• Open Access

Activation domains for controlling plant gene expression using designed transcription factors

Authors


Correspondence (fax 317 337 3228; email jfpetolino@dow.com)

Summary

Targeted gene regulation via designed transcription factors has great potential for precise phenotypic modification and acceleration of novel crop trait development. To this end, designed transcriptional activators have been constructed by fusing transcriptional activation domains to DNA-binding proteins. In this study, a transcriptional activator from the herpes simplex virus, VP16, was used to identify plant regulatory proteins. Transcriptional activation domains were identified from each protein and fused with zinc finger DNA-binding proteins (ZFPs) to generate designed transcriptional activators. In addition, specific sequences within each transcriptional activation domain were modified to mimic the VP16 contact motif that interacts directly with RNA polymerase II core transcription factors. To evaluate these designed transcriptional activators, test systems were built in yeast and tobacco comprising reporter genes driven by promoters containing ZFP-binding sites upstream of the transcriptional start site. In yeast, transcriptional domains from the plant proteins ERF2 and PTI4 activated MEL1 reporter gene expression to levels similar to VP16 and the modified sequences displayed even greater levels of activation. Following stable transformation of the tobacco reporter system with transcriptional activators derived from ERF2, GUS reporter gene transcript accumulation was equal to or greater than those derived from VP16. Moreover, a modified ERF2 domain displayed significantly enhanced transcriptional activation compared with VP16 and with the unmodified ERF2 sequence. These results demonstrate that plant sequences capable of facilitating transcriptional activation can be found and, when fused to DNA-binding proteins, can enhance gene expression.

Introduction

Differences in many key traits relevant to agricultural crop productivity, for example, fruit size in tomato (Cong et al., 2008), branching in maize (Doebley et al., 1997), seed dehiscence in rice (Li et al., 2006), etc., appear to be based on specific changes in transcriptional regulatory circuits that operate in the plant genome. As a consequence, the ability to manipulate such circuits represents an important direction of effort in crop improvement (Century et al., 2008; Petolino and Davies, 2013). A major technological platform to attain such endogenous gene control is represented by designed transcriptional regulators: fusions of (i) engineered DNA-binding domains designed to recognize specific cis-acting sequences in regulatory regions of target genes, (ii) functional domains that can impose specific transcriptional states, thereby driving the selective activation or repression of target genes.

For general utility, designed transcriptional regulators require, first, the ability to develop DNA-binding proteins with predetermined sequence specificity. The majority of effort in this regard has focused on the ubiquitous DNA recognition motif in metazoa, the Cys2His2 zinc finger (Miller et al., 1985; Pavletich and Pabo, 1991; Tupler et al., 2001). Work over the past two decades has shown that zinc finger proteins (ZFPs) can be engineered to bind to virtually any locus of one's choosing in vitro (Pabo et al., 2001). Moreover, of critical importance have been extensive data showing that fusions of such ZFPs to various functional domains have resulted in the desired outcomes in plants (Holmes-Davis et al., 2005). More recently, transcriptional activator-like (TAL) effector domains from plant pathogenic bacteria have also been used to target functional domains to predetermined genomic loci (Doyle et al., 2012). In a broad sense, several solutions are available to address the problem of obtaining an engineered DNA-binding domain against a given genomic region.

The second component of a designed transcriptional regulator technology is a series of domains that evoke, at the target gene of interest, the desired functional outcome. While the general principles of transcriptional regulation (i.e. an integration of inputs from the chromatin template along with locus-specific and general trans-acting factors) appear to be conserved between all eukarya (Ravasi et al., 2010), in contrast to protein-DNA recognition, which operates according to essentially universal principles, transcriptional control employs protein–protein interactions that are, inevitably, species-specific (Walhout, 2006). Most studies on designed transcriptional activators in plants have used the acidic activation domain, VP16, derived from a human pathogen, the herpes simplex virus (Guan et al., 2002; Gupta et al., 2012; Sanchez et al., 2006). In these studies, constructs comprising DNA-binding proteins, engineered against a region of interest and fused to the VP16 domain, were introduced into the plant genome via stable transformation, and the desired change in transcript levels of the target gene was observed. In this manner, APETALA3 (Guan et al., 2002) and 4-COUMARATE:COENZYME-A LIGASE (Sanchez et al., 2006) were activated by ZFP-VP16 fusions resulting in expected phenotypes, that is, altered floral morphology and modified lignin content, respectively, in Arabidopsis. In addition, oil content in leaves and seed of Brassica napus was modified following expression of a ZFP-VP16 targeted to β-KETOACYL-ACP-SYNTHASE II (Gupta et al., 2012).

