Nucleosome dynamics during transcription elongation

Abstract

The nucleosome is the basic packing unit of the eukaryotic genome. Dynamic interactions between DNA and histones in the nucleosome are the molecular basis of gene accessibility regulation that governs the kinetics of various DNA-templated processes such as transcription elongation by RNA Polymerase II (Pol II). Based on single-molecule FRET measurements with chemically modified histones, we investigated the nucleosome dynamics during transcription elongation and how it is affected by histone acetylation at H3 K56 and the histone chaperone Nap1, both of which can affect DNA-histone interactions. We observed that H3K56 acetylation dramatically shortens the pause duration of Pol II near the entry region of the nucleosome, while Nap1 induces no noticeable difference. We also found that the elongation rate of Pol II through the nucleosome is unaffected by the acetylation or Nap1. These results indicate that H3K56 acetylation facilitates Pol II translocation through the nucleosome by assisting paused Pol II to resume and that Nap1 does not affect Pol II progression. Following transcription, only a small fraction of nucleosomes remain intact, which is unaffected by H3K56 acetylation or Nap1. These results suggest that (i) spontaneous nucleosome opening enables Pol II progression, (ii) Pol II mediates nucleosome reassembly very inefficiently, and (iii) Nap1 in the absence of other factors does not promote nucleosome disassembly or reassembly during transcription.

INTRODUCTION

The nucleosome is the most basic eukaryotic DNA packaging unit^{1, 2}. A nucleosome core particle contains ~147 base pairs (bp) of DNA wrapped around a histone octamer core consisted of two H2A-H2B dimers and one (H3-H4)₂ tetramer^3–5. Many studies have explored the structural dynamics and the function of the nucleosome as a platform for gene regulation^6–11. It has been established that nucleosomes impose physical barriers on enzymes translocating along nucleosomal DNA such as RNA Polymerase II (Pol II)^12–17. Nucleosomes induce major pauses of Pol II at two regions – around the 15^th and the 45^th nucleotides from the nucleosome entry^{18, 19}. Recently published cryo-EM structures of Pol II-nucleosome complexes also confirm the two major pause locations¹⁹. The pause around the 15^th nucleotide (nt) is at the super-helical location −5 from the dyad or SHL(−5) and the other at SHL(−1). The structures also revealed weaker pauses at SHL(−6) and SHL(−2). The pause dynamics of Pol II have critical implications in the kinetics of transcription elongation as the pause duration is the main determinant of the overall efficiency and rate of transcription elongation through the nucleosome^20–22. Dynamic histone-DNA interactions are one of the major players in regulating the pause dynamics of Pol II.

The lysine residue K56 of histone H3 (H3K56) is at the entry and exit regions of the nucleosome and provides a positive charge for the histone core to interact with DNA²³. Acetylation of this lysine residue facilitates spontaneous DNA dynamics that should increase DNA accessibility^23–30. The number of nucleosomes with its DNA unwrapped at the entry region increases by two folds when acetylated at H3K56²⁶. While this large extent of spontaneous DNA unwrapping has not been observed under other experimental conditions, data demonstrating transient unwrapping in the range of a few angstroms has been published based on single-molecule FRET and cryo-EM measurements^{19, 29}. However, this extent of unwrapping is not large enough to directly affect Pol II binding or progression. Decreased pause densities and durations in early regions of the nucleosome including the entry region were reported when all histone lysine residues were mutated to glutamine^{20, 31}. However, the effect cannot be attributed solely to acetylation as glutamine mutation is limited in modeling acetylation³² and such extensive glutamine mutation may result in an altered nucleosome structure. Due to the lack of direct investigations based on a highly refined and pure experimental system, the consequences of H3K56 acetylation on the kinetics of Pol II progression through the nucleosome remain largely unknown.

