Short activation domains control chromatin association of transcription factors
- Department of Molecular and Cell Biology, University of California, United States
- Thomas C. Jenkins Department of Biophysics, Johns Hopkins University, United States
- Biohub, United States
- Center for Computational Biology, University of California, United States
Figures
Figure 1
A synthetic transcription factor (TF) for quantifying chromatin association and reporter activity.
(A) Schematic of the synthetic TF used in this study: activation domains (39–60 amino acids) are fused to HaloTag, an estrogen-binding domain, six synthetic C2H2 zinc fingers (ZF1), and a V5 epitope tag. Black line indicates linkers between domains. (B) AlphaFold2 model of the synthetic TF. (C) Schematic of the synthetic reporter locus (not to scale) installed at AAVS1. Four zinc-finger binding sites are engineered upstream of a minimal promoter, minCMV. These elements are upstream of GFP, which is produced in response to estradiol induction. (D) Overlay of GFP and phase-contrast images of reporter-bearing cells expressing VP16-synthetic TF at 0 hr (left) and 72 hr (right) after estradiol induction. Quantification of the percentage of GFP-positive cells over time from live imaging is shown in the right plot. (E) Left: Trajectories extracted from single-molecule tracking (SMT) movies from cells expressing two example proteins. Thousands of these trajectories are pooled to generate inferred diffusion spectra, right. Lower panel indicates the cumulative distribution function of the diffusion spectra. We consider the shaded gray region (D<0.1 µm2/s) the chromatin-associated fraction of a protein.
Figure 2 with 4 supplements
Short activation domains are sufficient to control the fraction of transcription factor (TF) molecules bound to chromatin.
(A) Diffusion spectra (top) and cumulative distributions (bottom) of synthetic TFs bearing various activation domains. Numbers indicate the fraction of molecules with diffusion coefficients below 0.1 µm2/s (shaded region), which we infer to be chromatin-bound. (B) Flow cytometry measurements of cells bearing various activation domains after 24 hr of estradiol treatment. Distributions are shown in the top panel, where the dashed line is an arbitrary gate for GFP-positive cells. (C) Bar plot for mean fluorescence intensity (MFI), the geometric mean of GFP values for cells in the positive gate. (D) For each synthetic TF, the MFI of GFP+ cells is plotted against its bound fraction. Error bars indicate bootstrapping standard deviations of the bound fraction (Materials and methods). (E) HIF1α and CITED2 fluorescence recovery after photobleaching (FRAP) recoveries over time plotted as rolling means (solid lines). Curve fits to a double-exponential equation (Materials and methods) are shown in dashed lines.
Figure 2—figure supplement 1
Jump length histograms and Spot-On fits for two synthetic transcription factors (TFs).
(A) Jump-length histograms of CITED2-synthetic TF and HIF1α-synthetic TF over increasing frame intervals. (B) Graphical representation of three-state Spot-On fits to the jump-length histograms shown in (A). Height of the bars indicates population fraction, and position along the x-axis denotes the diffusion coefficient of the fit. Both fractions bound and diffusion coefficients are consistent with the analyses shown in the main text.
Figure 2—figure supplement 2
Analysis of a single-molecule tracking (SMT) dataset for sources of variance.
(A) Key statistics from one example SMT dataset for VP16. Each cell line was imaged on 2 days, and the data were combined for analyses. (B) Cell-wise mean posteriors from state array inference (Materials and methods, saspt). Each row represents the diffusion spectrum from one cell as a normalized heatmap. (C) Jump resampling experiment (Materials and methods), sampling 1–100,000 jumps from the whole dataset over 1000 bootstrapping replicates. Orange lines: medians; boxes: Q1-Q3; whiskers: Q1–1.5 IQR and Q3+1.5 IQR. (D) Variance of jumps from jump resampling experiment. Each point represents the variance of the 1000 bootstrapping replicates, sampled in three different ways (Materials and methods). (E) Minimum values from (D), summarizing sources of variance in this dataset. Cell-to-cell variability is about two orders of magnitude larger than day-to-day variability in this dataset.
