## Introduction

In eukaryotes, genomic processes such as DNA replication, transcription and repair are strongly influenced by chromatin architecture. The fundamental constituent of chromatin is the nucleosome, which comprises 147 base pairs of double stranded (ds)DNA wrapped ~1.7 times around a histone octamer, resulting in an inner and an outer turn of the DNA1,2. The canonical histone octamer consists of four homodimers of the histone proteins H2A, H2B, H3 and H4 and binds to the DNA via strong electrostatic interactions, occluding one complete face of the DNA helix from its environment1. Consequently, nucleosomes greatly affect how accessible the genome is for proteins to bind. It is therefore of great importance for genomic processes that the organization of nucleosomes is actively regulated. In vivo, this is enzymatically regulated by histone-modifying enzymes and ATP-dependent chromatin remodelers3,4. The latter can alter chromatin architecture through different mechanisms, including nucleosome sliding5, nucleosome unwrapping6, nucleosome eviction (i.e., removal of the histone octamer)7 and exchange of histone dimers8.

ATP-dependent chromatin remodelers belong to the SNF2 family of proteins and share a highly conserved ATPase domain9, which facilitates binding to the minor groove of dsDNA10. Through a cycle of ATP binding and hydrolysis, the domain alternates between closed and open states, resulting in helical translocation of the protein along the DNA minor groove5. This translocation activity is thought to be the key mechanism of nucleosome remodeling. Although histone-DNA interactions could, in principle, be disrupted directly by collision between a moving translocase and a static nucleosome, in many cases translocation-induced remodeling is known to also involve the generation of torsional stress10,11. This usually occurs when the translocating remodeler encounters a ‘roadblock’ that hinders its helical translocation along the DNA. As a result, a DNA loop is extruded, which destabilizes nucleosomes indirectly by altering the DNA conformation. Roadblocks occur when the remodeler undergoes dimerization with another monomer12, or in the form of other protein-protein or protein-nucleosome interactions10. To facilitate loop extrusion, many SNF2 proteins carry accessory domains that allow for protein-protein interactions and/or nucleosome binding13. In addition, these accessory domains are important for recruitment of other co-factors, such as actin-related proteins, specific histone variants or DNA-repair proteins13.

While many SNF2 remodelers are primarily involved in the regulation of gene access and transcription9, the role of the CSB/ERCC6 subfamily of remodelers in DNA-repair processes has recently received increased attention14,15,16,17,18,19,20,21,22,23,24,25. An intriguing example of such a protein is PICH, also known as ERCC6L, which plays a key role in the dissolution of ultrafine anaphase bridges (UFBs) during mitosis16,17,25. UFBs arise from sister chromatid entanglements and become apparent during anaphase when the mitotic spindle starts pulling the chromatids to opposite poles of the cell25,26. To ensure faithful sister chromatid segregation, it is essential that UFBs are resolved in a timely fashion27,28. PICH is thought to initiate the process of UFB dissolution both by stimulating topoisomerase TopoIIα17,29 and by recruiting the BTRR dissolvasome17,26,29,30,31,32. Furthermore, bulk biochemistry and single-molecule studies have demonstrated that PICH displays many of the hallmark capabilities of chromatin remodelers, such as DNA translocation16, DNA looping33,34 and mono-nucleosome remodeling30. Although the DNA translocation and looping properties of PICH are similar to those of other SNF2 remodelers16,33,34,35,36, its nucleosome remodeling activity has been reported to be orders of magnitude less efficient than that of other SNF2 remodelers16.

Another difference from many other SNF2 proteins is that PICH mainly associates with the genome during mitosis, i.e., after breakdown of the nuclear envelope25,29,37, when chromosomes are highly compacted. Curiously, while PICH exhibits a strong affinity for bare DNA16, it lacks a nucleosome-binding domain25, which suggests that it is unable to bind to mitotic chromosomes. It has therefore been hypothesized that the binding of PICH to chromatinized bridges (such as arising UFBs) only commences when chromatid entanglements result in a build-up of tension on the chromatin (via pulling forces of the mitotic spindle)38,39,40. This would result in local nucleosome unwrapping16, exposing small regions of dsDNA to which PICH could bind, and thus act (indirectly) as a sensor for mitotic chromatin under tension. Subsequently, PICH could generate more bare dsDNA through tension-dependent nucleosome remodeling, which, in turn, would allow for a recruitment of repair complexes such as BTRR or TopoIIα16. This model would explain the weak chromatin-remodeling activity of PICH in ensemble biochemistry assays since the model implies nucleosome remodeling under tension, which cannot be emulated in ensemble chromatin-remodeling assays. Furthermore, the proposed mechanism of action of PICH would allow for specific remodeling of anaphase bridge-associated nucleosomes, while maintaining the integrity of non-entangled chromatin. However, to date, due to the limitations of ensemble chromatin remodeling assays, there is no direct experimental evidence to support this model.

