Tailor-Made Zinc-Finger Transcription Factors Activate FLO11 Gene Expression with Phenotypic Consequences in the Yeast Saccharomyces cerevisiae Authors: Jia-Ching Shieh, Yu-Che Cheng, Mao-Chang Su, Michael Moore, Yen Choo, Aaron Klug Published: 2007, PLOS ONE DOI: 10.1371/journal.pone.0000746 ABSTRACT Cys 2 His 2 zinc fingers are eukaryotic DNA-binding motifs, capable of distinguishing different DNA sequences, and are suitable for engineering artificial transcription factors. In this work, we used the budding yeast Saccharomyces cerevisiae to study the ability of tailor-made zinc finger proteins to activate the expression of the FLO11 gene, with phenotypic consequences. Two three-finger peptides were identified, recognizing sites from the 5′ UTR of the FLO11 gene with nanomolar DNA-binding affinity. The three-finger domains and their combined six-finger motif, recognizing an 18-bp site, were fused to the activation domain of VP16 or VP64. These transcription factor constructs retained their DNA-binding ability, with the six-finger ones being the highest in affinity. However, when expressed in haploid yeast cells, only one three-finger recombinant transcription factor was able to activate the expression of FLO11 efficiently. Unlike in the wild-type, cells with such transcriptional activation displayed invasive growth and biofilm formation, without any requirement for glucose depletion. The VP16 and VP64 domains appeared to act equally well in the activation of FLO11 expression, with comparable effects in phenotypic alteration. We conclude that the functional activity of tailor-made transcription factors in cells is not easily predicted by the in vitro DNA-binding activity. INTRODUCTION The Cys 2 His 2 zinc finger (ZF) domain, initially discovered in TFIIIA from Xenopus laevis [1] , [2] , is the most common DNA-binding (DB) motif in eukaryotes, and the second most frequently-encoded protein domain in the human proteome, being found in approximately 2% of all our proteins [3] – [5] . Individual fingers of about 30 amino acids can bind RNA and protein, but they generally function by binding DNA. Each ZF module holds both an anti-parallel β-sheet and a recognition α-helix that interacts directly with ∼3 bps of DNA in the major groove. This structurally-simple ββα domain is stabilized by hydrophobic interactions and the chelation of single zinc ion between a pair of cysteines from the β-sheet and a pair of histidines from α-helix [6] . Significantly, Cys 2 His 2 ZF domains are very suitable for the construction of artificial transcription factors (TFs) as they usually bind as covalent tandem repeats, allowing the recognition of extended asymmetrical DNA sequences. The modularity of the ZF proteins, both in structure and function, serves greater advantage in comparison with other types of DNA-binding domains (DBDs) that normally recognize DNA as dimers and use non-modular recognition domains [7] , [8] . The Cys 2 His 2 ZF of Zif268 has recently become known as a versatile scaffold for use in engineering novel DBDs with sequence-specificity [8] – [10] and has shown the ability to bind a variety of DNA sequences [11] – [13] . In addition, arrays of linked Zif268 finger-variants allow binding of longer asymmetric sequences [14] , [15] . To date, customized TFs based on Zif268 have been constructed by a number of groups and shown to effectively activate or repress individual target genes in diverse biological systems [8] , [10] , [16] . Recently, we have shown that inhibition of HSV and HIV replication can be achieved by repression of HSV and HIV genes in their host cells, with tailor-made ZF TFs [17] , [18] . The possibilities for therapeutic applications have been supported by reports that engineered ZF proteins can induce angiogenesis, suggesting a treatment for cardiovascular disease in a multi-cellular animal model [19] , [20] . Beyond therapeutic applications, ZF proteins have also been used in the fields of drug discovery [21] , stem cell research [22] , [23] , and gene targeting with sequence-specific integrases, nucleases, and methylases [17] , [24] – [28] . In addition to developing methods of selection and applications in therapeutics, we are particularly interested in extending the use of ZF-based TFs to functional genomics. To gain insights on how the regulation of the gene expression is achieved by engineered TFs, such that altered gene expression leads to specific phenotypes for selection, we have taken advantage of the well-studied yeast Saccharomyces cerevisiae . This is a convenient model to study the behavior of tailor-made TFs to regulate the expression of genes with phenotypic consequences. We have chosen the FLO11 (also known as MUC1 ) gene that is a downstream target on pathways of invasive growth in haploid