Spontaneous histone exchange between nucleosomes

The nucleosome is the fundamental gene-packing unit in eukaryotes. Nucleosomes comprise ∼147 bp DNA wrapped around an octameric histone protein core composed of two H2A-H2B dimers and one (H3-H4)2 tetramer. The strong yet flexible DNA–histone interactions are the physical basis of the dynamic regulation of genes packaged in chromatin. The dynamic nature of DNA–histone interactions also implies that nucleosomes dissociate DNA–histone contacts both transiently and repeatedly. This kinetic instability may lead to spontaneous nucleosome disassembly or histone exchange between nucleosomes. At high nucleosome concentrations, nucleosome–nucleosome collisions and subsequent histone exchange would be a more likely event, where nucleosomes could act as their own histone chaperone. This spontaneous histone exchange could serve as a mechanism for maintaining overall chromatin stability, although it has never been reported. Here we employed three-color single-molecule FRET (smFRET) to demonstrate that histone H2A-H2B dimers are exchanged spontaneously between nucleosomes on a time scale of a few tens of seconds at a physiological nucleosome concentration. We show that the rate of histone exchange increases at a higher monovalent salt concentration, with histone-acetylated nucleosomes, and in the presence of histone chaperone Nap1, while it remains unchanged at a higher temperature, and decreases upon DNA methylation. These results support the notion of histone exchange via transient and repetitive partial disassembly of the nucleosome and corroborate spontaneous histone diffusion in a compact chromatin context, modulating the local concentrations of histone modifications and variants.

The nucleosome is the fundamental gene-packing unit in eukaryotes. Nucleosomes comprise 147 bp DNA wrapped around an octameric histone protein core composed of two H2A-H2B dimers and one (H3-H4) 2 tetramer. The strong yet flexible DNA-histone interactions are the physical basis of the dynamic regulation of genes packaged in chromatin. The dynamic nature of DNA-histone interactions also implies that nucleosomes dissociate DNA-histone contacts both transiently and repeatedly. This kinetic instability may lead to spontaneous nucleosome disassembly or histone exchange between nucleosomes. At high nucleosome concentrations, nucleosome-nucleosome collisions and subsequent histone exchange would be a more likely event, where nucleosomes could act as their own histone chaperone. This spontaneous histone exchange could serve as a mechanism for maintaining overall chromatin stability, although it has never been reported. Here we employed three-color single-molecule FRET (smFRET) to demonstrate that histone H2A-H2B dimers are exchanged spontaneously between nucleosomes on a time scale of a few tens of seconds at a physiological nucleosome concentration. We show that the rate of histone exchange increases at a higher monovalent salt concentration, with histone-acetylated nucleosomes, and in the presence of histone chaperone Nap1, while it remains unchanged at a higher temperature, and decreases upon DNA methylation. These results support the notion of histone exchange via transient and repetitive partial disassembly of the nucleosome and corroborate spontaneous histone diffusion in a compact chromatin context, modulating the local concentrations of histone modifications and variants.
The nucleosome is the basic gene-packing unit in eukaryotes. Nucleosomes further compact into chromatin and eventually into a chromosome. A nucleosome core particle comprises a 147 base-pair (bp) DNA fragment wrapped around a histone protein core in 1.67 left-handed superhelical turns (1,2). The histone protein core is composed of two H2A-H2B dimers and one (H3-H4) 2 tetramer, forming the characteristic octameric structure (1). The negatively charged DNA backbone interacts strongly with the positively charged histone surface to compact genes, casting high energy barriers for DNA access (2). On the other hand, the nucleosome structure is flexible to allow for dynamically regulated DNA access. For instance, the DNA termini of the nucleosome undergo "breathing" motions, rendering transient and repetitive DNA accessibility (3)(4)(5).
These strong yet flexible DNA-histone interactions are necessary to enable dynamic gene regulation while maintaining the structural integrity of chromatin. However, the dynamic nature of the nucleosome structure leads to its spontaneous disassembly at a sub-nanomolar concentration (3,6). The kinetic instability suggests that nucleosomes undergo constant, repetitive, and transient partial disassembly (6). A cryo-EM structural study showed that many nucleosomes have 15 bp of DNA unwrapped at the entry/exit region and that this partially disassembled state can be stabilized by internal rearrangement of histone H2A-H2B dimers (7). At a high concentration of nucleosomes, such transiently disassembled nucleosomes may interact with each other to form internucleosomal DNA-histone contacts. To support this hypothesis, a recent study reported dinucleosome stacking between two partially disassembled nucleosomes through internucleosomal DNA-histone interactions (8,9). Under this hypothesis, nucleosomes would function as their own histone chaperone, stabilizing the entire population of nucleosomes at a high nucleosome concentration as in chromatin.