This study was motivated by a desire to engineer transcriptional activation domains that would be superior to VP16 in two specific respects. First, as transcriptional activation appears to be a graded, as opposed to a binary, phenomenon (Mapp and Ansari, 2007), the question arises as to whether a plant-derived domain might, in fact, be more efficacious at interacting with plant transcriptional machinery, that is, activating plant gene transcription, compared with a viral domain. Second, the engineering of plant traits using designed transcriptional regulators involves stable introduction of an ectopic transgene into the plant genome and hence, the expression of a heterologous protein. Although the ZFP portion of a designed transcriptional regulator would not be recognized as non-native due to the wealth of autochthonous Cys2His2 DNA-binding domains in the genomes of all metazoa, the VP16 functional domain is highly immunologically xenogenic (Spatz et al., 2000). Thus, in addition to an entirely legitimate scientific need for superior activation domains, there is also the pragmatic consideration of identifying plant-derived functional domains for use in crop trait engineering (Rommens, 2004).

The present work represents an integration of a bioinformatics-guided approach to the identification of novel activation domains, a protein engineering effort driven by extensive information on protein–protein interactions during transcriptional activation, a novel high-throughput proxy screening system to identify optimal candidate activation domains and, finally, an in planta validation step. Several domains in the proteomes of crop species were identified with motif similarity to that of the VP16 acidic activation domain; structural information on activator–basal transcriptional machinery interactions was used to introduce specific, targeted changes into the plant domains thus identified; a budding yeast-based system was built and used for screening and ranking the resulting activation domains for efficacy; and finally, domains active in a plant system, tobacco, were identified. The effort has yielded a panel of plant-derived transcriptional activation domains bearing highly localized point mutations that yield a higher level of targeted gene activation in plants than the VP16 domain. The domains identified will represent a critical component of a platform to build designed transcriptional activators for plant trait engineering.

Results and discussion

Evolutionary conservation of an acidic transcription activation motif in the plant proteome

In facing the challenge of developing a superior successor to the acidic transcriptional activation domain of the herpes simplex virus VP16 protein, the features that make it so robust were considered. Discovered nearly 30 years ago (Campbell et al., 1984), this protein is known to have evolved for rapid, efficient transcriptional activation of early viral genes. Two foundational sets of studies that followed shortly after its identification established VP16 a source of a broadly applicable activation domain for both basic and applied research. An acidic stretch was identified in the COOH-terminus of the VP16 protein essential for its activity (Triezenberg et al., 1988). It was then shown that this domain, when fused to the DNA-binding domain of the yeast transcriptional regulator Gal4p, activates transcription of a reporter construct in hamster cells (Sadowski et al., 1988). The multi-heterology of the system (i.e. an activator from a virus that infects humans, fused to a DNA-binding domain from budding yeast, active in hamster cells) suggested that VP16 acts via a molecular circuit that is evolutionarily conserved.

Two decades of subsequent work have revealed the mechanism of action in detail. The activation domain of VP16 contains two subdomains, residues 410–452 and 453–490 (Figure 1), each with robust transcription activation potential. These domains interact with a cohort of intranuclear factors that function at several steps of gene activation, including chromatin remodelers (Tumbar et al., 1999), key components of the basal transcriptional machinery such as TFIID (Stringer et al., 1990) and TFIIH (Xiao et al., 1994) and the Mediator complex (Mittler et al., 2003). Both subdomains contain a combination of acidic and hydrophobic amino acid residues essential to their function (Cress and Triezenberg, 1991). The acidic residues are known to engage their targets in an initial electrostatic interaction, after which the domain becomes structured, that is, α-helical, establishing a more stable interface with its target proteins thereby activating transcription (Jonker et al., 2005). Of significant note, the human tumour suppressor protein, TP53, contains a ~75 amino acid transcriptional activation domain that relies on a similar mechanism (Fields and Jang, 1990), indicating that such a mechanism is not limited to virally-encoded genes.

Figure 1.

Sequence alignment of transactivation domains. Asterisks denote the presumed VP16 contact amino acids with TFIIH (Langlois et al., 2008). Putative contact motif is highlighted in grey. Sequence modifications to mimic VP16 are highlighted in red.

While the genomes of vascular plants encode neither VP16 nor any known analogue of the p53 protein, we hypothesized that the plant proteome may, in fact, contain trans-acting factors that bear short transcription activation motifs analogous to that found in VP16 and p53. The strategy pursued in this study involved first identifying genes for such factors in plants bioinformatically and then determining whether the larger domain within which they occur can directly, or after targeted engineering, be useful as a transcription activation domain for plant gene control.

An iterative, customized search algorithm of the plant proteome was performed (see 'Experimental procedures') using 79 amino acid residues of the VP16 transactivation domain as input to search for putative transcription activator domains. Following additional alignment checking, seven distinct genes were identified. They represented annotated regulatory genes the products of which are known to function in genome control (Table 1). Their open reading frame contained a specific, confined stretch of homology to the VP16 activation domain (Figure 1), in which acidic amino acid residues (aspartic acid and glutamic acid) were interspersed with aromatic residues (phenylalanine, tryptophan and tyrosine). Of note, the domains identified were enriched for proline, serine/threonine and/or glycine residues, pointing to a potential for remaining disordered in their apo state. Taken together, the presence of these domains, in the context of known regulators of genome function and the similarity at the amino acid level, pointed to these seven sequences as suitable candidates for evaluation of their transcriptional activation potential in a heterologous context, that is, as ZFP fusions.