Here, we employ a combination of dual-trap optical tweezers and confocal fluorescence microscopy to determine if PICH exhibits tension-dependent chromatin remodeling activity. To this end, we apply force to a dsDNA construct containing an array of nucleosomes to mimic chromatin under tension (such as is the case with anaphase bridges). In this way, we reveal that PICH can invade nucleosome arrays at tensions ≥3 pN, and can efficiently facilitate nucleosome unwrapping at tensions of ~5–10 pN in an ATP-dependent manner. Moreover, after nucleosome unwrapping has occurred, we observe that PICH can slide remaining histones along the DNA. We propose an updated model for the role of PICH in UFB dissolution, whereby PICH induces unwrapping of the inner turn of nucleosomes and subsequently slides histones to generate bare dsDNA. This allows additional PICH molecules to bind, providing an efficient substrate for TopoIIα and/or the BTRR dissolvasome to bind to, and thus promoting the timely dissolution of the UFB.

## Results

### PICH is able to invade nucleosome arrays at tensions ≥ 3 pN

We utilized combined dual-trap optical tweezers and confocal fluorescence microscopy to determine whether PICH exhibits force-dependent chromatin-remodeling activity. To this end, we reconstituted nucleosomes on a DNA construct containing an array of 12 repeats of the 601 nucleosome positioning sequence41, each flanked by 25 base pairs of identical linker DNA42. The nucleosome array was located close to the center of a longer dsDNA molecule, such that the array was flanked by ~4 and ~4.5 kilo base pairs of dsDNA. These flanking DNA handles were biotinylated and thus allowed us to optically manipulate the construct using streptavidin-coated beads. To verify successful nucleosome positioning on our constructs, we first fluorescently labelled the nucleosome array with Anti-H3-Alexa647 antibodies. As shown in Fig. 1a, (i), this confirmed that reconstituted nucleosomes are located in the middle of the DNA construct. Next, to determine how PICH interacts with such a substrate, the nucleosome constructs were incubated (at 10 pN) in a microfluidic channel containing 25 nM PICH-eGFP and 2 mM ATP (Supplementary Fig. 1). As expected from the high affinity of PICH for dsDNA16, PICH was found to readily bind to these constructs (Fig. 1a, (ii, iii)).

Nucleosomes positioned in an array are well-known to form higher-order chromatin structures, via nucleosome-nucleosome interactions between neighboring octamers, which can be disrupted by applying tensions of ≥3 pN43,44. When monitoring the locations of PICH bound to the nucleosome-array constructs at lower tensions (e.g., ~1 pN), PICH was found predominantly on the nucleosome-free DNA handles (Fig. 1b, (i)) and not the nucleosome array itself. This is consistent with PICH lacking a nucleosome-binding domain and thus a specific affinity for nucleosomes. In addition, we excluded the possibility that the relative absence of PICH in the nucleosome array at ~1 pN was merely due to the anti-H3-Alexa647 antibody blocking its access (Supplementary Fig. 2). Moreover, our findings suggest that, in the absence of tension, nucleosome-nucleosome interactions hinder the ability of PICH to access the regions of linker DNA (i.e., the DNA between neighboring nucleosomes). In contrast, when forces of ≥3 pN were applied to the construct, PICH was observed to also bind to the region of the construct containing the nucleosome array (Fig. 1a, b, (ii)). We attribute this observation to the release of nucleosome-nucleosome interactions43,44, combined with a loss of nucleosome outer turns (of ~60 base pairs), which are well-known to start to unwrap at these forces (Fig. 1b, (ii), top)38,39,40. This results in increased accessibility of the linker DNA, allowing PICH to bind to these regions. The above observations support the previously proposed model in which PICH is only recruited to UFBs once the pulling forces from the mitotic spindle disrupt nucleosome-nucleosome interactions16.

### PICH facilitates tension-induced nucleosome unwrapping using ATP hydrolysis

Given our observation that PICH can bind to linker DNA regions at forces ≥3 pN, we next examined whether PICH can alter tension-dependent unwrapping of nucleosomes. To this end, we monitored nucleosome unwrapping at different constant forces, using a force clamp ranging from 3 to 15 pN, in the absence and presence of PICH and ATP. At these elevated tensions, we assume that the outer turn is already unwrapped and that we mainly detect unwrapping of inner-turn nucleosomes38,39,40. We identified unwrapping events by monitoring step-wise increases in the end-to-end length of the nucleosome-array constructs (Fig. 2a, b). In the absence and presence of PICH and ATP, the average step size was found to be 26.4 ± 0.2 nm and 25.8 ± 0.2 nm respectively, which is consistent with the unwrapping of the remaining inner turn of ~80 base pairs (Supplementary Fig. 3)2,42. Notably, in the presence of PICH and ATP, transient and reversible shortening events of the DNA were also observed, for all forces tested (Fig. 2b). These events occurred throughout the constant-force measurements, independent of the presence of nucleosomes, and are indicative of the loop-extrusion activity of PICH that has previously been reported on bare DNA under tension by Bizard et al.34. This observation of loop extrusion, which is likely the result of oligomerization-induced roadblocks12, is in agreement with the previously reported oligomerization of PICH16 and the observation that the majority of PICH appeared to be present as oligomers during our constant-force measurements (Supplementary Fig. 4). Nevertheless, unwrapping events could still be determined, as the extruded loops persist only temporarily (Fig. 2b and Methods).