Histone chaperones mediate DNA-histone interactions to help assemble or disassemble nucleosomes depending on the context and eventually drive a DNA-histone mixture to its thermodynamic equilibrium (10). Histone chaperones contain an acidic surface to compete against DNA for histone binding (11)(12)(13). Nucleosome assembly protein 1 (Nap1) is a member of the NAP family of histone chaperones and is universally conserved among eukaryotes. Nap1 is involved in histone trafficking and nucleosome assembly in vivo (14). As for some of its functions in vitro, Nap1 mediates nucleosome assembly, exchanges nucleosomal H2A-H2B with free H2A-H2B, induces H2A-H2B dissociation from the nucleosome, and facilitates nucleosome sliding to a thermodynamically more favorable location (15,16). These functions of Nap1 are based on its strong binding with histones likely via some of the acidic residues of Nap1 and some of the basic residues of histones although the molecular nature of their interactions remains unclear. A crystal structure of a yNap1 2 -H2A-H2B complex made with N-and C-terminal truncated yNap1 and N-terminal truncated H2A-H2B manifests some acidic residues of Nap1 interacting with the basic residues of H2A-H2B that are involved in nucleosome wrapping (14). Interestingly, the stoichiometry of yNap1:H2A-H2B is 2:1 in the complex while 1:1 stoichiometry has been reported based on a binding assay (16), leaving a possibility that the truncated tails of the proteins may play important roles in their interactions. A recent molecular dynamics simulation study suggests strong yet fuzzy interactions between the acidic C-terminal tail of Nap1 and H2A-H2B (17). All these results point to interactions between the acidic C-terminal tail and the globular residues of Nap1 and the basic residues of H2A-H2B that are involved in nucleosome wrapping. These interactions suggest that Nap1/ H2A-H2B binding would be facilitated by partial disruption of the DNA/H2A-H2B interfaces (18,19). To support this notion, a recent biochemical study demonstrated that Nap1 dissociates H2A-H2B out of a partially unwrapped nucleosome, resulting in a hexasome and Nap1-H2A-H2B (17).
As charge-charge interactions are important in the interactions between histone chaperone and histones, naked DNA and the DNA region that is partially unwrapped in the nucleosome may also act as histone chaperones (20). Such a histone chaperone function of the nucleosome would result in stabilized nucleosomes at the cost of freely and spontaneously exchanged histones, although spontaneous histone exchange between nucleosomes has never been reported. To support this hypothesis, spontaneous exchange of nucleosomal H2A-H2B with free H2A-H2B has been reported (16,21,22). Histone exchange among nucleosomes on a relevant timescale would imply that histone modifications and variants can diffuse and redistribute themselves spontaneously in chromatin after serving their purpose.
Here we investigated spontaneous histone H2A-H2B dimer exchange between nucleosomes under various conditions with three-color single-molecule FRET (smFRET). Our results indicate that histone H2A-H2B dimers are exchanged spontaneously between nucleosomes at a rate corresponding to a few tens of seconds exchange timescale at physiological nucleosome and monovalent salt concentrations. The rate of histone exchange increases at a higher monovalent salt concentration or with a thialysine analog of H3K56ac (6,22), supporting spontaneous, transient, and repetitive partial disassembly of the nucleosome as the underlying mechanism. We investigated the effect of histone chaperone Nap1 to support that it facilitates histone exchange, thereby catalyzing histone equilibrium among nucleosomes in chromatin. We also found that CpG methylation significantly lowers the rate of histone exchange, suggesting its role in suppressing DNA accessibility and maintaining chromatin integrity. The rate constants remain unchanged at 25 C from those at 4 C, indicating that the DNA-histone interactions are not disturbed further enough to make a difference under these two conditions. These results support a model where nucleosomes are stabilized in chromatin by chaperoning themselves, and histone modifications and variants in H2A-H2B are spontaneously redistributed and equilibrated on a timescale of a few tens of seconds in chromatin.

