Engineering Signal Transduction Pathways

Leading Edge

Review Engineering Signal Transduction Pathways Christina Kiel,1 Eva Yus,1 and Luis Serrano1,2,*

EMBL-CRG Systems Biology Unit, Design of Biological Systems, Centre de Regulació Genòmica, Dr. Aiguader 88, 08003 Barcelona, Spain Intitució Catalana de Recerca i Estudis Avançats (ICREA), 08010 Barcelona, Spain *Correspondence: [email protected] DOI 10.1016/j.cell.2009.12.028

1 2

Cells respond to their environment by sensing signals and translating them into changes in gene expression. In recent years, synthetic networks have been designed in both prokaryotic and eukaryotic systems to create new functionalities and for specific applications. In this review, we discuss the challenges associated with engineering signal transduction pathways. Furthermore, we address advantages and disadvantages of engineering signaling pathways in prokaryotic and eukaryotic cells, highlighting recent examples, and discuss how progress in synthetic biology might impact biotechnology and biomedicine. Introduction A basic property of living systems is the ability to respond to extracellular signals by evoking an internal response, which often leads to changes in gene expression and phenotypic alterations. Synthetically modifying an organism to respond differently to a signal, or to respond to an artificial signal, is accomplished through rewiring existing pathways or adding new modules that in some cases may be a complete pathway (for instance, the transplantation of a pathway to another cell type or species). The two main objectives of engineering signaling pathways are to understand how natural networks function and to build synthetic networks with specific applications or functionalities. Progress in this field may have a major impact on biomedical engineering, such as gene therapy or tissue engineering, and industrial biotechnology. Synthetic biology has naturally evolved as a field in biotechnology that has a broader engineering scope: to modify entire systems. This would not have been possible without the development of systems biology, which has benefited tremendously from high-throughput technologies (DNA microarrays, ultrasequencing, mass spectrometry, automated microscopy, and computation). Initially, synthetic biology focused on the engineering of genetic circuits, which were first designed to study negative and positive feedback loops, oscillations, noise, and robustness within a system (Becskei and Serrano, 2000; Becskei et al., 2001; Elowitz and Leibler, 2000; Gardner et al., 2000; Fung et al., 2005; Swinburne et al., 2008). Later, gene circuits were designed to couple gene expression with metabolism, to understand quorum-sensing pathways (Bulter et al., 2004; Balagaddé et al., 2008), and to synthesize new chemical compounds (for example, the production of terpenoids in engineered Escherichia coli; see Martin et al., 2003). They have also been used to study memory (Friedland et al., 2009), spatial patterning (Basu et al., 2005), and network evolution (Isalan et al., 2008). The design of synthetic circuits in eukaryotes began with the accessibility of a wide range of molecular tools (many coming from work in bacterial engineering, such as inducible expression systems; reviewed in Gossen and Bujard, 2002), which

have since evolved tremendously in terms of their complexity. For example, the design of an inducible gene silencing module based on RNA interference and repressor proteins (Deans et al., 2007) as well as the first tunable synthetic mammalian oscillator in hamster ovary cells (CHO) and human embryonic kidney (HEK293) cells (Tigges et al., 2009) have recently been reported. Several recent reviews have covered the engineering of synthetic gene circuits (see Dueber et al., 2004; Andrianantoandro et al., 2006; Drubin et al., 2007; Greber and Fussenegger 2007; Serrano, 2007; Michalodimitrakis and Isalan, 2009). However, not covered in these reviews is a detailed analysis of the challenges and differences of engineering gene circuits as compared to signaling systems, and when using prokaryotic versus eukaryotic cells. The aim of this review is to address these two features of synthetic biology. The first part describes the conceptual differences between engineering signaling pathways as compared to genetic circuits. We highlight the different properties of the components and the general properties of the circuits. In the second part, we compare signaling pathway engineering in prokaryotes as opposed to eukaryotes, contrasting the differences in architecture and components of these two classes of systems and how these differences impose different design considerations. This is complemented by reviewing examples of pioneering and the latest engineering approaches. The review concludes with our perspective on how the field might evolve and what remaining challenges are left to be faced. Engineering of Gene Circuits versus Signaling Pathways What are the differences and what should be considered when engineering signal transduction pathways as opposed to gene circuits? There are two main points to consider, one being the general properties of the system and the other being its components (Figure 1). Regarding the general properties of the system: (1) Signaling pathways operate fast (milliseconds to minutes). In contrast, transcriptional responses can range from minutes (prokaryotes) to hours (eukaryotes) (see the recent review by Pryciak, 2009) (although expression changes might be faster if the system involves noncoding RNA, as it would not require Cell 140, January 8, 2010 ©2010 Elsevier Inc.  33

Figure 1. The Engineering of Signal Transduction Pathways and Gene Networks When engineering signal transduction pathways as opposed to gene circuits, there are two main points to consider: the general properties of the system and its components. Regarding the components, nucleotides can usually be exchanged without having a large impact on neighbor nucleotides and DNA structure (“independence approximation”). In contrast, amino acid substitutions in proteins can lead to structural changes and misfolding. With respect to the circuit properties, gene circuits are modular, they usually respond slowly, and the responses can be stochastic. In contrast, signaling pathways are highy modular, respond fast, often involve signal amplication, and make use of spatial localization.

nuclear export and translation. (2) Signal transduction pathways usually depend on subcellular localization, and therefore elicit spatially restricted, context-specific responses. (3) Operations at the level of protein activity, unlike protein levels, allow for a larger degree of control and tuning capability. Thus, engineering signal transduction pathways could allow for more versatility and design options, but the predictability of the engineered pathway, at least in eukaryotes, could be low because of the complexity of signal propagation and the greater number of different molecules involved. (4) Genetic circuits tend to be noisy because mRNA and protein synthesis occur in bursts and are the major source of biological noise (Pedraza and Paulsson, 2008), whereas signaling pathways usually involve larger number of molecules and thus tend to be less stochastic. (5) Signaling systems employing amplification cascades need to avoid spontaneous activation, which is usually achieved by negative feedback regulation (such as in epidermal growth factor signaling; see Amit et al., 2007) and also in many cases by the requirement of a double triggering signal (as in B cell signaling; reviewed in Kurosaki, 2002). Regarding the components of the two systems, there are some fundamental differences (Figure 1): (1) Gene circuit engineering is done on the level of DNA and DNA binding 34  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

proteins. DNA is easy to modify, given that it has a modular structure and nucleotides can be exchanged without having a large impact on DNA structure (Benner and Sismour, 2005). Therefore, placement of DNA sequences where they can be recognized by DNA binding proteins in regulatory regions of a gene is, in principle, relatively easy in prokaryotes or, when plasmids are used, in eukaryotes. However, when inserting a construct in eukaryotic chromosomes, one needs to consider context effects (such as heterochromatin versus euchromatin, and epigenetic regulation). On the other hand, there have been some advances for DNA binding proteins, in rationally modifying their DNA binding specificities (Ashworth et al., 2006; Redondo et al., 2008), but this is not yet a well-established methodology. However, engineering DNA binding can be greatly facilitated by taking advantage of the extensive work on modular zinc-finger proteins to create transcription factors with new DNA binding specificities (Townsend et al., 2009). (2) Engineering of signal transduction networks requires the modification of proteins alone, or protein-protein interactions, either by mutagenesis, insertion of unstructured recognition sequences, or alteration of domain composition. Although in some cases polypeptides (linear motifs) can be exchanged as easily as DNA modules, introduction of mutations inside domains or globular proteins is more complicated given that the conformation of an amino acid side chain greatly depends on the conformation of its neighboring residues and vice versa. As a consequence, their replacement could cause structural changes that can lead to protein misfolding and aggregation (López De La Paz et al., 2002; Chiti and Dobson, 2006). Thus, manipulating DNA can often be done without compromising its structural composition and the function of the various components, whereas manipulating proteins requires the use of protein design tools (Dahiyat, 2006), or a combination of directed evolution with selection (Looger et al., 2003). In summary, the three main differences between engineering signal transduction as opposed to gene circuits are as follows: (1) signaling systems operate fast, and thus designing interconnections and feedbacks requires accurate prediction of the system behavior, (2) subcellular localization plays an important role in signaling and must be considered, and (3) engineering of proteins is more difficult than DNA because protein structure and folding is less understood and less predictable. Hence, the main challenges in engineering genetic systems are (1) coping with the inherent stochasticity of transcriptiontranslation (in prokaryotes this can be diminished by engineering negative feedback of the transcription factors on their own promoters), (2) generating enough variants of selected DNA binding proteins to ensure recognition of almost all DNA sequences, (3) finding more transcription factors that can be modulated by chemical compounds (for example, the TET repressor), or postranslational modifications, to ensure a high dynamic range of regulation and the possibility of combining more than one transcription factors, (4) getting a better grasp of the regulatory role of chromatin structure in eukaryotes and possibly in prokaryotes, (5) incorporating other regulatory molecules (such as riboswitches and small RNAs), and (6) taking advantage of the possibility of playing with the histone code.