If PICH would assist nucleosome inner-turn unwrapping, we would expect that unwrapping occurs, on average, at lower forces in the presence of PICH and ATP than in their absence. To investigate this, we monitored the number of unwrapping events that occurred within 10 min under constant tension (Fig. 2a, b and Methods). The number of nucleosomes that had not yet unwrapped was then determined by stretching the construct and counting the number of force ruptures corresponding to the unwrapping of remaining nucleosomes (Fig. 2c)2,42. The unwrapped fraction was defined as the number of unwrapped nucleosomes (within the 10 min duration under constant tension, Nunwrapped) divided by the total amount of nucleosomes (unwrapped + remaining nucleosomes, Ntotal) (Eq. 1).

$$\alpha=\frac{{N}_{{{{{{\rm{unwrapped}}}}}}}}{{N}_{{{{{{\rm{total}}}}}}}}$$
(1)

As expected from previous studies2,43,45,46, the fraction of unwrapped nucleosomes increased as a function of applied force. More importantly, we observed larger unwrapped fractions in the presence of PICH and ATP than in its absence (Fig. 2d). To investigate whether the PICH-induced stimulation of inner-turn unwrapping of nucleosomes under tension is dependent on ATP hydrolysis, we repeated the above experiment at a force of 10 pN using AMP-PNP, a non-hydrolysable analog of ATP. Note that, in this case, the PICH concentration had to be increased by at least four-fold (100 nM instead of 25 nM) in order to achieve a similar coverage of PICH on the nucleosome-array construct as in the presence of ATP, indicating a stimulatory effect of ATP on PICH binding to DNA (Supplementary Fig. 5). Nevertheless, when using AMP-PNP instead of ATP, PICH did not increase the unwrapped fraction (+0.002, with a 95% confidence interval (CI) ranging from −0.16 to +0.17). In contrast, in the presence of ATP, a relative increase of +0.48 was observed (with 95% CI of +0.37 to +0.58) (Fig. 2d). Together, the above findings demonstrate that the PICH-induced stimulation of nucleosome inner-turn unwrapping is an active process and therefore rule out that the increase in the probability of tension-induced unwrapping is induced passively by binding, as has been recently demonstrated for HMGB proteins46.

### ATP-dependent PICH activity reduces the lifetime of nucleosomes under tension

In order to assess the kinetics of nucleosome unwrapping under tension in the absence and presence of PICH and ATP, we determined the lifetime of inner-turn nucleosomes. We define this lifetime as the duration that an individual nucleosome on the array remains wrapped within a 10 min period at a constant force (Fig. 3a and Methods). It should be noted that we were not able to determine reliable lifetimes at tensions below 10 pN due to the low fraction of unwrapped nucleosomes in the absence of PICH (Fig. 2d), and extensive loop-extrusion in the presence of PICH and ATP. We therefore only quantified and compared nucleosome lifetimes at 10, 12.5 and 15 pN (Fig. 3a). In total, 139 and 186 unwrapping events were collected in the absence and presence, respectively, of PICH and ATP, and these were used to calculate cumulative probability distributions (CPDs) (Fig. 3b, c). As mentioned above, in the absence of PICH, at 10 and 12.5 pN, not all nucleosomes unwrap within a 10 min period (Fig. 2d), resulting in remaining fractions of 0.48 and 0.07, respectively. We accounted for this by normalizing the distributions from the remaining fraction to 1.

As expected from the previously determined unwrapped fractions (Fig. 2d), this analysis confirmed that nucleosomes unwrap faster at increasing forces (Fig. 3b). Furthermore, at all tensions tested, the presence of PICH and ATP significantly reduced the nucleosome lifetimes (Fig. 3c). Remarkably, we found that the CPDs of the lifetimes do not follow a mono-exponential decay, but can best be described by bi-exponential decay functions, which we attribute to the presence of two distinct nucleosome populations; with 74 ± 1% belonging to a slower unwrapping population and the remainder to a faster population. This behavior was independent of the presence of PICH and ATP (slower populations of 76 ± 1% in the absence and 72 ± 1% in the presence of PICH and ATP). The faster population unwraps more than two orders of magnitude quicker than the slower one at a force of 10 pN (Supplementary Fig. 6), indicating that it represents a subset of nucleosomes with a weaker histone-DNA interaction. We account for this observation by assuming the presence of tetrasomes in our assay. Tetrasomes comprise a nucleosome-intermediate, consisting of a H3/H4 tetramer (but no H2A/H2B dimers) and a single dsDNA wrap, and are known to be highly dynamic47. Furthermore, it is known that tetrasomes typically constitute ~20–40% of assembled histone complexes in reconstituted nucleosome arrays48,49,50. In line with these earlier observations, we speculate that our minor, faster population could reflect the unwrapping of tetrasomes.

