Synthetic gene expression perturbation systems with rapid, tunable, single-gene specificity in yeast

Abstract

A general method for the dynamic control of single gene expression in eukaryotes, with no off-target effects, is a long-sought tool for molecular and systems biologists. We engineered two artificial transcription factors (ATFs) that contain Cys(2)His(2) zinc-finger DNA-binding domains of either the mouse transcription factor Zif268 (9 bp of specificity) or a rationally designed array of four zinc fingers (12 bp of specificity). These domains were expressed as fusions to the human estrogen receptor and VP16 activation domain. The ATFs can rapidly induce a single gene driven by a synthetic promoter in response to introduction of an otherwise inert hormone with no detectable off-target effects. In the absence of inducer, the synthetic promoter is inactive and the regulated gene product is not detected. Following addition of inducer, transcripts are induced >50-fold within 15 min. We present a quantitative characterization of these ATFs and provide constructs for making their implementation straightforward. These new tools allow for the elucidation of regulatory network elements dynamically, which we demonstrate with a major metabolic regulator, Gcn4p.

Figure 1.

Schematic of hormone-based gene expression system. ATFs contain a DNA-binding zinc-finger array, the ligand binding domain of the human estrogen receptor and the VP16 activation domain. In the presence of β-estradiol (1), ATFs dissociate from Hsp90 (2), translocate to the nucleus (3) and activate transcription of a gene of interest (GOI) (4). Once produced (5) the gene products can be detected using a variety of methods.

Figure 2.

(A) The structure of the ATFs Z₄EV, GEV and Z₃EV. (B) The DNA-binding motifs of Gal4p, Zif268 and the Z₄ array.

Figure 3.

Dose response curves. Strains yMN5 (GEV), yMN7 (Z₃EV) and yMN14 (Z₄EV) were grown to early log-phase and then different amounts of β-estradiol were added to the growth medium. GFP was quantified following 12 h of induction by flow cytometry. Error bars represent ±1 SD from three independent cultures.

Figure 4.

(A) Hierarchically clustered gene expression responses of continuous cultures containing GEV, Z₃EV or Z₄EV upon addition of 1 µM β-estradiol (data are time-zero normalized in each experiment). Time points in each experiment are 0, 0.5 and 3 h. Each culture was maintained at a doubling time of 4.07 h. (B) Induction of GFP reporters in cultures from (A) measured by qRT–PCR. Error bars represent ±1 SEM of three technical replicates. Strains are yMN5 (GEV), yMN7 (Z₃EV) and yMN14 (Z₄EV).

Figure 5.

Non-yeast DNA-binding domains eliminate growth defect conferred by GEV induction. (A) Growth curves of GEV-, Z₃EV- or Z₄EV-containing strains in the presence of different amounts of β-estradiol. (B) The level of growth reduction conferred by ATFs.

Figure 6.

Induction of GFP reporters driven by promoters of different affinity by Z₃EV following addition of 1 µM β-estradiol. Error bars represent ±1 SD of three independent cultures.

Figure 7.

(A) Schematic of strains used for inducing genomic GCN4 allele. Z₄EV is driven by the ACT1 promoter. In the presence of β-estradiol, Z₄EV activates transcription of the native GCN4 allele driven by the synthetic promoter Z₄EVpr. KanMX is expressed in the reverse orientation. (B) Measuring growth Z₄EV-containing haploids containing GCN4 or Z₄EVpr-GCN4 in the presence or absence of 5 nM β-estradiol. Cells were grown in low phosphate chemostat medium. (C) The transcriptional response of Z₄EVpr-GCN4 (strain = DBY12423) in response to 1 µM β-estradiol in a continuous controlled maintained under phosphate limitation. The culture was maintained at a doubling time of 6.3 h. (D) The global transcriptional response of cells in (C) measured out to 2 h. We denote the presence of Gcn4p binding to a promoter by a blue line [(38) P-value <0.001]. (E) Expression data from (D) were divided into five expression clusters using K-means clustering with the Euclidean distance similarity metric. The optimal number of clusters over-represented and under-represented motifs within these clusters was determined using the FIRE algorithm (F).