All cells respond to signals from the environment. Extracellular stimuli activate intracellular signal transduction pathways that make decisions about cell identity, behavior, and survival. A nascent field aims to design and construct new signaling pathways beyond those found in nature. Current strategies exploit the structural modularity of many signaling proteins, which makes them inherently amenable to domain-swapping tactics that exchange their input and output connections. The results reveal a remarkable degree of functional plasticity in signaling proteins and pathways, as well as regulatory logic that can be transported to new proteins. Modified adaptor and scaffold proteins can reroute signal traffic and adjust the response behavior of the pathway circuit. These synthetic biology approaches promise to deepen our understanding of existing signaling pathways and spur the development of new cellular tools for research, industry, and medicine.
Decades of research have provided biologists with an impressive understanding of how cell fate and behavior are controlled by external signals. Many intracellular signaling pathways have been dissected to an extent that the complete “parts list” is known and the activity of each individual protein component is understood. Consequently, researchers have begun to harness this knowledge to engineer new signaling pathways. Although this emerging field is still in its early stages, some initial efforts have been stunningly successful, in ways suggesting that conceptually simple regulatory strategies may be transportable between different signaling proteins. As a result, repeated iteration of pretested modules, motifs, and tactics may allow signaling proteins to be controlled in increasingly predictable ways. Ultimately, cellular engineers may be able to use relatively straightforward principles to design new signaling circuits with predetermined properties. This commentary will explore some of the underlying concepts, recent progress, and future directions.
Why engineer new signaling pathways? First, it provides a way of testing whether we really understand how signaling pathways work or whether critical aspects remain undiscovered. Second, it could foster the development of new research tools that allow cellular events to be probed with new or more precise control. Third, synthetic signaling pathways could have industrial or therapeutic applications, such as biosensors that detect and report on the presence of toxins, modified industrial microbes that execute desired metabolic activities only when conditions meet predetermined set points, or cell-based delivery systems that seek out a target niche for localized drug delivery. Fourth, it is conceivable that engineered cells could be developed as computational devices that rival electronic microprocessors. Finally, by attempting to mimic how sophisticated signaling pathways emerged in nature, synthetic approaches may help test ideas about the mechanisms of evolution.
This commentary will focus on synthetic signaling pathways, as opposed to synthetic gene expression circuits (constructed using transcriptional activators, repressors, and promoters) that have already yielded many interesting circuit behaviors, such as switches, oscillators, and memory. What advantages or differences can be offered by engineering signaling pathways? One important difference is speed, because some signal transduction responses (both activation and inactivation) can occur within seconds, which is too dynamic to wait for transcriptional and translational synthesis. Posttranslational responses may also pose a lower energetic burden on the cell. Another difference is that signaling pathways can mediate spatially restricted responses that are confined to a localized region of the cell. Finally, direct regulation of protein activity, rather than just protein levels, offers extra variety and precision of control over cellular events. However, the complexity of signal transduction pathways may make computational prediction of circuit behavior considerably more difficult than it has been for gene expression circuits.
Advances in “rational design” may eventually allow new signaling proteins to be created de novo, but in the near future it will be considerably less laborious and more promising to modify existing signal transduction components in ways that co-opt their functions for new purposes. Importantly, it will be advantageous if cellular engineers do not have to start from scratch for each new protein or pathway. Rather, it will be preferable to develop methods that are generalizable and portable, so that common principles and reagents can be used repeatedly and predictably. This strategy would lend itself toward standardization of parts and practices that is a foundation of other engineering disciplines, where designers can build systems and devices with minimal concern for the inner workings of the component parts. Below we will consider how some fundamental properties of cell signaling could be harnessed toward these goals.
Many signal transduction proteins have a modular architecture, such that their ultimate function derives from the combined properties of multiple independent domains. Often, one domain (the “activity” or “output” domain) harbors a catalytic activity (e.g., kinase, phosphatase, or nucleotide exchange factor), and this is linked with other motifs, such as protein-protein interaction domains that dictate the connections to upstream regulators and downstream targets. Because these domains and motifs are often structurally autonomous and independently folding, they can confer their individual functional properties in a context-independent fashion. This arrangement is thought to make signaling pathways inherently evolvable through domain shuffling mediated by genetic recombination. Thus, if evolution has exploited the modular nature of signaling proteins to increase the diversity of natural pathways, then perhaps cellular engineers can use similar shuffling approaches to create new signaling proteins and pathways.
