Comparing the Assembly and Handedness Dynamics of (H3.3-H4)₂ Tetrasomes to Canonical Tetrasomes

Single-molecule Instrumentation

The traces monitoring NAP1-assisted nucleosome and tetrasome assembly in real time via changes in extension and linking number, as well as any subsequent dynamics in linking number, were measured using the FOMT [27]. The torque measurements were carried out using the eMTT [28]. All measurements were performed at 21°C and acquired at an acquisition frequency of 100 Hz.

Protein expression and purification

Recombinant Drosophila core histones were expressed in E. coli BL21(DE3) Rosetta (Novagen) and purified as described in [29], with the distinction that the purification procedure for the H3.3-H4 tetramers was identical to that of the H2A/H2B dimers. Expression plasmids were a kind gift of J. Kadonaga. Concentrations of core histones were determined by SDS PAGE and Coomassie staining as well as calculated from A₂₈₀ measurements and the H3-H4 extinction coefficient (S1 Fig). Recombinant Drosophila NAP1 was purified according to [30].

Flow cell passivation and buffer conditions

In all experiments, we employed a buffer consisting of 50 mM KCl, 25 mM Hepes-KOH pH 7.6, 0.1 mM Ethylenediaminetetraacetic acid (EDTA), 0.025% Polyethylene Glycol (PEG) and 0.025% Polyvinyl Alcohol (PVOH) as crowding agents, and 0.1 mg/ml bovine serum albumin (BSA) both as crowding agent and for the prevention of nonspecific binding of histones to the surface. For the tetrasome assembly we used the histone chaperone NAP1. Although in vivo NAP1 is known as a histone chaperone for H2A and H2B, in vitro it has been shown that NAP1 assembles complete nucleosomes [7, 31–35]. The used protein concentrations were: 200 nM NAP1, 67 nM H3.3 and 67 nM H4 were preincubated for 30 min on ice. The pre-incubation buffer contained 50 mM KCl, 25 mM Hepes pH 7.6, 0.1 mM EDTA, 0.25% PEG, 0.25% PVOH and 1 mg/ml BSA. Just prior to flushing in, the protein concentration was reduced ~4000-fold. To achieve nucleosome assembly following tetrasome formation, 270 nM NAP1, 268 nM H2A, 268 nM H2B were preincubated for 30 min on ice. Just before flushing in, the protein concentration was reduced ~300-fold.

DNA constructs

We used 1.9 kilo-base-pair (kbp) double-stranded DNA (dsDNA) molecules in the FOMT experiments and 3.4 kbp DNA molecules in the eMTT experiments, both without positioning sequences (sequence available in the S1 File). To attach the DNA molecules to the glass surface and the bead, their extremities contained multiple digoxigenin molecules at one end and multiple biotin molecules at the other end. The DNA molecules did not contain any nucleosome-positioning sequences. In the FOMT experiments, we used 0.5 μm diameter beads (Ademtech) and in the eMTT experiments we used 0.7 μm diameter beads (MagSense).

NAP1-assisted assembly of (H3.3-H4)₂ tetrasomes

We directly monitored tetrasome formation upon flushing in H3.3 and H4 pre-incubated with the histone chaperone NAP1 into the flow cell using FOMT, a technique that allows one to simultaneously measure dynamical changes in the end-to-end length and linking number of single tethered DNA molecules. In this approach, a vertically oriented magnetic field is used to apply a stretching force, without constraining the free rotation of the DNA molecule (Fig 2A). The DNA molecules employed (1.9 kbp in length) did not contain specific nucleosome-positioning sequences. We limited the applied stretching force to 0.8 pN, well below the 3 pN above which DNA begins to peel off from the nucleosome [36]. Upon flushing in NAP1/histone complexes, this experimental configuration allowed us to observe a distinct, stepwise decrease in the end-to-end length z of the DNA, indicating compaction, accompanied by a clockwise rotation θ of the bead, reflecting a decrease in the linking number of the DNA tether (left panels in Fig 2B and 2C). From several independent (H3.3-H4)₂ assembly experiments, we obtained an average extension change <Δz> = -25 ± 6 nm (Fig 2B, right) and linking number change <Δθ _assembly> = -0.8 ± 0.2 turns (Fig 2C, right). By changing the histone concentration, we assembled varying numbers of tetrasomes per DNA molecule. The total degree of compaction Δz and the overall change in linking number Δθ _assembly following assembly were found to be linearly correlated with a slope Δz/Δθ _assembly of 32 ± 2 nm/turn (Fig 2D).