yeast, because the pathways [29] – [31] and the gene itself are well characterized. FLO11 is a adhesin-encoding gene whose product is a large, cell wall-associated, glycosylphosphatidylinositol (GPI)-anchored threonine/serine-rich protein, with structural similarity to mammalian mucins and yeast flocculins [32] , [33] . FLO11 is required for either pseudohyphal formation in diploid yeast upon nitrogen depletion [34] or invasive growth and cell-cell adhesion in haploid yeast when depleted of carbon source [31] , [35] . The ability of haploid cells to invade and to initiate flocculation is achieved by overexpressing FLO11 without the need for carbon source depletion [35] . To date, many transcriptional activators of FLO11 have been shown to be needed for the induction of filamentous or invasive growth, through various signaling pathways, including Kss1 and cAMP-regulated as well as glucose repression cascades [30] , [36] . These parallel pathways converge on the FLO11 promoter, with at least four upstream activation sequences (UASs) and nine repression elements, which together span a minimum of 2.8 kb, in one of the largest promoters in S. cerevisiae genome [37] . Deletion of the repression region of the FLO11 promoter appears to increase FLO11 expression [38] . Intriguingly, FLO11 promoter-dependent epigenetic variegation of FLO11 expression has recently been revealed [39] . To add even more complexity, the presence of internal tandem repeats within the coding region of FLO11 have been shown to maximize its expression by the THO complex at the level of transcriptional elongation [40] . Notably, invasive growth in haploid yeast, induced by transcriptional activation of FLO11 , has been shown by overproduction of transcriptional regulators such as Msn1 and Mss11 [41] as well as Flo8 and Ste12 [37] . In this report, we show that three-finger (3F) peptides, selected by the bipartite approach of phage display [42] , recognize two 9-bp sites from the 5′ un-translated region (5′ UTR) of the FLO11 gene. The combined six-finger (6F) peptides (derived by linking the 3F peptides) recognize an 18-bp site with a 2-bp gap. Recombinant TFs (rTFs) were made by fusing these peptides to the activation domain (AD) VP16, from herpes simplex virus, or its derivative VP64. Despite finding several candidates that work in vitro, we found that only one peptide can competently activate the expression of FLO11 in haploid yeast cells. The activation of FLO11 expression is not dependent on carbon source depletion and can advance invasive growth of cells in semi-solid medium, biofilm formation of cells in liquid culture, and tighter adhesion of cells to polysterene. The inability of the rest of rTFs to efficiently activate FLO11 expression is presumably due to exclusion from the binding site by endogenous factors binding to the site for the rTFs. Our results suggest that artificial TFs based on ZFs can be used as a general gene-switch tool for modulating the expression of specific genes that render specific selectable phenotypic alterations. RESULTS To select phage clones expressing 3F peptides, a phage display approach was adopted. The Lib12 phage library was used directly to screen the 3F peptides against the target site TFLO11/F1 ( Figure 1B ). One phage clone from Lib12, binding to the 9 bp of TFLO11/F1, was thus obtained. Three paired targets, against 5 bp “half sites” from TFLO11/F2, TFLO11/R1, or TFLO11/R2 ( Figure 1B ), were selected in parallel using two complementary phage libraries, Lib12 and Lib23 (43). Thus, phage clones were obtained that bound to all the “half sites”: 12TFLO11/F2 and 23TFLO11/F2; 12TFLO11/R1 and 23TFLO11/R1; 12TFLO11/R2 and 23TFLO11/R2. All of the initially-selected clones were screened by phage ELISA to determine which ZFPs bound DNA with an affinity comparable to that of Zif268 to its cognate target. (A) 5′UTR sequences of FLO11 used in identification of sequence-specific 3-finger DNA-binding motifs. Nucleotides are numbered relative to the translational start site. The 9-bp core sequences of the target sites TFLO11/F1, TFLO11/F2, TFLO11/R1 and TFLO11/R2 are shown with arrows indicating the direction of the sequence from 5′ to 3′. Sequences of Ste12 ( PRE ) , and Tec1 ( TCS ) binding sites are shown above. (B) DNA oligos used in the phage selection, including 2-bp flanking sequences on each side. The shaded areas represent the core bound sequences. Individual fingers bind triplet nucleotides, indicated by F1, F2, and F3, respectively. For direct selection of TFLO11/F1 from Lib12, the oligo CTGCGGTATCTTC is shown shaded light grey. Targets of TFLO11/F2, TFLO11/R1, and TFLO11/R2 were subjected to bipartite selection: targets were divided into two “half sites” that comprise a gcgg anchor (in lower case) and a 5-bp