Our aim was to investigate the interplay between PICH and canonical (i.e., properly assembled) nucleosomes. We therefore focused on the nucleosome lifetimes of the major and more stably bound population. The influence of PICH on these nucleosome lifetimes depended on the applied force and was most pronounced at 10 pN (the lowest force we were able to examine), with a 6.1-fold reduction of the mean lifetime (from 874 ± 51 to 143 ± 10 s). This decrease in lifetime was absent when AMP-PNP was used instead of ATP (Fig. 3c, (i)), consistent with our earlier finding that ATP hydrolysis is required to stimulate unwrapping. At higher tensions (12.5 and 15 pN), the effect of PICH on the mean nucleosome lifetime was less pronounced, with a 3.5- and 1.7-fold reduction, respectively (Fig. 3d). The decreased effect of PICH on nucleosome lifetimes at higher tensions can be explained by the intrinsic effect force has on nucleosome lifetime: sufficiently high tensions (≥15 pN) result in fast nucleosome unwrapping, even in the absence of PICH (Fig. 3b, d)2,45.

### PICH lowers the force required to achieve inner-turn unwrapping

The presence of a minor, faster population (assumed to be tetrasomes) has major implications for the previously determined fractions of unwrapped nucleosomes shown in Fig. 2d. As we are interested only in canonical nucleosomes, we corrected the unwrapped fractions such that they comprise only canonical nucleosomes (see Methods). This allowed us to extract additional quantitative information regarding inner-turn nucleosome unwrapping. Hence, we describe this process as a transition comprising two states (wrapped versus unwrapped). The force dependence of the unwrapping probability P in the absence and presence of PICH and ATP can then be fitted with an Arrhenius-type equation (Eq. 2).

$$P(F)=\frac{1}{1+\,{e}^{-\frac{\varDelta x}{{k}_{B}T}\left(F-{F}_{1/2}\right)}}$$
(2)

Here, Δx reflects the effective distance to the transition state and kBT is the thermal energy, while F and F1/2 correspond to the applied and half force, respectively. The latter reflects the average force at which nucleosomes undergo inner-turn unwrapping.

When the above function was fitted to the corrected unwrapped fractions, it revealed that PICH lowers F1/2 by about two-fold (from 10.6 ± 1.7 pN in the absence of PICH and ATP to 5.5 ± 0.8 pN in its presence, Fig. 3e). The value of F1/2 in the absence of PICH is in good agreement with the ~8.9 pN reported previously51. Furthermore, the fit yields similar Δx values of ~5 nm for the absence (5.7 ± 0.7 nm) and presence (4.4 ± 0.5 nm) of PICH and ATP. From these values the rate-limiting free energy barrier associated with inner-turn unwrapping, $${\Delta G}^{{{\ddagger}} }$$, can be calculated52: $${\Delta G}^{{{\ddagger}} }=\,{F}_{1/2}\,\varDelta x$$, yielding ~15kBT in the absence and ~6kBT in the presence of PICH and ATP. Again, the $${\Delta G}^{{{\ddagger}} }$$ in the absence of PICH is consistent with previously reported values of ~15–16 kBT51,53,54. Our analysis thus reveals that, under our experimental conditions, the presence of PICH and ATP lowers the rate-limiting energy barrier associated with inner-turn unwrapping by ~9 kBT; i.e., by more than 50%. The influence of PICH on the force-dependent probability of unwrapping can also be demonstrated graphically by subtracting the probabilities for the absence and presence of PICH (Fig. 3f). The force range within the full width at half maximum of this curve indicates that PICH facilitates unwrapping of nucleosomes most efficiently at forces between 5.3 and 10.7 pN. The peak of this curve additionally indicates that the effect of PICH on nucleosome unwrapping is strongest at forces of ~8.2 pN. At this tension, PICH increased the probability of unwrapping (within 10 min) by more than 90% (from ~4% to ~95% in the absence and presence of PICH and ATP, respectively). Taken together, these data suggest that, under our experimental conditions, PICH displays a maximum activity of force-induced nucleosome unwrapping in the force range of ~5–10 pN.

### PICH translocation can result in histone sliding along the DNA

We have shown previously that, in the absence of PICH, histones (at least histone H3) often remain bound to the DNA after force-induced nucleosome unwrapping has occurred42. In the present study, this binding stability of histones was confirmed by non-specifically labelling amine groups of the histone proteins bound to the nucleosome-array constructs with Atto-647N (see Methods). Although a decrease in the intensity of Atto-647N fluorescence was often observed during nucleosome unwrapping, substantial fluorescence signal persisted long after nucleosomes had unwrapped. To examine if PICH can affect the binding stability of histones, we next studied the interaction of PICH-eGFP with the Atto-647N-labelled nucleosome-array constructs. The correlation of fluorescence and nucleosome unwrapping data revealed that also in the presence of PICH and ATP, histones often remained bound to the DNA for at least several minutes after nucleosome unwrapping (Fig. 4a). This indicates that nucleosome unwrapping does not result in complete histone eviction, even in the presence of PICH. Notably, these results suggest a difference in the underlying mechanisms of nucleosome remodeling between PICH and the structurally related SNF2 protein Rad549, for which direct evidence of histone eviction has been recently reported55.