Results
Histone H2A-H2B dimers are spontaneously exchanged between nucleosomes Monitoring H2A-H2B exchange between nucleosomes as a function of time is challenging with an ensemble averaging method due to various sources of heterogeneity such as contaminants and impurities from sample preparation and labeling. We constructed two types of differently labeled nucleosomes to investigate the rate of histone H2A-H2B exchange between nucleosomes (Fig. 1). One type of the nucleosomes is labeled with Cy3 and Cy5 at DNA as a FRET pair (Fig. 1B). This nucleosome is also labeled with biotin for surface immobilization. The fluorescence signals from Cy3 and Cy5 were divided into three spectral channels and imaged on a camera as is shown in Figure 1C. The FRET histograms in Figure 1B constructed with the fluorescence intensities recorded in the three spectral channels confirm 0.6 FRET Cy5 and zero FRET Cy5.5 . The other type of the nucleosomes has H2A-H2B labeled with Cy5.5 and no labels on the DNA (Fig. 1B). After equal amounts of both nucleosomes were mixed to a final concentration of 200 nM each at 10 mM NaCl, the fraction of nucleosomes with H2A-H2B exchanged at 4 C was monitored based on the smFRET signals ( Figure 2). Upon H2A-H2B exchange between the two types of the nucleosomes, a FRET Cy5.5 signal arises. The Cy5.5 fluorophore labeled at the H2A-H2B proximal to the nucleosome entry would result in a high FRET Cy5.5 and strong fluorescence from Cy5.5, which leads to its fast premature photobleaching before proper observation (Fig. 2B). We constructed a mimetic 3color smFRET system with DNA (Table S1 and Fig. S1) to confirm the relative intensity levels of Cy3, Cy5, and Cy5.5 in case of proximal H2A-H2B exchange. The Cy5.5 fluorophore labeled at the H2A-H2B distal to the nucleosome entry will result in a low Cy3 signal and a mid-high Cy5 and Cy5.5 signals (Fig. 2B), as is confirmed with another DNA mimetic (Table S1 and Fig. S1). The histograms of FRET Cy5 and FRET Cy5.5 in the case of distal H2A-H2B dimer exchange (Fig. 2B) show non-zero FRET Cy5.5 counts and lower FRET Cy5 values than those from the intact nucleosome before exchange (Fig. 1B). We counted the distal dimer exchanged nucleosomes since counting the proximal dimer exchanged nucleosomes results in a large error due to their premature photobleaching. Histograms of FRET Cy5 and FRET Cy5.5 from the nucleosomes showing zero intensities after photobleaching (single nucleosomes for accurate background correction), longer photobleaching lifetimes (>10 s), and a decent signal-to-noise ratio (>4) before and after distal H2A-H2B exchange are shown in Figures 2C and S2. Note that the histograms cannot be used as a measure of the efficiency of histone exchange as Cy5.5 photobleaches much faster than Cy5. According to our experience, it is typical that fluorophores labeled on protein photobleach much faster than those on nucleic acids. Instead, we counted the number of nucleosomes displaying the sign of distal H2A-H2B exchange (i.e. non-zero FRET Cy5.5 with concomitantly lowered FRET Cy5 ) at various time points during the reaction and computed the fractions of H2A-H2B exchanged nucleosomes. The fractions are plotted against time and fitted with Equation 1 to obtain the exchange rate constant ( Fig. 3A and Table S2). The rate constant at 10 mM NaCl at 4 C is 356 ± 64 M −1 s −1 . The error is the standard error of fitting from a Levenberg-Marquardt algorithm that can be used as the standard deviation to compute the confidence intervals of the fit value (37). This rate corresponds to 40 ± 7 s H2A-H2B exchange time at a 70 μM nucleosome concentration as in a human nucleus (3.0 × 10 6 nucleosomes per 690 μm 3 nucleus volume) (38).
To investigate the effect of a higher monovalent salt concentration, we measured the exchange rate at 50 mM NaCl and 150 mM KCl (Fig. 3A). The rate constant is 686 ± 111 M −1 s −1 , corresponding to 21 ± 3 s H2A-H2B exchange time at a 70 μM nucleosome concentration. The difference between the low and high salt cases (= 330 M −1 s −1 ) is larger than the sum of the errors (= 175 M −1 s −1 ), making the difference statistically significant. The significant increase in the rate supports a mechanism where a stronger ionic condition facilitates transient nucleosome partial disassembly, resulting in a higher fraction of nucleosomes eligible for internucleosomal DNA-histone interactions upon collision and subsequent histone exchange. This mechanistic model is depicted in Figure 4.
We also monitored H2A-H2B exchange at an elevated temperature of 25 C from 4 C. A temperature increase may destabilize DNA-histone interactions and facilitate nucleosome diffusion and nucleosome-nucleosome collisions. However, the exchange rate constant did not change much at 414 ± 64 M −1 s −1 which is within error and thus statistically insignificant (i.e., the difference is smaller than the sum of the errors) from the value at 4 C (Fig. 3A). This result suggests that the anticipated effects at 25 C are not sufficient to induce a statistically significant change in the exchange rate constant.
