Identifying functional transcriptional regulators from a library of targeted CRs

A large body of work has identified correlations between the expression of specific CRs and global transcriptional activity. However, it is often unclear which CRs are causative of or merely associated with changes in transcriptional activity at specific loci. Therefore, we fused a library of 223 full-length putative CRs, comprising 45 chromatin-regulating complexes (Lenstra et al., 2011), to the targeting ZF and individually tested each protein's ability to activate or repress transcription from the pCYC1 reporter (Figure 2A, top). As shown in Figure 2B, numerous repressors and activators were identified from the library, spanning approximately 20-fold changes in repression and activation (Figures 2B, S1A, S1D and Table S1). We also observed expected changes in histone modifications at the reporter locus, measured by chromatin immunoprecipitation-quantitative PCR, upon expression of 17 CRs chosen for their predicted histone modifying catalytic domains/activities (Figure S2B). To confirm the changes in reporter expression were not simply a product of CR overexpression, we fused the 27 strongest repressors, 48 strongest activators, and all CRs with histone-modifying catalytic domains to a truncated, non-binding ZF protein. The vast majority of these fusions generated negligible changes in yEGFP expression (Figure S6B). Rsc3, Ldb7, Sum1 and Tod6 were exceptions, exhibiting changes in yEGFP levels regardless of targeting, suggesting either ZF-independent recruitment to the locus or global transcriptional regulation. We observed no correlation between transcriptional activation and the level of CR expression as measured by western blot (Figure S2C).

Clustering CRs by chromatin complex and function

We next asked if the targeted library could identify relationships between transcription and CR protein functions or complexes. Using gene ontology annotations, we first clustered all CRs by macromolecular complex (Table S2). Individual CRs were then conservatively classified as activators (Figure 2B, red bars, > 2 fold change yEGFP) and repressors (blue bars, < 0.7 fold change yEGFP). This classification excluded all non-targeted CRs, aside from the exceptions noted above (Figure S6B). When the percentage of activators in each chromatin complex was plotted against the percentage of repressors (Figure 2C, Table S2, Figure S1B), we discovered a number of clear patterns. Histone acetyltransferase (blue dots), H3K4 methyltransferase (“COMPASS/Set1”), and RNA PolII transcription-related complexes (red dots) were mostly composed of activating CRs. Histone deacetylase complexes were primarily composed of repressive CRs (pink dots). Nucleosome remodeling complexes trended weakly towards having more activators than repressors (green dots). When clustered by protein function terms (Figure S2A, Table S2, Figure S1C), groups associated with the transcriptional complex (red dots), histone acetyltransferase (blue dot), and histone methyltransferase (brown dots) terms contained primarily activators, while groups associated with chromatin binding (orange dots) and histone deacetylase (pink dots) terms contained primarily repressors. These results largely agree with the regulatory roles assigned to various complexes and protein functions through previous genome-wide and knockout/mutant strain studies (Lenstra et al., 2011; Ram et al., 2011; Venters et al., 2011).

Engineering combinatorial transcriptional logic

Native genes are simultaneously regulated by multiple proteins with different functions and activities, often giving rise to combinatorial transcriptional logic. Therefore, we next explored how co-recruitment of factors affects transcription. In particular, we were interested in how different CRs modulate the activity of a co-recruited VP16 domain. We fused VP16 to a second, orthogonal ZF protein (97-4, TTATGGGAG) (Khalil et al., 2012), which could be independently recruited to an operator placed directly downstream of the ZF-CR operator (Figure 3A). As expected, upon co-recruitment of VP16 with the ZF-CR library, we found that transcriptional outputs generally increased as compared to recruitment of CRs alone (Table S3). The CRs divided into six distinct classes of combinatorial regulators based on transcriptional logic: CRs capable of (1) dominant (Sir2 and Mig1) and (2) partial (Ash1 and Dot1) inhibition of VP16-mediated activation; CRs with no regulatory roles on their own and either (3) no (Eaf7 and Rvb2) or (4) enhanced (Ies6 and Cdc73) effect on VP16-mediated activation; finally, CRs that act (5) additively (Taf14 and Med4) or (6) synergistically (Cac2 and Set1) with VP16 to increase yEGFP expression (Figures 3B, S3A, and S3B). Synergy is the “cooperation” of factors to produce a total output, and here it was defined as the fraction of total output in excess of summing the outputs from the individual components (Figures 3C, 3D, S3C, and S3D).