Despite the lack of complete histone eviction by PICH, we did observe an effect on the behavior of the remaining histones. Whereas in the absence of PICH we found that histones remained immobile after nucleosome unwrapping (N = 21 constructs, 10–12.5 pN), in the presence of PICH and ATP, displacements of histones were occasionally observed (N = 3/18). These histone displacements only occurred when there was a clear co-localization of histone and PICH fluorescence, and seemed to halt after collision with other bound PICH proteins (Fig. 4b). This suggests that PICH can slide histones along the DNA in an ATP-dependent manner and that histone displacements over larger distances might be possible with a less dense PICH coating (where the number of PICH proteins acting as roadblocks is minimized). We tested this idea by reducing the PICH coating (7 ± 1 bound PICH proteins) and monitoring the interaction of PICH-eGFP with the Atto-647N-labelled histones at a constant tension of 12.5 pN. Under these low PICH-coverage conditions, continuous PICH-mediated histone sliding (typically over hundreds of nanometers) was observed in 65% of constructs (N = 11/17). In the majority of these cases, all nucleosomes had unwrapped before the onset of histone sliding (N = 9/11). In the other cases, histone sliding was observed when only a single inner-turn nucleosome remained wrapped. We also note that the observation of a histone sliding event does not necessarily mean that all histones present on the construct are being translocated by PICH. This is substantiated by the observation that often the Atto-647N signal was split: some of the labelled histones were displaced by PICH, while the rest remained stationary (Fig. 4c). Taken together, these results suggest that PICH only slides fully unwrapped histones.

Similar to the high-PICH coverage condition, histone sliding was only observed in cases where clear co-localization of the fluorescence signals of PICH and histones was observed (N = 11), which is consistent with the notion that SNF2 proteins typically slide nucleosomes through collision5,7,55. As determined from the fluorescence intensities, in the majority of cases histone sliding was driven by a single PICH protein that began translocating when bound between, or directly adjacent to, the labelled histone(s) (Fig. 4c). In other cases, the translocating PICH appeared as an oligomer, consisting of either two or three proteins, and sometimes, histone sliding resulted from PICH that was bound to the nucleosome-free DNA handle and translocated toward the histones (Fig. 4d). Dissociation of a PICH-histone complex from the DNA was observed only once in eleven trajectories during the tested timeframe. This is consistent with the strong affinity of PICH for DNA, as observed by Biebricher et al.16. Clear dissociation of PICH from a histone was not observed in any of the eleven trajectories. A more detailed analysis of the trajectories of histone sliding (Fig. 4e, (i) and Methods) revealed that most consisted of episodes of directed translocation (defined as ≥3 nm s−1, N = 28 episodes) interspersed by pausing events (<3 nm s−1, N = 6 episodes). In almost all cases, the histones and PICH co-translocated, with or without pauses, in a single direction until a roadblock (additional PICH or a bead) was reached (Fig. 4e (i)). The mean histone sliding velocity was found to be 10.1 ± 0.8 nm s−1. This velocity is similar to that reported for PICH alone (9.8 ± 0.2 nm s−1)16, suggesting that the translocation rate of PICH is not reduced by sliding histone(s). Moreover, we detected no significant difference between the velocities of PICH-histone trajectories involving PICH monomers and oligomers, respectively (9.2 ± 0.7 versus 11.9 ± 2.0 nm s−1, Fig. 4e, (ii)). Together, these results indicate that after nucleosome unwrapping, ATP-dependent translocation of PICH monomers and oligomers can result in sliding of histones along the DNA, without affecting the translocation rate of PICH.

## Discussion

By exploiting the combination of dual-trap optical tweezers, confocal fluorescence microscopy and in vitro reconstituted nucleosome-array constructs, we reveal that the SNF2 protein PICH is a nucleosome remodeler that is force dependent, and thus optimized for its specific cellular role. We demonstrate that PICH can invade nucleosome arrays and use the free energy of ATP hydrolysis to facilitate inner-turn unwrapping at tensions of ≥3 pN. Moreover, we find that, after nucleosome unwrapping, collision between a translocating PICH and unwrapped histones can result in histone sliding along the DNA.

The observation that PICH is only able to invade nucleosome arrays when they are under tension provides additional evidence for the previous predictions that, in vivo, PICH acts as a (indirect) sensor for chromatin under tension16. Thus, our results support the notion that PICH will only bind to sister chromatid entanglements when slightly elevated forces (exerted via the mitotic spindle) locally destabilize the chromatin, yielding small regions of nucleosome-free dsDNA. This efficient, early recognition of chromatin bridges by PICH would furthermore provide an explanation for the in vivo observation that, already in early mitosis, PICH can be observed on very short (i.e., weakly stretched) chromatin bridges26. Based on our analysis of the probability of inner-turn nucleosome unwrapping and mean nucleosome lifetimes, PICH facilitates unwrapping most efficiently at forces between ~5–10 pN. As forces acting on separating sister chromatids have been reported ranging over three orders of magnitude56,57,58, we are unable to compare this observed optimal force range for PICH to forces present on UFBs in vivo. However, our results do raise the possibility that PICH might encounter a force range of ~5–10 pN in vivo.

