Synthetic approaches to understand, rewire, and construct complex transcriptional networks, particularly in eukaryotes, have been severely hindered by a lack of reliable components and a framework for designing and assembling them. We have developed an extensible synthetic biology framework for regulating eukaryotic transcription, whereby artificial ZF proteins are used as core building blocks from which to construct complex transcription functions and circuitry. The use of a context-dependent ZF selection scheme allows us to rapidly alter and program the ZF-DNA interaction specificity, and identify orthologous pairs of ZF arrays-DNA sites that can be engineered into sTFs for wiring networks within yeast. This work brings additional forms and levels of connectivity to synthetic transcriptional circuits, beyond that which is achievable with the few, classical prokaryotic TF-promoter pairs. Using our methodology, one should be able to create a virtually unlimited number of sTF-promoter pairs, with which to make transcriptional circuit connections. In this regard, we note that three-finger arrays have been engineered for more than 500 different nine-bp sites using the OPEN (Maeder et al., 2008
) and context-dependent assembly (CoDA) (Sander et al., 2011
) methods (J. Sander, M. Maeder, C. Khayter, E. Dahlborg, and J.K.J, unpublished data).
Engineered transcription activator-like (TAL) effectors have recently emerged as an important alternative to customized ZFs for programming DNA specificity. Naturally occurring TAL effectors encoded by Xanthamonas
bacteria bind to target DNA sequences using arrays of highly conserved 33–35 amino acid repeat domains. A single TAL effector repeat binds to one nucleotide of DNA with specificity of binding associated with the identities of amino acids at two positions known as repeat variable di-residues (RVDs). TAL effector repeats bearing different RVDs have been described for specific binding to each of the four possible DNA nucleotides, and these repeats can be simply joined together to create arrays capable of binding to extended target DNA sequences (Bogdanove and Voytas, 2011
). The simplicity and modularity of the TAL effector repeats suggests that nearly any DNA sequence should be targetable, and recent work has demonstrated that engineered TAL effector nuclease (TALEN) fusions can be robustly generated for a wide variety of different sequences (Reyon et al., 2012
). Nonetheless, engineered ZFs have several important advantages, including their considerably smaller size, their less repetitive coding sequence (potentially important for packaging into viral vectors), and a greater understanding of their biochemical properties, structure, and function, which is important for creating variations in affinity, specificity, and tunability.
We constructed a synthetic transcriptional cascade in yeast that was used as a test bed for systematically tuning sTF component properties. For instance, we tuned the strength of transcriptional outputs through modifications of ZF binding sites in the promoter architecture as well as through structure-guided modifications to the ZF protein to alter ZF-DNA interactions. Additionally, our framework provides the ability to engineer and tune transcriptional cooperativity. To date, there exists no simple way of building cooperative transcriptional systems, even though their importance is well-documented in both natural and synthetic gene regulation. As a result, in most synthetic studies, researchers have used TFs with integrated, cooperative properties. In contrast, our approach establishes a modular framework for constructing cooperative transcriptional activation schemes de novo, through the multimerization of weakly-activating ZF-TFs using low-affinity protein interaction domains (i.e., PDZ domains). This has important consequences for constructing higher-order complexes that more accurately mimic eukaryotic transcription regulation schemes, lead to sharper switch-like responses, and modulate cooperativity within circuits. Indeed, multimerization and cooperativity are ubiquitous molecular regulation schemes that underlie complex gating and decision-making in cells. For example, the yeast GAL1 promoter is able to integrate coactivator proteins in specific temporal order by utilizing the cooperativity of certain interactions to gate subsequent recruitment events (Bryant and Ptashne, 2003
). It is of great interest to understand how activators function cooperatively to assemble specific initiation complexes and regulate transcription. Our bottom-up and modular approach to transcriptional cooperativity could be used to synthetically recapitulate such phenomena so as to study these fundamental mechanisms of regulation. This type of approach has been used to understand transcriptional synergy in prokaryotes (Joung et al., 1993
), and our platform should now enable this strategy to be used to model more combinatorially complex eukaryotic promoters. Furthermore, our extensible and modular framework for multimeric and cooperative transcriptional systems may allow for the implementation of expanded computational operations in eukaryotes, such as logic devices with more input possibilities.
