Ruler elements in chromatin remodelers set nucleosome array spacing and phasing

Elisa Oberbeckmann; Vanessa Niebauer; Shinya Watanabe; Lucas Farnung; Manuela Moldt; Andrea Schmid; Patrick Cramer; Craig L. Peterson; Sebastian Eustermann; Karl-Peter Hopfner; Philipp Korber

Remodeler-specific rulers set spacing in a densityindependent or -dependent way. To compare spacing generated by different remodelers at different nucleosome densities, we focused on the averaged length of linker 1 ( Figure 2B ), which was most accessible across all nucleosome densities. Chd1 generated the shortest (12-13 bp) and ISW1a a bit longer (21-26 bp) linker 1 lengths without significant effects by nucleosome density ( Figure 2D ,E). ISW2 generated rather constant spacing (54-58 bp) at low and medium but tighter spacing (38 bp) at high density. For INO80, linker lengths steadily increased with decreasing density from 33 to 82 bp. We concluded that linker lengths and their dependencies on nucleosome density were remodeler-specific and interpreted this as follows. Spacing activity of a remodeler has two aspects. On the one hand, the remodeler equalizes linker lengths leading to regularity in arrays, which is the classical definition of spacing activity (Ito et al., 1997; Varga-Weisz et al., 1997) . On the other hand, the resulting linkers have a certain length. In our purified system, this may either be determined by nucleosome density and/or by a remodeler-intrinsic feature. Following (Yamada et al., 2011) , we call a remodeler feature that sets nucleosome spacing a "ruler". We use this term also for the feature that sets the distance to barriers (see below). Indicative for a remodeler ruler is remodeler-specific clamping, i.e., if constant spacing is generated at different nucleosome densities (= clamping) and different remodelers generate different spacing (= remodeler-specific), which shows that spacing depends on remodeler-intrinsic and not nucleosome-intrinsic properties (Lieleg et al., 2015) . We saw remodeler-specific clamping for Chd1 at all, for ISW1a at high versus medium and for ISW2 at medium versus low densities ( Figure 2C -E). As none of the remodelers with spacing activity can disassemble nucleosomes (Clapier and Cairns, 2009) and thereby affect nucleosome density, their rulers can only set their respective linker lengths if these are shorter than or equal to the density-determined linker length at equidistant nucleosome distribution. Accordingly, Chd1 and ISW1a set their rulerspecified linker lengths at all and ISW2 at medium and low densities. ISW2 had to generate shorter linkers at high density and INO80 either did not have a ruler or the ruler responded to changes in nucleosome density. In vitro mononucleosome assays suggested that INO80 requires at least 40 bp of nucleosome-free DNA for nucleosome sliding (Zhou et al., 2018) , while it generated 30 bp linkers in tri-nucleosomes (Udugama et al., 2011) . Here, at high nucleosome density, INO80 generated linkers of about 33 bp consistent with previous observations. We tried to enforce even tighter spacing by increasing nucleosome density. This did not decrease spacing and phasing distances but peak heights ( Figure S2B ,C), probably due to increased aggregation without effective increase in nucleosome density of soluble chromatin.