SNF2 proteins generally display two different mechanisms for the coupling of ATP hydrolysis to nucleosome remodeling: via DNA-translocation induced collision or indirectly through DNA loop extrusion and the accompanying torsional stress10,11. While our current assay does not allow us to definitely answer which mechanism PICH uses to unwrap nucleosomes, it is likely that ATP-dependent loop extrusion plays an important role for the following two reasons. First, during the PICH-induced destabilization of nucleosomes under tension, we observed extensive loop extrusion. Second, the finding that the collision-induced sliding of histones typically happened minutes after unwrapping suggests that PICH-induced nucleosome unwrapping is not simply caused by direct contact between PICH and the nucleosome, but is the result of an indirect long-range effect, such as loop extrusion. Therefore, we propose that, in vivo, after binding to the emerging bare dsDNA binding sites on the arising UFB (Fig. 5, (i)), PICH oligomerizes and utilizes ATP hydrolysis to promote nucleosome unwrapping via loop extrusion (Fig. 5, (ii)). Previous studies have shown that nucleosome inner-turn unwrapping coincides with the dissociation of H2A/H2B dimers49,59,60, which is consistent with the in vivo observation that PICH-coated UFBs are devoid of histone H2B26,29,30. H3 and H4 histones have, however, been found to be more resistant to dissociation from the DNA upon nucleosome unwrapping in vitro42,49,59. This chimes with our findings that, even in the presence of PICH and ATP, many of the labelled histones remained bound long after nucleosomes had unwrapped. Together, this raises the possibility that UFBs in vivo may still exhibit bound H3 and/or H4 histone proteins (Fig. 5, (iii)).

Since every nucleosome unwrapping event will yield additional regions of free dsDNA, progressive unwrapping of nucleosomes on a UFB allows increasing numbers of PICH to bind (Fig. 5, (iii)). Consequently, the region of the entanglement (i.e., the catenane) will become bound by PICH, which is then able to slide the remaining histones (e.g., H3 and/or H4) (Fig. 5, (iv)). Histone sliding is generally thought to facilitate exposure of DNA segments5,7. Consequently, we postulate that the function of histone sliding by PICH is to enhance catenane accessibility, as has also been previously hypothesized by Ke et al.30. This would allow additional PICH proteins to interact with the catenane (Fig. 5, (v)), which, in turn, enables stimulation of TopoIIα and efficient recruitment of the BTRR dissolvasome localized at the DNA entanglement (Fig. 5, (vi)). Both of these processes are thought to play a key role in the timely dissolution of UFBs17,26 (Fig. 5, (vii)).

Our results have characterized PICH as a tension-dependent nucleosome remodeler that is uniquely adapted for initiating the dissolution of UFBs in vivo. More generally, our findings may have implications for the study of related remodelers. A two-step remodeling mechanism of loop extrusion-induced unwrapping followed by collision-induced histone sliding might also apply to other CSB/ERCC6 proteins. We note that a similar nucleosome remodeling mechanism was proposed for the subfamily member CSB, in which CSB actively wraps and unwraps DNA to remodel nucleosomes22. Furthermore, CSB was found to directly interact with histones18, which would be consistent with histone-sliding activity. Although direct evidence supporting our proposed two-step mechanism has not yet been found for related proteins, the versatile assay presented here could be used to study chromatin remodeling by other SNF2 proteins at the single-molecule level. We anticipate that future work will seek to unravel the similarities and differences between related SNF2 proteins, highlighting how their function is specialized for their distinct cellular roles.

## Methods

### Reagents

All chemicals were purchased from Sigma-Aldrich. Plasmids were purified using the QIAprep Spin Miniprep kit (Qiagen). Restriction digestion and biotinylation were performed in NEbuffer 3.1 (New England Biolabs). The restriction enzyme BglII was purchased from New England Biolabs. Klenow Fragment (3′−5′ exo-), biotin-14-dATP, biotin-14-dCTP, dTTP and dGTP were purchased from Thermo Fisher Scientific. Double stranded competitor DNA of 147 base pairs was purchased from Integrated DNA Technologies (for sequence, see ref. [42]). EDTA-free protease inhibitor tablets were purchased from Roche. Recombinant Human Histone Octamers (EpiCypher) were used for the reconstitution of nucleosomes. To visualize histones on the DNA, either recombinant monoclonal rabbit histone H3 antibody conjugated with Alexa Fluor 647 (referred to here as Anti-H3-Alexa647; RRID: AB_2663072, Lot #: VC294388, Thermo Fisher Scientific) or Atto-647N NHS ester (ATTO-TEC) was used.

### Preparation of nucleosome-array constructs

An Escherichia coli expression vector was used to express a pKYB1 plasmid containing 12 × 601 repeats. For plasmid construction, see Spakman et al.42. The plasmid was linearized by BglII and biotinylated using previously established protocols61. The restriction enzyme BglII ensured that the nucleosome positioning sites are located roughly in the centre of the construct (flanked by dsDNA handles of ~4 kilo base pairs and ~4.5 kilo base pairs, respectively). Nucleosome reconstitutions were performed by gradient salt dialysis using a peristaltic pump (Watson Marlow) and a 6–8 kD MWCA micro dialysis system (Hampton Research), following a procedure described by ref. [62]. Reconstitution was performed using different Human Histone Octamer:601 motif ratios, 116 ng/µL of the 12 × 601-pKYB1 plasmid (~200 nM 601 motifs), and 19 ng/µL of competitor dsDNA (~200 nM)42. Reconstituted nucleosome-array constructs were stored at 4 °C and were stable for at least 7 weeks.