Table 1. Comparing Signal Transduction in Prokaryotes and Eukaryotes Prokaryotes

Eukaryotes

  One-component systems

yes

no

  Two-component systems

yes

yes (only in yeast and plants)

  Histidine-Aspartic acid phosphotransfer

yes (mainly)

yes (rare)

  Serine/threonine kinases

yes (rare)

yes (mainly)

  Tyrosine kinases

yes (rare)

yes (not in plants, rare in yeast)

Other posttranslational modification (such as ­methylation, acetylation)

yes

yes

  Nucleotide derivatives

yes

yes

  Lipids

no

yes

Allosteric regulation

yes

yes

  Multidomains

fewer

majority

  Autoinhibition

yes

yes

  Anchoring

yes

yes

  Oligomerization

yes

yes

  Receptor clustering

yes

yes

Proteolysis and degradation

yes

yes (complex regulation)

Active transport

yes (simple)

yes (complex)

Pathway length

short

long cascades

  Subcellular localization

yes

yes

  Scaffolds

poorly studied

yes

  Gradients

unknown

important

Regulation by scaffolds

yes (but not well established)

yes

Noise resistance/robustness

yes, often required for function

yes

Negative and positive feedback regulation

yes (mainly transcriptional)

yes, important (mainly at protein level)

Crosstalk between pathways

yes (rare, and physiological r­ elevance often unknown)

yes (very important)

Mechanism of Signal Transduction Phosphorrelay

Kinases and phosphatases

Second messengers and alarmones

Modularity

Complex formation

Network Properties

Regulation by localization

Tools for Signaling Pathway Engineering Signal transduction relies on a series of mechanisms, which to some degree are conserved in both prokaryotes and eukaryotes (Table 1). Essentially, they include all or some of the following mechanisms: localization, complex assembly, competition, activation-deactivation, diffusion and/or active transport, modularity, degradation, negative and positive feedbacks, specificity, and crosstalk (reviewed in Teruel and Meyer, 2000; Jordan et al., 2000; Aravind et al., 2003; Galperin, 2004; Galperin and Gomelsky, 2005). Thus, in order to design or modify signal transduction pathways, one could tackle any of the characteristics above. Depending on which segment of the pathway one would like to target, different tools can be employed (Figure 2). Below, we will describe both the design tools that have been used already (some examples are highlighted in the next section) and those

that are available but have not yet been tested in synthetic approaches. In every case, we will point out the peculiarities of prokaryotes and eukaryotes that could constrain the use of particular tools. Localization As signal transduction pathways in eukaryotes are often spatially restricted, modifying the location of proteins is a powerful tool for rewiring a pathway. A protein can be retargeted by using scaffold proteins that recognize a protein domain, adaptor proteins, or lipids (Harris et al., 2001; Park et al., 2003; Bashor et al., 2008), as well as by introducing posttranslational modifications recognized by other proteins or by the membrane (for instance, lipid anchoring; see Kamalakkannan et al., 2004). Another tool for rewiring signal transduction routes by localization in single cells is the spatially restricted expression of proteins using localized transfection, Cell 140, January 8, 2010 ©2010 Elsevier Inc.  35

Figure 2. Signaling Pathway Engineering in Prokaryotes and ­Eukaryotes Signal transduction can be engineered on three different levels, illustrated here using three examples: (1) pathway modulation, for example by protein mutagenesis, (2) the rewiring of existing pathways, which can be achieved by creating chimeric proteins that have new input-output domains, and (3) the creation of artificial signal transduction pathways and cells.

or using the localized application of a stimulus with magnetic beads (Kempiak and Segall, 2004; Isalan et al., 2005; Santori et al., 2006). Furthermore, there is the possibility of changing localization with bifunctional small molecules. Using the so called “anchor-away technique” (in the yeast Saccharomyces cerevisiae), a protein of interest, which is shuttling between the nucleus and cytosol, is depleted from the nucleus and consequently becomes associated to an anchor membrane protein (Haruki et al., 2008). 36  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

Subcellular localization exists in prokaryotes as well, with receptor clustering and signaling protein enrichment in cell poles being critical in bacteria chemotaxis (see Porter et al., 2008) as well as for cell division and differentiation (reviewed in Jenal and Stephens, 2002). However, in general, there are no isolated compartments, and therefore the number of localizations and the possibilities of using localization to engineer signaling in bacteria are greatly reduced. In prokaryotes, the lack of knowledge of how to restrict proteins to particular regions of the membrane and the lack of compartments makes this difficult. Recent advances in understanding the role that positive membrane curvature plays in selective localization of bacterial proteins at the poles, as well as a growing list of bacterial proteins that exhibit specific membrane compartmentalization (Ramamurthi and Losick, 2009), could open the way for using localization as a tool in bacteria. Complex Assembly and Dynamics Signaling requires the transient assembly of protein complexes, which in many cases results in a restricted conformational flexibility of the proteins involved, as well as in the conformational changes that affect protein function. Altering the affinities and the kinetics of complex formation (dynamics), as well as creating artificial binding sites (complex assembly), can be a powerful way of modifying or changing signal transduction pathways in a rational manner. Mutagenesis, either driven by computer design or by directed evolution, can be used to alter the specificity and the affinity of an interaction (in eukaryotes, see Reina et al., 2002, and Kortemme et al., 2004; in bacteria, see Lyon et al., 2000, and Looger et al., 2003) as well as the kinetics of complex formation (Selzer et al., 2000; Kiel et al., 2004). Recent advances in protein design suggest that this is achievable, provided that a high-resolution structure of the protein-protein or proteinDNA complex is available (see Aloy et al., 2004, and Dahiyat, 2006; reviewed in Kiel et al., 2008). The main limitations at the moment are the capability of both predicting and engineering large conformational changes upon interaction, as well as the changes in the dynamics and flexibility of regions of a protein upon interaction with another. Future developments in protein design tools, which could tackle the above, will pave the way for using changes in protein flexibility and conformational rearrangements for engineering signaling pathways. Posttranslational modifications, such as phosphorylation, allow the possibility of modifying the assembly of protein complexes. For example, introducing a target site for a kinase can create binding sites for proteins that contain SH2 domains (which bind phosphorylated tyrosines), thereby influencing the assembly of its complexes and spatially restricting them (Dueber et al., 2003). In this case, the challenge is to know whether the kinase recognition motif is sufficient for bringing two proteins together or whether additional recognition is needed from other domains or scaffolds to ensure that the proteins colocalize in the cell. This is especially important in eukaryotic cells. Modularity The association of functional domains within a single polypeptide chain is not exclusive in every domain of life. Although protein modularity is present along the tree of life, the complexity of domain composition and the variety of protein folds is larger

in nucleated organisms (Aravind et al., 2003), with multiple domains having weak binding affinities present in one protein. In addition, the average protein length is larger in eukaryotes, (Brocchieri and Karlin, 2005). This provides a greater flexibility in engineering signal transduction pathways in eukaryotes, as it is possible to replace or add modules in proteins to change their binding partners and localizations (Dueber et al., 2003; Haruki et al., 2008). Of course, this is not trivial. Even if domains fold autonomously, domains can change, and unfold the structure of the full-length protein into which they are inserted. Such approaches often require several steps of domain optimization. These domain prediction and optimization steps can be supported by prediction programs, like SMART, Pfam, or InterPro (Finn et al., 2008; Letunic et al., 2009; Hunter et al., 2009). If folding problems persist when placing two domains on a single chain, another option is to express the domains separately in fusion with dimerization motifs, such as leucine zippers (Rieker and Hu, 2000), so that they fold independently. Although in principle one could design more complex and versatile, multidomain proteins in bacteria, the question remains whether such large polypeptides will be correctly folded and functional, because of the lack of folding machinery to cope with such large multidomain proteins (Seeliger et al., 2005). In the future, a database collection of protein domains that have been shown to fold robustly and to tolerate fusion to other domains could help enormously in the engineering of new multidomain proteins. In this respect, help may come from initiatives like the Biobricks repository (http://www.biobricks.org/) (Shetty et al., 2008), which plans to have a large collection of protein domains experimentally tested as building blocks. Activation-Deactivation A key characteristic of signal transduction is the need to activate components for only a brief period of time followed by deactivation. Activation can be done by binding chemical messengers or ions, or by posttranslational modifications (such as phosphorylation, acetylation, and methylation). Deactivation involves degradation of the chemical messengers, sequestration of the ions, or reversal of the respective posttranslational modifications. It is important to mention that many signal transduction proteins are regulated by autoinhibition: a part of the protein binds to or changes the conformation of its own active site, thereby inhibiting its activity. This property has been used by different investigators to regulate engineered networks (reviewed in Pryciak, 2009). Autoinhibition can be released by a competitive ligand or posttranslational modification (reviewed in Bhattacharyya et al., 2006). Other groups also have replaced small molecule activators (such as Ca2+, cAMP, and phosphoinositides), by drugs activating the corresponding enzyme (such as forskolin, for cAMP), giving an additional level of external control (Metzger and Lindner, 1981). Recently, a new generation of synthetic modulators, namely bifunctional small molecules (see Corson et al., 2008), have been designed with the ability to act as bridges between two domains or proteins, giving rise to different biological effects, such as receptor activation in the absence of its ligand (useful for orphan receptors) or changing activity via subcellular targeting (Spencer et al., 1993; Belshaw et al., 1996; Rivera et al., 1996). Furthermore, we would like to mention tools to control

phosphorylation in a light-dependent manner through the use of chimaeras of photoreceptors and kinases (Levskaya et al., 2005; Airan et al., 2009), as well as regulation by methylation (Kollmann et al., 2005). With respect to posttranslational modifications, phosphorelay systems are probably the most commonly used during evolution (Alm et al., 2006) and are found in both prokaryotic and eukaryotic systems. In prokaryotes, histidine and aspartic acid-phosphotransfer systems dominate, whereas serine, threonine, and tyrosine kinases are rarely found. The opposite holds for eukaryotic phosphorylation cascades that primarily involve serine/theonine or tyrosine kinases. It is worth mentioning that histidine/aspartic acid-kinases mainly transfer the phosphoryl group to only one protein, whereas serine/threonine and tyrosine kinases can phosphorylate multiple proteins. This might explain why in bacteria there is less feedback at the protein level, with negative feedback predominating at the transcriptional level (reviewed in Smits et al., 2006). Furthermore, phosphorylated histidine and aspartic acid residues are often unstable and hydrolyze spontaneously. As a result, phosphatases are not always required to turn off a signal in bacteria (Batchelor and Goulian, 2003). Another thing to consider is that the interaction between the histidine kinase and the aspartic acid-response protein specifically involves a three-dimensional folded surface. This is the opposite of serine/threonine and tyrosine kinases, which recognize linear unstructured motifs, as well. The same is true of the domains that recognize phosphorylated amino acids in eukaryotes (such as SH2, WW, and FHA). As a result, it is easier to engineer a serine, threonine, or tyrosine phosphorylation site in a protein, as well as a recognition site for domains that bind phosphorylated residues (Yeh et al., 2007). Recently, a photolytic method to cleave peptides has been designed, whereby an unnatural amino acid (2-nitophenylalanine) is incorporated into proteins expressed in E. coli, (Peters et al., 2009), a method similar to earlier work based on the incorporation of 2-nitrophenylglycine in oocytes from the frog Xenopus (England et al., 1997). This tool could open the exciting possibility of designing autoinhibited proteins, which could be photoactivated by cleaving off one domain. Light-induced activation of signaling is clearly an exciting field and one that could develop into new applications. Thinking further, one might be able to develop a system that uses the emission of green fluorescent protein as a source of light, which would open the possibility of engineering more complicated regulatory circuits. Competition After the recent avalanche of high-throughput protein-protein interaction data sets, many proteins are found to interact with a very high number of binding partners. It is evident that many of these interactions cannot happen simultaneously and that spatial and temporal aspects are needed to curate these large scale networks. Thus, a new field is emerging, aiming to understand network properties and their underlying principles using structure-based design of signaling pathways (see Kim et al., 2006; reviewed in Campagna et al., 2008, and Kiel et al., 2008). Structural information can be used to determine whether a protein can interact with two partner proteins at the same Cell 140, January 8, 2010 ©2010 Elsevier Inc.  37