Table 1. Seven plant proteins with sequences homologous to the VP16 transcriptional activation domain
Abr.Plant proteinAlignment scores with VP16References
CBF1 C-Repeat–DRE-Binding Factor—associated with dehydration stress response.85Achard et al. (2008)
DOF1 DNA Binding with One Finger—associated with light-regulated gene expression.97Yanagisawa (2001)
DREB1 Drought Responsive Element—associated with low temperature stress response.74Maruyama et al. (2004)
ERF1 Ethylene Response Factor 1—associated with stress response.90Fujimoto et al. (2000)
ERF2 Ethylene Response Factor 2—associated with stress response.84Nakano et al. (2006)
ORCA Octadecanoid-Derivitive Responsive Catharanthus AP2 Domain—associated with stress response.69Menke et al. (1999)
PTI4 Pseudomonas syringae cv Tomato Interacting—associated with disease response.59Gu et al. (2002)

Transcriptional activation domain engineering based on VP16/basal machinery interaction

While it was possible that one, or several, of the domains identified would fulfil the requirement of exhibiting enhanced transcriptional activation relative to the VP16 domain, before embarking on an extended effort in which these domains would be extracted from their native context, transferred to novel expression constructs to produce fusions with designed ZFP domains, then tested in various settings, we reasoned that subtle modifications could further enhance their potential utility. So in addition to simply using the VP16 transcriptional activator sequence to identify potential plant activator domains, we used the sequence as a basis for modifying plant domains for enhanced activity.

An examination of the detailed alignment between each individual plant domain and VP16 revealed an approach that ultimately proved successful: specifically, to rely on the manner in which a short motif within VP16, DDFEFEQMF, interacts with a key subunit of a major component of the basal transcriptional machinery, TFIIH. Studies have shown that this interaction requires (on the VP16 side of the complex) an α-helix harbouring the motif whereby aromatic side chains in the two phenylalanine residues interact with what is described as ‘two adjacent shallow pockets’ (Langlois et al., 2008) on the surface of TFIIH. Importantly, these same studies have shown that the p53 activation domain appears to use a similar structural motif (IExWF). Indeed, the design of an efficient artificial transcription activation domain based on the structural details of the VP16/p53 interaction with TFIIH was recently reported (Langlois et al., 2012).

In examining the precise alignment between the seven putative plant activation domains and the VP16 activation domain (Figure 1), we noted that minimal targeted alterations in each of the former could make them conform to a greater extent to the latter. Specifically, the full-length transactivation domain of VP16 (Residues 412–490) was aligned with the identified plant transcription activators. The interaction motif of the VP16 transactivation subdomain II was used to locate a putative interaction motif for each plant protein sequence (Figure 1). These were further modified to include proposed contact residues identified from the interaction motif of subdomain II of the VP16 transactivation domain (Langlois et al., 2008). Amino acids were introduced within the interaction motif of the putative plant transcription activators to produce variant sequences (Figure 1). Thus, for instance, the VP16-like motif in CBF1, a protein associated with plant stress response (Achard et al., 2008), is DEETMFGMP. One conservative amino acid substitution (aspartic acid to glutamic acid), the insertion of two aromatic amino acids and two additional substitutions—minimal changes by the standards of genetic variation introduced via crop breeding approaches (Huang and Han, 2012)—produce the motif DDFFETMFTD, which contains a perfect match to the VP16/TP53 activation domain consensus of FExMF (underlined). Similarly, in ERF2, the region from N53 to A85 aligned to subdomain II of VP16 transactivation domain and was thereby identified as a putative activation domain sequence of ERF2. Modifications were introduced into the short region from D66 to D76 that corresponded to the interaction motif VP16. Only the amino acid residues that were different from the identified direct contacts in the interaction motif VP16 were modified, while the remaining sequences were left unchanged. The resulting modified interaction motif of ERF2 (which included A67D, D71E, S73M and S74F) is similar to that of subdomain II of the VP16 transactivation domain (Figure 1). Modifications introduced into PTI4, ERF1, ORCA2, DREB1A and DOF1 interaction motifs followed a similar logic to those made for ERF2. At the conclusion of this effort, a panel of 14 domains to be evaluated for function in the context of an engineered transcription factor were obtained: seven obtained directly from plant open reading frames and seven derived from each one of these by site-directed modification (Figure 1). Other than the focused modifications introduced, the sequences of the modified allelic forms of the domains were otherwise identical to the native form.

Establishing a budding yeast-based system for transcription activation domain screening

Given the fact that entirely heterologous transcription activation domains, for example, derivatives of the mammalian glucocorticoid receptor (Schena and Yamamoto, 1988), can enhance transcription in yeast, we chose to validate the ability of our plant transcription activation domains to function in budding yeast. The aim was as the following: (i) to establish reporter system settings under which reproducible and rankable transcription activation is observed, (ii) to transfer the entire panel of activation domains to budding yeast expression vectors, (iii) to rank the plant-derived activators for activity as ZFP fusions, (iv) to transfer selected activator domains to plant expression vectors, (v) screen the selected activator domains as ZFP fusions in a plant expression system. At all stages of the process, the goal was to compare the activity of the plant-derived domains to that derived from VP16.