We have reported that spontaneous DNA opening motion at the nucleosome termini is facilitated upon H3K56 acetylation while the overall structural integrity of the nucleosome is unaffected²⁶. The spontaneous DNA opening rate at 50 mM NaCl is 1.9-fold higher in H3K56 acetylated nucleosomes compared to non-acetylated ones while the closing rate is kept constant. This change results in a 1.5-fold longer dwell time in the open state upon acetylation²⁶, and can be interpreted as a 1.5-fold increase in the probability of finding a nucleosome in an open state at any random moment. However, the rate of the histone chaperone Nap1 binding escalates by 5.9 folds³⁰ that is far larger than 1.5 folds. The greater increase in the Nap1 binding rate is because H3K56 acetylation (H3K56ac) not only increases the opening rate, but also makes the average opening slightly larger and that Nap1 binding requires a much larger opening than the average opening. According to our rationale, the probability of DNA opening large enough for Nap1 binding dramatically increases upon H3K56 acetylation while that of opening by any distance becomes merely 1.5-fold higher. This principle can also be applied to other processes that require spontaneous DNA opening. For instance, a paused Pol II in the nucleosome would need opening of the histone-DNA interface to resume. A large effect of H3K56ac would be expected if a paused Pol II requires a large opening to resume, which it does.

Transcription factors and histone chaperones can also influence nucleosome dynamics by promoting or disrupting DNA-histone and histone-histone interactions^7–11. The well-studied Nap1 belongs to a class of histone chaperones that can interact with positively charged histone proteins via its acidic patch, thus plays roles in histone transport, nucleosome assembly and disassembly^33–41. Nap1 facilitates the thermodynamic equilibrium of a DNA-histone mixture, mediating nucleosome assembly or eliminating DNA-histone interactions that do not lead to nucleosome formation^{38, 39}. Despite its nanomolar affinity to histone H2A-H2B, Nap1 cannot bind H2A-H2B in canonical nucleosomes because its binding regions on the dimer are shielded by DNA³⁹. Only when DNA-histone interaction is weakened, Nap1 can interact with histones in the nucleosome. As such, Nap1 can mediate both nucleosome assembly and disassembly, although it remains unclear how on a molecular level Nap1 can mediate DNA-histone dynamics during transcription elongation by Pol II.

Here, we investigated the effects of H3K56ac and Nap1 on the dynamics of transcription elongation through the nucleosome. We employed single-molecule FRET experimental systems⁴² with chemically modified H3K56ac nucleosomes in order to monitor the nucleosome dynamics during transcription elongation by Pol II in real time in a time-resolved manner. Our results indicate that H3K56ac shortens the pause duration of Pol II at the nucleosome entry region by 4.0 folds and that H3K56ac has no effect on the next pause duration or elongation rates. Surprisingly, Nap1 has no effect on the Pol II progression kinetics. We did not observe any effect of H3K56ac or Nap1 on the already low efficiency of nucleosome reassembly upon Pol II passage. Our results strongly support that a large spontaneous opening of the nucleosome enables a paused Pol II to resume and that H3K56ac greatly facilitates this process. This effect would be very much pronounced on the overall rate of nucleosomal transcription as the pause durations are on the order a few to a few hundreds of seconds.

METHODS

Nap1 and Nucleosome reconstitution

6xHis-Yeast nucleosome assembly protein 1 (Nap1) was expressed in E. coli and purified with Ni-NTA beads (Thermo Fisher Scientific, Waltham, MA) as reported in a previous publication⁴³. Histone proteins were purchased from Histone Source at the Colorado State University. Histone H3 with K56 acetylation was prepared following published protocols⁴⁴ and confirmed by mass-spectrometric analysis (Fig. S1). A brief description can be found in the supplemental methods. The resulting histone H3 has a thioether linked sidechain of acetylated lysine at the residue C56, often denoted by H3K_C56ac. We will denote this acetylation by H3K56ac for simplicity. See the supplemental methods for histone H2B labeling. Five DNA fragments (see Fig. S2 for sequences) were purchased from Integrated DNA Technologies (Coralville IA). The fragments were annealed and ligated to form the 601-DNA sequence⁴⁵. The EC-42 strands (see Figs. 1 and S2 for sequences) with a 13 nt extension to a G-less cassette were ligated to the nucleosome entry site. The amine functionalized C6 linkers coupled to the thymine nucleotides at the +34 and +112 positions from the entry site of the nucleosome, were labeled with NHS-ester functionalized Cy5 and Cy3, respectively. Histone proteins were mixed in stoichiometric amounts and eluted through a gel filtration column to produce H2A-H2B dimers and (H3-H4)₂ tetramers separately^{46, 47}. Nucleosomes were reconstituted by dialyzing a stoichiometric mixture of histones and DNA in a dialysis device (Slide-A-Lyzer MINI Dialysis Device, 7K MWCO, Thermo Fisher Scientific) against 1X TE (pH 8.0) buffer and decreasing salt concentrations of 850, 650, 500, 300, and 2.5 mM NaCl stepwise^{46, 47}. The assembled transcription templates containing the nucleosome were confirmed by native PAGE (Fig. S3).