Figure 2—figure supplement 3
Diffusion spectra of activation domain-synthetic transcription factor (TF) constructs in WT U2OS cells.
WT U2OS cells without an integrated reporter locus were electroporated (Materials and methods) with plasmids carrying different activation domain-synthetic TF constructs and imaged. Diffusion spectra are shown as in Figure 2.
Figure 2—figure supplement 4
Fluorescence recovery after photobleaching (FRAP) curves for histone H2B, HIF1α-synthetic transcription factor (TF), and CITED2-synthetic TF.
(A) Aggregate FRAP recovery curves showing a rolling-mean average (black line) and fits to a single-exponential line in blue and double-exponential line in red (Materials and methods) for cells expressing H2B-HaloTag. (B) As in (A) for HIF1α-WT. (C) As in (A) for CITED2-WT. Single-exponential fits to synthetic TF recoveries were poor. (D) Normalized radial bleach profiles (Materials and methods) for cells expressing an H2B control. For the five frames immediately after the bleaching period, a radial profile centered at the bleach spot center is plotted and normalized between 0 and 1. The pixel size is 105.46875 nm. There are no obvious bleach spot shape changes as fluorescence recovers during photobleaching. (E) As in (D) for cells expressing HIF1α-WT synthetic TF. (F) As in (D) for cells expressing CITED2 AD synthetic TF.
Figure 3 with 5 supplements
Mutations in individual activation domains modulate the fraction of molecules bound to chromatin.
(A) The amino acid sequence of CITED2 and its mutant constructs is shown above their diffusion spectra and the cumulative distributions. Numbers indicate the fractions bound, the probability density in the shaded region of the diffusion spectra. (B) As in A for VP16. (C) As in A for HIF1α. (D) Summary of activation domain activity and chromatin-bound fractions. Mean fluorescence intensity of GFP+ cells plotted against bound fractions. Error bars indicate standard deviations of a cell-wise bootstrapping scheme (Materials and methods, Figure 3—figure supplement 2).
Figure 3—figure supplement 1
An additional allelic series with cholesterol transcription factor (TF) SREBP’s activation domain.
(A) Top: Sequence of the WT SREBP activation domain appended to the synthetic TF, and alanine mutations for three mutants of the SREBP activation domain. Bottom: Diffusion spectra of cells expressing the labeled SREBP activation domain-synthetic TF constructs. (B) Flow cytometry distributions of GFP reporter expression in cells expressing various SREBP activation domains after 24 hr of estradiol induction. (C) Mean (geometric) fluorescence intensity of GFP-positive cells versus fraction bound for the SREBP constructs. Horizontal error bars denote standard deviation of bootstrapping replicates. Fraction bound correlates with mean fluorescence intensity in these four alleles.
Figure 3—figure supplement 2
Example outputs from a cell-wise bootstrapping scheme to estimate error in single-molecule tracking (SMT) diffusion spectra.
Left panels: Plots showing all 96 bootstrapping replicate spectra (Materials and methods) as individual lines. Right panels: Plots showing the mean of the bootstrapping replicates (solid lines), their standard deviations (dark gray shading), and their 95% confidence intervals (light gray shading). Numbers are the bootstrapping bound-fraction means and their standard deviations. (A) CITED2-WT; (B) CITED2-motif mut; (C) CITED2-L>F; (D) CITED2-K>A.
Figure 3—figure supplement 3
Cell-wise relationships between chromatin-bound fraction, transcription factor (TF) expression, and reporter output.
(A–D) Scatterplots of fraction bound measured by single-molecule tracking (SMT) versus TF abundance, mean intensity in the HaloTag-ligand dye channel. Each point represents data from a single cell and is colored according to the legends. There is no obvious relationship between TF abundance and fraction bound on a cell or population level. (E–H) Scatterplots of mean intensity in the GFP channel versus fraction bound. Each point represents data from a single cell plotted on the same axes and colored according to the legends in the upper right. Cell lines expressing synthetic TFs adopt different distributions for both fraction bound and reporter expression, but neither parameter predicts the other in individual cells.