Two ZFPs, ZFP-G, a six-finger protein designed to bind to a GFP reporter gene sequence (Geurts et al., 2009) and ZFP-C, a four-finger protein designed to bind a human CCR5 gene sequence (Perez et al., 2008), were used in this study. Recognition sites for ZFP-G (19 bp) and ZFP-C (13 bp) were placed upstream of a minimal, TATA-containing promoter. Immediately, downstream of the ZFP-target/minimal promoter construct was placed the open reading frame for the α-galactosidase MEL1 gene encoding an enzyme that is efficiently secreted by budding yeast, and its levels can be easily, rapidly and accurately measured in liquid culture by a simple colorimetric ELISA assay. The overall outline of the screening procedure was as follows: (i) the reporter plasmid was transformed into budding yeast, (ii) 48 h later, the reporter strain was transformed with the expression construct encoding the ZFP-TF of interest, (iii) 48 h after that, synthesis of the ZFP-TF was induced by a change in growth medium, (iv) reporter construct activity was assayed at defined times (typically 4 h) following induction. In all cases, activation of ZFP-G or ZFP-C fusions was compared with the same activator fused to a ZFP-NT, that is, which recognizes a sequence that does not occur anywhere in the reporter construct.

Preliminary evaluations of the yeast system involved fusions between the two ZFPs and the VP16 activation domain. When the reporter construct was integrated into the URA3 locus, ZFP-TF synthesis-dependent increases in the levels of MEL1 activity were observed but were accompanied by significant signal variation between biological replicates, for example, assays performed on two separate clonal cultures of the reporter construct and high background MEL1 expression (data not shown). The background MEL1 expression was minimized by switching to a W303a strain of yeast. We hypothesized that the replicate variability of the reporter at the URA3 locus could be due to some interaction between the polymerase machinery transcribing the URA3 gene and the reporter construct. We thus re-engineered the plasmid harbouring the reporter construct to target it to the HO locus, which is transcriptionally silent in the W303a strain (thereby allowing for the generation of multiple biological replicate clones harbouring the reporter), transformed both ZFP-G-VP16 and ZFP-C-VP16 expression constructs into these clones and assayed MEL1 expression at defined time points following ZFP-TF induction. In these settings, minimal background levels of reporter construct expression were observed; robust activation in the presence of ZFP-TF was seen, and essential congruent data were obtained on multiple (n = 4–6) biological replicates (data not shown). Thus, reporter integration into the HO locus in the W303a strain was used for screening the candidate activation domains.

Screening plant transcription domains using a budding yeast system

The screening of plant-derived activation domains in the yeast system is presented in Figure 2a–b. Mock-transformed cells yielded no signal above background (data not shown). Cells transformed with all activation domains fused to ZFP-NT produced minimal levels of reporter gene activity (Figure 2a–b, grey bars). Both ZFP-C and ZFP-G, fused to the VP16 activation domain, drove robust activation of the reporter gene (Figure 2a–b). While three of the nonmodified plant-derived domains, ERF1, ORCA and CBF1, failed to activate expression above background, clear activation of the MEL1 reporter was observed with ZFP-C and ZFP-G fused to PTI4, DREB1a and ERF2 (Figure 2a–b). Of note, activation by ZFN-G fused to the unmodified ERF2 domain approached that driven by the same ZFP when fused to the VP16 domain (Figure 2b).

Figure 2.

Use of a budding yeast-based chromosomal reporter system to screen plant transcription activation domains. (a) The indicated domains (native—listed by name; modified—listed with a ‘m’ suffix), fused to ZFP-C (blue bars) or ZFP-NT (grey bars), were assayed for transcription activation potential of a reporter construct ('Experimental procedures' for details). MEL1 reporter gene activity was assayed as described (McCammon et al., 2011); the error bars represent the standard deviation in technical replicates. ZFP-C/VP16 activation level is depicted by dashed red line. No colonies were obtained for ZFP-C/DOF1 m or ZFP-NT/DOF1 m. (b) The identical set of domains as in panel a was assayed for transcription activation potential when fused to ZFP-G (green bars) or ZFP-NT (grey bars). Screening was performed exactly as in panel a. ZFP-G/VP16 activation level depicted by dashed red line. No colonies were obtained for ZFP-NT/DOF1 m. (c) ZFP-C/ERF2 m and ZFP-C/VP16 expression in yeast. Whole-cell extracts were analysed by Western blot using an antibody against the FLAG epitope (top panel) and GAPDH as a loading control (bottom panel). (d) ZFP-G/ERF2 m and VP16 expression in yeast. Western blot was performed as in panel (c). Bands pertaining to full-length ZFP-TFs denoted with asterisks.