time (“compatible interaction”) or not (“exclusive interaction”) by examining protein complex interfaces. One could imagine rewiring signal transduction pathways by modifying the relative affinity of compatible and exclusive binding sites, using structure-based design or by selectively blocking one binding site. Alternatively, overexpression of one partner could block one of the branches in a signal transduction pathway. Again, the limitations of this method are the availability of X-ray structures of the complexes and reliable protein design tools. Degradation Another mechanism to achieve a dynamic response and a quick adaptation to signals (aside from fast phosphorylation and dephosphorylation) is the ability to control the half-life of proteins by proteolysis or degradation. Eukaryotes mainly use ubiquitin as a molecule to target proteins for destruction, which allows very complex regulation involving multiple proteins ensuring the specificity of ubiquitination. Bacteria also contain complex degradation machineries, and the regulators of signal transduction are usually unstable proteins (reviewed in Jenal and Hengge-Aronis, 2003). However, no system analog to ubiquitin has being found so far and degradation of proteins follows simpler rules such as the identity of the N-terminal amino acid or some sequence specific signals. Tunable degradation (for example, phosphorylation dependent) has been successfully developed in yeast and bacteria and is therefore expected to be a valuable tool for both prokaryotes and eukaryotes (McGinness et al., 2006; Grilly et al., 2007). Groups have taken advantage of our increasing knowledge in protein degradation as a tool for engineering signaling pathways. For example, a specific heterobifunctional small molecule, called PROTAC (proteolysis-targeting chimeric molecule), induces ubiquitin-dependent proteolysis that targets a given protein for degradation in a controlled manner and timing (Sakamoto et al., 2001). Protac1 is a chimeric molecule that can simultaneously bind the target protein (in this case, methionine aminopeptidase-2) and the F-box protein β-TRCP as part of the ubiquitin ligase complex, and thus Protac1 induces ubiquitination and degradation of the target protein. Diffusion and Transport In general, the topology and regulation of signal transduction networks are largely distinct in eukaryotes and prokaryotes. The different cell size and the existence of a nuclear membrane that isolates the genetic material and delays the responses in eukaryotes, as well as the uncoupling of transcription-translation, have profound effects on the efficiency of signal propagation. Therefore, signaling cascades are usually longer in eukaryotes where the increased number of intermediates, exemplified by the mitogen-activated protein kinase (MAPK) pathways, presumably plays a role in signal amplification and in the reduction of input noise (Mayawala et al., 2004). As a result, in many cases, the system cannot rely on diffusion to transmit a signal; it needs active transport that is ensured by the cytoskeleton and molecular motors. Prokaryotes, on the other hand, because of their small volume, can rely on diffusion as the way to transmit the signal, although there are exceptions, such as the Min system (Meinhardt and de Boer, 2001), which localizes proteins to the division plane in cell division. 38  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

Feedback Regulation and Computational Modeling In both prokaryotes and eukaryotes, robustness is achieved by negative and positive feedback regulation (Kollmann et al., 2005; Ferrell, 2008). The smaller size in general of protein phosphorylation cascades in bacteria, as well as the smaller degree of amplification requires fewer safeguards against spontaneous activation than in eukaryotes, and thus the complexity of negative feedback regulation is lower. The opposite is true in Eukarya, and negative feedback is expected to be larger and occur on multiple levels. The presence of negative feedbacks makes it difficult in many cases to modify signaling by altering kinetic or equilibrium binding properties (Kitano, 2007; Kiel and Serrano, 2009). By combining all these molecular engineering techniques, we now have the capability to not only rewire signaling pathways, but to control the way we rewire them. For instance, positive and negative feedback loops can prolong or shorten the output of a signaling cascade, and interconnected feedback loops can result in sophisticated network behaviors such as oscillations (for example, oscillations in the p53/MDM2 system; see Geva-Zatorsky et al., 2006). The presence of negative and positive feedbacks in signal transduction makes it difficult for the human brain to predict the consequences of altering or engineering the molecules of the network and thus requires the use of computer simulation. Even simple networks can show nonintuitive responses, and computer simulations are essential for predicting the effect of interlinked positive and negative feedbacks (reviewed in Di Ventura et al., 2006). Powerful modeling requires information on concentration of species and rate constants for reactions. Therefore, it is worth stressing the necessity of having tools for acquiring rate constants (in vitro and in vivo) (Slaughter et al., 2007), as well as tools for determining absolute concentrations (Picotti et al., 2009). Also, sensitivity analysis, which is the changing of parameters for reactions and concentrations, can be helpful in guiding experiments and finding the sensitive/crucial points in the network prior to engineering (Chen et al., 2009). Specificity and Crosstalk There is little evidence of crosstalk in prokaryotes (Bourret, 2008), whereas eukaryotic signaling pathways are highly complex and interconnected, which is a requirement for responding to and integrating multiple signals (reviewed in Jordan et al., 2000). Part of the complexity in signal transduction in eukaryotes arises from plasticity (pathway switching), meaning that signals can be switched into alternative pathways, while achieving nearly identical outputs (Amit et al., 2007). This could be a problem when engineering signaling in eukaryotes as it is difficult to isolate the engineered circuit from the rest of the cell. The documented presence of scaffolds and adapters can be used for isolating the signal, as well as for linking the engineered network to other cellular networks (reviewed in Bhattacharyya et al., 2006). Scaffolds in bacteria exist in the form of cytoskeletal proteins, like MreB, MinCDE, and FtsZ (Vats et al., 2009), and in the form of adaptors, such as those for targeted proteolysis, which improve the kinetics and specificity of substrate recognition (Dougan et al., 2002). In summary, in signaling systems, the main challenges lie in improving the protein design tools for rational engineering of protein dynamics, for allosteric conformational changes,

Figure 3. Modulation of Signal Transduction (A) A Ras mutant (V12, G37) that is unable to activate the effector protein c-Raf has been used to study c-Raf-independent effector pathways in human T cells (Czyzyk et al., 2003). (B) When human growth hormone (GH) is modified with a G120R mutation, it induces conformational changes in the growth hormone receptor (GHR) such that it can activate the Jak/STAT pathway but not the ERK pathway (Rowlinson et al., 2009). (C) Using computational design, a TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) mutant has been created such that it can only bind to the DR5 receptor, and not to TRAIL’s other receptors. This specificity induces faster kinetics of receptor activation and induction of apoptosis in colon carcinoma (Colo205) cells (E. Szegezdi, L.S., and A. Samali, unpublished data). (D) Using structure-based design, various c-Raf mutants have been made that alter the kinetics of its association with Ras. The effect on signal transduction of these mutants strongly depends on the network topology: the effect is minor under conditions of strong negative feedback in human embryonic kidney cells (HEK293), but a strong effect is observed in a cell line with reduced negative feedback (rabbit kidney cells, RK13) (Kiel and Serrano, 2009). (E) A system for controllable degradation of proteins has been devised in the yeast Saccharomyces cerevisiae that involves the expression of a modified ClpXp protease from the bacterium Escherichia coli. Target proteins are expressed with a ssrA tag, which is specifically recognized by the ClpXp proteasome (Grilly et al., 2007). EGFP, enhanced green fluorescent protein. (F) A system for controllable protein degradation in E. coli also uses the ClpXp system and target proteins tagged to ssrA. For better control of degradation, an additional feature is the inclusion of the adaptor protein SspB, which increases the efficiency of degradation, is expressed under an inducible promoter (McGinness et al., 2006).

and for linker properties, such as flexibility and effective distance (that is, the distance between two domains depends on the linker properties and not necessary only on its length). Of special importance is the possibility of manipulating existing receptors so that they can either recognize molecules not present in the biological system, or be connected (via the cytoplasmic part of the receptor) to newly engineered pathways in a specific manner. This also requires improvements in tools for protein design, and a better understanding of the conformational changes of receptors upon ligand binding. Simulation tools that are fast and take into account cell localization, diffusion, and active transport, will be needed to exploit spatial localization in signal transduction engineering. Finally, we need to control the kinetic parameters of association and dissociation, in order to fine tune our engineered pathways. Engineered Pathways in Eukaryotes and Prokaryotes Examples of engineered signal transduction pathways can be separated into engineering exercises that involve modulation of signaling pathways (that is, changing the properties of the pathway), those that rewire existing signal transduction pathways, and those that create artificial signaling modules. In the following, we will discuss recent engineering work accomplished in both eukaryotic and prokaryotic signal transduction pathways.

Modulation of Signaling When modulation is used to engineer signal transduction, one can distinguish between changes that selectively eliminate one or more particular branches in a pathway and those where the properties of proteins are altered, thus affecting interactions or half-lives of the involved proteins. Partial loss-of-function mutations can be used to specifically dissect the contribution of different branches within a network to a particular output. This has been done to study the contribution of c-Raf-independent effector pathways of Ras in T cells (Czyzyk et al., 2003). Various effector proteins were expressed together with Ras mutants impaired in their ability to bind and activate certain Ras effectors (Figure 3A). Improving these partial loss-of-function mutations in Ras or creating them for other key cellular proteins using rational protein design would provide the framework for dissecting single branches in the context of larger networks (“branch pruning”). In a recent study, growth hormone (GH) binding to its receptor was altered by mutagenesis (based on the X-ray structure) to determine the choice between two signaling pathways (Rowlinson et al., 2009). The designed growth hormone mutant shows altered conformational changes that induce one downstream pathway (Jak/STAT) but can no longer activate the other (ERK) (Figure 3B). Cell surface receptors and ligands are often involved in disease and have been the targets of structure-based design to block or change binding specificity. This has been successCell 140, January 8, 2010 ©2010 Elsevier Inc.  39

