GAGA zinc finger transcription factor searches chromatin by 1D-3D facilitated diffusion

Abstract

The search for target sites on chromatin by eukaryotic sequence-specific transcription factors (TFs) is integral to the regulation of gene expression but the mechanism of nuclear exploration has remained obscure. Here we use multicolor single-molecule fluorescence resonance energy transfer and single-particle imaging to track the diffusion of purified Drosophila GAGA factor (GAF) on DNA and nucleosomes. Monomeric GAF DNA-binding domain (DBD) bearing one zinc finger finds its cognate site through one-dimensional (1D) or three-dimensional (3D) diffusion on bare DNA and rapidly slides back and forth between naturally clustered motifs for seconds before dissociation. Multimeric, full-length GAF also finds clustered motifs on DNA through 1D-3D diffusion but remains locked on target for longer periods. Nucleosome architecture effectively blocks GAF-DBD 1D sliding into the histone core but favors retention of GAF-DBD once it has bound to a solvent-exposed motif through 3D diffusion. Despite the occlusive nature of nucleosomes, 1D-3D facilitated diffusion enables GAF to effectively search for clustered cognate motifs in chromatin, providing a mechanism for navigation to nucleosomal and nucleosome-free sites by a member of the zinc finger TF family.

Fig. 1. DNA sequence specificity of GAF-DBD is kinetically defined.

a, Domain map of GAF. ZF, zinc finger. b, Schematics of two-color smFRET experiment to measure cognate-specific binding of GAF-DBD. c, Single-molecule trajectories showing Cy3–GAF-DBD binding to Cy5–DNA containing a GAF cognate site. Boxed binding events are magnified in e. Examples of dwell time measurements (t) are shown on the third trajectory. d, Representative single-molecule trajectories of Cy5–cognate DNA bound by Cy3–GAF-DBD showing an initial donor-only period before acceptor signal increases. The donor-only period is highlighted in gray. e, Zoomed-in view of binding events boxed in c. The asterisk indicates a transient Cy3-only fluorescence spike during binding. f, Single-molecule trajectories showing Cy3–GAF-DBD nonspecifically binding to Cy5–DNA where the cognate site was substituted with a noncognate sequence. g, Binding frequency of GAF-DBD to cognate or noncognate DNA. h, Dwell time of GAF-DBD on cognate or noncognate DNA. Error bars show the s.e.m. from mean binding frequencies of technical replicates (cognate DNA, three replicates, total n = 317; noncognate DNA, two replicates, total n = 282). The concentration of GAF-DBD used in these experiments was 0.2 nM. Fluor., fluorescence.

Fig. 2. GAF-DBD explores free DNA by 1D diffusion.

a, Three-color smFRET experiment distinguishes whether GAF-DBD is bound to Cy7-labeled site 1, Cy5-labeled site 2 or a nonspecific site on Drosophila hsp70 promoter DNA. b, A single-molecule trajectory shows GAF-DBD sliding back and forth on the DNA between two cognate motifs. Binding site assignment is shown below as a colored ribbon. c, A total of 50 binding events shown as a rastergram. d, Zoomed-in view of the boxed region in b. e, GAF-DBD dwell times for motif-specific binding (τ_msb) on site 1, site 2 or nonspecific site on the DNA with regular (6.25 mM, n = 51 traces) or low (3 mM, n = 32 traces) MgCl₂. The concentration of GAF-DBD used in these experiments was 0.1 nM. Error bars represent 95% confidence intervals. f, Schematic of GAF-DBD sliding on DNA. Circles represent GAF-DBD. Green circles, GAF-DBD binding to a nonspecific site; blue circle, GAF-DBD binding to site 1; red circle, GAF-DBD binding to site 2.

Fig. 3. Nucleosome blocks GAF-DBD 1D sliding beyond SHL7.

a, Schematics of nucleosome constructs where the nucleosomal Cy7-labeled cognate site (blue) is placed at SHL7, SHL5 or SHL3 and the other Cy5-labeled cognate site (red) is on the linker DNA. The nucleosome is positioned by the Widom 601 sequence flanked by 40-bp linker DNA on both sides (40-N-40). Seq., sequence. b–d, Representative single-molecule trajectories of GAF-DBD binding to linker or nucleosomal sites at SHL7 (b), SHL5 (c) and SHL3 (d). e–g, Binding site rastergrams for GAF-DBD on nucleosome constructs where the Cy7-labeled motif is located at SHL7 (e), SHL5 (f) or SHL3 (g). h, GAF-DBD binding categories for each construct; n = 485 for SHL7, n = 260 for SHL5 and n = 272 for SHL3. i, Categories of binding events on SHL7 construct. Data are represented as the mean values ± s.d. of three independent imaging sessions, with a sum of n = 485 binding events. j, GAF-DBD motif dwell times on SHL7, linker site and nonspecific sites on the SHL7 construct (n = 34 traces). Error bars represent 95% confidence intervals. Nucl., nucleosome.