target (shaded grey or black). The targets were selected from two complementary libraries, Lib12 and Lib23, and recombined to bind a full 9 bp sequence, made from the two grey and black half-targets. The sequences flanking the core target sequences are underlined. (C) The apparent K d s of the individual clones bound to the respective 12TFLO11/F2 and 23TFLO11/F2 “half sites”, as well as the individual clones of TFLO11/F1 or TFLO11/F2 “full sites”. (D) Corresponding amino acid residues of 3ZfFLO11/F1 and 3ZfFLO11/F2 and their proposed mode of contact with DNA targets. The amino acids are shown with circles numbered relative to their helical positions. Due to a cross-strand interaction of amino acid 2 of finger 1, each of the 3F domains potentially binds to an extra base, making a 10-bp DNA site. The linker peptide between the two 3F domains is Thr-Gly-Gly-GlyGly-Ser-Gly-Gly-Ser-Glu-Arg-Pro. The amino acid residues with proposed contacts to DNA bases are shown with arrows. The shaded dotted arrows indicate that the contact is either stereochemically unfavorable (in black), favorable (in white), or neutral (in gray) according to “recognition-code” rules [43] , [44] . To engineer clones capable of binding to their full-length 9 bp targets, those binding with paired overlapping 5 bp “half sites” were recombined according to the bipartite protocol, such that the recombinant clones would recognize the full-length targets (43). Four clones binding to the full-length target of TFLO11/F2 were identified after recombination from a mixture of phage clones recognising the respective 12TFLO11/F2 and 23TFLO11/F2 “half sites”. The apparent K d s of the individual clones are shown in Fig. 1C . The TFLO11/R1 and R2 clones gave no detectable binding after recombination, even though their “half site” modules bound DNA with an apparent affinity equal of better than Zif268 by phage ELISA (data not shown). Therefore only the four recombinant clones binding to TFLO11/F2 and the one clone binding to TFLO11/F1 were subjected to further analysis. To determine the binding affinity of each selected or recombinant clone, the apparent K d of the individual clones were assessed by phage ELISA. The recombinant clones, expressing 3F peptides, were tested for binding against serial dilutions of their target sites of either TFLO11/F1 or TFLO11/F2, and were found to display apparent K d comparable to that of Zif268 to its cognate site, which is in the expected range of 10 −9 to 10 −10 M ( Figure 1C ). These clones were then sequenced. The amino acid sequences, deduced from DNA sequences of the four clones 3ZfFLO11/F2, showed that they were identical. The amino acid sequence of the helical region from either 3ZfFLO11/F1 or 3ZfFLO11/F2 is shown in Fig. 1D . We note that the DNA-protein contacts do not entirely follow the principle of recognition-code stereochemistry [43] , [44] , even though they have survived a selection pressure for both affinity and specificity, by adding competitor DNA to the phage binding mixture (43). TAILOR-MADE RECOMBINANT TRANSCRIPTION FACTORS DERIVED FROM SELECTED ZINC FINGERS ARE CAPABLE OF BINDING TO DNA SITES IN VITRO To generate functional rTFs, the 3F peptides were fused to either VP16 or VP64 AD. These domains have been shown to be potent trancription activators, used in diverse biological systems including yeast S. cerevisiae [45] , [46] . Four constructs were established, based on pcDNA3.1(−), that allowed expressing the rTFs of VP16-3ZfFLO11/F1, VP64-3ZfFLO11/F1, VP16-3ZfFLO11/F2, and VP64-3ZfFLO11/F2 in an in vitro T7 expression system. To increase the affinity and stability of the protein-DNA complex and to try to improve the in vivo function of the rTFs, a 6F rTF was also generated [14] , [47] – [49] . The 6 fingers contained 3ZfFLO11/F2 in the N-terminus and 3ZfFLO11/F1 in the C-terminus, with a central linker containing TGGGGSGGSERP, able to span a gap of 2-bp between two 9-binding sites (3-bp gap, if the cross-strand position is counted as being in the gap; Moore, M., unpubl.; Figure 1D ). This was then subcloned into pVP16 and pVP64 to allow expressing VP16-6ZfFLO11/F2-F1 and VP64-6ZfFLO11/F2-F1 in vitro . To assess whether fusing zinc fingers with ADs had any deleterious effect on DNA-binding affinity, gel-shift assays were conducted. TFLO11/F1-F2 DNA was tested against individual rTFs, produced from pV16-rTFs and pVP64-rTFs, in a transcription-translation extract. The results showed no difference in binding affinity of rTFs with either VP16 or VP64 ( Figure 2 ), suggesting that the ADs in such constructs performed equally well, at least with respect to DNA-binding ability. Importantly, the binding affinity of 3ZfFLO11/F1 and 3ZfFLO11/F2 to the target sites appeared to be comparable to that of Zif268 to its cognate