Figure 4. Rewiring of Signal Transduction Pathways (A) Combining different guanine nucleotide exchange factor (GEF) domains with an artificial regulatory motif has been used to rewire GTPase signaling. The regulatory motif consists of a PDZ domain and a PDZ binding motif, rendering the GEF inactive by autoinhibition. Activation is achieved by cAMP induction, PKA activation, and phosphorylation of the PDZ binding motif (Yeh et al., 2007). (B) In RAT-2 rat fibroblasts epidermal growth factor (EGF) signaling has been rewired by fusing SH2 or PTB domains of the adaptor proteins Grb2 and Shc to the dead domain (DED) of FADD. Consequently, cell death is triggered after stimulation by EGF (Howard et al., 2003). (C) A chimera between the light-sensing part of the G protein coupled receptor (GPCR) of Rhodopsin and two other GPCRs, the β2 and α1 adrenergic receptors, results in light stimulated β2 and α1 specific cellular responses in human embryonic kidney (HEK293) cells and in adult mice (Airan et al., 2009). (D) Light-dependent activation in the bacterium Escherichia coli is achieved by a chimera of a cyanobacterial photoreceptor (PCB) with the EnvZ-OmpR two-component system. The reporter is modified to produce a black compound, so E. coli cells when grown on agar, produce a light dependent image (“bacterial photography”) (Levskaya et al., 2005). (E) Saccharomyces cerevisiae cells were engineered for cell-to-cell communication with elements from the model plant Arabidopsis thaliana. Sender cells produced a plant cytokinin (IP), which could diffuse to receiver cells and bind to the plant receptor AtCRE1, and consequently induced the yeast YPD1 signaling cascade (Chen and Weiss, 2005). (F) A cell-to-cell communication system was engineered in the marine bacterium Vibrio fischeri, which produced a diffusible acyl-homoserine lactone (AHL). Upon a critical cell density, AHL together with a transcriptional regulator (LuxR) induced the production of a “killer gene” (E) (You et al., 2004).

fully achieved for tumor necrosis factor (TNF) and TNF-related apoptosis-inducing ligand (TRAIL) signaling (van der Sloot et al., 2006; Tur et al., 2008; Mukai et al., 2009; Reis et al., 2009). In another approach, a kinetically controlled inhibition through prevention of heterodimerization, leads to faster kinetics of receptor activation and induction of apoptosis (E. Szegezdi, L.S., and A. Samali, unpublished data). In this case, computer simulations of the activation kinetics guided the design of mutants that were otherwise counterintuitive (Figure 3C). In the above discussed design examples, the successes were achieved by a combination of available X-ray structures, powerful design algorithms, and a bounty of prior knowledge on these systems. Protein design can also be used to specifically modify affinities or kinetic properties of protein complexes to provide knowledge about the underlying network topologies. Recently, subtle mutations were designed to unveil the role of electrostatic interactions in the interface of the Ras-c-Raf on the magnitude of signal transduction stimulated by epidermal growth factor (EGF) (Kiel and Serrano, 2009). By a combination of experimental and modeling approaches, it was shown that the rate constants for the Ras-c-Raf association have a greater effect on signal transduction than the rate constants for 40  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

dissociation. However, this effect is strongly cell-type specific, with its behavior being critically dependent on the underlying network topology (strong/weak negative feedback) (Figure 3D). Here, the success was achieved without prior knowledge of the system; the different underlying feedback regulation of the two different cell lines only became clear later, after combining design, simulations, and experimental tools. The tunable degradation of a tagged protein has been achieved in a S. cerevisiae strain by expressing a modified E. coli protease (ClpXP) under the control of a repressible promoter (Grilly et al., 2007) (Figure 3E). A similar technique has been used in baceria for engineering controllable protein degradation (McGinness et al., 2006). In that case, the degradation system was optimized using modifications in the tag and by induction of the adaptor SspB, to allow better control of degradation (Figure 3F). Rewiring of Signal Transduction Pathways The modularity of proteins involved in signal transduction enables domain shuffling to make chimeras with new connectivity. Rewiring of signaling pathways using domains and peptides involved in autoinhibition has been successfully introduced into N-WASP (Neural Wiskott-Aldrich syndrome protein), a protein that regulates actin assembly (Dueber et al., 2003).

Figure 5. Artificial Signal Transduction and Cells (A) Introduction of a minimal human p53 signaling module into yeast, as a model system to study context independent human p53-Mdm2 interactions (Di Ventura et al., 2008). (B) In the bacterium Escherichia coli, partial artificial modules from the mevalonate pathway of the yeast Saccharomyces cerevisiae have been engineered to produce a terpenoid precursor for artemisinin, an antimalarial drug (Martin et al., 2003). (C) In the model plant Arabidopsis thaliana, a cytokinin response has been coupled to an E. coli twocomponent system (PhoB and OmpR) (Antunes et al., 2009). GFP, green fluorescent protein. (D) Transplantation of the genome of Mycoplasma mycoides into M. capricolum, via cloning of the M. mycoides genome as a yeast centromeric plasmid. This opens the possibility of engineering a bacterial genome in yeast, before transplanting in into the host bacterium (Lartigue et al., 2009).

Autoinhibitory interactions in the natural variant of N-WASP involve the GTPase binding domain (GBD), and this inhibition is relieved after binding of the GTP-bound form of CDC42, a Rho family GTPase. Rewiring of this pathway is achieved by replacing the relief of autoinhibition through the GBD-CDC42•GTP interaction with the relief of autoinhibition by a PDZ domain/ PDZ-ligand interaction. As a consequence, activation of this artificial N-WASP protein is achieved by increasing the amount of free PDZ ligand. This work has been extended in a remarkable follow-up study, in which GTPase signaling is rewired by recombining catalytic guanine nucleotide exchange factors (GEFs) for Rho proteins with new regulatory motifs (Yeh et al., 2007). These synthetic RhoGEFs responded to nonnative inputs that ultimately evoke morphological changes in the cells (Figure 4A). These studies are promising because they demonstrate a very high evolutionary plasticity of this large family of modular RhoGEF proteins. In another study, signal transduction by the receptor tyrosine kinase ErbB2 could be rewired by fusing SH2 domains of the adaptor proteins Grb2 and Shc to the death domain of FADD, which consequently induces cell death (Howard et al., 2003) (Figure 4B). Interestingly, a potential therapeutic application of this finding is that these hybrid adapters have the ability to selectively kill oncogenic cells. A large body of research on yeast signaling pathways has demonstrated that modifying signaling can be achieved by changing scaffold interactions (see Harris et al, 2001; reviewed in Dueber et al., 2004). In one study, pathways were rewired to have artificial input-output properties, such that α-factor, which normally activates the yeast mating pathway, instead triggers the osmolarity response (Park et al., 2003). More recently, the Ste5 scaffold has been used as a platform to modify overall pathway kinetics and dynamics in a predictable way (Bashor et al., 2008). Moreover, rewiring of kinase specificities is possible in bacterial two-component systems. This has been accomplished by investigating all residues involved in binding for both

the families and species and identifying the residues that have coevolved. Mutating these highly specific residues changes the output of the signal (Skerker et al., 2008). Other researchers, using computational methods, have successfully rewired quorum-sensing pathways and predicted outputs that could be confirmed experimentally (Haseltine and Arnold, 2008). A very exciting light-dependent regulation of a signaling cascade has been developed for eukaryotes (Airan et al., 2009). In this work, the authors generated a chimera between the light-sensing part of rhodopsin, a G protein-coupled receptor (GPCR), and another GPCR (the β2 and the α1 adrenergic receptor) (Figure 4C). This was used to study β2 and α1 GPCR signaling in human embryonic kidney 293 cells after induction with light and to investigate light-dependent behaviors in adult mice. Light-dependent activation and control of protein phosphorylation has also been achieved in E. coli (Levskaya et al., 2005). In this case, cyanobacterial photoreceptor (PCB) was fused with the E. coli EnvZ-OmpR two-component system (Figure 4D). The lacZ reporter was modified to produce a black compound, and hence these engineered E. coli when grown on agar could be used for “bacterial photography.” Inouye and colleagues were the first to create of chimeric chemoreceptors in E. coli (Utsumi et al., 1989). In this pioneering work, the cytoplasmic domain of the aspartate-sensing membrane receptor, Tar, was replaced by the C-terminal part of the osmosensor EnvZ, which uses OmpR as the transcriptional activator. Since then, many functional chimeric receptors have been built in E. coli. Among those are the chimeras between the E. coli aspartate receptor and the human insulin receptor (Moe et al., 1989), the chemoreceptor Trg and EnvZ in E. coli (Baumgartner et al., 1994), and the NarX sensor kinase and the Tar chemoreceptor in E. coli (Ward et al., 2002). Rewiring methods can also be used to modify cell-to-cell communication. A recent study used synthetic networks to induce cell-cell communication between engineered yeast Cell 140, January 8, 2010 ©2010 Elsevier Inc.  41

cells (Chen and Weiss, 2005). Here, “sender” cells synthesize a plant cytokinin, isopentyladenine (IP), which diffuses to neighboring receiver cells, that the plant IP receptor AtCRE1 (Figure 4E). Upon binding of IP, the activated receptor phosphorylates the yeast response regulator YPD1, further downstream proteins of this signaling cascade, and ultimately transcription is activated (and GFP production). Interestingly, both sender and receiver cells contained signaling elements from the model plant Arabidopsis thaliana. In another exercise in the design of cell-to-cell communication systems (in this case in the marine bacterium Vibrio fischeri), gene expression changes were coupled to cell survival or cell death, which allows for control of the population dynamics by “regulated killing” (You et al., 2004). Here, a diffusible acyl-homoserine lactone (AHL) is synthesized by the LuxI protein (Figure 4F). With increasing cell density AHL accumulates in the cytosol and binds to the LuxR transcriptional regulator, which in turn leads to the production of a killer gene (E). In another example, a directed evolution procedure has been applied to the quorum-sensing response (Collins et al., 2006). Here, the sensitivity of different chemoatractants is altered by dual positive-negative selection, tackling two goals: the stabilization of a desired interaction and the destabilization of the other. Artificial Signaling Networks and Cells One aim in synthetic biology is to build synthetic pathways that do not crosstalk with the endogenous machinery, and, ultimately, to create completely artificial cells. This can be done at two levels: (1) by transplanting whole pathways from other organisms, thus in principle minimizing crosstalk, and (2) by engineering different proteins in order to create de novo pathways designed to minimize context effects. The first steps toward complete artificial networks have been achieved by the successful introduction of a higher eukaryotic p53 signaling module into yeast (Di Ventura et al., 2008) (Figure 5A). The aim of this work was to study p53-Mdm2 functional interactions, in the absence of other functional regulators in the p53 signaling pathways. Therefore, the authors have chosen to analyze these interactions in yeast as a model organism, given that many regulators are missing in this system. Partial artificial modules have been created in the field of biotechnology, such as that for artemisinin biosynthesis, by exporting the precursor synthesis pathway from S. cerevisiae to E. coli (Martin et al., 2003) (Figure 5B). Moreover, the interkingdom transfer of signaling modules has been accomplished for the two-component systems from plants to bacteria (Antunes et al., 2009) (Figure 5C) and vice versa (Suzuki et al., 2001). The above examples suggest that in the future it may be possible to transplant full pathways from one species to another while preserving functionality. Progress can also be expected toward the assembly of fully artificial organelles, which would independently perform their functions in the cytosol (Purnick and Weiss, 2009). Currently, computational tools provide a new array of possibilities for protein engineering, such as computer-designed receptor variants for pre-existing ligands (Looger et al., 2003). This approach has also been used to convert a ribose-binding protein into a receptor for extracellular Zn2+, which in turn could be coupled to the activation of a target gene (Dwyer et al., 2003). Eventually, we will be able to design signaling motifs 42  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