Figure 3—figure supplement 4
Activation domain net charge and fractions bound for constructs in Figure 3.
Net activation domain charges are defined as the number of K and R residues minus the number of D and E residues. Net charges are plotted against fractions bound for each activator, as in Figure 3. There is no obvious relationship between net charge and bound fraction.
Figure 3—figure supplement 5
Mutations in activation domains change reporter activity and fluorescence recovery after photobleaching (FRAP) recovery.
(A) Flow cytometry measurement of GFP reporter expression after 24 hr of estradiol induction for cells expressing CITED2 alleles appended to the synthetic transcription factor (TF). The CITED2-WT curve is repeated from Figure 2B. (B) As in (A) for VP16 alleles. The VP16-WT curve is repeated from Figure 2B. (C) As in (A) for HIF1α alleles. The HIF1α-WT curve is repeated from Figure 2B. (D) Aggregate FRAP recovery curves (rolling averages, solid lines) for HIF1α and superactive allele of HIF1α (QN>E). Double-exponential fits are plotted as dashed lines. HIF1α-WT FRAP curve also appears in Figure 2D.
Figure 4 with 1 supplement
Activation domain alleles control the fraction of molecules bound to chromatin when fused to an alternative DNA-binding domain (DBD) or in a full-length transcription factor (TF).
(A) Left: We replaced the synthetic zinc finger DBD with the human SOX2 DBD. Right: Diffusion spectra and cumulative distributions for three alleles of CITED2. Shaded region indicates density contributing to the bound fraction. (B) Top: Schematic of the HIF1α and HIF2α intrinsically disordered region (IDR) swap chimeras from Chen et al., 2022. Bottom: Diffusion spectra and cumulative distributions of the constructs shown. The HIF1α IDRs control the fraction of molecules bound to chromatin under hypoxic conditions. (C) Top: We introduced three superactivating point mutations (QN>E) into full-length HIF1α’s second activation domain. Bottom: Diffusion spectra and cumulative distributions of the constructs shown. These three mutations in HIF1α increase the fraction of molecules bound to chromatin, recapitulating the differences seen in synthetic TFs.
Figure 4—figure supplement 1
AAVS1-integrated GFP reporter with synthetic ZF-binding sites is not activated by SOX2DBD-activation domain constructs.
Fluorescence intensities from the GFP channel of microscope images were quantified from experiments in Figure 4A and compared to those in Figure 3A. Orange lines: medians; boxes: Q1-Q3; whiskers: Q1–1.5 inter-quartile range and Q3+1.5 inter-quartile range. Left: GFP intensities from cells with SOX2DBD fused to various mutants of CITED2. Right: GFP intensities from cells with the synthetic ZF DBD fused to various mutants of CITED2.
Figure 5 with 1 supplement
Proximity-assisted photoactivation (PAPA) of synthetic transcription factors (TFs) shows that stronger activation domains promote interaction with p300.
(A) Schematic of our PAPA experiment using p300 endogenously tagged with HaloTag and an exogenous SNAP-tagged synthetic TF. First, red SNAP-tag2-liganded dyes are shelved (left, Materials and methods). A pulse of violet light (405 nm) indiscriminately reactivates these shelved dyes, while green light (561 nm) excites the sender, HaloTag-liganded dye, which reactivates nearby SNAP-tag dyes selectively. (B) Normalized PAPA ratio quantifying the proximity between p300-HaloTag and the constructs labeled. Error bars are 95% confidence intervals derived by bootstrapping (Materials and methods). (C) Diffusion spectra and cumulative distributions after segregation of green-reactivated (p300 proximal molecules) and violet-reactivated (direct reactivation) trajectories for cells expressing HIF1α WT. Density in the gray shaded region is quantified as the bound fraction. Light green and purple shading denotes the standard deviation of bootstrapping trials. (D) As in (C) for HIF1α RK >D. (E) As in (C) for HIF1α QN >E.
Figure 5—figure supplement 1
Proximity-assisted photoactivation (PAPA) analysis intermediates.