Figure 1. A single-molecule FRET experimental system reports the dynamics of DNA unwrapping and rewrapping during transcription elongation by Pol II.

(A) Nucleosomes with a FRET pair were immobilized on a surface-passivated microscope slide via a series of conjugations shown, which is biotinylated poly(ethylene glycol) – streptavidin – biotinylated protein A – antibody of Pol II (C-terminal domain of the Rpb1 subunit) – Pol II/nucleosome (elongation complex). Since the nucleosomes are immobilized via Pol II, they are always bound with Pol II during our observation. Transcription-initated elongation complexes were formed by mixing Pol II, transcription template, ATP, CTP, and UTP. The complexes have Pol II stalled at the end of a G-less cassette. Nucleosomal DNA was labeled with a FRET pair, Cy3 and Cy5, at the +112^th and +34^th nucleotides from the entry site, respectively. Dynamic changes in the FRET pair intensities represent the DNA dynamics in the nucleosome. (B) DNA sequence of the transcription template including the Widom 601 sequence. The lower strand is the template strand. Each strand of the 601 sequence is linked to a 66-nucleotide DNA fragment (highlighted in blue) that contains a G-less cassette. The template strand had a 13-nucleotide extension to the 3’ end (highlighted in orange). (C) Sample time trajectories of Cy3 and Cy5 fluorescence intensities and their FRET efficiency reveal mainly three stable FRET states, indicating three or more paused states of Pol II. The pause locations +10, +20, +50, and +60 are counted from the nucleosome entry site (see panel A). The FRET time trajectories were manually cut into pieces such that each piece include only a single stable FRET state (see the vertical dotted lines in the FRET trajectory). These FRET events were grouped into three according to their FRET efficiency – high-, mid-, and low-FRET states. Each group was used to construct the corresponding FRET efficiency histogram shown in D. (D) Distributions of FRET efficiencies for the three states (HF, MF, and LF) from the wild-type and H3K56ac nucleosomes. Based on Gaussian fittings, HF (0.73 ± 0.00₂), MF (0.37 ± 0.00₂), LF (0.07 ± 0.00₂) for wild-type nucleosomes, and HF (0.74 ± 0.00₃), MF (0.41 ± 0.00₂), LF (0.09 ± 0.00₂) for H3K56ac nucleosomes were reported at 95% confidence interval. A subscript in the errors indicates that the number is not a significant figure.

Formation of transcription elongation complex

Transcription was performed in a buffer containing 60 mM KCl, 1 mM MnCl₂, 50 mM K-HEPES (pH 7.8), 0.5 mM DTT, and 10 % glycerol. The transcription templates were added to 50 uL transcription buffer to 10 nM along with 9.5 nM Pol II, 0.3 mM UpG primer, 0.1 mg/mL BSA, and 2 unit/μL RNasin® RNase inhibitor to make the total volume of 91.5 μL. Pol II was prepared following published protocols^{18, 42}. The mixture was incubated for 5 min at 30°C, followed by the addition of ATP, CTP, and UTP to 1 mM to initiate transcription. The transcription elongation complexs were stopped and synchronyzed at a G-tract located at the 13^th nucleotide position into EC-42. The reaction mixture was diluted to 150 μL, and incubated for 30 min at 30°C.

Single-molecule FRET measurements

Microscope slides for Pol II immobilization and single-molecule FRET measurements were prepared as previously published⁴². See supplemental methods for further descriptions on surface preparation and data analysis.

Nucleosome dynamics during transcription elongation by Pol II with and without Nap1

After adding the imaging buffer, Cy3 was excited with a 532 nm laser and fluorescence signals from both Cy3 and Cy5 were recorded with an electron multiplying CCD (EMCCD) camera. Recording started at around 100 sec after injecting the imaging buffer. Recording started at around 300 sec in the presence of Nap1. Nap1 (600 nM) was added to the imaging buffer. FRET data was collected for 30 min at a 250 ms time resolution. A total of 15 2-minute movies from different slide spots were recorded. The time window of measurements was limited to 30 minutes in order to avoid any spontaneous disassembly or decay of the nucleosomes. A total of 39 independent measurements were made on a total of 6 separate days.