### Preparation of recombinant PICH-eGFP

Recombinant human PICH-eGFP was expressed in Sf21 insect cells using pFastback-based baculoviruses and subsequently purified, essentially as described previously16,17. In brief, cells were infected for ~60 h, after which they were harvested at 4 °C and stored at −80 °C. For cell lysis, the pellet was resuspended on ice in lysis buffer (50 mM Tris-HCl pH 7.5, 500 mM NaCl, 10% glycerol, 50 mM Imidazole, 1 mM PMSF, 1 protease inhibitor tablet (PI)) and cells were dounce homogenized and sonicated. In order to remove cell debris, the lysate was cleared by centrifugation at 20,000 g for 1 h. The lysate supernatant was passed through a 1 mL HisTrap HP affinity column. Unbound protein was washed-out using three different wash buffers (WB) with varying salt concentrations (WB1–WB3). WB1 includes lysis buffer without PMSF and PI; WB2 includes WB1 with 1 M NaCl; and WB3 includes WB1 with 200 mM NaCl. Bound proteins were eluted (using WB3 with 500 mM imidazole) and fractions with PICH-eGFP were pooled and loaded onto a 1 mL HiTrap Heparin HP column. Protein was eluded using a linear salt gradient. Again, fractions with PICH-eGFP were pooled and, subsequently subjected to gel filtration chromatography on a 120 mL HiLoad 16/600 Superdex 200 column equilibrated in buffer (50 mM Tris-HCl pH 7.5, 200 mM NaCl, 10% glycerol, 0.1 mM EDTA, 1 mM DTT). The appropriate fractions were pooled and dialyzed against storage buffer (50 mM Tris-HCl pH 7.5, 200 mM NaCl, 20% glycerol, 0.1 mM EDTA, 1 mM DTT) for 6 h. The dialyzed protein was centrifuged at 20,000 g for an hour before snap freezing in dry ice and was stored at −80 °C in aliquots (12.5 µL of 1 µM). All the steps were done at 4 °C, unless otherwise stated.

### Labelling of histones on nucleosome-array constructs

To verify the location of nucleosomes on the DNA construct, ~0.7 nM nucleosome-array construct was incubated with 200 µg/mL Anti-H3-Alexa647 (1:5 dilution of 1 mg/mL stock) for 3 h in measurement buffer (see below) at 4 °C. As the labelling efficiency using Anti-H3-Alexa647 was low, for constant-force experiments we switched to labelling the histones on our construct with Atto-647N using NHS ester labelling42, which yields a much higher degree of labelling. Here, ~0.7 nM nucleosome-array construct was incubated with 200 µM Atto-647N NHS ester (1:10 dilution of 2 mM stock) for ~1 h at room temperature. The incubation buffer consisted of 20 mM Hepes-NaOH pH 7.5, 100 mM NaCl, 2 mM MgCl2, 0.02% (v/v) Tween-20 and 0.2% (w/v) BSA.

### Single-molecule experiments

A commercially available combined dual-trap optical tweezers and confocal fluorescence microscopy instrument (C-trap, LUMICKS B.V.) was used. The instrument was controlled using a custom-written program in LabVIEW 2011. Experiments were performed in a multi-channel laminar flow cell containing 5 separate channels (Supplementary Fig. 1). Channels 1 and 2 were used to optically trap streptavidin-coated polystyrene beads (1.76 μm diameter, Spherotech BV) and tether biotinylated nucleosome-array constructs, respectively63,64. This dumbbell-like construct could then be rapidly moved, via channel 3, to either channel 4 or 563,64,65. All channels contained the measurement buffer, which consists of 20 mM Hepes-NaOH (pH 7.5), 100 mM NaCl, 2 mM MgCl2, 0.02% (v/v) Tween-20, 0.2% (w/v) casein, 0.1 mM EDTA and 1 mM sodium azide. The presence of 0.2% casein in the measurement buffer ensured minimal dissociation/degradation of the nucleosomes from the DNA template, as previously shown for the blocking agent BSA42,66. Channels 4 and 5 were used to study nucleosome unwrapping in the absence and presence of PICH, respectively. To this end, the buffer in channel 5 additionally contained 25 nM PICH-eGFP, 2 mM ATP and 1 mM DTT, unless stated otherwise. For studying PICH-induced histone sliding, Atto-647N labelled nucleosome-array constructs were briefly incubated in the channel containing PICH (at a tension of 3 pN) until several PICH proteins had bound (as determined by visual inspection of the fluorescence images). These nucleosome-array constructs were then moved to channel 4, which in this case, was supplemented with 2 mM ATP and 1 mM DTT. The Force-Distance data were acquired at 50 Hz in all experiments. Data were obtained at room temperature.

### Fluorescence imaging

Fluorescence snapshots of the tethered nucleosome-array construct were generated by consecutive confocal scans along a line to generate a confocal image of 39 × 144 pixels. A pixel size representing 75 nm in the sample was used for all fluorescence imaging experiments. Kymographs were created by means of confocal line scans between the center of the two beads at each time point. Imaging settings for visualizing PICH-eGFP and histones (labelled with either Anti-H3-Alexa647 or Atto-647N) are summarized in Supplementary Table 1. Together, these settings allowed detection of single PICH-eGFP proteins, as determined by single photo-bleaching steps and blinking of the fluorophore. Since Atto-647N NHS ester is a non-specific dye that can covalently bind to any free amine in a protein, the intensity of the Atto-647N fluorescence signal could not be used to determine the number of histones on the nucleosome-array construct. Therefore, in our analysis, the Atto-647N fluorescence signal is only used as a probe for the presence of histones.

