(A) Kernel density estimation plots (KDE, blue) and empirical cumulative density functions (cdfs, orange) of initial FRET values for 3/78 nucleosomes alone (in the presence of 1 mM ATP, but in the absence of remodeler; previous work with ISWI family remodelers has shown the addition of remodeler does not affect these initial FRET [Blosser et al., 2009; Deindl et al., 2013]). Left panel, wild-type H2A-containing nucleosomes; right panel, H2A/E64R-containing nucleosomes. N indicates the number of nucleosomes included in the cdf. KDEs are more intuitive—the y-axis is analogous to the frequency axis of a histogram—but cdfs have the advantage of not requiring any smoothing or binning. Note that here nucleosomes were imaged at a significantly higher laser power (20 mW) than for measuring remodeling (11.5 mW), to ensure that the majority of both donor and acceptor dyes photobleached within the 5 min imaging interval, so that dyes with two-step photobleaching events could be excluded. Only nucleosomes with an initial FRET value greater than 0.775 (black dashed lines) were included for further analysis. KDEs have Gaussian kernels with bandwidths ~ 0.01. (B) Our ability to measure remodeling of E64R nucleosomes by SNF2h is limited by the photobleaching rate. Previous work has shown remodeling in ensemble FRET assays to be comparable to remodeling in surface-immobilized smFRET assays (Blosser et al., 2009; Hwang et al., 2014), as well as to that of surface-attached dinucleosome constructs (Hwang et al., 2014). However, as shown here, remodeling of surface-attached E64R nucleosomes (red data) appears to go to completion in about 4 min (inset), whereas remodeling of these same nucleosomes in ensemble FRET assays (blue data) takes over two hours to go to completion. This discrepancy is not due to an effect of the surface immobilization, but rather to the competition between remodeling and photobleaching. In the presence of SNF2h alone, without ATP (black data), nucleosomes photobleach on roughly the same timescale as the apparent remodeling of surface-attached E64R nucleosomes. Thus we are able to put only a lower bound on pause durations for remodeling of E64R nucleosomes, because slower remodeling events are undetected due to the relatively fast photobleaching of the dyes. The smFRET data shown here corresponds to the summed Cy5 intensities of all individual nucleosomes in the indicated data set, binned in 1 s intervals to simulate an ensemble measurement.