Histone chaperone Nap1 mediates H2A-H2B exchange Histone chaperones mediate the interactions between histone and DNA by competing against DNA for histone binding, which facilitates nucleosome assembly or disassembly depending on the context (11). According to this mechanism, histone chaperones should mediate inter-nucleosomal histone exchange as well, facilitating histone redistribution among nucleosomes. We monitored H2A-H2B exchange in the presence of Nap1 at 400 nM at 4 C. The exchange rate constant is 683 ± 94 M −1 s −1 (Fig. 3A) which is statistically significantly higher than the rate of spontaneous exchange without Nap1 at 356 ± 64 M −1 s −1 . As the action of Nap1 would be enhanced by partially disrupted DNA/H2A-H2B contacts, this result supports the existence of the partially disassembled nucleosomes for facilitated Nap1-histone interactions, mediating histone exchange. This result also supports that histone  Acetylated H3K56 analog H3K S 56ac facilitates H2A-H2B exchange Histone acetylation is typically associated with elevated gene activities (7,23,35,39,40). In particular, acetylation at H3K56 (H3K56ac) has been coupled to active transcription and increased termini breathing dynamics of the nucleosome (7,34,39). These results lead to a hypothesis where H3K56ac increases the fraction of transiently disassembled nucleosomes eligible for histone exchange upon collision, thereby facilitating histone exchange. We tested this hypothesis with nucleosomes containing H3K S 56ac under various conditions (Fig. 3B).
Under all the conditions tested, the rate of H2A-H2B exchange is very high with H3K S 56ac, reaching the measurement Figure 2. Experimental scheme to monitor the kinetics of histone H2A-H2B exchange between nucleosomes. A, two types of nucleosomes differently labeled are mixed at the total concentration of 400 nM under various conditions to monitor the fraction of nucleosomes with H2A-H2B exchanged. B, the scenarios leading to H2A-H2B exchange detectable with the three-color smFRET setup are shown. The reaction will result in nucleosome-entry proximal or distal H2A-H2B exchanged to display observable changes in the FRET signals as shown in the fluorescence intensity traces that are corrected for interchannel leakage. The fluorescence intensities are in arbitrary units (a.u.). The histograms of FRET Cy5 and FRET Cy5.5 are also shown on the side of the traces. In the case of entry proximal H2A-H2B exchange, high FRET Cy5.5 and low FRET Cy5 signals are detected, which leads to fast photobleaching of Cy5.5. At the end of an observation time window, the laser is turned off and subsequently, a 635 nm laser is turned on to ensure that the intensity change is due to Cy5.5 photobleaching. In the case of distal H2A-H2B exchange, low-mid FRET Cy5, and non-zero FRET Cy5.5 signals are observed. The other nucleosomes shown in the parentheses do not generate any Cy5.5 signals. C, histograms of FRET Cy5 and FRET Cy5.5 from the nucleosomes showing zero intensities after photobleaching (single nucleosomes for accurate background correction), longer photobleaching lifetimes (>10 s), and a decent signal-to-noise ratio (>4) at 0 (intact nucleosomes labeled with Cy3 and Cy5) and the 18th hour time points (nucleosomes with Cy3-Cy5 and the entry-distal H2A-H2B exchanged) (eighth-hour time point for H3 K56 acetylated nucleosomes) time points after mixing. The histograms show the growth of non-zero FRET Cy5.5 and lower FRET Cy5 counts as marked in the red circled area, confirming histone exchange. Note that the histogram counts cannot be used as a measure of the percent histone exchange as Cy5.5 photobleaches much faster than Cy5.   (25 C), and in the presence of Nap1, respectively (Fig. 3B). The only condition that did not show a statistically significantly higher rate is at the high salt concentration likely because the rate is already high without H3K S 56ac (916 ± 139 M −1 s −1 and 686 ± 111 M −1 s −1 for acetylated and unacetylated nucleosomes, respectively) (Fig. 3). Under all the conditions tested, the rates reached nearly the limit of the measurements as the change is almost complete at the third time point (i.e. 2 h time point with 1 h minimum separation between two time points). As a result, no statistically significant difference could be detected among the reactions under various conditions. Regardless, it is clear that H3K S 56ac promotes H2A-H2B exchange likely via the same mechanism as is proposed in the above cases. Under this mechanism, H3K S 56ac would weaken the DNA-histone interactions near the termini by removing a positive charge on the histone surface, thus facilitating spontaneous partial disassembly of the nucleosome.

CpG methylation inhibits H2A-H2B exchange
CpG methylation has been coupled to gene silencing (41). Multiple studies reported that CpG methylation reduces the conformational flexibility of DNA (27,28,(42)(43)(44) although the effects of such a change on the overall thermodynamic stability of the nucleosome remains unresolved. Regardless, the conformational rigidity induced upon CpG methylation may lead to a decreased rate of H2A-H2B exchange. Supporting this notion, a recent molecular dynamics study demonstrated higher kinetic stability of nucleosome termini upon CpG methylation which will decrease the exchange rate according to our proposed model (Fig. 4) (43). To this end, we monitored histone exchange with CpG methylated nucleosomes under the same salt and temperature conditions as in those for the unmodified nucleosomes (Fig. 3C).