With the exception of modified DOF1 domain that yielded no transgenic colonies, six plant-derived activation domains modified for a closer alignment with the VP16 contact motif were significantly more active in reporter gene activation than their parental nonmodified versions (Figure 2a–b). Significantly, this effect was observed when the modified domains were screened either in the context of ZFP-E or ZFP-G. In these settings, the modified activation domains ranked as follows: ERF2 > PTI4 CBF1 = DREB1 = ERF1 > ORCA2. Of significant note, the modified ERF2 domain activated transcription of the reporter nearly threefold more effectively than the VP16 domain, while the PTI4 domain did so approximately twofold more effectively. Western blot analysis confirmed that the full-length fusion protein was produced in the yeast cells (Figure 2c–d). In addition to the full-length fusion ZFP-C/ERF2 and ZFP-G/ERF2, several additional bands were observed presumably due to post-translational modification, for example, ubiquitination (Kornitzer et al., 1994). Taken together, these results provide a resounding validation of the protein engineering strategy that aimed to meet or exceed the activity driven by the VP16 domain. Based on these results, a decision was made to advance native and modified ERF2 and PTI4 for testing in a plant test system.

Establishing a plant-based reporter for transcription activation domain screening

A six-finger ZFP designed to bind a human Pedf gene sequence was used in this study (Yokoi et al., 2007). To test the activator function in plants, a reporter system was built involving a GUS gene driven by a constitutive actin-2 promoter into which eight tandem ZFP-binding sites were integrated 548–749 bp upstream of the transcriptional start site (Figure 3a). The reporter construct also included a PAT selectable marker gene for stable plant transformation. A total of 51 BASTA®-resistant plants were generated, of which 24 were found to be low complexity (1–2 copies of PAT) based on hydrolysis probe analysis of copy number. Of these low-complexity events, 18 displayed an intact GUS expression cassette, as determined by PCR analysis. Following Southern blot analysis, one of the single-copy, GUS-expressing events were used to generate homozygous T2 plants for testing ZFP activators.

Figure 3.

Activator testing in tobacco. (a) The GUS reporter construct contains 8 repeated six-finger ZFP recognition sites upstream of a constitutive promoter, AtAct2. (b) VP16 activation in tobacco. Mean (±SD) ratio of GUS transcript to that of an internal standard gene, BYEEF, in events expressing a ZFP-VP16-FL activator vs. a ZFP-only control. (c) ZFP activators in tobacco. Mean (±SD) ratio of GUS transcript to that of an internal standard gene, BYEEF, in events expressing ZFP activators vs. ZFP-VP16 and ZFP-only controls. Dashed line represents average ZFP-VP16 activation. N = number of independent events generated per construct. Column values with the same letter were not significantly different (P > 0.01) using a Welch's ANOVA.

To validate the tobacco system, transgenic events were generated using Agrobacterium strains harbouring either ZFP-VP16-FL or ZFP-only constructs and GUS transcript levels were determined in leaves of regenerated plants (Figure 3b). Averaged across all 24 transgenic events for each construct, ZFP-VP16-FL displayed ~3X higher GUS transcript accumulation compared with ZFP-only controls. The difference between the construct means was highly significant (P > 0.01) using a Welch's ANOVA. Moreover, individual events from ZFP-VP16 exhibited GUS transcript levels as high as ninefold above the ZFP-only average (data not shown). ZFP-VP16 transcript accumulation was observed in all events, although there appeared to be no significant correlation between ZFP-VP16 transcript and GUS activation (R2 = 0.33). This result is consistent with those reported previously (Van Eenennaam et al., 2004) and suggests that the activator-binding sites may become saturated such that elevated levels beyond a certain point may not enhance activation further. Nonetheless, these results show that GUS reporter gene transcription in tobacco can be activated following stable transformation with, and expression of, a ZFP activator.

Transcription activator domains that exhibit higher in planta efficacy than VP16

Figure 3c summarizes the resulting ratio of GUS transcript levels for the native and modified ERF2 and PTI4 activation domains, normalized by endogenous BYEEF gene expression levels, as compared to ZFP-VP16 and ZFP-only in the tobacco test system. Both native plant domains displayed GUS activation equal to or greater than ZFP-VP16. The modified version of ERF2, ZFP-ERF2 m, displayed significantly greater activation (P > 0.01 using Welch's ANOVA) than did the unmodified sequence, ZFP-ERF2 (Figure 3b). In fact, several individual events from ZFP-ERF2 m displayed GUS transcript levels greater than fourfold the highest individual ZFP-VP16 event (i.e. GUS/BYEEF transcript ratio of 11.3 vs. 2.6, respectively). The results observed with tobacco are consistent with those observed in yeast (Figure 2 vs. 3) and, taken together, clearly demonstrate that plant sequences capable of facilitating transcriptional activation can be found and, when fused to appropriate DNA-binding proteins, can enhance gene expression.

It is anticipated that additional modifications to enhance transcriptional activation of these domains may be possible. Recent studies with Arabidopsis ERF98, a transcriptional activator associated with defence signalling pathways, identified an acidic activation domain, EDLL, comprising glutamic/aspartic acid and leucine residues (Tiwari et al., 2012). Such domains may be used to make and test additional designs.