Previous reports have described various frameworks for creating dimeric ZF proteins. In all of these studies, elements derived from naturally occurring TFs (Pomerantz et al., 1998
; Wolfe et al., 2000
) or ones selected from combinatorial peptide libraries (Wang et al., 2001
; Wang and Pabo, 1999
) were used to dimerize two-finger units. A disadvantage of this strategy is that a dimerized two-finger complex would have a maximum specificity of 12 bps (assuming that each of the four fingers in the dimer specifies 3 bps). Our approach differs by utilizing three-finger monomers that have had their binding activities reduced by mutagenesis of nonspecific phosphate-contacting residues. This strategy creates dimeric proteins that can have specificities as high as 18 bps, a sequence long enough to be potentially unique in a mammalian genome. Furthermore, the use of modular protein-protein interaction domains for multimerization is advantageous for various reasons. For example, the interaction is orthogonal and tunable, allowing us to “match” affinities of all the component interactions making up the sTF complex, and it provides extensibility for further expanding the complex architecture and dynamically increasing or decreasing the DNA binding specificity of the complexes.
We also showed that complex signal processing behaviors can arise when control of TF cooperativity is combined with the ability to engineer promoters of multimerized ZF binding sites. Cellular signal processing is a mechanism by which environmental and other signals are integrated to modulate transcription and thus critical cellular processes, such as growth and stress responses. We constructed a simple, synthetic two-input transcriptional system that allowed us to decompose contributions from the sTF component properties to the system’s processing behavior. We showed that, with the same two core TFs and promoter operator sites, a cell could process and integrate signals in a variety of ways. For example, subtle changes, such as reversing promoter operators and disrupting protein-protein interactions, can have striking effects on the output of the system. This led to the construction of not just varied digital logic behavior, but a range of analog tunability. In an inhibitory system (), we arrived at an interesting Boolean logic gate that produced a positive signal only in the presence of a single input: A > B (A does not imply B). A broad observation from our studies was that specific perturbations to an sTF’s component properties (DNA affinity, multimerization with other species, location of operator, etc.) could allow it to interconvert between different transcriptional roles within the complex, such as activator, cooperative factor, noncontributory, and inhibitor. This synthetic approach could be utilized to explore the diversity of behaviors that can be programmed by even just a few transcriptional components; furthermore, our findings provide simple strategies for reprogramming the signal processing behavior of a cell. Similar strategies are undoubtedly employed naturally, where there are many examples of individual proteins that can take on either activating or repressing roles depending on the cellular and environmental states (Ma and Ptashne, 1988
; Maxon and Herskowitz, 2001
; Rubin-Bejerano et al., 1996
Given that TFs containing ZFs play a central role in eukaryotic promoter regulation (Pabo et al., 2001
), our system represents a promising means for synthetic recapitulation of eukaryotic promoter function, and thus will significantly enhance our ability to construct synthetic gene networks in mammalian cells, an area of tremendous potential in synthetic biology (Weber and Fussenegger, 2009
). Indeed, yeast may serve as a well-characterized testbed for the design and validation of synthetic gene circuits that can be subsequently ported to higher organisms. Synthetic transcriptional regulators based on the sTFs described here can be used to create classifier circuits to identify cell state (Nissim and Bar-Ziv, 2010
; Xie et al., 2011
), memory devices to record cellular events, and logic gates for cellular signal processing (Kramer et al., 2004
), which can aid in the study and control of stem cell differentiation, therapeutics, and complex human diseases. Additionally, ZF-based proteins have been shown to be powerful targeting elements of endogenous genomic loci in many mammalian cells, including cancer, immune, and stem cells. These proteins include ZF nucleases (Zou et al., 2009
), which are being tested in therapeutic applications for modifying/ disrupting disease-causing alleles and genes, and artificial TFs (Blancafort et al., 2005
), which can be used to modify the expression of native genes for reprogramming purposes. Together with these ZF technologies, our work may lead to the construction of integrated gene circuits—artificial circuitry that seamlessly and specifically integrates into endogenous gene networks and/or leads to the modification of endogenous genes—for more dynamical and sophisticated genetic control in cell-based therapeutic applications.
Synthetic biology is helping us to understand how organisms behave and develop through the forward engineering of molecular circuitry with well-understood genetic components (Elowitz and Lim, 2010
). The present work expands the synthetic biology toolkit with new genetic components, beyond re-purposed bacterial transcriptional components, to program eukaryotic cells. Additionally, it provides a bottom-up framework for exploring the complexity of eukaryotic promoters and their combinatorial regulation by TF complexes and circuitry. This framework can be a starting point for determining the transcriptional components, modules, and circuitry needed to implement the sophisticated behaviors that control the development and function of eukaryotic cells.