(A) For one PAPA replicate imaging cells expressing SNAP-tag2, the number of localizations as a function of frame number is plotted as a black line. Localizations immediately following a violet or green pulse (and considered reactivated for downstream analyses) are plotted as bold violet and green lines, respectively. (B) As in (A) for cells expressing HIF1α WT. (C) As in (A) for cells expressing HIF1α QN>E. (D) Cell-wise normalized PAPA ratios (Materials and methods) are plotted across three replicates for cells expressing SNAP-tag2 and colored according to the legend. A line whose y-intercept is zero is fit to these points and plotted as a dashed red line. (E) As in (D) for cells expressing HIF1α WT. (F) As in (D) for cells expressing HIF1α QN>E.
Figure 6 with 1 supplement
Perturbing coactivator function alters transcription factor (TF) bound fraction.
(A) VP16-synthetic TF diffusion spectra and cumulative distributions for cells treated for 2 hr with norstictic acid or vehicle (DMSO). Density in light gray is quantified as the bound fraction. Shading around the spectra and cumulative distributions indicates standard deviation of bootstrapping trials. (B) Diffusion spectra and cumulative distributions of cells expressing the synthetic TF with no activation domain treated for 2 hr with norstictic acid or vehicle (DMSO). Density in light gray is quantified as the bound fraction. Shading around the spectra and cumulative distributions indicates standard deviation of bootstrapping trials. (C) As in (A) for A485, a p300 inhibitor. (D) As in (B) for A485, a p300 inhibitor. The DMSO curve is repeated from (B). (E) As in (A) for BRM-014, a BAF inhibitor (BAFi). (F) As in (B) for BRM-014, a BAF inhibitor (BAFi). The DMSO curve is repeated from (B).
Figure 6—figure supplement 1
Drug perturbations on the synthetic transcription factor (TF) bearing mutants of the VP16 activation domain.
(A) Diffusion spectra (top) and cumulative distributions (bottom) for VP16 F442A-synthetic TF-expressing cells treated for 2 hr with norstictic acid or vehicle (DMSO). Shading indicates standard deviation of bootstrapping trials. (B) As in (A) for VP16 7As-synthetic TF.
Figure 7 with 1 supplement
CUT&RUN on synthetic transcription factors (TFs) with different activation domains reveals most binding occurs at open chromatin.
(A) We performed CUT&RUN on the parental cell line bearing the GFP reporter locus and five cell lines, each expressing synthetic TFs with a variable activation domain. Cells for all conditions were probed with an anti-V5 antibody, which should bind specifically only to the V5 epitope of the synthetic TFs. (B) Genomic binding signal at the engineered AAVS1 locus, shown to scale. All the synthetic TFs, including the empty TF, bound the cognate binding sites in the reporter construct. (C) Mean binding signal of the various synthetic TFs between the dashed lines of (B), which include the ZF1 binding sites and minimal promoter. (D) Binding of synthetic TFs and ATAC-seq signal at an exemplary locus. TF binding signal is strongest at ATAC-seq peaks, with stronger activation domains conferring more genomic binding in some cases. Not all ATAC-seq peaks are bound by the synthetic TFs. (E) ATAC-seq (left) and synthetic TF factor binding (right plots) signals plotted across 80118 ATAC-seq peaks, sorted by decreasing ATAC-seq peak strength. The five synthetic TFs share a color scale; the ATAC-seq signal uses a separate scale. Binding patterns for the synthetic TFs generally follow the strength of the ATAC-seq signal.
Figure 7—figure supplement 1
Synthetic transcription factors (TFs) bearing various activation domains generally bind to similar, open regions of the genome.
(A) Stacked bar plot quantifying ATAC-seq and CUT&RUN peaks overlapping (blue) and not overlapping (orange) ATAC-seq peaks for various activation domain-synthetic TFs. (B) ATAC-seq signal and synthetic TF binding signal is plotted in descending peak strength order for the TF shown labeled on the left. The five synthetic TFs share a color scale; the ATAC-seq signal uses a separate scale. The synthetic TFs with various activation domains generally bind to similar regions of the genome, especially those within ATAC-accessible regions.