Experimental procedures

VP16 homology search and modification of interaction motifs

The transactivation domain of VP16 (residues 412–490) was used as the query sequence to BLAST the plant protein database in NCBI (http://blast.ncbi.nlm.nih.gov/Blast.cgi) resulting in the identification of seven plant proteins: CBF1 (GenBank Accession No. NP_567721.1), DOF1 (GenBank Accession No. NP_001105709.1), DREB1A (GenBank Accession No. NP_567720.1), ERF1 (GenBank Accession No. NP_567530.4), ERF2 (GenBank Accession No. NP_199533.1), ORCA2 (GenBank Accession No. CAB93940.1) and PTI4 (GenBank Accession No. ACF57857.1)—all of which contain an acidic domain similar to the transactivation domain of VP16 and an ‘alignment score’ above 50 (Table 1). Using the VP16 sequence, putative transactivation domains and contact motifs were located for each protein (Figure 1). New variants of the presumed plant protein domains that contained the amino acid contact residues of the contact motif of subdomain II of the VP16 transactivation domain (Langlois et al., 2008) were generated (Figure 1) as described in the Results and Discussion section.

Testing transcriptional activation in yeast

Reporter strain

A yeast strain for testing transactivation was produced based on a previously reported MEL1 reporter gene (Doyon et al., 2008) by first amplifying two separate fragments from pNorMEL1: a yeast KanMX expression cassette and a fragment with outward-facing homology arms to the yeast HO locus (Voth et al., 2001) separated by a bacterial origin of replication. These two fragments were digested with EcoRI/SpeI and ligated to generate a KanMX-selectable HO-targeting vector. The MEL1 expression cassette from pMELα2 (Melcher et al., 2000) was amplified and cloned into the KpmI site of the KanMX-selectable HO-targeting vector. ZFP-binding sites were synthesized de novo (DNA2.0, Menlo Park, CA). These sites, comprised of sequences from either the GFP (Geurts et al., 2009) or the human CCR5 gene (Perez et al., 2008) for ZFP binding, were cloned into the BamHI/PacI site of the KanMX-selectable HO-targeting vector upstream of the MEL1 reporter gene to create pHO-zBG-MEL1 which could be linearized with NotI to expose the flanking homology arms for targeting to the yeast HO locus and transformed into Saccharomyces cerevisiae strain BY4741 MATα (Invitrogen, Carlsbad, CA) using the manufacturer's suggested protocol. Briefly, 3 mL of a log phase culture was pelleted and washed in TEL buffer (10 mm Tris–HCl pH 8.0, 1 mm EDTA and 100 mm lithium acetate). The yeast cell pellet was resuspended in 360 μL of ‘yeast transformation solution’ consisting of 33% PEG-3350 (Sigma-Aldrich, St. Louis, MO), 0.1 m lithium acetate and 0.2 mg/mL salmon sperm DNA (Stratagene, La Jolla, CA) in 1X TE along with 3 μg of linearized pHO-zBG-MEL1 and then heat-shocked at 42 °C for 40 min. Yeast cells were pelleted and washed prior to selection on YPD plates containing 1 mg/L Genticin® (Life Technologies, Carlsbad, CA). Resistant clones were restreaked on YPD + Genticin® and used for subsequent transformations. The MEL1 reporter constructs are shown in Figure 2a and b.

Transactivation constructs

Expression constructs containing transcriptional activators were generated by mobilizing transactivation domains (Figure 1) as BamI/HindIII restriction enzyme fragments and cloning them directly downstream of sequences encoding ZFP-G (Geurts et al., 2009), ZFP-C (Perez et al., 2008) or ZFP-NT (i.e. ‘no target’—a ZFP with a recognition sequence which does not occur anywhere in the reporter construct). The resulting ZFP transcriptional activator expression cassettes utilized a GAL1,10 promoter (West et al., 1987) and a CYC1 terminator (Osborne and Guarente, 1988) and were based on the yeast pRS315 series vector. The resulting vectors contained transactivation domains as in-frame fusions with the ZFP-G, ZFP-C or ZFP-NT.

Transcriptional activation in yeast

Overnight cultures were grown in YPD + Genticin®, and 1 μg of vector containing a ZFP transcriptional activator expression cassette was delivered using a standard yeast transformation protocol in a 96-well format. All transformations were run in duplicate. Transformed yeast cells were recovered in ‘synthetic dextrose medium’ lacking leucine (SD—leu) to select for the vector containing the ZFP transcriptional activator expression cassette. After 72 h, the yeast cells were enriched by a 1 : 10 dilution of the transformants in ‘synthetic dextrose medium’ lacking leucine and grown for an additional 24 h. The yeast cells were then diluted 1 : 10 into ‘synthetic raphinose medium’ lacking leucine to de-repress the GAL1,10 promoter. After 24 h, yeast cells were pelleted and resuspended in ‘synthetic galactose medium’ lacking leucine. At time points of 0, 3 and 6 h postgalactose induction, 110 μL of yeast cells were harvested for MEL1 activity analysis. MEL1 activity was assayed by first diluting 100 μL of yeast cells with 100 μL of water and obtaining an OD600 using a spectrophotometer. The remaining 10 μL of yeast cells were incubated in 90 μL of MEL1 buffer consisting of 77 mm Na2HPO4, 61 mm citric acid and 2 mg/mL PNPG (Sigma-Aldrich) for 1 h at 30 °C. The reaction was stopped by the addition of 100 μL 1 m Na2CO3. MEL1 activity was assessed at OD405, and mU was calculated using the ratio of the OD405 and OD600 measurements (Doyon et al., 2008). Methods used for Western blot analysis are described in the Supplemental Material.