These single-molecule assembly experiments revealed that NAP1 is capable of assembling (H3.3-H4)₂ tetrasomes. The obtained average extension change <Δz> = -25 ± 6 nm for assembly of (H3.3-H4)₂ tetrasomes agrees well with previous results on canonical tetrasomes and nucleosomes [7, 35, 36], indicating that the alterations in the histone fold domain of H3.3 do not affect the overall tetrasome structure formed. The linear correlation between the total degree of compaction Δz and the overall change in linking number Δθ _assembly following assembly indicated that the conformation of the tetrasomes on the DNA is independent of the number of protein complexes assembled, excluding large effects of inter-tetrasomal interactions.

(H3.3-H4)₂ tetrasomes exhibit spontaneous dynamic changes in linking number

We also carried out a separate set of experiments to determine whether there were further dynamics to be observed on the single-molecule tethers following assembly. For reference, bare DNA has a constant length z and linking number θ, apart from Brownian fluctuations (Fig 3A and 3B). Following the assembly of a single (H3.3-H4)₂ tetrasome (which compacted the DNA by 22 ± 1 nm), the resulting reduced DNA length z stayed constant (Fig 3A and 3C left panels). Concomitant with the step in z, a step in the linking number of -0.81 ± 0.25 turns occurred (Fig 3A and 3C right panels). However, subsequently the linking number did not stay constant but was rather observed to change between -0.80 ± 0.05 and +0.86 ± 0.39 turns (Fig 3A and 3C right panels). From this, we concluded that (H3.3-H4)₂ tetrasomes exhibited spontaneous flipping between a preferentially occupied left-handed state and a right-handed state in a manner that left the DNA extension unchanged. On average, the (H3.3-H4)₂ tetrasome showed spontaneous fluctuations in the linking number with a typical ΔL _k = 1.68 ± 0.14 turns, a mean change in linking number very similar to that observed for the canonical tetrasome [7].

To determine whether this flipping tetrasome could accommodate the assembly of a complete nucleosome, we subsequently added histones H2A-H2B (Methods). This led to a decrease in the mean linking number by 0.55 ± 0.25 turns in a single step, together with an arrest of the flipping behavior (Fig 3D, right panel). The total amount of compaction due to the assembly of the H2A-H2B was 31 nm. Adding histones H2A-H2B thus led to the assembly of a left-handed nucleosome with a total linking number of -1.36 ± 0.2 turns and total compaction of 54 ± 7 nm. We find that the handedness of nucleosomes containing only H3.3 remained stable.

We observed flipping signatures in the linking number for every DNA molecule that was loaded with H3.3-containing tetrasomes (but never for bare DNA nor for nucleosome-loaded DNA). A second example of a single tetrasome is shown in Fig 4A, whereas the behavior of four assembled tetrasomes, together with cartoons illustrating the number of tetrasomes in the left- and right-handed configuration, is shown in Fig 4B. To exclude any potential effect of NAP1 on the dynamics of (H3.3-H4)₂ tetrasomes, we also performed an experiment in which we removed the free proteins. Under these conditions, DNA molecules with tetrasomes, both for conditions with (black) and without (grey) free proteins in solution, displayed similar-sized angular steps between the discrete levels, <Δθ _flipping > = 1.6 ± 0.1 turns (Fig 4C). These results, taken together with a similar independence of flipping dynamics on NAP1 observed for canonical tetrasomes assembled by salt dialysis [7], leads us to conclude that NAP1 does not induce the change in handedness of the (H3.3-H4)₂. Collectively, these experiments show that the inherent flipping behavior of the handedness of tetrasomes is not limited to H3-containing tetrasomes, but also applies to tetrasomes that contain H3.3. Moreover, they demonstrate that the dynamically flipping (H3.3-H4)₂ tetrasome is a viable intermediate in the assembly of stable, left-handed, nucleosomes.

The flipping of the canonical tetrasomes loaded onto DNA by NAP1 can be analyzed in the framework of a binomial model in which a single tetrasome occupies either the left- or right-handed state with probabilities p and (1-p), respectively [7]. A value of p close to 1 indicates that tetrasomes are much more likely to occupy the left-handed state over the right-handed state, whereas a shift towards lower values of p indicates a more positively wrapped tetrasome. For each experiment, we determined the relative occupancies of each state from the ratios of the respective peak areas in the linking number histograms. We fitted p for distinct DNA molecules loaded with different numbers of tetrasomes resulted in a <p> = 0.91 ± 0.03 (N = 12) in the presence of free proteins (Fig 4D, dark blue crosses). Using ΔG = -k _B T ln (1/p– 1) to compute the free energy difference between the left- and right-handed states, we deduced a free energy difference between the left- and right-handed states of 2.3 ± 0.4 k _B T (D, dark red squares), similar to the 2.3 k _B T value found for canonical tetrasomes and the 2.5 k _B T value determined via electrophoretic mobility analysis of nucleosome populations [5, 7]. We note that that we measured a slightly reduced probability for the occupancy of the left-handed state <p> = 0.84 ± 0.09 (N = 7) after flushing out free proteins (Fig 4D, green plus signs), corresponding to a decreased free energy difference between the states of 1.6 ± 0.8 k _B T (Fig 4D, filled pink circles). Flushing out of the proteins thus mildly increases the probability to occupy the right-handed state of the (H3.3-H4)₂ tetrasome. This finding, together with the observation that Δθ _flipping is unaffected by the removal of free proteins, suggests that NAP1 may stimulate the left-handed wrapping slightly while leaving the linking number of the left- and right-handed states unchanged.