target site ( Figure 2 ). This is consistent with results in the phage-ELISA where only the DBDs of 3F peptides were expressed. These observations indicated that additional domains fused to the C-terminus of ZFs, including nuclear localization signal and AD, did not change the binding affinity of the proteins tested. In addition, the binding affinity of the rTF with a 6F DBD was approximately 10-fold stronger to those of the rTFs with 3Fs ( Figure 2 ). This is modest when compared with most of the previous observations that 6F proteins increase affinity for DNA from 20- to 6,000-fold, relative to 3F proteins [14] , [15] , [49] . Nonetheless, we decided to compare the relative functioning of these constructs in vivo . Gel-shift assays were carried out using the full target DNA site TFLO11/F2-F1, with rTFs produced from an in vitro transcription-translation system, using a 5-fold dilution series of protein. The binding of Zif268 with its cognate site (left) was used as a reference to determine the relative binding affinity of rTFs and TFLO11/F2-F1. To estimate Kd values, it was assumed that protein expression in each TNT reaction, and each subsequent dilution, was the same. Triangles indicate the concentration of rTFs where half the amount of target DNA bound with rTFs. THE INDUCTION OF FLO11 MRNA DEPENDS ON THE PRESENCE OF RECOMBINANT TRANSCRIPTION FACTORS IN S. CEREVISIAE To generate constructs capable of expressing functional rTFs in S. cerevisiae , the coding sequences of rTFs from pV16-rTFs and pVP64-rTFs were PCR amplified and cloned into pCM185 such that the expression of rTFs is tetracycline-repressible. The resulting six pCM185-rTF constructs allow expressing the rTFs of VP16-3ZfFLO11/F1, VP64-3ZfFLO11/F1, VP16-3ZfFLO11/F2, VP64-3ZfFLO11/F2, VP16-6ZfFLO11/F1-F2, and VP64-6ZfFLO11/F1-F2. These pCM185-rTFs constructs were introduced into wild type haploid yeast strain W303. To assess the expression of rTFs in yeast cells, the transformants with individual constructs were grown exponentially in selective medium, in non-induced conditions, followed by growing in either induced or non-induced conditions for various times. Samples were taken to measure the levels of rTFs expression by Western Blot analysis. The levels of rTFs began to accumulate after 3 hrs and reached a maximum within 12 hrs of induction (data not shown), consistent with the observed kinetics of induction in the system [50] . By contrast, no detectable rTFs were observed when cells were grown in non-induced conditions, as compared with those in induced conditions ( Figure 3A ). This regulatable mode of transcription control therefore provides a useful way to determine the dependency of gene expression by tailor-made rTFs targeting specific sequences of genes. (A) DNAs capable of encoding rTFs were constructed into pCM185, a tetracycline-repressible yeast vector. Cells expressing rTFs, as indicated, were grown either in the presence or absence of 1 µg/ml tetracycline for 12 hrs, after which the cells were collected for Western blotting to detect the expression of rTFs. (B) The same cells were grown in the absence of tetracycline (inducing conditions) and collected at the times indicated, for Northern blotting to detect FLO11. The empty plasmid pCM185, designated as pCont., was used as a negative control. Cells with pCM185 growing in the absence of tetracycline and 0.05% glucose, indicated as LGlu, instead of 2% glucose were used to examine the expression of FLO11 under derepressed conditions (C) The northern blots were quantitated using ImageQuant v1.2 and the ratios of induced FLO11, relative to ACT1 control RNA, are plotted as a histogram. To test the dependency of the FLO11 expression on rTFs, yeast cells carrying constructs of pCM185-rTFs were cultured exponentially in selective medium with glucose, in non-induced conditions. Samples were subsequently grown in induced or non-induced conditions for various times, and then the levels of FLO11 mRNA examined by Northern Blot analysis. As shown in Figure 3B and C , while the FLO11 mRNA only stayed at the basal level in cells carrying control pCM185, FLO11 mRNA accumulated in cells expressing rTFs. This occurred in a time-dependent manner with the maximal level after 24 hrs of induction. Interestingly, such induction was particularly apparent in cells expressing rTFs with 3ZfFLO11/F2, although the induction could also be seen in cells expressing rTFs with 6ZfFLO11/F2-F1. Unexpectedly, the level of FLO11 mRNA in cells expressing rTFs with 3ZfFLO11/F1 was similar to that in cells with the control, suggesting that this domain was not functioning in vivo . It seemed that rTFs with VP64 activate the expression of FLO11 better than those with VP16 ( Fig. 3B, C ). However, differences in activation between