Testing transcriptional activation in tobacco

Reporter construct

A DNA sequence comprising eight tandem repeats of a six-finger ZFP-binding site (TGTGGTGGGAGAGGAGGGTGG) was synthesized de novo (IDT, Coralsville, IA) with SacII sites added to the 5′ and 3′ ends to facilitate cloning. The entire 8X-ZFP-binding domain was subsequently cloned into a pre-existing Gateway® entry vector containing plant expression elements. The 8X-ZFP-binding domain was mobilized on a SacII fragment and cloned immediately upstream of the Arabidopsis thaliana actin-2 promoter (An et al., 1996) driving a β-glucuronidase (GUS) coding sequence (Jefferson et al., 1987) using unique NcoI/SacI sites with the ATG codon of the NcoI site forming the initiation codon. An Agrobacterium tumefaciens open reading frame-23 3′ untranslated region was used to terminate transcription (Gelvin, 1987). The final transformation vector, p8X-ZFP-Actin-GUS, was the result of a Gateway® ligation with a destination vector containing an A. thaliana ubiquitin-10 promoter (Callis et al., 1995), phosphinothricin acetyl transferase (PAT) coding region (Wohlleben et al., 1988) and an A. tumefaciens open reading frame-1 3′ untranslated region (Huang et al., 1990). p8X-ZFP-Actin-GUS was confirmed via sequencing and transformed into A. tumefaciens strain LBA4404 (Invitrogen, Carlsbad, CA). The GUS reporter construct, p8X-ZFP-Actin-GUS, is shown in Figure 3a.

Production of transgenic reporter events

Leaf discs (1 cm2) cut from Petit Havana tobacco plants were incubated in an overnight culture of A. tumefaciens strain LBA4404 harbouring plasmid p8X-ZFP-Actin-GUS grown to OD600 ~1.2, blotted dry on sterile filter paper and then placed onto MS medium (Phytotechnology Labs, Shawnee Mission, KS) and 30 g/L sucrose with the addition of 1 mg/L indoleacetic acid and 1 mg/L benzylaminopurine in 60 × 20 mm Petri dishes (5 discs per dish) sealed with Nescofilm® (Karlan Research Products Corporation, Cottonwood, AZ). Following 48 h of co-cultivation, leaf discs were transferred to the same medium with 250 mg/L cephotaxime and 5 mg/L BASTA® (Bayer Crop Sciences, Kansas City, MO). After 3–4 weeks, plantlets were transferred to MS medium with 250 mg/L cephotaxime and 10 mg/L BASTA® in PhytaTrays™ (Sigma, St. Louis, MO) for an additional 2–3 weeks prior to leaf sampling and molecular analysis. Green plants displaying shoot elongation and root growth on medium with 10 mg/L BASTA® were then sampled for GUS reporter gene expression and molecular analysis. For GUS reporter gene expression analysis, leaf discs (~0.25 cm2) were cut and placed into a 24-well tray (1 leaf disc per well) containing 250 μL of GUS assay solution (Jefferson et al., 1987). The 24-well dish was wrapped with Nescofilm® (Fisher Scientific, Pittsburgh, PA) and incubated at 37 °C for 24 h. After 24 h, the GUS assay solution was removed from each well and replaced with 250 μL of 100% ethanol. The dish was wrapped with Nescofilm® and incubated at room temperature for 2–3 h. The ethanol was removed and replaced with fresh ethanol. The leaf discs were then viewed under a dissecting microscope. Leaf discs which were stained blue were scored positive for GUS expression. Sampling for molecular analysis involved cutting leaf tissue with a sterile scalpel and placing either 1–2 cm2 into 1.2 mL cluster tubes (Fisher Scientific, Nazareth, PA) for PCR analysis or 3–4 cm2 into 2.0 mL Safe Lock tubes (Eppendorf, Hauppauge, NY) for Southern blot analysis surrounded by dry ice for rapid freezing. The tubes were then covered in 3M micropore tape (Fisher Scientific) and lyophilized for 48 h in a Virtual XL-70 (VirTis, Gardiner, NY). Once the tissue was lyophilized, the tubes were capped and stored at 8 °C until analysis. Following molecular analysis (see below), selected BASTA-resistant plants were grown to maturity in the greenhouse and allowed to self-pollinate. T1 seed was collected, surface-sterilized for 3 min in 20% bleach followed by two sterile water rinses and germinated on MS medium (Phytotechnology Labs) with 30 g/L sucrose in PhytaTrays™ (Sigma). Following zygosity screening via pat gene copy number analysis (see below), homozygous T1 plants were grown to maturity in the greenhouse and allowed to self-pollinate. T2 seed was then collected, surface-sterilized and germinated as previously described and used to generate reporter plants for plant transactivation testing. Methods used for the molecular analysis of reporter events are described in the Supplemental Material.

