. Proceedings of the National Academy of Sciences 94(3): 814-819.

Weiss, R. 2001. Cellular Computation and Communi-cations Using Engineered Genetic Regulatory Networks. Ph.D. thesis, Massachusetts Institute of Technology, September 2001.

Weiss, R., and S. Basu. 2002. The device physics of cellular logic gates. Pp. 54-61 in NSC-1: The First Workshop of Non-Silicon Computing, Boston, Massachusetts, February 2002. Available online at: www.hpcaconf.org/hpca8.

Weiss, R., S. Basu, A. Kalmbach, S. Hooshangi, D. Karig, R. Mehreja, and I. Netravali. 2003. Genetic circuit building blocks for cellular computation, communications, and signal processing. Natural Computing, an International Journal 2(1): 47-84.

Weiss, R., G. E. Homsy, and T. F. Knight, Jr.. 1999. Pp. 275-295 in Dimacs Workshop on Evolution as Computation. Princeton, N.J.: Springer Verlag.

Weiss, R., and T.F. Knight, Jr. 2000. Engineered Communications for Microbial Robotics. Pp. 1-16 in DNA6: Sixth International Workshop on DNA-Based Computers, DNA2000, Leiden, The Netherlands, June 2000. New York: Springer-Verlag.

Yokobayashi, Y., R. Weiss, and F.H. Arnold. 2002. Directed evolution of a genetic circuit. Proceedings of the National Academy of Sciences 99(26): 16587-16591.

About the Author:Ron Weiss is an assistant professor in the Departments of Electrical Engineering and Molecular Biology at Princeton University.