Minute torques can drive structural transitions within (H3.3-H4)₂ tetrasomes

We next studied the response of (H3.3-H4)₂ tetrasomes to physiologically relevant applied torques [37] at an applied stretching force of 0.8 pN by using eMTT [28] (Fig 5A). For these experiments, we utilized 3.4 kbp DNA (again without specific nucleosome-positioning sequences) loaded with tetrasomes by NAP1. Reference measurements on bare DNA showed that the application of turns to torsionally relaxed bare DNA initially left the DNA extension unchanged as the DNA twist increased, resulting in a linear build-up of torque (black squares, Fig 5B). At a critical buckling torque, a decrease in the DNA extension z was observed as DNA buckled to form plectonemic supercoils, and beyond this, no further torque build-up occurred (plateau in black squares for > 6 turns and < -6 turns in Fig 5B). The torque response following NAP1-mediated assembly of ~5 (H3.3-H4)₂ tetrasomes (deduced from the total length decrease of 135 nm given the average length decrease of 25 nm per tetrasome, Fig 2B) is shown in Fig 5B. Starting at positively induced supercoiling, the torque response was first measured from +17 turns to -17 turns (red triangles). Consecutively, the measurement direction was reversed (green diamonds). At the center of both torque response curves, a plateau at nearly zero torque was clearly visible. In this region, the induced turns did not lead to build up of twist (and hence torque) in the tethered molecule; instead, changes in the tetrasome conformation likely occurred that prevented such build-up. Alternately stated, a negligibly low torque could be used to drive a tetrasome into a left-handed configuration (when negative turns were imposed) or into a right-handed configuration (when positive turns were imposed). The widths of the near-zero torque plateaus for the negative (red) and positive (green) rotation directions were 7.5 ± 1.0 and 5.9 ± 1.0 turns, respectively (Fig 5B). Therefore the ∆L_k/tetrasome in the plateaus is 1.5 and 1.2 for the negative and positive rotation direction respectively. Once a sufficient number of turns was applied to force all tetrasomes to occupy either left- or right-handed states, torque build-up ensued as in the case of bare DNA. Finally, saturation of torque build-up occurred beyond a torque of +10 (-12) pN nm (accompanied by a concomitant decrease in extension, consistent with plectoneme formation), also similar to the case of bare DNA.

We briefly examine the regions in which torque is built up for DNA loaded with tetrasomes (Fig 5B). In all cases, these slopes are shallower than those measured for bare DNA; this could reflect gradual changes in the conformation of the loaded tetrasomes. An example of this would be a change in the angle of the tetrasome’s entry and exit DNA. Additionally, the torque response curves display hysteresis: neither the slopes of the torque response nor the widths of the plateaus around zero rotations are identical upon reversal of the direction of rotation. From the constant length of the molecule (data not shown), we can conclude that this hysteresis is not induced by tetrasome dissociation/rebinding events. Instead, it appears that the induced conformational changes from right- to left-handed tetrasomes is more gradual than vice versa. For example, for the molecule shown in Fig 5we expect that ~5 tetrasomes have been assembled as deduced from the length decrease upon assembly. We therefore expect the width of the zero-torque plateau to comprise 5 x (θ _flipping =) 1.7 = 8.5 turns Fig 4C). Given the measured plateau widths of 7.5 turns (from negative to positive rotations) and 5.9 turns (from positive to negative rotations), it appears that 0.2–0.5 turns per tetrasome are absorbed by more gradual conformational changes that occur as the magnitude of the torque in the tetrasome-loaded DNA is decreased. Summing up, these experiments directly demonstrate that the application of only very weak positive torques can drive tetrasomes from left- into right-handed states. Furthermore, the torque response displays hysteresis as a function of the direction of rotation, indicating that the sudden conformational changes as shown by handedness flipping at low torques (e.g., conformational change at the H3.3-H3.3 interface as suggested in (18)) are accompanied by more gradual conformational changes (e.g., due to slight changes in direction of the entry and exit DNA) under the influence of torque) at increased levels of torque.