VP16 and VP64 were observed as being rather limited (data not shown). The potency of transcriptional activation between VP16 and VP64 is therefore by and large the same. These results indicated that the expression of FLO11 mRNA was mediated by the induction of rTFs targeting the 5′UTR of FLO11 gene. However, significant differences in transcriptional activation among rTFs were evident in that rTFs of 3ZfFLO11/F2 seemed to be the more potent transcriptional activators when compared with the rest of rTFs. Overall, this suggests that rTFs containing fingers targeting TFLO11/F1 were ineffective in activating the expression of FLO11 in vivo . THE PROMOTION OF INVASIVE GROWTH IN S. CEREVISIAE REQUIRES RECOMBINANT TRANSCRIPTION FACTOR-DEPENDENT EXPRESSION OF FLO11 MRNA Invasive growth in haploid S. cerevisisae , induced by transcriptional up-regulation of FLO11 , has been shown by growing in medium depleted with glucose [35] . To investigate the enhancement of invasive growth in haploid S. cerevisisae being mediated by rTF-dependent expression of FLO11 mRNA, haploid yeast cells carrying constructs of pCM185-rTFs were grown exponentially in selective medium with glucose, in either induced or non-induced conditions, and the improvement of invasiveness of cells was assessed by their ability to remain inside semi-solid YEPD agar, after being washed off from the surface of the plate. As shown in Figure 4 , in non-induced conditions, only a restricted amount of new colonies formed from the remaining cells after washing. Under induced conditions, cells with control plasmid pCM185 alone showed limited invasive growth, as indicated in Figure 4 . Similarly, cells expressing rTFs with either the 3ZfFLO11/F1 or 6ZfFLO11/F2-F1 DBDs exhibited only some degree of invasive growth, suggesting that those rTFs were not effective in up-regulating the expression of FLO11 gene for a clear phenotypic alteration. By contrast, cells expressing rTFs with 3ZfFLO11/F2 became far more invasive, to an extent even greater than the derepressed conditions of low glucose ( Figure 4 ). The three-finger 3ZfFLO11/F2, with either VP16 or VP64, was effective in triggering the expression of the FLO11 gene to induce the phenotypic change. Sequences encoding rTFs were subcloned into pCM185. Cells capable of expressing rTFs, as indicated, were grown in the presence of 1 µg/ml tetracycline, after which the cells were spotted onto semi-solid YEPD agar plates, using 10 µl of cells at an optical density at 600 nm (OD 600 ) of 0.01. Plates were grown in induced (−Tet) or a non-induced (+Tet) conditions for two days at 30°C to form colonies. Colonies were then rinsed off the surface of the agar with a gentle stream of sterile dH 2 O. Invasive growth was shown by cells that remained inside the agar and that were able to repopulate the plate. The empty plasmid pCM185 (pCont.) was used as a negative control. Cells with pCM185 were grown and spotted on to semi-solid YEP agar plates with either 2% glucose (Glu) or 0.05% glucose (LGlu) in the absence of tetracycline to examine invasiveness under derepressed conditions. THE ENHANCEMENT OF AGGREGATION OF S. CEREVISIAE CELLS REQUIRES RECOMBINANT TRANSCRIPTION FACTOR-DEPENDENT EXPRESSION OF FLO11 MRNA To investigate further the differences in potency among engineered TFs, via up-regulation of FLO11 , a quantitative assay was adopted to measure the aggregation of yeast cells for biofilm formation [51] , [52] . Haploid S. cerevisisae cells carrying constructs of pCM185-rTFs were grown exponentially in selective medium with glucose in non-induced conditions. These cells were then induced for 16 hrs prior to dispensing into wells of polysterene 96-microtiter plates, to allow cell attachment over a time-course. After staining with crystal violet for 30 mins, and after several washes, the plates were subjected to microscopic examination or determination of optical density at 570 nm. The cells with a better ability to form biofilms attach tighter to the wells of polystyrene plate, and thus stain with more crystal violet. The crystal violet on these cells, when eluted, displays a higher value of optical density. As shown in Figure 5 , cells with control pCM185 plasmid were weakest in terms of their ability to adhere to polystyrene as compared with those containing rTFs under induced conditions (in the presence of glucose as carbon source). It appeared that cells expressing rTFs with 3ZfFLO11/F1 displayed only an insignificant ability to adhere than those of the control (pCont., p>0.1), with a fold-of-induction no more than 1.06 ( Figure 5 ; Table S1 ; Table S2 ). Compared to those of the control (pCont.), cells with 6ZfFLO11/F2-F1 had only a fold-of-induction between 1.13 and 1.17 but enhanced adhesion with statistical significance (0.0051