Figure 8
Model for how activation domains contribute to chromatin binding by interactions with coactivators.
(A) A solitary transcription factor (TF) diffusing in the nucleoplasm, a lone wolf. (B) A TF bound to a coactivator with its activation domain diffusing through the nucleoplasm. (C) Three TFs bound simultaneously to a coactivator diffusing in the nucleoplasm as a complex or wolfpack. (D) A TF nonspecifically bound to random DNA. (E) A TF specifically bound with its DBD to a cognate DNA motif at the reporter locus. (F) TFs specifically bound to motifs at the reporter locus and simultaneously bound to a coactivator. (G) Proposed model for how the activation domain tethers the synthetic TF to chromatin through the coactivator. This situation occurs at active loci where the coactivator has been recruited by the orange and blue TFs.
Author response image 1
Conceptual model for how a coactivator can lead to apparent cooperativity between transcription factors.
(A) A transcription factor nonspecifically-bound to random DNA. (B) A transcription factor specifically-bound to a cognate DNA motif at our reporter. (C) A transcription factor non-specifically-bound to random DNA and simultaneously bound to a coactivator. The coactivator is also bound to another transcription factor specifically bound to its cognate motif. This situation represents other active loci. (D) A transcription factor specifically-bound to a cognate DNA motif and simultaneously bound to a coactivator tethered to the DNA by another transcription factor. This binding mode could occur at our reporter locus, but the identity of the dark blue transcription factor is unknown.
Author response image 2
Videos
Video 1
Exemplary single-molecule tracking (SMT) movie of a cell expressing VP16-synthetic transcription factor (TF).
SMT movie shown in real time of a cell expressing VP16-synthetic TF. Trajectories overlaid and colored according to their mean jump length, where brighter-colored spots belong to faster-moving trajectories. Initially, frames are dense with detections and are rejected for analysis. Detections outside a curated nuclear mask are also rejected.
Video 2
Exemplary fluorescence recovery after photobleaching (FRAP) movie of a cell expressing CITED2-synthetic transcription factor (TF).
FRAP movie, nuclear mask, photobleach spot mask, and an overlay of the movie and spot mask are shown above the normalized intensity of the FRAP recovery over time. Movies of many cells expressing CITED2-synthetic TF are aggregated and fit to generate a recovery curve, like that in Figure 2E (green).
Video 3
Exemplary fluorescence recovery after photobleaching (FRAP) movie of a cell expressing HIF1α-synthetic transcription factor (TF).
FRAP movie, nuclear mask, photobleach spot mask, and an overlay of the movie and spot mask are shown above the normalized intensity of the FRAP recovery over time. Movies of many cells expressing HIF1α-synthetic TF are aggregated and fit to generate a recovery curve, like that in Figure 2E (purple).
Additional files
-
Supplementary file 1
Fluorescence recovery after photobleaching (FRAP) parameters extracted from a double-exponential recovery model.
FRAP population fractions and recovery times for cells expressing various synthetic transcription factors (TFs) and H2B. Stronger activation domains have longer slow-recovery times, indicating longer residence times bound to chromatin. *Note that >95% of H2B molecules do not recover on the timescale of the FRAP experiment.
- https://cdn.elifesciences.org/articles/105776/elife-105776-supp1-v1.csv
-
Supplementary file 2
Summary of single-molecule tracking (SMT) measurements.
Fraction bound, number of cells imaged across all replicates, number of trajectories, number of jumps, and number of detections for all SMT experiments reported in this article. All tabulated values are post-filtering (Materials and methods: SMT and PAPA-SMT analysis). Number of trajectories and number of detections include singlets: detections that were not linked to any other detection and do not contribute any data to downstream analyses but have their own ‘trajectory’ index. * indicates datasets acquired without an automated nucleus-finding macro.
- https://cdn.elifesciences.org/articles/105776/elife-105776-supp2-v1.csv
-
MDAR checklist
- https://cdn.elifesciences.org/articles/105776/elife-105776-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Short activation domains control chromatin association of transcription factors
eLife 15:e105776.
https://doi.org/10.7554/eLife.105776