Transcriptional activation in tobacco

Transactivation domains (Figure 1) flanked by restriction enzyme sites as BamHI/SacI were synthesized de novo (DNA2.0, Menlo Park, CA) and cloned immediately downstream of a six-finger ZFP (Yokoi et al., 2007) using unique BamHI/SacI sites found in an existing Gateway® Entry backbone vector (Invitrogen) whereby the ZFP activator was under the control of the constitutive cassava vein mosaic virus promoter (Verdaguer et al., 1996) and terminated with the Atumefaciens open reading frame-23 3′ untranslated region (Gelvin, 1987). A control vector, pZF only, which contained the ZFP and no activator domain was included. Additionally, for early assay development, a positive control vector, pZFP-VP16-FL was constructed which placed the VP16 residues 412–490 under the control of a strong constitutive promoter derived from the mannopine/octopine synthase genes (Ni et al., 1995). Transformation vectors resulted from a Gateway®-mediated ligation (Invitrogen) with a destination vector containing an A. thaliana ubiquitin-3 promoter (Callis et al., 1995), hygromycin phosphotransferase selectable marker (Waldron et al., 1985) and A. tumefaciens open reading frame-24 untranslated region (Gelvin, 1987). The final binary vector was confirmed via DNA sequencing and transformed into A. tumefaciens strain LBA4404 (Invitrogen).

To produce transgenic plant events expressing the ZFP activator constructs, leaf discs cut from T2 reporter events were transformed as previously described except that following Agrobacterium co-cultivation, and tissue was transferred to MS medium (Phytotechnology Labs) and 30 g/L sucrose with the addition of 1 mg/L indoleacetic acid, 1 mg/L benzyaminopurine, 250 mg/L cephotaxime and 10 mg/L hygromycin. Multiple independent transgenic events were generated for each of the ZFP activator constructs and the ZFP-only control and compared using Welch's analysis of variance (ANOVA) using JMP software (SAS Institute, Cary, NC).

Transcript analysis in tobacco

Expanding leaf tissue was sampled from hygromycin-resistant plants and flash frozen on dry ice in 96-well collection plates. RNA was isolated using an RNEasy® 96-well extraction kit (Qiagen, Valencia, CA) according to the manufacturer's instructions, and a Model 2-96A Kleco™ tissue pulverizer (Garcia Manufacturing, Visalia, CA) was used for tissue disruption. Resulting mRNA was quantified using a NanoDrop™ 8000 spectrophotometer and software (Thermo Scientific, Wilmington, DE) using a standard RNA measurement method. cDNA was prepared from diluted mRNA using the Quantitect® RT kit (Qiagen) following manufacturer's instructions whereby 1 μg total mRNA was used in each reaction. Resulting cDNA was stored at −20 °C until analysis could be performed. GUS and ZFP activator gene transcript analyses each used two DNA hydrolysis probe assays, both of which were analogous to TaqMan® assays (Applied Biosystems, Carlsbad, CA). Steady-state levels of GUS and ZFP activator mRNA for each individual transgenic event were estimated using sequence-specific primers and probes. The mRNA was normalized using the steady-state level of mRNA for an endogenous tobacco reference gene BYEEF (GenBank Accession No. GI927382). BYEEF is a constitutively-expressed ‘housekeeping’ gene that is commonly used for normalization in transgene expression studies (Nicot et al., 2005). Assays for all three genes were designed and evaluated using Light Cycler® Probe Design Software 2.0 and the real-time PCR 480 system (Roche Applied Science, Indianapolis, IN), respectively. For GUS and ZF activator amplification, Light Cycler® 480 Probe Master Mix was prepared at 1X final concentration in a 10 μL volume multiplex reaction containing 0.4 μm of each primer and 0.2 μm probe. A two-step amplification reaction was performed with an extension at 56 °C for 40 s with fluorescence acquisition. All samples were run undiluted in triplicate, and the average Ct values were used for analysis of each sample. For BYEEF amplification, Light Cycler® 480 Probe Master Mix was prepared at 1X final concentration in a 10-μL volume multiplex reaction containing 0.25 μm of each primer and 0.1 μm probe. A two-step amplification reaction was performed with an extension at 56 °C for 25 s with fluorescence acquisition. All samples were diluted 1 : 10 in triplicate, and the average Ct values were used for analysis of each sample. Analysis of real-time PCR data was performed using Light Cycler® software (Roche Applied Science, Indianapolis, IN) via the relative quant module and is based on the ΔΔCt method.

Ancillary