### Identification and characterization of inner-turn nucleosome unwrapping events

Distance-Time and Force-Distance curves were extracted using a custom-written LabVIEW program. A previously described MatLab-based (ML 2016b) step-fitting algorithm that detects horizontal segments in graphs67,68 was used to identify inner-turn nucleosome unwrapping events and obtain nucleosome lifetimes and step sizes from Distance-Time traces (obtained during constant-force measurements). All unwrapping events identified by the algorithm were visually verified by monitoring the end-to-end length of the nucleosome arrays (an increase in length of ~26 nm should be observed for each unwrapping event). To exclude transient PICH-induced shortening events, likely caused by loop-extrusion activity34, only steps that resulted in an overall increase in length were included in the analysis. The duration between the start of the constant-force measurement and the start of each horizontal segment determined the lifetime of a single nucleosome on the array. The vertical distance between two horizontal segments determined the step size. Step sizes of ≥40 nm were assumed to reflect double steps, where the inner turns of two nucleosomes are unwrapped simultaneously. Therefore, the associated lifetimes of these steps were counted twice. Step sizes indicating that three or more steps occurred simultaneously were not encountered (Supplementary Fig. 3). The obtained nucleosome lifetimes for each condition were subsequently fitted to a bi-exponential decay function using OriginPro 2019. To determine the number of nucleosomes that remained wrapped during the constant-force measurements, the Force-Distance curves (obtained by stretching the nucleosome-array construct up to at least 35 pN, after the constant-force measurement) were visually inspected to identify additional force-ruptures and ~26 nm shifts in DNA contour length, associated with inner-turn nucleosome unwrapping2,42. The number of nucleosomes that unwrapped under constant tension and those that remained wrapped were used to calculate the fraction of unwrapped nucleosomes (Eq. 1). To correct this fraction for the presence of tetrasomes, we subtracted the observed tetrasome population (of ~26%) from the previously determined unwrapped fractions α, using the following equation, αcorrected = (α – 0.26)/(1 – 0.26). A Python (3.7) algorithm was used to apply a logistic regression analysis to the unwrapped fractions in the absence or presence of PICH and ATP. This resulted in sigmoidal curves that approximate the force dependency of the probability of inner-turn nucleosome unwrapping. All errors represent a standard error of the mean (SEM), unless indicated otherwise. 95% confidence intervals (CIs) for the proportions and probabilities of nucleosomes that unwrapped during the constant force measurement were calculated using the Clopper-Pearson method69,70.

### Fluorescence image analysis

In order to extract the localization pattern of PICH on the nucleosome-array constructs, intensity profiles were made by measuring the background-corrected photon counts over a line scan between the two beads in the kymograph, averaged over the first 6 s of the constant-force measurement. To determine the coverage of PICH proteins on a nucleosome-array construct, and their oligomeric states, a Gaussian function was fitted to the average background-corrected intensity profile of a single PICH-eGFP monomer. For this, ≥50 intensity profiles originating from ≥5 different PICH-eGFP monomers were used for each set of imaging settings (Supplementary Table 1). The Gaussian-fitting was performed using OriginPro 2019. The obtained Gaussian distribution corresponds to the expected point spread function (PSF) of a monomer and was used to determine the Gaussian distributions of PSFs of apparent PICH-eGFP oligomers (up to pentamers). Next, for each nucleosome-array construct, background-corrected intensity profiles over line scans of the first 10 s of constant-force measurements were averaged. Then, by using the Peak Analyzer tool of OriginPro 2019, peaks of the obtained intensity profiles were fitted using width and area constraints derived from the Gaussian distributions of the PSFs of PICH-eGFP monomers and ‘apparent’ oligomers. Note that due to the diffraction-limited resolution of our fluorescence imaging setup, we cannot distinguish between true oligomers and monomers that are sufficiently close that their intensity peaks fully overlap. However, for our purposes, it is not essential to differentiate between true oligomers and neighboring PICH monomers. For this reason, sufficiently close PICH monomers are referred to as apparent oligomers. To determine the apparent oligomeric state of bound PICH-eGFP that is responsible for histone sliding, the integrated density of background-corrected intensity profiles (at N = 5 or N = 10 time points, depending on the length of the track) of these proteins were compared to the area under the curve of the PSF of PICH-eGFP monomers. Within all timescales where PICH-eGFP coating and oligomeric states were determined, the effect of photo-bleaching was negligible, as experimentally determined from individual bleaching traces. For estimating the average velocity of PICH-mediated histone sliding, trajectories of translocating PICH-eGFP and co-localized Atto-647N-labelled histones were manually tracked using the open-source software ImageJ (v1.53c). Segments of tracks with a velocity of <3 nm s−1 were defined as pauses, whereas segments with a velocity of ≥3 nm s−1 were defined as active translocation. All active translocation segments with a duration of ≥3 s were included in average velocity calculations. All errors represent an SEM.

### Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.