and domains with the desired structure, which has been shown for a small protein (Kuhlman et al., 2003) or for a new enzyme (Jiang et al., 2008), allowing synthetic biologists to build new artificial circuits that will not interfere with the endogenous network. The number of possibilities could be further expanded by using nonnatural amino acids. For example, orthogonal aminoacyl-tRNA synthetases and tRNA suppressors have been used to implement the genetic code with new, synthetic amino acids (Liu et al., 2007), and this might open the possibility for increasing the structural variability and binding properties of proteins. For instance, one can include amino acids that induce the dimerization of a given protein with its binding partner upon light exposure, which would be a means of controlling or inducing protein-protein interactions (Chin et al., 2002). The limitation for these kinds of exercises is the problem of synthesizing large chromosomes or plasmids that contain multiple genes plus regulatory elements in an organism. However, recent advances in DNA synthesis (Gibson et al., 2008) and the possibility of replacing a whole bacterial chromosome with another (Lartigue et al., 2007; Lartigue et al., 2009) (Figure 5D) show that this problem may be overcome soon. Toward Applications in Biotechnology and Medicine There have been synthetic signaling networks that have been successfully designed for specific applications in biomedicine and industrial biotechnology, but there is still more to be done. Here, we describe some examples highlighting the promising future applications of engineered signaling pathways. Some researchers have been engineering simple peptide building blocks to assemble new proteins through functional screening of combinatorial libraries (“motif reprogramming”; see Saito et al., 2007), which has proven successful in reprogramming cancer cells to undergo apoptosis (Saito et al., 2004). In addition, the use of synthetic intracellular antibodies (“intrabodies”; see Lobato and Rabbitts, 2003) could lead to an unlimited battery of surfaces for protein-protein interactions. For example, single-domain antibody fragments have proven useful in preventing Ras signaling in oncogenic cells (Tanaka et al., 2007) and could potentially set the basis for macromolecule-based drugs in cancer therapy. A different approach is exemplifid by an engineered strain of the pathogenic bacterium Yersinina sp. that can sense the microenvironment of a tumor and invade cancer cells (Anderson et al., 2006). Similarly, attenuated Salmonella strains have been engineered to target malignant cells (reviewed in Bermudes et al., 2002). In the future, these bacteria might also be used as vehicles for drug delivery. A synthetic network introduced into mammalian cells has been used to evaluate sensitivity to an antituberculosis drug (Weber et al., 2008). With further developments, this could become a system that aids in drug discovery. In another exciting approach, programmable cells have been constructed using combinations of synthetic gene networks and regulatory cellular circuits (Kobayashi et al., 2004). In this work, the engineered bacterial strains act as whole-cell biosensors—that is, one strain stores a detection event for later interrogation, while another strain expresses a target gene upon reaching a critical cell density. Security agencies have also found applications of bacterial signaling. For instance, a syn-

thetic signal transduction pathway regulating gene expression has been engineered so that it could function as a biosensor for new ligands, which could have applications for the detection of biological or chemical hazards (Looger et al., 2003). This system was based on a computational engineering approach, with the designed receptor acquiring the capacity to bind new ligands such as trinitrotoluene, l-lactate, or serotonin with high selectivity and affinity. Programmed pattern formation and engineering of complex behaviors is a first step toward tissue engineering (see reviews by Sia et al., 2007, and Purnick and Weiss, 2009). Currently, designed systems have succeeded in producing programmed multicellular pattern formation (Basu et al., 2005). Future progress in pattern formation and tissue engineering will primarily depend on the creation and modification of circuits for cell-tocell communication (Brenner et al., 2007). Another important application for industrial biotechnology might be engineering plant signal transduction (reviewed in Bowen et al., 2008). For example, in an exciting approach, a synthetic circuit allows chlorophyll levels to be placed under the control of a specific input (Antunes et al., 2006). Conclusions and Outlook In principle, in order to create fully artificial living systems that are able to respond to their environment, evolve, and reproduce, we first need a deep level of understanding of how cells work. However, we are still far away from reaching this goal. Systems biology has contributed significantly by uncovering many general systems properties, such as feedback regulation and noise. Further, large-scale experiments have revealed genetic and protein interaction networks in different organisms. However, our understanding is incomplete concerning the cellular functions and regulation of all proteins, and how cells adapt to a different environments. Even for a small bacterium like Mycoplasma pneumoniae, a full analysis of protein-protein interactions, gene expression, and metabolism (Güell et al., 2009; Kühner et al., 2009; Yus et al., 2009) reveals greater complexity than previously thought. For example, the promiscuity of many protein interactions and crosstalk between signaling pathways makes it difficult to engineer highly specific interactions. This could be minimized by using modules or proteins from other species, but this does not guarantee spurious interactions (for instance, the expression of Abl or Src kinases in yeast is toxic; see Trager and Martin, 1997, and Seeliger et al., 2005). Finally, the lack of knowledge in regards localization signals, degradation mechanisms, and protein turnover are drawbacks for full automatization and success of engineering projects. In spite of this, engineering has been successful in both prokaryotes and eukaryotes. Engineering in prokaryotes has been dominated by rewiring of two-component systems and the generation of chimeric receptors. In eukaryotes, rewiring strategies are mainly based on domain shuffling or using chimeric receptors with scaffolds rewiring being a very prominent engineering tool. The main conclusion from all of these examples is that crosstalk does not appear to be a major problem as has been feared. However, a general lack of understanding of negative feedback regulation could create problems, given that such regulation could dampen desired signaling outcomes

in engineered systems (Kiel and Serrano, 2009). Thus, in the same way that we are able to engineer proteins provided that we have a structure even when we do not yet understand how it folds, it appears that we can engineer living systems provided that we understand the basic features of how they are organized and regulated. Another problem that could hamper engineering of signaling pathways is the unique characteristics of proteins and domains that makes them difficult to engineer and extrapolate from one to another (for example, an SH3 domain can be folded on its own in many instances, but in some cases it will be fully unfolded; see Northey et al., 2002). Similarly, folding and aggregation issues when fusing domains from different proteins or spurious cell localization signals could hamper engineering of signaling pathways. Beyond these biological obstacles, there are other limitations, including the design of suitable linkers when shuffling domains, the efficient transformation and tight and controlled regulation of the expression of engineered proteins, and the inherent stochasticity and variability found in cells (Sigal et al., 2006). However, in practical terms, groups have found that domain reshuffling works very well in many systems. In respect to evolution, this clearly makes sense: multidomain proteins with rewired functionalities, and changed input-output behaviors, can be generated easily (and with not too many trials on optimal domain boundaries). Similarly, it seems that it is not difficult to find the right length of a linker, and neither aggregation nor misfolding has been reported as a major obstacle. Thus, improving algorithms for protein design to include protein flexibility and large conformational changes and the design of selections systems with directed evolution seems to be the main limitation at the moment. To tackle these issues, we will need to correctly identify proteins, domains, and interaction modules that behave in a robust and predictive manner to allow for assembly in different combinations. The success of engineering signal transduction pathways, and whether it will become a discipline like DNA engineering, will greatly depend on how many reliable standardized parts we can generate and how we can minimize context effects (such as going from a neuron to a muscle cell). Finally, aside from manipulating phosphorylation signals, it is expected that we will also witness the harnessing of signaling through second messengers [such as AppppA, (p) ppGpp, cAMP, and Ca2+] and other posttranslational modifications (such as acetylation or methylation). The combination of modules that can respond to these molecules or new ones engineered to recognize synthetic molecules (for example, the engineering of artifical receptors; reviewed in Mathonet and Fastrez, 2004), with classical phosphorylation relays, will lead the way to more complex designs and behaviors. The examples we have discussed above suggest that in principle many of the possible hurdles and problems one could predict to hamper synthetic biology are probably smaller than we thought. Even the limitation of making small chromosomes is on its way to being solved by new DNA synthesis capabilities. Perhaps the biggest challenge at the moment is to have a large collection of well-characterized protein modules that could be used for building new protein networks. We need modules that can be activated selectively by small molecules Cell 140, January 8, 2010 ©2010 Elsevier Inc.  43

that can diffuse through the membrane, or by other external perturbations (such as light, temperature, and pH). Also, we require modules that can undergo allosteric conformational changes in a controlled way, so that hidden surfaces can be exposed in a controlled manner to allow localization or interaction with other proteins. Furthermore, we need linear motifs that can be postranslationally modified and then recognized in a highly specific manner by other modules. In addition, it is necessary to have more scaffolds, which could help insulate signals, and, more importantly, we need to have large repositories with all the information of these modules connected to tools for network simulation, so that one could work like an engineer designing a new machine. Of course, on top of this, we require a good knowledge of the system to be engineered, but this does not need to be exhaustive at least for many biotechnological applications. There have been many discussions on whether there is a limitation to what we can learn about biology. Although there will always be something new to learn about biological systems, like other disciplines of science, the pace of discovery may eventually slow down. On the other hand, provided that we have the tools, the engineering of living systems appears to be limited only by our imagination. Acknowledgments We thank J.A. Leigh-Bell, K. Michalodimitrakis, and T. Ferrar for their contributions and their comments on the manuscript. Research on signal transduction engineering in L.S.’s lab is supported by the European Union (Proteomics Specification in Time and Space [PROSPECTS], grant agreement number HEALTH-F4-2008-201648). Research on engineering in Mycoplasma is supported by Celldoctor-ERC advanced grant number 232913. References Airan, R.D., Thompson, K.R., Fenno, L.E., Bernstein, H., and Deisseroth, K. (2009). Temporally precise in vivo control of intracellular signalling. Nature 458, 1025–1029. Alm, E., Huang, K., and Arkin, A. (2006). The evolution of two-component systems in bacteria reveals different strategies for niche adaptation. PLoS Comput. Biol. 2, e143. Aloy, P., Böttcher, B., Ceulemans, H., Leutwein, C., Mellwig, C., Fischer, S., Gavin, A.C., Bork, P., Superti-Furga, G., Serrano, L., and Russell, R.B. (2004). Structure-based assembly of protein complexes in yeast. Science 303, 2026–2029. Amit, I., Citri, A., Shay, T., Lu, Y., Katz, M., Zhang, F., Tarcic, G., Siwak, D., Lahad, J., Jacob-Hirsch, J., et al. (2007). A module of negative feedback regulators defines growth factor signaling. Nat. Genet. 39, 503–512. Anderson, J.C., Clarke, E.J., Arkin, A.P., and Voigt, C.A. (2006). Environmentally controlled invasion of cancer cells by engineered bacteria. J. Mol. Biol. 355, 619–627. Andrianantoandro, E., Basu, S., Karig, D.K., and Weiss, R. (2006). Synthetic biology: new engineering rules for an emerging discipline. Mol. Syst. Biol. 2, 2006.0028. Antunes, M.S., Ha, S.B., Tewari-Singh, N., Morey, K.J., Trofka, A.M., Kugrens, P., Deyholos, M., and Medford, J.I. (2006). A synthetic de-greening gene circuit provides a reporting system that is remotely detectable and has a re-set capacity. Plant Biotechnol. J. 4, 605–622. Antunes, M.S., Morey, K.J., Tewari-Singh, N., Bowen, T.A., Smith, J.J., Webb, C.T., Hellinga, H.W., and Medford, J.I. (2009). Engineering key components in a synthetic eukaryotic signal transduction pathway. Mol. Syst. Biol. 5, 270.