The results indicate that CpG methylation under a physiological salt condition at 50 mM NaCl and 150 mM KCl inhibits histone exchange considerably at 397 ± 51 M −1 s −1 as compared to the rate without methylation at 686 ± 111 M −1 s −1 , which is statistically significant and in line with previous studies suggesting enhanced kinetic stability of nucleosome termini upon CpG methylation (27). The difference is statistically insignificant at a low salt concentration of 10 mM NaCl (277 ± 42 M −1 s −1 and 356 ± 64 M −1 s −1 for the methylated and non-methylated cases) likely due to the already low rate of exchange without methylation. No difference is detected at an elevated temperature at 25 C (400 ± 53 M −1 s −1 and 414 ± 64 M −1 s −1 for the methylated and non-methylated cases), suggesting that the anticipated changes are not large enough to affect the exchange rate. Despite CpG methylation suppressing histone exchange, Nap1 mediates histone exchange among CpG methylated nucleosomes as efficiently as among unmodified ones (626 ± 72 M −1 s −1 and 683 ± 94 M −1 s −1 for the methylated and unmethylated cases). This result suggests that the action of histone chaperone in mediating histone exchange in chromatin is not inhibited by CpG methylation.

Discussion
It has been reported that nucleosomes are stable under physiological ionic conditions only at above certain threshold concentration in vitro (3,6). At a lower concentration, nucleosomes spontaneously disassemble within a few hours (6,45). An unanswered question is how these fragile nucleosomes are stable at a high concentration such as in chromatin, which allows for tight packaging of genes while they are still accessible when needed. We set out to test if nucleosomes can efficiently act as their own histone chaperone and protect themselves from spontaneous disassembly, while the DNAhistone contacts are repeatedly, transiently, and partially disrupted all the time.
Our results indicate that nucleosomes spontaneously exchange histone H2A-H2B. The rate is 356 M −1 s −1 at 10 mM NaCl which might have been underestimated as it is the rate of exchange of the entry-distal H2A-H2B. It has been reported that the nucleosome wrapping in the entry-distal H2A-H2B region is stronger than that in the proximal region when wrapped in the 601 sequence (44,46). Although it is out of reach to quantify the extent of underestimation with the currently available data, it can be stated that the rate of H2A-H2B exchange should be higher than what we report here. Consequently, the timescale of H2A-H2B exchange and equilibrium in chromatin should be faster than the estimated rates reported here, especially considering that the 601 sequence is extremely strong in nucleosome positioning. The exchange rate is elevated to 686 M −1 s −1 at 50 mM NaCl and 150 mM KCl. This increase in the rate supports that histone exchange takes place via partially unwrapped nucleosomal DNA serving as histone chaperone for another nucleosome with partially disrupted DNA/H2A-H2B contacts. A stronger ionic condition will increase the fraction of partially unwrapped nucleosomes and, subsequently, the chance of histone exchange upon their collisions. To support the di-nucleosomal state of this exchange mechanism (Fig. 4), stacking between a partially disassembled nucleosome and an intact nucleosome has been reported (9). Our results with Nap1-mediated histone exchange also support this mechanism where Nap1 binding with H2A-H2B would be enhanced by partial disruption of the DNA/H2A-H2B interfaces as supported by a recent biochemical study (17). Of note, when mediated by Nap1 the path to histone exchange could be different from what is proposed in Figure 4 because the histone acceptor in this case is Nap1 instead of another nucleosome particle. As such, Nap1-mediated histone exchange may require another step in the mechanism that likely involves Nap1-H2A-H2B dissociated from nucleosomes and the resulting hexasomes (17). In this scenario, the Nap1-H2A-H2B complexes would deliver H2A-H2B to hexasomes, eventually leading to histone exchange. This alternative mechanism is based on the reported functions of Nap1 that competes against DNA for histone binding and helps a DNA-histone mixture properly assemble the nucleosome by inhibiting non-productive DNA-histone interactions and delivering histones only to the right intermediates (10,16).
Our results indicate that CpG methylation and H3K S 56ac meaningfully change the rate of histone exchange under a physiological ionic condition. CpG methylation is strongly tied to gene silencing (47,48). According to our results, CpG methylation may contribute to keeping quiet regions of the genome intact by suppressing histone exchange. As for the effect of H3K S 56ac, a thialysine analog of H3K56ac, it promotes H2A-H2B exchange as is supported by our previous reports on facilitated nucleosome termini opening motions by this modification (28,36). All of these results support that histone exchange takes place via a di-nucleosomal state where partially disassembled nucleosomes form internucleosomal DNA-histone interactions (Fig. 4). A direct test of this proposal would involve visualization of partially disassembled nucleosomal states and concomitant di-nucleosome formation, which is likely too transient for currently available techniques (4).
