Single-molecule imaging of DNA curtains reveals intrinsic energy landscapes for nucleosome deposition

Transcription

1 SUPPLEMENTARY INFORMATION Single-molecule imaging of DNA curtains reveals intrinsic energy landscapes for nucleosome deposition Mari-Liis Visnapuu 1 and Eric C. Greene 1 1 Department of Biochemistry & Molecular Biophysics, Columbia University, and the Howard Hughes Medical Institute, 650 West 168th Street, Black Building Room 536, New York, NY To whom correspondence should be addressed: ecg2108@columbia.edu Supplementary Video 1. Nanofabricated DNA curtains. Electron-beam lithography was used to nanofabricate chromium diffusion barriers on the fused silica surface of a microfluidic sample chamber. λ-dna molecules stained with YOYO1 (green) were then anchored via a biotinneutravidin linkage to a fluid lipid bilayer deposited on the surface of the slide. Application of buffer flow aligns the DNA molecules along the leading edges of the nanofabricated barriers and extends the DNA parallel to the sample chamber surface, confining the molecules within the excitation volume defined by the penetration depth of the evanescent field. When buffer flow is transiently paused the DNA molecules collapse due to entropic forces and move away from the surface of the sample chamber demonstrating that the molecules do not interact nonspecifically with the lipid bilayer. Supplementary Video 2. Visualizing DNA curtains and fluorescent nucleosomes. A YOYO1- stained DNA curtain (green) was assembled as in Video S1, with the exception that the DNA was pre-bound by nucleosomes deposited through salt dialysis. The nucleosomes were labeled in situ using fluorescent quantum dots (magenta). When buffer flow is transiently paused the nucleosomebound DNA molecules collapse due to entropic forces and move away from the surface of the sample chamber demonstrating that the molecules do not interact nonspecifically with the lipid bilayer and verifying that the nucleosomes are bound to the DNA.

2 Supplementary Fig. 1. Quantitative Analysis of Resolution Limits. Panel (a) shows examples of DNA that were labeled at the end or at an internal location with a single fluorescent QD. The DNA was stained with YOYO1 and is shown in green, and the QDs are shown in magenta. The top panels show images acquired during application of buffer flow to stretch the DNA, and the bottom panels show control images taken after transiently terminating buffer flow to verify that neither the DNA molecules nor the QDs were nonspecifically bound to the surface. End-labeled DNA substrates were made by annealing digoxigenin (DIG)-labeled oligonucleotides to the 12-nucleotide overhang at the right end of the λ-dna. Internally labeled DNA substrates were constructed by treating biotinylated λ-dna with the nicking enzymes Nb.BsmI and Nt.BstNBI (NEB) to generate nicks flanking a 16-bp region from 26,151-26,166 bps. A DIG-tagged oligonucleotide complementary to the 16-bp gap was added to a final concentration of 500 nm and incubated at 55 C for 30 min to allow exchange. The reaction was cooled to room temperature followed by the addition of T4 DNA ligase. The DIG-tagged DNA substrates were assembled into DNA curtains, stained with YOYO1 and labeled with anti-dig QDs. A histogram illustrating the locations of the engineered tags is shown in (b). The data was collected under conditions identical to those of nucleosome position assay, and the positions of the tags were measured as described in the Methods. The internal tag was localized to within 37-bp of its actual position, and the end tag was localized to within 153-bp of its known position.

3 Supplementary Fig. 2. Characterization of recombinant nucleosomes. Panel (a) shows a Coomassie-stained SDS-PAGE of the purified histone octamer made with FLAG-H2B. For single molecule assays nucleosomes were assembled by 48.5-kb λ-dna or 23-kb PCR fragment of the human β-globin locus made using the Expand 20kb PLUS PCR system (Roche). The nucleosomes were assembled using salt dialysis (see below). The ratio of DNA to histone octamer was varied, and micrococcal nuclease assays were used to verify nucleosome assembly (b). Panel (c) shows gel shift assays and antibody specificity controls using a smaller PCR fragment. The 280-bp DNA substrate containing a copy of the 601 nucleosome positioning sequence was prepared by PCR. Nucleosome assembly reactions contained 3 µg of DNA and 1.5 to 2-fold molar excess of histone octamer in 1.5 M NaCl, 10 mm Tris-Cl [ph 7.8], 1 mm EDTA. The reactions were dialyzed against a step-wise salt gradient (2h at 1 M NaCl, 2h at 0.8 M NaCl, 2h at 0.6 M NaCl, 2h at 0.4 M NaCl, and 12h at 0.2 M NaCl) at 4 C and nucleosome formation was verified by gel shift on 4% polyacrylamide. Antibody binding and specificity was also verified by gel shift after incubating µg of nucleosomes with either anti-flag or anti-ha antibodies at 1:1 ratio for 15 minutes at 4 C. Panel (d) shows Coomassie-stained SDS PAGE of the reconstituted and purified histone octomers made with either H2AZ, Cse4, or Cse4/Scm3 (hexasomes), as indicated. Gel shift assays and antibody specificity controls using the 280-bp 601 PCR fragment are shown in (e). Micrococcal nuclease footprint assays were used to verify nucleosome formation on λ-dna (f).