Open in a separate window

Figure 3

Combinatorial recruitment reveals distinct classes of regulators for engineering transcriptional logic

(A) An engineered two-input system enabling the co-recruitment of CRs and VP16 transactivating domain (ZF 43-8, grey; ZF 97-4, blue) (Khalil et al., 2012). (B) Representative transcriptional logic outputs of the two-input system divide CRs into six distinct classes (top to bottom): VP16-independent dominant repressors, repressors, CRs with no effect, VP16 enhancers, additive activators (purple), and synergistic activators (red). (C) Activating CRs clustered by complex and plotted by level of transcriptional synergy. Transcription/preinitiation complex regulators generated weak synergy, while chromatin assembly/remodeling, chromatin-modifying, and transcription-elongating regulators generated strong synergy. Synergy is the “cooperation” of factors to produce a total output, and here is defined as the fraction of total output not accounted for by summing the outputs from the individual components. Synergy = [(A – 1) – (B – 1) – (C – 1)]/(A – 1) where A = CR and VP16, B = CR only, and C = VP16 only. (D) Activators clustered by gene ontology function terms and plotted as percentage of CRs in each term group with “strong synergy” (greater than the average synergy of 0.2). Error bars are standard deviations of three isogenic strains. See also S3.

To develop insight into CR functions that may underlie these different, combinatorial logic behaviors, we clustered CRs by complex (Figures 3C and S3C) and function (Figures 3D and S3D) and calculated the percentage of CRs in each cluster with strong synergy. When clustered by complex, we found that the majority of Mediator and TFIID subunits exhibited weak synergy with VP16 (Figures 3C and S3C, purple bars). In contrast, complexes that remodel and assemble chromatin (Swr1, RSC, CAF-1), promote transcriptional elongation (Paf1), or modify histones to open chromatin structure (NuA4, Set1) were comprised primarily of CRs that synergistically enhanced activation (Figures 3C and S3C, red bars).

We observed the same general trend when we clustered activating CRs by function, as opposed to complex (Figures 3D and S3D, Table S4); that is, CRs related to transcription factor and RNA PolII terms exhibited weak synergy with VP16 (purple), while those associated with chromatin remodeling, modifying, and binding exhibited strong synergy (red). VP16 is believed to activate transcription by recruiting preinitiation and transcription complex factors along with the Mediator complex (Milbradt et al., 2011). Thus, additive activation might occur through a co-recruited CR that functions similarly to VP16 or is part of either the transcription complex or Mediator. Importantly, we observed a simple additive relationship when we co-recruited two identical VP16 domains (Figure 3C, “VP16”). In contrast, other functions such as remodeling nucleosomes, modifying histones to alter chromatin accessibility, and promoting transcriptional elongation may synergistically amplify the output by increasing access of transcriptional machinery to DNA (Lam et al., 2008; Mirny, 2010).

Revealing spatially-encoded regulatory modes

While transcriptional regulation is canonically focused at promoter regions, there is also considerable evidence for chromatin-mediated regulation at other locations relative to ORFs: (1) nucleosomes are arrayed over entire genes with distinct positioning at promoter and terminator regions (Lam et al., 2008); (2) native CRs and transcription factors are often localized to spatially-specific, non-promoter regions to regulate genes (Groner et al., 2010); and (3) histone mark gradients have been observed over genes (Li et al., 2007; Pokholok et al., 2005). These observations suggest that CRs may asymmetrically and differentially regulate genes depending on their relative location to an ORF.