Fig. 4. 3D diffusion dominates for inner nucleosomal targets.

a, Drosophila hsp70 promoter nucleosome construct (0-N-40, same DNA as in Extended Data Fig. 3f) for investigating GAF-DBD search when two binding sites lie within the nucleosome core. b,c, Representative single-molecule trajectories for site A (b) and site B (c) binding (0.4 nM GAF-DBD used). d, Binding event categories on the nucleosome. e, A 1 − CDF plot of motif dwell times for nucleosome (orange) compared to free DNA (blue). Fitted τ values are provided in Supplementary Table 5. Free DNA data are aggregated from Cy7–dyad site and Cy5–edge site dwell times on the same DNA template (Extended Data Fig. 3f). f, EMSA for GAF-DBD binding to Cy5–hsp70 NCP. g, EMSA showing GAF-FL binding to Cy5–hsp70 NCP. h, EMSA showing GAF-FL binding to Cy5–hsp70 DNA (the same DNA for Cy5–hsp70 NCP). EMSA experiments in f–h were independently repeated at least once with similar results. Source data

Fig. 5. GAF-FL multimer undergoes 1D diffusion during target search.

a, Schematic of construct design. Plasmids with or without the hsp70 promoter sequence were digested with HindIII and concatenated using T4 DNA ligase. Exposed ends were then biotinylated with adaptors, resulting in dual-end biotinylated DNA carrying repeating hsp70 promoters. b, Setup of dual optical tweezers for confocal microscopy of stretched DNA tethered between two streptavidin-coated polystyrene beads. c,d, Representative kymographs show AF488–GAF-FL fluorescence signal (cyan) on DNA over time in the absence (c; vector only) or presence of hsp70 promoter (d; vector + hsp70). The white arrowhead shows GAF-FL binding directly to target in 3D; the black arrowheads show GAF-FL undergoing 1D search before finding target; the white asterisk shows direct 3D dissociation. Data are representative of at least three independent imaging sessions. e, Compiled plots of position versus time for GAF-FL trajectories on vector-only DNA. A total of 100 traces were collected for each condition and arranged to start from t = 0. f, Compiled plot for GAF-FL positions over time on vector + hsp70 DNA. g, Average MSD over time lag of all collected traces for vector only (purple) and vector + hsp70 DNA (blue). The shaded area reports the s.e.m. (n = 100). h, Pie charts categorizing GAF-FL traces by DNA binding at onset of video (left) and targeting directly (3D binding) or indirectly (1D sliding) (right). i, A representative kymograph of GAF-ΔPOZ on vector + hsp70 DNA. j, A 1 − CDF plot comparing dwell times of GAF-ΔPOZ on vector + hsp70 DNA (red; τ = 1.67 s), GAF-FL on vector + hsp70 (blue; τ = 43.4 s) and GAF-FL on vector only (purple; τ = 21.0 s). k, Single-trace diffusion coefficients for GAF-FL on vector + hsp70 and vector-only DNA and GAF-ΔPOZ on vector + hsp70 DNA. Error bars represent the s.e.m. Statistical differences were determined using a two-tailed unpaired t-test with Welch’s correction (n = 100 GAF molecules; **P = 0.0012 and ****P < 0.0001). These experiments were conducted using 80 nM GAF-FL and 25 nM GAF-ΔPOZ.

Fig. 6. Model for GAF-DBD and GAF-FL target search.

a, GAF-DBD uses two search modes to locate its target on chromatin. In the 1D sliding mode, GAF-DBD lands on an off-target location on free DNA and then slides back and forth to locate the cognate motif (GAGAG). It can escape the cognate site to search for the next site nearby. This 1D search mode allows GAF-DBD to invade into the nucleosome edge but no further. Alternatively, GAF-DBD can also directly associate with a solvent-exposed cognate motif in the nucleosome core from 3D space. This 3D search mode allows GAF to effectively target nucleosomal motifs that are inaccessible by 1D sliding. b, GAF-FL uses both 3D and 1D diffusion to locate cognate motif clusters on free DNA. If the motif cluster is inside a nucleosome, GAF-FL can use 3D diffusion for target location. c–h, Hypothetical stepwise model for GAF–remodeler collaboration to mobilize targeted nucleosome for PIC assembly, based on this study and prior literature. c, GAF localizes cognate targets on closed chromatin through a combination of 1D sliding and 3D diffusion. d, Nucleosome-bound GAF multimer may recruit chromatin remodelers, such as NURF and PBAP. e, The recruited ATP-dependent chromatin remodelers shift nucleosomes away from the cognate sites. It is unclear whether GAF dissociates from chromatin during this process. f–h, Once the cognate sites become nucleosome free, GAF multimer rapidly locates the free DNA target through 3D and 1D diffusion (f), enabling downstream PIC recruitment, transcription initiation (g) and binding of other TFs such as HSF (h). HSE, heat shock element; NDR, nucleosome-depleted region.