Our results suggest that a histone H2A-H2B dimer is exchanged between two nucleosomes within 21 ± 3 s under a physiological ionic condition at 70 μM nucleosome concentration which is an approximate level of nucleosome concentration in a human nucleus. This is a high rate likely underestimated by which histones diffuse among nucleosomes in a compact chromatin context and locally concentrated histone variants and modifications of H2A-H2B will be diluted away. This spontaneous histone exchange might be a mechanism to remove gene-specific modifications of H2A-H2B after they serve their purpose. In this mechanism, H2A-H2B modifications and variants will be diffused over a large area of chromatin and eventually reverted by enzymes with no gene specificity.

Construction of nucleosomal DNA
We used the Widom 601 nucleosome positioning sequence for the sequence of our nucleosomal DNA (49). Singlestranded DNA (ssDNA) oligonucleotides were purchased from Integrated DNA Technologies (IDT Inc, Coralville, IA). The sequences are listed in Table S3. Nucleosomal DNA was constructed by following previously published protocols (35,50). Briefly, each nucleosomal DNA construct was prepared by annealing and ligating the oligonucleotides. Oligonucleotides F2 and F3 (Table S3) were labeled with Cy5 and Cy3 fluorophores respectively, via conjugation between a C6-amino linker and an NHS-ester functionalized fluorophore.

Histone H3K56 acetylation and histone chaperone Nap1 preparation
Histone H3 K56C/C110A double mutant (Xenopus laevis) was purchased from The Histone Source (Colorado State University). The histone was dissolved in 200 μl of the reaction buffer containing 0.2 M sodium acetate (pH 4), 6 M guanidine-HCl, 7 mM L-glutathione, 50 mM N-vinyl acetamide, 100 mM dimethyl sulfide, and 5 mM VA-044 (2,2-[azobis(dimethyl methlene)]bis(2-imidazoline)-dihydrochloride) to the final concentration of 1 mM as was previously published (51). The reaction was carried out in complete darkness for 2 h at 70 C in a water bath. The resulting histone H3 is a thialysine analog of acetylated H3 at K56 that has a thioether-linked sidechain of acetylated lysine at residue C56, often denoted by H3K S 56ac. A mass-spectrometeric analysis confirmed the modification (Fig. S3). This mimetic has been successfully used to reproduce the chromatin array decompaction activity of native histone acetylation and has been employed to investigate histone acetylation in other contexts (36,40,51,52). The product was dialyzed against deionized water three times for 2 h, 4 h, and overnight, and then lyophilized. The solid form of the acetylated histone protein was stored at −80 C. 6xHis-Yeast nucleosome assembly protein 1 (Nap1) was expressed in E. coli BL21(DE3) pLysS and purified with Ni-NTA beads (Thermo Fisher Scientific) as was reported in a previous publication (11).

Preparation of CpG methylated DNA
CpG methylation on the DNA was carried out by following standard protocols for the CpG methyltransferase enzyme M.SssI (New England Labs Inc). Briefly, fluorophore-labeled or unlabeled nucleosomal DNA was mixed with S-adenosylmethionine in the methyltransferase reaction buffer (50 mM NaCl, 10 mM Tris-HCl (pH 7.9), 10 mM MgCl 2, 1 mM DTT) and the M.SssI enzyme. The reaction mixture was kept at 37 C for 4 h. The reaction was stopped by heating the mixture at 65 C for 20 min. Digestion of DNA with BstUI restriction enzyme (New England BioLabs, Inc) confirmed successful methylation (Fig. S4).
Nucleosome reconstitution was carried out by mixing stoichiometric amounts of DNA and histones according to the previously reported protocols (53). In the case of H3K S 56ac, H2A-H2B and (H3-H4) 2 were used instead of the octamer form. We used the salt gradient dialysis method in a dialysis cup (Slide-A-Lyzer MINI Dialysis Device, 7K MWCO, Thermo Fisher Scientific) using 1 × TE (pH 8.0) buffer with stepwise decreasing salt concentrations of 1200, 850, 600, 400, 200, and 10 mM NaCl. The purity of all nucleosome sets was confirmed with native-PAGE (Fig. S6).