We sought to explore spatially-dependent regulatory behaviors using site-specific CR recruitment. We moved the ZF operators in our reporter locus from upstream of the coding sequence to downstream of the terminator (Figures 4 and S4). Our library of ZF-CRs was then inducibly recruited to the downstream element (Figure S4, blue and gold bars). No CRs were able to activate transcription from the downstream position, suggesting the importance of preinitiation/transcription complex assembly at promoters for activation. However, many CRs were able to repress transcription from the downstream position. Interestingly, several of these CRs exhibited “asymmetric” regulatory modes; in other words, they had opposite regulatory functions when targeted upstream versus downstream (Figure S4, grey bars). To develop insight into CR functions that may underlie these spatially-encoded behaviors, we grouped CRs by their spatial regulatory profile (i.e., upstream-activating or -repressive and downstream-activating or -repressive) and obtained associated gene ontology function terms for each group (Figure 4, Table S5). Subsets of these terms were unique to each grouping (Figure 4). Interestingly, while many upstream-activating/downstream-neutral CRs were associated with regulation of the transcriptional complex, factors that were upstream-activating/downstream-repressive appeared enriched in ATPase remodeling and DNA translocase activity. This suggests that remodeling activities can influence transcription from both ends of a gene, potentially by increasing RNA PolII accessibility at upstream regions while disrupting transcriptional elongation at downstream regions.

Open in a separate window

Figure 4

Engineering spatial regulation by targeting CRs upstream and downstream of a gene

A gene locus was engineered to recruit 223 CR fusions to operators either upstream or downstream (downstream of a CYC1 terminator) of a reporter gene. CRs were grouped according to their upstream- and downstream-targeted regulatory profiles. Gene ontology function terms unique to each group are listed along with the number of CRs in the group associated with each term. See also Figure S4.

Simultaneous and differential regulation of multiple genes

The spatial qualities of chromatin-based regulation could be exploited to engineer the simultaneous regulation of multiple genes. For example, based upon our comparison of upstream versus downstream targeting of the CR library (Figure 4), a single CR might simultaneously (and differentially) regulate two genes if recruited upstream of one ORF and downstream of another. To identify CRs capable of such simultaneous regulation, we constructed a dual-gene reporter system (Figure 5A) and recruited a small subset of our ZF-CRs to it. The CRs exhibited a variety of dual-gene regulation profiles (Figures 5A, S5A, and S5E). Notably, these included factors that could activate expression of one reporter gene while repressing the other. To confirm that the local contexts of the genes were not responsible for the observed behaviors, the ZF operators were swapped between upstream and downstream positions at both genes (Figures 5B, S5B, and S5F). As expected, the regulatory profiles were correspondingly inverted. Moreover, fusions of CRs to non-binding ZFs did not appreciably modulate transcription, strongly suggesting that these engineered regulatory modes are the result of site-specific targeting (Figures S5A, S5B, S5E, and S5F, left).

Open in a separate window

Figure 5

Simultaneous and distinct regulation of two genes by individual CRs

(A) Schematic of the engineered, dual-gene reporter locus (CYC1 promoters and terminators used throughout) (top). Fold change in GFP (green bars) and mCherry (red bars) expression for six targeted CR fusions (bottom). (B) Swapping operator locations results in inversion of transcriptional outputs. (C) Schematic of the same locus architecture as in (A) but containing two different promoters and terminators (BIO2 promoter and ADH1 terminator in purple). Error bars are standard deviations of three isogenic strains. See also Figure S5.

To test if we could shift the dynamic ranges of both reporter genes while qualitatively maintaining the same dual-gene regulatory profiles, we co-recruited VP16 to upstream positions in the dual-gene architecture (Figure S5, right columns). We were able to engineer increased dynamic ranges of both reporter genes while maintaining similar regulatory trends of the CRs. Since future applications may require the simultaneous regulation of two distinct promoters, we next added a second, different promoter to the reporter construct. Specifically, we replaced the downstream CYC1 promoter that drives the expression of mCherry with the full-length BIO2 promoter, which has a similar intermediate basal level of expression as the CYC1 promoter (Figures 5C, S5C, and S5D) (Blount et al., 2012). Overall, we found that the regulatory output profiles were consistent with those from the reporter harboring two repeated CYC1 promoters, suggesting conservation in these forms of regulation.