Extended Data Fig. 1. Purification of GAF-DBD protein.

a, Subtractive Ni-NTA purification after “one-pot” reaction which cleaved off 6xHis-SUMO and labeled the N-terminus of GAF-DBD with Cy3. The calculated molecular weights are 34.4 kDa for 6xHis-SUMO-Cys-GAF-DBD, 21.0 kDa for Cys-GAF-DBD and 13.4 kDa for 6xHis-SUMO. This experiment was independently repeated once with similar results. b, EMSA showing GAF-DBD binds to hsp70 promoter NCP DNA with K_1/2 ~ 12 nM. GAF-DBD concentrations are, from left to right, 0, 4, 8, 12, 16, 20, 24, 28, 32 nM and the DNA concentration is 3 nM. This experiment was independently repeated once with similar results. c, The binding frequency of GAF-DBD on cognate DNA as a function of protein concentration (For each concentration, N = 40 traces from two imaging sessions were used to quantify the binding frequency). The binding constant is determined from a linear fit to the data (R² = 0.99) to be 0.27 s⁻¹ nM⁻¹. To calculate K_D, we divide the measured k_on (3.1 s⁻¹, Fig. 1) by this number to get 11.5 nM which agrees well with the 12 nM apparent K_D (K_1/2) determined from the EMSA in b. Data are represented as mean values +/− SD. d, Schematics of GAF-FL short isoform, GAF-DBD and NURF-binding regions to scale. e, Native hsp70 promoter DNA sequence. GAF cognate sites are highlighted in red. TATA box in bold. Source data

Extended Data Fig. 2. Donor-only dwell time is inversely correlated with ionic strength of buffer solution.

a, Representative trajectories showing Cy5-cognate DNA bound by Cy3-GAF-DBD in 0, 25, 50, or 100 mM NaCl. Grey-highlighted durations indicate donor-only dwell times. b, Representative trajectories showing Cy5-cognate DNA bound by Cy3-GAF-DBD in 0, 1, 5, 7.5, or 10 mM MgCl₂. The sensitivity of FRET dynamics to salt concentration suggests that the transient donor-only periods are not due to photo-induced blinking. c, Donor-only dwell time (τ from fitting 1-CDF to a single-exponential decay; see Methods) as a function of NaCl concentration (N = 119, 147, 109, 23 for 0 mM, 25 mM, 50 mM, and 100 mM, respectively), and d, MgCl₂ concentration (N = 177, 128, 116, 149 for 0, 1, 5, and 7.5 mM, respectively). Data are represented as mean values +/- standard errors for a single-component exponential decay fit. DNA is the same as cognate DNA in Fig. 1. e, 1-CDFs for dwell times reported in Fig. 1h with fits to a single-component exponential decay function (ExpDec1 in OriginPro). Replicates were fit individually and their average τ is reported as the final dwell time.

Extended Data Fig. 3. GAF sliding kinetics on Cy5 & Cy7 dual-labeled DNA as a function of salt concentration, cognate site and motif orientation.

a, Representative single-molecule trajectories of Cy5 & Cy7 DNA bound by Cy3-GAF-DBD in low MgCl₂ (3 mM). b, Rastergram of 31 Cy5 & Cy7 DNA molecules bound by Cy3-GAF-DBD at 3 mM MgCl₂. c, Schematic of Site 2 Only construct where Site 1 was replaced with non-cognate sequence. d, Representative single-molecule trajectories of Site 2 Only DNA bound by GAF-DBD. e, Rastergram of 40 Site 2 Only DNA molecules bound by GAF-DBD. f, Schematic of the flipped motif DNA construct where Site 1 is replaced with its complementary sequence, Site 1’. g, Representative single-molecule trajectories of flipped motif DNA bound by GAF-DBD. h, Rastergram of 57 flipped motif DNA molecules bound by GAF-DBD. DNA same as Fig. 4. i, Categories of GAF-DBD arrival landing site and departure launching site.