Microscope slide preparation
Pre-drilled microscope quartz slides were purchased from G. Finkenbeiner Inc. The cleaned slides were constructed with five flow channels and passivated with a lipid bilayer following published protocols (35,54). Briefly, a sub-monolayer of biotin-PEG-silane was coated (MW 3400, Laysan Bio) on a carefully and thoroughly cleaned quartz microscope slide followed by the deposition of a lipid bilayer. For preparation of the lipid solution, dried lipid vesicles made of 1,2-Dioleoyl-sn-Glycero-3-phosphoethanolamine-N-[Methoxy(Polyethylenegl ycol)-5000] (ammonium salt) (DOPC) (Avanti Polar Lipids, product number 880230) were suspended in a buffer containing 10 mM Tris-HCl (pH 7.8) and 100 mM NaCl. The suspension was subjected to 1 min sonication/cooling cycles on ice with a tip-based (Branson Ultrasonics, part no #101-148-062) ultrasonic cell disruptor (Sonifier 550, Branson Ultrasonics) until the lipid suspension became transparent. Once the suspension became transparent, the lipid vesicles were extruded through a 100 nm pore-sized polycarbonate membrane filter (Avanti Polar Lipids). Flow channels on the surface of a biotin-PEG coated microchannel were injected with 1 mg/ml concentrated lipid solution and incubated for 45 min to deposit a bilayer (Fig. 1C). Following the lipid bilayer deposit, 40 μl of 0.1 mg/ml streptavidin solution was injected into the flow channels and incubated for 15 min for biotin-streptavidin conjugation. This conjugation is to immobilize the nucleosome samples on the slide surface for subsequent smFRET measurements.

Sample preparation and smFRET measurements for dimerexchange reaction
Dimer exchange reaction(s) were carried out at 4 C after mixing DNA-labeled nucleosomes with histone-labeled nucleosomes at a 1:1 ratio in a test tube. The final concentration of the nucleosomes in the stock was 200 nM each, resulting in a total concentration of 400 nM. These are the reaction stock(s) for the dimer-exchange reactions. The reaction kept taking place in the test tube. Upon preparing a reaction stock, a 1 μl aliquot was taken out at 0, 1,2,4,6,8,12,18, and 24 h time points to measure the fraction of histone exchanged nucleosomes. For the measurement at each time point, the aliquot taken from the reaction stock was loaded on to the slide after diluting to a concentration of 70 to 100 PM in an imaging buffer containing 10 mM Tris-HCl (pH 8.0), 1 mM Trolox, 0.1 mg/ml of BSA, 2 mM protocatechuic acid and 0.2 U/ml protocatechuate-3,4-dioxygenase. This severe dilution quenches the exchange reaction. A total of 4 to 10 independent measurements were made per time point. For the low salt reaction condition, 10 mM NaCl was used, whereas for the high salt condition, 50 mM NaCl and 150 mM KCl were used. Room temperature exchange reactions were carried out at 25 o C at 10 mM NaCl. Nap1-mediated exchange reactions were carried out at 400 nM Nap1 (nucleosome:Nap1 = 1:1) at 10 mM NaCl.

Three-color smFRET setup
Three-color smFRET measurements were taken with a custom-built EMCCD-based TIRF setup as was reported previously (18,55). Briefly, a 532 nm green laser (CrystaLaser, Reno, NV) and a 635 nm red laser (CrystaLaser, Reno, NV) were used for FRET donor and acceptor excitation respectively. The surface of a flow channel was illuminated with a green laser in a prism-coupled TIR geometry and the fluorescence images were recorded with an EMCCD camera (IXON Ultra 897, Oxford Instruments), followed by a brief dark period and excitation with 635 nm red laser to verify the existence of Cy5 and/or Cy5.5. The signal integration time for fluorescence imaging was 200 ms. In this one-donor (Cy3) two-acceptor model (Cy5 and Cy5.5), the fluorescence signals from the nucleosomes were spectrally separated into three signals (Cy3, Cy5, and Cy5.5 channels) (Fig. 1C). The relative intensities from a Cy3-Cy5 pair in an intact nucleosome in the three spectral regions are shown in Figure 1C. The three-color scheme ensures that we count only properly assembled nucleosomes before and after H2A-H2B exchange and that the measurements are free from any errors due to fluorescent contaminants and impurities originated from sample preparation Background correction and interchannel leakage correction for single-molecule fluorescence intensities Time traces of the fluorescence intensities in the three spectral channels (Cy3, Cy5, and Cy5.5 channels) were obtained from the single-molecule experiments (Figs. 1 and 2). These raw intensities are compounded by a constant background and inter-channel leakages mainly between the two acceptors Cy5 and Cy5.5 channels. To correct the constant background, we obtained the background intensities in each channel after the fluorophores were photobleached. These constant background values were subtracted from the intensities of the fluorophore.