New parts for synthetic biology and cellular engineering

Synthetic biology offers a bottom-up approach for exploring the design and function of biological systems, and for engineering cells and organisms to address a range of biomedical and industrial applications (Fischbach et al., 2013; Khalil and Collins, 2010; Purnick and Weiss, 2009; Weber and Fussenegger, 2010; Ye et al., 2013). Here we decomposed chromatin-based transcriptional regulation into minimal components – minimal promoters and individually-targeted CRs – to provide a useful framework of parts and behaviors for broad applications in synthetic biology and cellular engineering.

Targeted CRs could be used as synthetic transcriptional activators and repressors in eukaryotic organisms. Many CRs matched or exceeded the activation or repression levels achieved by commonly used regulatory domains, such as VP16 and Mig1 (Figure 2A and 2B). While the behavior of any individual CR may vary for different genomic contexts, the general regulatory properties revealed by this library-based approach will streamline selection and testing of relevant CRs. Moreover, we observed strong correlation between the relative activities of CRs in alternative loci (Figures S1, S3, S5E-F, S7A, S7D and Tables S1 and S4), suggesting some conservation or robustness in the function of these factors across different genomic contexts.

This work also has interesting implications for the design of new synthetic gene circuits. First, our work demonstrates that chromatin-based components can vastly extend the regulatory potential of an individual genetic locus, thus expanding the regulatory possibilities of circuit nodes. A diverse range of transcriptional logic can be programmed by designing a genetic locus to recruit different combinations of CRs. As a result, circuits composed of CRs may represent a more efficient solution to information processing than those composed of canonical transcriptional components, like bacterial transcription factors. In other words, a minimal number of CRs targeted to a single locus may perform similar logical or computational tasks as a network composed of many interacting transcription factors. This feature could be useful for biotechnology applications by helping to reduce the size of gene expression cassettes to be delivered into a cellular host. Yields from bioprocesses could also be increased by the reduction in metabolic load on production organisms. Second, quantitative control of transcriptional outputs, including the ability to program synergistic activation, could be useful in controlling the expression levels of enzymes in engineered metabolic pathways and of regulatory proteins in synthetic circuits. Chromatin-based control schemes could be used to tune the sensitivity of cellular sensors to multiple environmental factors, or to tune the expression range of signaling factors such as chimeric antigen receptors in T-cell adoptive immunotherapy. The properties of synthetic circuits such as induction threshold, cycle period, and entrainment strength are known to be sensitive to expression levels (Atkinson et al., 2003), which could in principle be tuned through co-recruitment of synergistic CRs. Finally, epigenetic regulation of specific sets of genomic loci fundamentally underlies the transition between distinct cellular states, including in response to stress (Crews et al., 2012) or differentiation into cells of distinct tissue types (Meissner, 2010). The ability to establish epigenetic states at defined loci may enable construction of simplified synthetic systems to study the regulatory principles governing these processes.

Chromatin-based systems also enable multi-gene regulation, providing interesting new strategies for precisely addressing individual genes within a locus. For example, an asymmetric spatial regulator could be used to simultaneously repress one gene while activating another, a property that could serve as the foundation for new bistable genetic switches. Furthermore, some of the components presented here (CRs, nucleosome-disfavoring sequences, etc.) may be used to mitigate undesired context effects of placing genes and regulatory elements in proximity to one another. Finally, long-range CR repressors could be deployed to stably silence entire genomic regions, for example, to inactivate a synthetic circuit or to regulate an entire secondary metabolite production cassette. Quantitative measurement of properties, such as the kinetics of activation or repression, distance-dependence of spatial regulators, and spreading kinetics of long-range regulators, would greatly enhance the utility of CRs for these purposes.