44  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

Aravind, L., Anantharaman, V., and Iyer, L.M. (2003). Evolutionary connections between bacterial and eukaryotic signaling systems: a genomic perspective. Curr. Opin. Microbiol. 6, 490–497. Ashworth, J., Havranek, J.J., Duarte, C.M., Sussman, D., Monnat, R.J., Jr., Stoddard, B.L., and Baker, D. (2006). Computational redesign of endonuclease DNA binding and cleavage specificity. Nature 441, 656–659. Balagaddé, F.K., Song, H., Ozaki, J., Collins, C.H., Barnet, M., Arnold, F.H., Quake, S.R., and You, L. (2008). A synthetic Escherichia coli predator-prey ecosystem. Mol. Syst. Biol. 4, 187. Bashor, C.J., Helman, N.C., Yan, S., and Lim, W.A. (2008). Using engineered scaffold interactions to reshape MAP kinase pathway signaling dynamics. Science 319, 1539–1543. Basu, S., Gerchman, Y., Collins, C.H., Arnold, F.H., and Weiss, R. (2005). A synthetic multicellular system for programmed pattern formation. Nature 434, 1130–1134. Batchelor, E., and Goulian, M. (2003). Robustness and the cycle of phosphorylation and dephosphorylation in a two-component regulatory system. Proc. Natl. Acad. Sci. USA 100, 691–696. Baumgartner, J.W., Kim, C., Brissette, R.E., Inouye, M., Park, C., and Hazelbauer, G.L. (1994). Transmembrane signalling by a hybrid protein: communication from the domain of chemoreceptor Trg that recognizes sugar-binding proteins to the kinase/phosphatase domain of osmosensor EnvZ. J. Bacteriol. 176, 1157–1163. Becskei, A., and Serrano, L. (2000). Engineering stability in gene networks by autoregulation. Nature 405, 590–593. Becskei, A., Séraphin, B., and Serrano, L. (2001). Positive feedback in eukaryotic gene networks: cell differentiation by graded to binary response conversion. EMBO J. 20, 2528–2535. Belshaw, P.J., Spencer, D.M., Crabtree, G.R., and Schreiber, S.L. (1996). Controlling programmed cell death with a cyclophilin-cyclosporin-based chemical inducer of dimerization. Chem. Biol. 3, 731–738. Benner, S.A., and Sismour, A.M. (2005). Synthetic biology. Nat. Rev. Genet. 6, 533–543. Bermudes, D., Zheng, L.M., and King, I.C. (2002). Live bacteria as anticancer agents and tumor-selective protein delivery vectors. Curr. Opin. Drug Discov. Dev. 5, 194–199. Bhattacharyya, R.P., Reményi, A., Yeh, B.J., and Lim, W.A. (2006). Domains, motifs, and scaffolds: the role of modular interactions in the evolution and wiring of cell signaling circuits. Annu. Rev. Biochem. 75, 655–680. Bourret, R.B. (2008). Signal transduction meets systems biology: deciphering specificity determinants for protein-protein interactions. Mol. Microbiol. 69, 1336–1340. Bowen, T.A., Zdunek, J.K., and Medford, J.I. (2008). Cultivating plant synthetic biology from systems biology. New Phytol. 179, 583–587. Brenner, K., Karig, D.K., Weiss, R., and Arnold, F.H. (2007). Engineered bidirectional communication mediates a consensus in a microbial biofilm consortium. Proc. Natl. Acad. Sci. USA 104, 17300–17304. Brocchieri, L., and Karlin, S. (2005). Protein length in eukaryotic and prokaryotic proteomes. Nucleic Acids Res. 33, 3390–3400. Bulter, T., Lee, S.G., Wong, W.W., Fung, E., Connor, M.R., and Liao, J.C. (2004). Design of artificial cell-cell communication using gene and metabolic networks. Proc. Natl. Acad. Sci. USA 101, 2299–2304. Campagna, A., Serrano, L., and Kiel, C. (2008). Shaping dots and lines: adding modularity into protein interaction networks using structural information. FEBS Lett. 582, 1231–1236. Chen, M.-T., and Weiss, R. (2005). Artificial cell-cell communication in yeast Saccharomyces cerevisiae using signaling elements from Arabidopsis thaliana. Nat. Biotechnol. 23, 1551–1555. Chen, W.W., Schoeberl, B., Jasper, P.J., Niepel, M., Nielsen, U.B., Lauffen-

burger, D.A., and Sorger, P.K. (2009). Input-output behavior of ErbB signaling pathways as revealed by a mass action model trained against dynamic data. Mol. Syst. Biol. 5, 239. Chin, J.W., Martin, A.B., King, D.S., Wang, L., and Schultz, P.G. (2002). Addition of a photocrosslinking amino acid to the genetic code of Escherichiacoli. Proc. Natl. Acad. Sci. USA 99, 11020–11024. Chiti, F., and Dobson, C.M. (2006). Protein misfolding, functional amyloid, and human disease. Annu. Rev. Biochem. 75, 333–366. Collins, C.H., Leadbetter, J.R., and Arnold, F.H. (2006). Dual selection enhances the signaling specificity of a variant of the quorum-sensing transcriptional activator LuxR. Nat. Biotechnol. 24, 708–712. Corson, T.W., Aberle, N., and Crews, C.M. (2008). Design and applications of bifunctional small molecules: why two heads are better than one. ACS Chem. Biol. 3, 677–692.

Galperin, M.Y., and Gomelsky, M. (2005). Bacterial signal transduction modules: from genomics to biology. ASM News 71, 326–333. Gardner, T.S., Cantor, C.R., and Collins, J.J. (2000). Construction of a genetic toggle switch in Escherichia coli. Nature 403, 339–342. Geva-Zatorsky, N., Rosenfeld, N., Itzkovitz, S., Milo, R., Sigal, A., Dekel, E., Yarnitzky, T., Liron, Y., Polak, P., Lahav, G., et al. (2006). Oscillations and variability in the p53 system. Mol. Syst. Biol. 2, 2006.0033. Gibson, D.G., Benders, G.A., Axelrod, K.C., Zaveri, J., Algire, M.A., Moodie, M., Montague, M.G., Venter, J.C., Smith, H.O., and Hutchison, C.A., 3rd. (2008). One-step assembly in yeast of 25 overlapping DNA fragments to form a complete synthetic Mycoplasma genitalium genome. Proc. Natl. Acad. Sci. USA 105, 20404–20409. Gossen, M., and Bujard, H. (2002). Studying gene function in eukaryotes by conditional gene inactivation. Annu. Rev. Genet. 36, 153–173.

Czyzyk, J., Brogdon, J.L., Badou, A., Henegariu, O., Preston Hurlburt, P., Flavell, R., and Bottomly, K. (2003). Activation of CD4 T cells by Raf-independent effectors of Ras. Proc. Natl. Acad. Sci. USA 100, 6003–6008.

Greber, D., and Fussenegger, M. (2007). Mammalian synthetic biology: engineering of sophisticated gene networks. J. Biotechnol. 130, 329–345.

Dahiyat, B.I. (2006). In silico protein design: fitting sequence onto structure. Methods Mol. Biol. 316, 359–374.

Grilly, C., Stricker, J., Pang, W.L., Bennett, M.R., and Hasty, J. (2007). A synthetic gene network for tuning protein degradation in Saccharomyces cerevisiae. Mol. Syst. Biol. 3, 127.

Deans, T.L., Cantor, C.R., and Collins, J.J. (2007). A tunable genetic switch based on RNAi and repressor proteins for regulating gene expression in mammalian cells. Cell 130, 363–372.

Güell, M., van Noort, V., Yus, E., Chen, W.-H., Leigh-Bell, J., Michalodimitrakis, K., Yamada, T., Arumugam, M., Doerks, T., Kühner, S., et al. (2009). Transcriptome complexity in a genome-reduced bacterium. Science 326, 1268–1271.

Di Ventura, B., Lemerle, C., Michalodimitrakis, K., and Serrano, L. (2006). From in vivo to in silico biology and back. Nature 443, 527–533.

Harris, K., Lamson, R.E., Nelson, B., Hughes, T.R., Marton, M.J., Roberts, C.J., Boone, C., and Pryciak, P.M. (2001). Role of scaffolds in MAP kinase pathway specificity revealed by custom design of pathway-dedicated signaling proteins. Curr. Biol. 11, 1815–1824.

Di Ventura, B., Funaya, C., Antony, C., Knop, M., and Serrano, L. (2008). Reconstitution of Mdm2-dependent post-translational modifications of p53 in yeast. PLoS ONE 3, e1507. Dougan, D.A., Mogk, A., Zeth, K., Turgay, K., and Bukau, B. (2002). AAA+ proteins and substrate recognition, it all depends on their partner in crime. FEBS Lett. 529, 6–10. Drubin, D.A., Way, J.C., and Silver, P.A. (2007). Designing biological systems. Genes Dev. 21, 242–254. Dueber, J.E., Yeh, B.J., Chak, K., and Lim, W.A. (2003). Reprogramming control of an allosteric signaling switch through modular recombination. Science 301, 1904–1908.