Figure 1 Modular Architecture of Signaling Proteins
(A) Natural signaling proteins often combine a catalytic (“output”) domain with interaction domains that determine its connections. Exchanging interaction domains can create synthetic chimeras that connect to new stimuli or targets.
(B) Interaction domains can link output domains to activators (i), substrates (ii), or subcellular locations (iii). They can also regulate protein activity by autoinhibitory binding (iv).
Indeed, existing evidence suggests that these structurally independent domains can be functionally independent and interchangeable. For instance, proteins in the MAP kinase family interact with their activators and targets by recognizing specific “docking sites”. These sites remain functional when moved to different positions in the partner protein and can be replaced with unrelated docking sequences from other partners. In another example, signaling in the yeast pheromone pathway requires two components (Ste5 and Ste20) to localize to the plasma membrane via both protein-membrane and protein-protein interactions. Yet each component remains functional when its polybasic membrane-binding motif is replaced with a structurally unrelated phospholipid-binding domain, and in one case even the protein-protein interaction can be replaced with a surrogate interaction. Finally, activation of Notch-family receptors triggers proteolytic release of the cytoplasmic tail; when this tail is replaced with heterologous domains (e.g., a transcription factor), new responses can be activated by Notch ligands. These and other examples illustrate how the modular architecture of signaling proteins confers an intrinsic degree of functional plasticity. Therefore, further elaboration of these domain swap strategies could readily alter the inputs, outputs, and/or subcellular locale of signaling events.
Signaling proteins can be activated by allosteric conformational changes that propagate from a regulatory site to the active site, but designing this form of regulation de novo is inherently difficult and protein specific. In contrast, many signaling proteins are controlled by fundamentally simpler “relief of inhibition” mechanisms that may be readily adapted for synthetic purposes. Here, binding interactions block the protein's function, and activating signals turn it on by disrupting the inhibitory interactions; the negative domains commonly occur in cis and hence are autoinhibitory. This class of regulatory mechanism includes both “intrasteric regulation,” in which inhibitory domains directly bind and occlude the catalytic site, as well as “modular allostery,” in which the active state is either sterically or conformationally prevented by interactions away from the catalytic site. In principle, any other modification or binding interaction that is mutually exclusive with the autoinhibited state (due to steric, electrostatic, or conformational incompatibility) could be appended to the protein and used to trigger its activation artificially.
Indeed, this strategy already has been spectacularly successful. The mammalian protein N-WASP regulates actin assembly via an output domain that is controlled by autoinhibition; it is then turned on when activating factors disrupt the inhibitory conformation. Lim and colleagues mimicked this mode of regulation with heterologous sequences, by attaching a common peptide-binding motif known as a PDZ domain at one end and its cognate target peptide at the other. The resulting intramolecular PDZ-peptide interaction inhibited the intervening N-WASP output domain, and the hybrid protein could be turned back on by the addition of soluble target peptide. In effect, this converted native N-WASP into a form that can be activated by a foreign signal, and in a way that is remarkably straightforward at both conceptual and technical levels. Similar regulation was achieved using another peptide-binding motif (an SH3 domain), and incorporation of both the PDZ-peptide and SH3-peptide pairs into the same molecule generated more sophisticated circuit behaviors, such as an “AND gate” in which protein activation required the addition of both peptide ligands. Separately, if the output domain was flanked with multiple tandem copies of the SH3 domain and its target peptide, the resulting cooperative binding changed the dose-response behavior from linear to sigmoidal (“ultrasensitive”).
Figure 2 Regulating Protein Activity with Foreign Autoinhibitory Interactions
(A) N-WASP stimulates the actin nucleation complex Arp2/3. The GTPase Cdc42 and the phospholipid PIP 2 activate N-WASP by relief of autoinhibition.
(B) The normal autoinhibitory interactions in N-WASP can be replaced with foreign sequences such as a PDZ domain and its binding peptide. (It is unclear whether the foreign interactions block catalytic activity by a steric or conformational effect.)
(C) Regulation of Rho GEF activity with foreign autoinhibitory interactions. Phosphorylation of the target peptide by PKA disrupts PDZ binding and hence activates the GEF.
In another remarkable example, the same core idea was applied to unrelated proteins. Starting with two g