4 Supplementary Fig. 3. Predicted distributions at differing nucleosome concentrations. The Segal et al. and Field et al. algorithms contain concentration parameters for adjusting the nucleosome density on the DNA (see Panels (a-d) contain predictions for either λ-dna (a and b) or the β-globin DNA (c and d) for each of the models under a range of concentration parameters as indicated at the top left corner of each panel. The prediction data sets used for correlation measurements with experimental data is indicated with asterisks (*). Correlation analysis of the Segal et al. and Field et al. predictions relative to one another for (e) λ-dna and (f) the human β-globin DNA fragment.

5 Supplementary Fig. 4. Predicted distribution analysis and observed data binned at same resolution. Panels (a) and (b) show the Segal et al. and Field et al. predictions and the observed nucleosome positions divided into 758-bp bins for λ-dna and the human β-globin DNA fragment, respectively.

6 Supplementary Fig. 5. Comparison of the theoretical distributions. Panels (a) and (b) show the theoretical distributions predicted by the Segal et al., and Field et al. models compared to the more recent model of Kaplan et al. for the λ-dna substrate and the human β-globin substrate, respectively.

7 Supplementary Fig. 6. Heat-shifted nucleosomes and comparison of all octamers. Panel (a) shows the positions of 1,247 canonical nucleosomes determined after 10-hour heat-shift incubation at 37 C. Panel (b) shows Pearson correlation analyses for the heat-shifted nucleosomes compared to the standard 4 C nucleosomes, and the Field, Segal, and Kaplan et al. models (clockwise from top left). Panel (c) shows the mean values for the canonical, H2AZ, and Cse4 octameric nucleosomes. The error bars represent the standard deviation of the three different data sets. The correlation between the mean distribution and the theoretical predictions from the Field et al. and Kaplan et al. models is shown in (d).

8 Supplementary Fig. 7. Comparison of canonical and H2AZ nucleosomes. Panel (a) shows the Segal et al. and Field et al. predictions for λ-dna along with the distribution histograms for canonical and H2AZ-containing nucleosomes, and the overlaid data sets. The correlation between the observed distributions and the theoretical predictions from the two models is shown in (b).

9 Supplementary Fig. 8. Anticorrelated predictions for the Segal and Field/Kaplan models with Yeast DNA. Panel (a) contains the Segal (light blue), Field (green) and Kaplan (purple) predictions for four randomly selected 20kb regions of the yeast genome. The Pearson correlation analyses for the Segal and Field models are shown in panel (b) for each of the four yeast genome regions. These data show that the Segal & Field/Kaplan models still yield anticorrelated results when testing yeast DNA substrates.

10 Supplementary Fig. 9. Scm3 does not dissociate from DNA after deposition of Cse4/H4. (a) The centromeric nucleosome complexes containing either Cse4/H4/H2A/H2B or His6- Scm3/Cse4/H4 were reconstituted by salt dialysis onto the 280-bp 601-containing DNA, as described in supplementary Fig. S2. (b) The nucleosomes were then incubated with Ni-NTA agarose beads for 30 minutes in 10 mm Tris [ph 7.8]. The beads were spun down and the supernatant was removed. The beads were washed twice with 20 mm imidazole and 10 mm Tris to remove any unbound nucleosomes. The bound nucleosomes were finally eluted with 300 mm imidazole. (c) The starting material, initial supernatant, washes and eluate were analyzed on 20% acrylamide SDS-PAGE gels and detected by SimplyBlue SafeStain (Invitrogen). The Cse4/H4/H2A/H2B complex did not bind the Ni-NTA agarose beads and remained in the supernatant. The His6-Scm3/Cse4/H4 complex was bound to the Ni-NTA agarose beads via the His6 tag on Scm3 and was eluted at 300 mm imidazole.

11 Supplementary Fig. 10. Correlation between the theoretical models and our experimental data plateaus at approximately 1000 data points. The positioning data were analyzed with an algorithm that selects a random subset of data points from an experimental data set for nucleosome positioning (the graph shown above used the data from Figure 2 of the manuscript), and the size of the randomly sampled bins ranges from i=1 to i=n, where N is the total number of experimental data points collected (N=2,458 in this example). The correlation coefficient was then calculated for each randomly selected subset of data for i=1 to i=n, this process was repeated 1000 times, and new random subsets data were selected for each iteration. The mean and standard deviation from the correlation coefficient was then calculated from values determined for each random subset of data, and the results were plotted as the number of data points versus the calculated correlation coefficient (± standard deviation, in red). As shown here, the coefficient value begins to converge around 1000 data points.