Haruki, H., Nishikawa, J., and Laemmli, U.K. (2008). The anchor-away technique: rapid, conditional establishment of yeast mutant phenotypes. Mol. Cell 31, 925–932. Haseltine, E.L., and Arnold, F.H. (2008). Implications of rewiring bacterial quorum sensing. Appl. Environ. Microbiol. 74, 437–445. Howard, P.L., Chia, M.C., Del Rizzo, S., Liu, F.F., and Pawson, T. (2003). Redirecting tyrosine kinase signaling to an apoptotic caspase pathway through chimeric adaptor proteins. Proc. Natl. Acad. Sci. USA 100, 11267–11272.

Dueber, J.E., Yeh, B.J., Bhattacharyya, R.P., and Lim, W.A. (2004). Rewiring cell signaling: the logic and plasticity of eukaryotic protein circuitry. Curr. Opin. Struct. Biol. 14, 690–699.

Hunter, S., Apweiler, R., Attwood, T.K., Bairoch, A., Bateman, A., Binns, D., Bork, P., Das, U., Daugherty, L., Duquenne, L., et al. (2009). InterPro: the integrative protein signature database. Nucleic Acids Res. 37 (Database issue), D211–D215.

Dwyer, M.A., Looger, L.L., and Hellinga, H.W. (2003). Computational design of a Zn2+ receptor that controls bacterial gene expression. Proc. Natl. Acad. Sci. USA 100, 11255–11260.

Isalan, M., Santori, M.I., Gonzalez, C., and Serrano, L. (2005). Localized transfection on arrays of magnetic beads coated with PCR products. Nat. Methods 2, 113–118.

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

Isalan, M., Lemerle, C., Michalodimitrakis, K., Horn, C., Beltrao, P., Raineri, E., Garriga-Canut, M., and Serrano, L. (2008). Evolvability and hierarchy in rewired bacterial gene networks. Nature 452, 840–845.

England, P.M., Lester, H.A., Davidson, N., and Dougherty, D.A. (1997). Sitespecific, photochemical proteolysis applied to ion channels in vivo. Proc. Natl. Acad. Sci. USA 94, 11025–11030.

Jenal, U., and Hengge-Aronis, R. (2003). Regulation by proteolysis in bacterial cells. Curr. Opin. Microbiol. 6, 163–172.

Ferrell, J.E., Jr. (2008). Feedback regulation of opposing enzymes generates robust, all-or-none bistable responses. Curr. Biol. 18, R244–R245.

Jenal, U., and Stephens, C. (2002). The Caulobacter cell cycle: timing, spatial organization and checkpoints. Curr. Opin. Microbiol. 5, 558–563.

Finn, R.D., Tate, J., Mistry, J., Coggill, P.C., Sammut, S.J., Hotz, H.R., Ceric, G., Forslund, K., Eddy, S.R., Sonnhammer, E.L., and Bateman, A. (2008). The Pfam protein families database. Nucleic Acids Res. 36 (Database issue), D2 81–D288.

Jiang, L., Althoff, E.A., Clemente, F.R., Doyle, L., Röthlisberger, D., Zanghellini, A., Gallaher, J.L., Betker, J.L., Tanaka, F., Barbas, C.F., 3rd, et al. (2008). De novo computational design of retro-aldol enzymes. Science 319, 1387–1391.

Friedland, A.E., Lu, T.K., Wang, X., Shi, D., Church, G., and Collins, J.J. (2009). Synthetic gene networks that count. Science 324, 1199–1202.

Jordan, J.D., Landau, E.M., and Iyengar, R. (2000). Signaling networks: the origins of cellular multitasking. Cell 103, 193–200.

Fung, E., Wong, W.W., Suen, J.K., Bulter, T., Lee, S.G., and Liao, J.C. (2005). A synthetic gene-metabolic oscillator. Nature 435, 118–122.

Kamalakkannan, S., Murugan, V., Jagannadham, M.V., Nagaraj, R., and Sankaran, K. (2004). Bacterial lipid modification of proteins for novel protein engineering applications. Protein Eng. Des. Sel. 17, 721–729.

Galperin, M.Y. (2004). Bacterial signal transduction network in a genomic perspective. Environ. Microbiol. 6, 552–567.

Kempiak, S.J., and Segall, J.E. (2004). Stimulation of cells using EGF-coated magnetic beads. Sci. STKE 2004, pl1.

Cell 140, January 8, 2010 ©2010 Elsevier Inc.  45

Kiel, C., and Serrano, L. (2009). Cell type-specific importance of ras-c-raf complex association rate constants for MAPK signaling. Sci. Signal. 2, ra38. Kiel, C., Selzer, T., Shaul, Y., Schreiber, G., and Herrmann, C. (2004). Electrostatically optimized Ras-binding Ral guanine dissociation stimulator mutants increase the rate of association by stabilizing the encounter complex. Proc. Natl. Acad. Sci. USA 101, 9223–9228. Kiel, C., Beltrao, P., and Serrano, L. (2008). Analyzing protein interaction networks using structural information. Annu. Rev. Biochem. 77, 415–441. Kim, P.M., Lu, L.J., Xia, Y., and Gerstein, M.B. (2006). Relating three-dimensional structures to protein networks provides evolutionary insights. Science 314, 1938–1941. Kitano, H. (2007). The theory of biological robustness and its implication in cancer. Ernst Schering Res. Found. Workshop 61, 69–88.

of terpenoids. Nat. Biotechnol. 21, 796–802. Mathonet, P., and Fastrez, J. (2004). Engineering of non-natural receptors. Curr. Opin. Struct. Biol. 14, 505–511. Mayawala, K., Gelmi, C.A., and Edwards, J.S. (2004). MAPK cascade possesses decoupled controllability of signal amplification and duration. Biophys. J. 87, L01–L02. McGinness, K.E., Baker, T.A., and Sauer, R.T. (2006). Engineering controllable protein degradation. Mol. Cell 22, 701–707. Meinhardt, H., and de Boer, P.A. (2001). Pattern formation in Escherichia coli: a model for the pole-to-pole oscillations of Min proteins and the localization of the division site. Proc. Natl. Acad. Sci. USA 98, 14202–14207. Metzger, H., and Lindner, E. (1981). The positive inotropic-acting forskolin, a potent adenylate cyclase activator. Arzneimittelforschung 31, 1248–1250.

Kobayashi, H., Kaern, M., Araki, M., Chung, K., Gardner, T.S., Cantor, C.R., and Collins, J.J. (2004). Programmable cells: interfacing natural and engineered gene networks. Proc. Natl. Acad. Sci. USA 101, 8414–8419.

Michalodimitrakis, K., and Isalan, M. (2009). Engineering prokaryotic gene circuits. FEMS Microbiol. Rev. 33, 27–37.

Kollmann, M., Løvdok, L., Bartholomé, K., Timmer, J., and Sourjik, V. (2005). Design principles of a bacterial signalling network. Nature 438, 504–507.

Moe, G.R., Bollag, G.E., and Koshland, D.E., Jr. (1989). Transmembrane signaling by a chimera of the Escherichia coli aspartate receptor and the human insulin receptor. Proc. Natl. Acad. Sci. USA 86, 5683–5687.

Kortemme, T., Joachimiak, L.A., Bullock, A.N., Schuler, A.D., Stoddard, B.L., and Baker, D. (2004). Computational redesign of protein-protein interaction specificity. Nat. Struct. Mol. Biol. 11, 371–379. Kuhlman, B., Dantas, G., Ireton, G.C., Varani, G., Stoddard, B.L., and Baker, D. (2003). Design of a novel globular protein fold with atomic-level accuracy. Science 302, 1364–1368. Kühner, S., van Noort, V., Betts, M.J., Leo-Macias, A., Batisse, C., Rode, M., Yamada, T., Maier, T., Bader, S., Beltran-Alvarez, P., et al. (2009). Proteome organization in a genome-reduced bacterium. Science 326, 1235–1240. Kurosaki, T. (2002). Regulation of B-cell signal transduction by adaptor proteins. Nat. Rev. Immunol. 2, 354–363. Lartigue, C., Glass, J.I., Alperovich, N., Pieper, R., Parmar, P.P., Hutchison, C.A., 3rd, Smith, H.O., and Venter, J.C. (2007). Genome transplantation in bacteria: changing one species to another. Science 317, 632–638. Lartigue, C., Vashee, S., Algire, M.A., Chuang, R.Y., Benders, G.A., Ma, L., Noskov, V.N., Denisova, E.A., Gibson, D.G., Assad-Garcia, N., et al. (2009). Creating bacterial strains from genomes that have been cloned and engineered in yeast. Science 325, 1693–1696.

Mukai, Y., Shibata, H., Nakamura, T., Yoshioka, Y., Abe, Y., Nomura, T., Taniai, M., Ohta, T., Ikemizu, S., Nakagawa, S., et al. (2009). Structure-function relationship of tumor necrosis factor (TNF) and its receptor interaction based on 3D structural analysis of a fully active TNFR1-selective TNF mutant. J. Mol. Biol. 385, 1221–1229. Northey, J.G.B., Di Nardo, A.A., and Davidson, A.R. (2002). Hydrophobic core packing in the SH3 domain folding transition state. Nat. Struct. Biol. 9, 126–130. Park, S.H., Zarrinpar, A., and Lim, W.A. (2003). Rewiring MAP kinase pathways using alternative scaffold assembly mechanisms. Science 299, 1061–1064. Pedraza, J.M., and Paulsson, J. (2008). Effects of molecular memory and bursting on fluctuations in gene expression. Science 319, 339–343. Peters, F.B., Brock, A., Wang, J., and Schultz, P.G. (2009). Photocleavage of the polypeptide backbone by 2-nitrophenylalanine. Chem. Biol. 16, 148–152. Picotti, P., Bodenmiller, B., Mueller, L.N., Domon, B., and Aebersold, R. (2009). Full dynamic range proteome analysis of S. cerevisiae by targeted proteomics. Cell 138, 795–806.

Letunic, I., Doerks, T., and Bork, P. (2009). SMART 6: recent updates and new developments. Nucleic Acids Res. 37 (Database issue), D229–D232.

Porter, S.L., Wadhams, G.H., and Armitage, J.P. (2008). Rhodobacter sphaeroides: complexity in chemotactic signalling. Trends Microbiol. 16, 251–260.

Levskaya, A., Chevalier, A.A., Tabor, J.J., Simpson, Z.B., Lavery, L.A., Levy, M., Davidson, E.A., Scouras, A., Ellington, A.D., Marcotte, E.M., and Voigt, C.A. (2005). Synthetic biology: engineering Escherichia coli to see light. Nature 438, 441–442.