Finally, recent advances in programmable DNA targeting technologies are providing new opportunities for inducing epigenomic alterations at any desired locus, for example, to correct disease-associated epigenomic changes. ZFs, transcription activator-like effector (TALE) repeat domains, and the recently described CRISPR/Cas system (Cong et al., 2013; Jinek et al., 2012) each provide unique benefits, and all are compatible with the approach outlined here. For example, ZFs are highly-specific, small, and efficient for gene/DNA delivery applications (Urnov et al., 2010). TALE proteins are easier to engineer, have a larger targeting range, and have been shown to enable the targeting of CR domains (Konermann et al., 2013; Mendenhall et al., 2013). Lastly, the CRISPR/Cas system can be used to promote multiplex recruitment of effectors to numerous loci simultaneously (Cong et al., 2013; Maeder et al., 2013b; Perez-Pinera et al., 2013).

Plasmid Construction

Reporter plasmids were constructed from integrative plasmid pRS406 (Stratagene) by cloning ZF (43-8 and/or 97-4) binding sequences at various locations within a previously described reporter construct (Khalil et al., 2012). ZF-CR and VP16 fusion proteins were expressed from previously described TetR- or LacI-regulated GAL1 promoters (Khalil et al., 2012). The ZF-CR expression constructs were cloned into single-integrating plasmid pNH603 (HIS3), and the VP16 fusion expression constructs into single-integrating plasmid pNH605 (LEU2).

Our host strain was generated by cloning an expression cassette that constitutively expresses TetR, LacI, and GEV into single-integrating plasmid pNH607 (HO). Constitutive expression of the repressors in glucose-containing media ensures low basal levels of expression of ZF-CRs from the engineered GAL1 promoters, which can be relieved by the respective addition of the chemical inputs, ATc and IPTG, along with beta-estradiol to the medium. The negative control, truncated (non-binding) ZF amino acid sequence is PRHLKTHLR. pNH603, pNH605, pNH607, and BFP were kind gifts from the Lim Lab (Zalatan et al., 2012).

Library Construction

Primer sequences were obtained from the Saccharomyces Genome Database (Cherry et al., 2012) (Table S6), synthesized (Integrated DNA Technologies), and used to amplify full length CR ORFs from wild-type yeast (BY4742). SbfI and NotI flanking restriction sites were used to ligate PCR products C-terminal to (3×FLAG) – (nuclear localization sequence) – (zinc finger array) – (17 amino acid glycine-serine linker).

Induction Experiments

Three single yeast colonies for each strain were picked after genomic integration and used to inoculate 500 μl of SD- media (synthetic drop-out media containing 2% glucose with defined amino acid mixtures) in Costar 96-well assay blocks (V-bottom; 2 ml max volume; Fisher Scientific). The cultures were grown at 30°C with 900 rpm shaking for 24-48 hr. Cultures, with and without inducers, were inoculated in SD-complete media to an OD600 of 0.05-0.1 and grown at 30°C with 900 rpm shaking for 12 hr. Cells were treated with 10 μg/mL cycloheximide to inhibit protein synthesis, and then assayed for yEGFP, mCherry, and BFP expression by flow cytometry.

We thank members of the Collins and Khalil labs for helpful discussions. Emily Egan and the Moazed Lab provided valuable technical advice on ChIP experiments. We thank Sylvain Meylan and Kwan T. Chow for engaging discussions and manuscript feedback. This work was supported by a NIH-NIGMS Ruth L. Kirschstein Postdoctoral Fellowship (1F32GM105272-01A1) (A.J.K.), a Boehringer Ingelheim Fonds PhD Fellowship (S.K.), a Defense Advanced Research Projects Agency grant (DARPA-BAA-11-23) (A.S.K. and J.J.C.), startup funds from the Department of Biomedical Engineering at Boston University (A.S.K.), a National Science Foundation CAREER Award (MCB-1350949) (A.S.K.), the Wyss Institute for Biologically Inspired Engineering (PRJ-0329) (J.J.C.), and the Howard Hughes Medical Institute (J.J.C.).