Extended Data Fig. 4. Nucleosomal motif accessibility depends on helical phasing.

a, Representative native PAGE gels for nucleosome constructs used in this study. All nucleosome constructs used in this study were assessed by native PAGE gels and showed similar results. b, Classification of Cy3-GAF-DBD binding events on ‘601’ nucleosomes (40-N-40) with a cognate site placed at SHL6.5, SHL4.5, and SHL2.5. Classification is based on Cy3-Cy7 FRET dynamics, see details in methods. c, Classification of Cy3-GAF-DBD binding events on same ‘601’ (40-N-40) nucleosomes with cognate site placed at SHL7, SHL5, and SHL3 (same dataset as Fig. 3, re-analyzed for direct comparison with a). d, Representative single-molecule trajectories for each category of 601-SHL6.5, -SHL4.5, and -SHL2.5 nucleosomes. Some scanning events show prolonged Cy3-Cy7 FRET (right) which appear distinct from scanning events on the 601-SHL7 nucleosomes (Fig. 3b). Note that scanning events (events that show dynamic Cy3-Cy7 FRET) may also come from free DNA which contaminates the nucleosome sample.

Extended Data Fig. 5. GAF-DBD preferentially re-visits the same cognate site on individual hsp70 nucleosomes.

a, Representative single-molecule trajectory for Site C binding. b, Representative single-molecule trajectories showing repetitive visits to the same binding site on a single nucleosome. Upper trace shows repetitive visits to Site A; middle trace, Site B; lower trace, Site C. The bottom trace shows repeated binding events with zero Cy3-Cy5 FRET and low (0.24) Cy3-Cy7 FRET, consistent with binding to a distal part of the Cy7-labeled motif similar to Site B. The difference in Cy3-Cy5 FRET could be due to heterogeneous phasing of the hsp70 nucleosome as suggested by the Cy5-Cy7 FRET histogram in c. Black asterisks mark transient Cy3-only periods, potentially caused by ultra-short-range 1D diffusion on the nucleosome, resulting in a temporary loss of FRET. Green asterisks indicate binding events to non-cognate sites on the nucleosome. c, Pie charts showing, for all binding events at site A (left pie chart, N = 65), B (middle, N = 45) or C (right, N = 25), the fraction of events that were followed by a second binding to site A, B or C. For example, if a nucleosome is visited by GAF-DBD three times – first at Site A, second also at Site A, and third at Site B – then an A-A rebinding event and an A-B rebinding event have occurred on this nucleosome. d, Cy5-Cy7 FRET histogram for the hsp70 nucleosome (N = 377).

Extended Data Fig. 6. Purification of SNAP-tagged GAF-FL and GAF-ΔPOZ.

a, GAF-FL purification workflow. b, SDS-PAGE gel of GAF-FL elution after 6XHis pulldown, stained with Coomassie Blue. c, SDS-PAGE gel of GAF-FL elution after MBP pulldown, stained with Coomassie Blue. d, SDS-PAGE gel of AF546-GAF-ΔPOZ, scanned for AlexaFluor 546 fluorescence, then stained with Coomassie Blue. e, EMSA for GAF-ΔPOZ binding to Cy5-labeled hsp70 NCP DNA. GAF-ΔPOZ concentrations were 0, 4, 8, 12, 16, and 20 nM from left to right. DNA concentration was 3 nM. SDS-PAGE and EMSA gels shown in b-e were independently repeated once with similar results. Source data

Extended Data Fig. 7. Optical tweezers experiment; design and confirmation of DNA assembly.

a, Agarose gel electrophoresis of concatenated plasmid DNA. This experiment was repeated once with similar results. b, Force versus distance plot reveals the length of double-stranded DNA tether. c, Diagram of flow cell constituents during imaging. Streptavidin coated beads, DNA, and Imaging Buffer were injected to the flow cell under laminar flow. Beads were first optically captured and moved to the DNA channel; once DNA was properly tethered to the trapped beads, the whole assemblage was moved to the protein channel containing AF-488 GAF-FL in Imaging Buffer. Imaging was performed in this channel to maximally visualize binding events. d, Representative kymographs where GAF-FL undergoes 1D search on vector+hsp70 DNA. e, Representative kymographs where GAF-FL binds to its target on vector+hsp70 DNA abruptly from 3D without 1D search. f, Representative kymograph where GAF-FL undergoes 1D diffusion from one target to another on vector+hsp70 DNA.

Extended Data Fig. 8. Behavior of full-length GAF depends on multimerization by POZ domain.

a, Representative kymographs show AF546-GAF-ΔPOZ binding to DNA over time on + hsp70 DNA. b, Representative fluorescence intensity trace for GAF-ΔPOZ showing an abrupt loss of fluorescence signal at 5.7 s due to either dissociation or photobleaching. c, Representative trace showing GAF-FL photobleaching. Arrows indicate stepwise photobleaching. d, Mass photometry spectrum for GAF-ΔPOZ and GAF-FL constructs used in optical tweezer measurements. Spectra show the distribution of protein mass in a sample solution containing the indicated concentration of GAF. Notably, at similar protein concentrations as those used for the optical tweezer experiments (80 nM GAF-FL and 25 nM GAF-ΔPOZ), GAF-FL exists predominantly as an octameric multimer whereas GAF-ΔPOZ exists as a monomer.