Pryciak, P.M. (2009). Designing new cellular signaling pathways. Chem. Biol. 16, 249–254.

Liu, W., Brock, A., Chen, S., Chen, S., and Schultz, P.G. (2007). Genetic incorporation of unnatural amino acids into proteins in mammalian cells. Nat. Methods 4, 239–244. Lobato, M.N., and Rabbitts, T.H. (2003). Intracellular antibodies and challenges facing their use as therapeutic agents. Trends Mol. Med. 9, 390–396. Looger, L.L., Dwyer, M.A., Smith, J.J., and Hellinga, H.W. (2003). Computational design of receptor and sensor proteins with novel functions. Nature 423, 185–190.

Purnick, P.E., and Weiss, R. (2009). The second wave of synthetic biology: from modules to systems. Nat. Rev. Mol. Cell Biol. 10, 410–422. Ramamurthi, K.S., and Losick, R. (2009). Negative membrane curvature as a cue for subcellular localization of a bacterial protein. Proc. Natl. Acad. Sci. USA 106, 13541–13545. Redondo, P., Prieto, J., Muñoz, I.G., Alibés, A., Stricher, F., Serrano, L., Cabaniols, J.P., Daboussi, F., Arnould, S., Perez, C., et al. (2008). Molecular basis of xeroderma pigmentosum group C DNA recognition by engineered meganucleases. Nature 456, 107–111.

López De La Paz, M., Goldie, K., Zurdo, J., Lacroix, E., Dobson, C.M., Hoenger, A., and Serrano, L. (2002). De novo designed peptide-based amyloid fibrils. Proc. Natl. Acad. Sci. USA 99, 16052–16057.

Reina, J., Lacroix, E., Hobson, S.D., Fernandez-Ballester, G., Rybin, V., Schwab, M.S., Serrano, L., and Gonzalez, C. (2002). Computer-aided design of a PDZ domain to recognize new target sequences. Nat. Struct. Biol. 9, 621–627.

Lyon, G.J., Mayville, P., Muir, T.W., and Novick, R.P. (2000). Rational design of a global inhibitor of the virulence response in Staphylococcus aureus, based in part on localization of the site of inhibition to the receptor-histidine kinase, AgrC. Proc. Natl. Acad. Sci. USA 97, 13330–13335.

Reis, C.R., van der Sloot, A.M., Szegezdi, E., Natoni, A., Tur, V., Cool, R.H., Samali, A., Serrano, L., and Quax, W.J. (2009). Enhancement of antitumor properties of rhTRAIL by affinity increase toward its death receptors. Biochemistry 48, 2180–2191.

Martin, V.J., Pitera, D.J., Withers, S.T., Newman, J.D., and Keasling, J.D. (2003). Engineering a mevalonate pathway in Escherichia coli for production

Rieker, J.D., and Hu, J.C. (2000). Molecular applications of fusions to leucine zippers. Methods Enzymol. 328, 282–296.

46  Cell 140, January 8, 2010 ©2010 Elsevier Inc.

Rivera, V.M., Clackson, T., Natesan, S., Pollock, R., Amara, J.F., Keenan, T., Magari, S.R., Phillips, T., Courage, N.L., Cerasoli, F., Jr., et al. (1996). A humanized system for pharmacologic control of gene expression. Nat. Med. 2, 1028–1032. Rowlinson, S.W., Yoshizato, H., Barclay, J.L., Brooks, A.J., Behncken, S.N., Kerr, L.M., Millard, K., Palethorpe, K., Nielsen, K., Clyde-Smith, J., et al. (2009). An agonist-induced conformational change in the growth hormone receptor determines the choice of signalling pathway. Nat. Cell Biol. 10, 740–747. Saito, H., Honma, T., Minamisawa, T., Yamazaki, K., Noda, T., Yamori, T., and Shiba, K. (2004). Synthesis of functional proteins by mixing peptide motifs. Chem. Biol. 11, 765–773. Saito, H., Minamisawa, T., and Shiba, K. (2007). Motif programming: a microgene-based method for creating synthetic proteins containing multiple functional motifs. Nucleic Acids Res. 35, e38. Sakamoto, K.M., Kim, K.B., Kumagai, A., Mercurio, F., Crews, C.M., and Deshaies, R.J. (2001). Protacs: chimeric molecules that target proteins to the Skp1-Cullin-F box complex for ubiquitination and degradation. Proc. Natl. Acad. Sci. USA 98, 8554–8559. Santori, M.I., Gonzalez, C., Serrano, L., and Isalan, M. (2006). Localized transfection with magnetic beads coated with PCR products and other nucleic acids. Nat. Protoc. 1, 526–531. Seeliger, M.A., Young, M., Henderson, M.N., Pellicena, P., King, D.S., Falick, A.M., and Kuriyan, J. (2005). High yield bacterial expression of active c-Abl and c-Src tyrosine kinases. Protein Sci. 14, 3135–3139. Selzer, T., Albeck, S., and Schreiber, G. (2000). Rational design of faster associating and tighter binding protein complexes. Nat. Struct. Biol. 7, 537–541. Serrano, L. (2007). Synthetic biology: promises and challenges. Mol. Syst. Biol. 3, 158. Shetty, R.P., Endy, D., and Knight, T.F., Jr. (2008). Engineering BioBrick vectors from BioBrick parts. J Biol Eng 2, 5. Sia, S.K., Gillette, B.M., and Yang, G.J. (2007). Synthetic tissue biology: tissue engineering meets synthetic biology. Birth Defects Res. C Embryo Today 81, 354–361. Sigal, A., Milo, R., Cohen, A., Geva-Zatorsky, N., Klein, Y., Liron, Y., Rosenfeld, N., Danon, T., Perzov, N., and Alon, U. (2006). Variability and memory of protein levels in human cells. Nature 444, 643–646.

transfer (Hpt) domains, signaling mediators in His-to-Asp phosphorelay, in Arabidopsis thaliana. Plant Cell Physiol. 42, 37–45. Swinburne, I.A., Miguez, D.G., Landgraf, D., and Silver, P.A. (2008). Intron length increases oscillatory periods of gene expression in animal cells. Genes Dev. 22, 2342–2346. Tanaka, T., Williams, R.L., and Rabbitts, T.H. (2007). Tumour prevention by a single antibody domain targeting the interaction of signal transduction proteins with RAS. EMBO J. 26, 3250–3259. Teruel, M.N., and Meyer, T. (2000). Translocation and reversible localization of signaling proteins: a dynamic future for signal transduction. Cell 103, 181–184. Tigges, M., Marquez-Lago, T.T., Stelling, J., and Fussenegger, M. (2009). A tunable synthetic mammalian oscillator. Nature 457, 309–312. Townsend, J.A., Wright, D.A., Winfrey, R.J., Fu, F., Maeder, M.L., Joung, J.K., and Voytas, D.F. (2009). High-frequency modification of plant genes using engineered zinc-finger nucleases. Nature 459, 442–445. Trager, J.B., and Martin, G.S. (1997). The role of the Src homology-2 domain in the lethal effect of Src expression in the yeast Saccharomyces cerevisiae. Int. J. Biochem. Cell Biol. 29, 635–648. Tur, V., van der Sloot, A.M., Reis, C.R., Szegezdi, E., Cool, R.H., Samali, A., Serrano, L., and Quax, W.J. (2008). DR4-selective tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) variants obtained by structure-based design. J. Biol. Chem. 283, 20560–20568. Utsumi, R., Brissette, R.E., Rampersaud, A., Forst, S.A., Oosawa, K., and Inouye, M. (1989). Activation of bacterial porin gene expression by a chimeric signal transducer in response to aspartate. Science 245, 1246–1249. van der Sloot, A.M., Tur, V., Szegezdi, E., Mullally, M.M., Cool, R.H., Samali, A., Serrano, L., and Quax, W.J. (2006). Designed tumor necrosis factor-related apoptosis-inducing ligand variants initiating apoptosis exclusively via the DR5 receptor. Proc. Natl. Acad. Sci. USA 103, 8634–8639. Vats, P., Yu, J., and Rothfield, L. (2009). The dynamic nature of the bacterial cytoskeleton. Cell. Mol. Life Sci. 66, 3353–3362. Ward, S.M., Delgado, A., Gunsalus, R.P., and Manson, M.D. (2002). A NarXTar chimera mediates repellent chemotaxis to nitrate and nitrite. Mol. Microbiol. 44, 709–719.

Skerker, J.M., Perchuk, B.S., Siryaporn, A., Lubin, E.A., Ashenberg, O., Goulian, M., and Laub, M.T. (2008). Rewiring the specificity of two-component signal transduction systems. Cell 133, 1043–1054.

Weber, W., Schoenmakers, R., Keller, B., Gitzinger, M., Grau, T., Daoud-El Baba, M., Sander, P., and Fussenegger, M. (2008). A synthetic mammalian gene circuit reveals antituberculosis compounds. Proc. Natl. Acad. Sci. USA 105, 9994–9998.

Slaughter, B.D., Schwartz, J.W., and Li, R. (2007). Mapping dynamic protein interactions in MAP kinase signaling using live-cell fluorescence fluctuation spectroscopy and imaging. Proc. Natl. Acad. Sci. USA 104, 20320–20325.

Yeh, B.J., Rutigliano, R.J., Deb, A., Bar-Sagi, D., and Lim, W.A. (2007). Rewiring cellular morphology pathways with synthetic guanine nucleotide exchange factors. Nature 447, 596–600.

Smits, W.K., Kuipers, O.P., and Veening, J.W. (2006). Phenotypic variation in bacteria: the role of feedback regulation. Nat. Rev. Microbiol. 4, 259–271.

You, L., Cox, R.S., 3rd, Weiss, R., and Arnold, F.H. (2004). Programmed population control by cell-cell communication and regulated killing. Nature 428, 868–871.

Spencer, D.M., Wandless, T.J., Schreiber, S.L., and Crabtree, G.R. (1993). Controlling signal transduction with synthetic ligands. Science 262, 1019–1024. Suzuki, T., Sakurai, K., Ueguchi, C., and Mizuno, T. (2001). Two types of putative nuclear factors that physically interact with histidine-containing phospho-

Yus, E., Maier, T., Michalodimitrakis, K., van Noort, V., Yamada, T., Chen, W.H., Wodke, J.A.H., Güell, M., Martínez, S., Bourgeois, R., et al. (2009). Impact of genome reduction on bacterial metabolism and its regulation. Science 326, 1263–1268.

Cell 140, January 8, 2010 ©2010 Elsevier Inc.  47