High-accuracy mapping of protein binding stability on nucleosomal DNA using a single-molecule method

Dear Editor, The conformation of nucleosomal DNA is significantly different from that of B-form double stranded DNA (dsDNA) (Richmond and Davey, 2003). In nucleosomal DNA, specific DNA sequences are less flexible and less accessible than in free dsDNA, which might be due to the tight association of histone cores. The allosteric effect via DNA has been documented recently (Kim et al., 2013), suggesting that DNA is not merely a solid rod providing recognition sequences. Previous studies of nucleosomal DNA–protein interactions only demonstrated the key mechanism involved—histone shielding (Li and Wrange, 1993; Hinz et al., 2010; Sahu et al., 2010). How conformational changes in nucleosomal DNA, compared with free dsDNA, affect nucleosomal DNA– protein interactions remains unknown. The glucocorticoid receptor (GR)-binding sequence GREs3 (5′-AGAACATCATGTTCT-3′) (Luisi et al., 1991) was inserted into a specific histone core-binding sequence W601 (Lowary and Widom, 1998) at the position from 2 16 to 4 bp around central dyad of the nucleosome (Supplementary Figure S1). Each nucleosomal construct was tethered onto a functional coverslip of a flowcell using a biotin-streptavidin connector (Figure 1A). Using a sensitive singlemolecule technique (Kim et al., 2013), we measured the stochastic residence time of Cy3B-labeled GR DNA-binding domain (GRDBD) on each nucleosomal GREs3. The average residence time (ART, t) for each construct was calculated by fitting the stochastic residence time to a singleexponential decay and extracting the decay constant (Figure 1B), which reflects the stability of GRDBD binding to each nucleosomal GREs3 (see Supplementary Figure S2 and Supplementary Methods for details). The ARTs of GRDBD binding to constructs without histone core, as control (labeled as DG instead of NG in Figure 1B), are very similar, suggesting inconspicuous influence of the flanking sequence of GREs3. Interestingly, the strongest binding (longest ART) occurred within the nucleosomal DNA–GRDBD complex (Figure 1B, NG-12) rather than the free DNA–GRDBD complex, demonstrating that the binding stability increased 1.5-fold on nucleosome. The main trend of the ART of all different rotationally positioned nucleosomal GREs3 was found modulated by the orientation (Red trendline in Figure 1C), agreed with previous ensemble data (Li and Wrange, 1993). Yet there are a few positions (NG-11, NG-10, and NG0) that could not be fitted on the trendline and showed a sudden reduction by 3.7-fold (Figure 1C). Not coincidently, these outliers were all positioned on the outward-facing segment of the nucleosome and were separated by 10–11 bps, which is identical with the DNA helical pitch on nucleosome (Richmond and Davey, 2003). These unexpected phenomena demonstrated that the orientation is not the only element to modulate the stability of nucleosomal binding protein. Two different types of dsDNA–GRDBD complex structures have been solved: the GREs3 sequence with two palindromic binding half-sites separated by three bases TCA (Figure 1D) and the GREs4 sequence (5′-AGAACATCGATGTTCT-3′, with four spacer bases TCGA in the middle) (Figure 1E). In these structures, longer GREs4, compared with GREs3, may cause tightened GRDBD dimer interface, so that the subunits of the GRDBD dimer could not specifically bind to both half-sites on the GREs4. Although the subunit may have missed interactions from its specific binding site, the dimer interface still remains (Figure 1E, blue arrows). The ART of the GRDBD binding on free GREs4 is only 38% + 2% of that on a free GREs3, indicating that breaking this interaction decreases the stability of the complex. The stability of GRDBD positioned on nucleosomal GREs3 at different locations is different from that of free GREs3, which may be due to the different interactions between GRDBD and various DNA conformations. A structural modeling with the strongest (NG-12) and weakest (NG0) interactional nucleosomal GREs3–GRDBD complexes using the HADDOCK approach (de Vries et al., 2010) well confirmed our hypothesis. Compression of the DNA structure in NG-12 not only conserved the interactions between two major grooves and GRDBD, but also produced the interaction between the minor groove and GRDBD (Figure 1F, red arrow), which increased the stability. In contrast, stretching of the DNA structure in NG0 mostly decreased the stability. Furthermore, bending of the histone core stretched the distance between two major grooves so much that one subunit of GRDBD cannot even contact the site, resulting in only one subunit interacting with the DNA (Figure 1G). Similar single-molecule binding assay was performed with the estrogen receptor DNA-binding domain (ERDBD), which is a structural homolog of GRDBD but binds to a different DNA sequence estrogen response element (ERE) (Schwabe et al., 1993). ERDBD binds to the nucleosomal DNA much weaker than to free ERE. Under similar conditions, the number of binding events (bright spots) between ERDBD 438 | Journal of Molecular Cell Biology (2014), 6(5), 438–440 doi:10.1093/jmcb/mju033 Published online July 17, 2014