The fluorescence from an acceptor (Cy5 or Cy5.5) leaks into the other acceptor channel. To compute their intensities without the contribution from the other acceptor, we first obtained two leakage factors r 5 (¼ I 5:5c I 5c þI 5:5c , where I 5C and I 5.5C are the intensities of Cy5 in the Cy5 and Cy5.5 channels, respectively) and r 5:5 (¼ I 5c I 5c þI 5:5c , where I 5C and I 5.5C are the intensities of Cy5.5 in the Cy5 and Cy5.5 channels, respectively). The value of r 5 represents the fraction of Cy5 emission registered in the Cy5.5 channel and can be obtained by measuring the intensities of Cy5 in the Cy5 and Cy5.5 channels. The value of r 5.5 represents the fraction of Cy5.5 emission registered in the Cy5 channel and can be obtained by measuring the intensities of Cy5.5 in the Cy5 and Cy5.5 channels. We made multiple measurements of the r 5 and r 5:5 values and used the average values of 0.31 ± 0.02 and 0.27 ± 0.04 for r 5 and r 5:5 , respectively for leakage correction. Let the true fluorescence intensities of Cy5 and Cy5.5 without leakage be I 5 and I 5:5 . Writing I 5 and I 5:5 in terms of the apparent intensities from the Cy5 (I a 1 Þ and Cy5.5 channels (I a 2 ), I 5 ¼ I a 1 −I 5:5 r 5:5 þI 5 r 5 I 5:5 ¼ I a 2 −I 5 r 5 þI 5:5 r 5:5 Solving these two equations for the true intensities I 5 and I 5:5 results in where I 3 , I 5 , and I 5.5 are the intensities of Cy3, Cy5, and Cy5.5 after background and leakage correction.

Identifying dimer-exchanged nucleosomes based on the fluorescence intensity signatures
To identify a dimer-exchanged nucleosome, Cy3/Cy5/Cy5.5 relative intensity levels were examined from the surfaceimmobilized nucleosomes (Fig. 2). In the entry-proximal H2A-H2B exchanged case, the distances between the three fluorophores result in near 100 % FRET funneling to Cy5.5 (Fig. S1A). We modeled this scenario with two 15-bp DNA fragments labeled with Cy3, Cy5, and Cy5.5 with similar interdye distances (Fig. S1A). The three fluorophore intensities are shown in Fig. S1C are similar to what we observed from some of the histone-exchanged nucleosomes, confirming proximal H2A-H2B exchange (Fig. 2B). In the entry-distal H2A-H2B exchanged case, the distances between the fluorophores should result in strong FRET to both Cy5 and Cy5.5 from Cy3 (Fig. S1, B and C). We used another pair of 15-bp DNA fragments labeled with the same fluorophores to simulate the scenario (Fig. S1B). The relative intensities of the three fluorophores are very similar to some of the histone-exchanged nucleosomes, confirming distal H2A-H2B exchange (Figs. 2B and S1C).

Extracting the rate constant of histone exchange
We counted the numbers of intact nucleosomes and distal H2A-H2B exchanged nucleosomes as described above and calculated the fractions of histone exchange at various time points. The time course of the fractions was fitted with an equation approximating the kinetics to extract the histone exchange rate constant. Considering partial labeling of H2A-H2B and near complete labeling of DNA, Fig. S7A illustrates nucleosome species with all possible fluorophore combinations. Fig. S7B lists histone exchange between all possible such labeled pairs of the nucleosomes with the same rate constant k. These reactions lead to the differential rate equation for the production of trackable H2A-H2B exchanged nucleosomes (i.e. species B in Fig. S7A) as follows: Let [B] be its mole fraction x, then the mole fractions of the other species are as follows in the beginning.
[  2 , where F is the H2A-H2B labeling efficiency that is 53.7 % according to our UV-Vis measurement. Writing the differential rate equation in terms of mole fractions is as follows: Note that, we approximate the mole fraction of both H2A-H2B exchanged as negligible (i.e. [B'] ≈ 0), which should be reasonable during an early time period.
We counted the number of species e, the distal H2A-H2B exchanged nucleosomes. Solving the differential equation for The mole fraction of species e at equilibrium is the probability of having the entry-proximal H2A-H2B unlabeled and the other H2A-H2B labeled in a nucleosome (= (1-0.537/ 2) × 0.537/2 = 0.196, where 0.537/2 is the mole fraction of labeled H2A-H2B in the reaction mix and 0.537 is the labeling efficiency of H2A-H2B). Therefore, the approximate mole fraction [e] as a function of time is as follows. mole fraction of e ðtÞ ¼ xðtÞ

Data availability
All data are contained within the manuscript. Conflict of interest-The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.