Nucleosome states after transcription by Pol II

After adding the imaging buffer, 2 fluorescence images with 250 ms signal integration (= 2 × 250 ms) every 7.5 sec were taken with 532 nm excitation followed by 2 images taken with 635 nm excitation. In order to ensure full frame excitation during imaging, we used 750 ms excitation instead of 500 ms. This excitation and imaging scheme was employed to avoid premature photobleaching of Cy5.5 so that we can observe H2A-H2B-Cy5.5 displacement without complete dissociation from the elongation complex. Each movie was taken for 3 minutes during which Cy5.5 was excited for a total of 18 seconds. According to the typical photobleaching lifetime of Cy5.5 (42 sec) with 635 nm excitation under our conditions, the majority of Cy5.5 is not photobleached during imaging. Ten 3-minute movies from different slide spots were recorded and analyzed. A total of 38 independent measurements were made on a total of 6 separate days.

RESULTS

A single-molecule FRET system reports nucleosome dynamics during transcription elongation by Pol II

We employed an in vitro transcription system based on single-molecule FRET (smFRET) to study the nucleosome dynamics during transcription elongation by Pol II as illustrated in figure 1. In the setup (Fig. 1A), the nucleosomes are immobilized via Pol II and therefore always bound with Pol II during our observation. The changes in the fluorescence intensities of a FRET pair Cy3 and Cy5 enable us to quantify the extent of DNA unwrapping and potential rewrapping during the progression of Pol II through the nucleosome. In the absence of Pol II, the nucleosome does not show FRET dynamics within our observation time window of 30 min. We designed the FRET locations according to the published Pol II pause positions^{18, 19} so that we can monitor the early dynamics of the nucleosome when Pol II translocates from the entry site through the region where a histone H2A-H2B dimer interacts strongly with DNA. Dynamic changes in the intensities of the FRET pair are evident (Fig. 1C). The initial decrease in the FRET efficiency (~10 sec time point in Fig. 1C) indicates that the DNA terminus is unwrapping. After initial DNA unwrapping, some Pol II stall, some proceed until the DNA is completely unwrapped, and some show signs of DNA rewrapping after polymerase passes. When Pol II stalls, the extent of DNA unwrapping also stalls, resulting in a stable FRET state whose lifetime is on the order of a few to few hundreds of seconds. Three stable FRET states were identified according to their FRET efficiencies – high-FRET (HF), mid-FRET (MF), and low-FRET (LF). The FRET efficiency distributions (Fig. S3) were fit with three Gaussian distribution functions. For more accurate fitting of the distributions, we separated the FRET states each of which was fit with a single Gaussian distribution function as is shown in figure 1D. A pause is a stochastic event with a low probability. As such, it is rare to find a nucleosome that shows all three pauses. This is further exacerbated by pause durations longer than the photobleaching lifetimes of the fluorophores, making it impossible to tell whether a stable FRET event is due to a paused or an arrested state. Nevertheless, the pause locations overlap with experimentally-determined arrest points, and therefore, we can assign the three FRET states to the pause locations reported by a Cryo-EM study¹⁹. We assigned the HF state to the combination of pauses at SHL(−6) and (−5), MF to SHL(−2), and LF to SHL(−1). We assigned the highest FRET state (= HF state) to the paused state at SHL(−6) or (−5) as there is no change in the FRET distance between the intact nucleosome and the nucleosome with paused Pol II at SHL(−6) or at SHL(−5) (Fig. S5). The estimated FRET efficiency from the Cyro-EM structure of the SHL(−2) paused state (= 0.4) is in good agreement with the MF state FRET efficiency (= 0.37 ± 0.00₂). As the only remaining and major pause is at SHL(−1) and the estimated (<0.1, Fig. S5) and measured FRET efficiencies (= 0.07 ± 0.00₂) are in good agreement, we assigned the LF state to the SHL(−1) paused state. Previous biochemical studies also support these FRET assignments^{18, 19}.

H3K56ac shortens the pause duration in the nucleosome entry region

We investigated whether H3K56 acetylation (H3K56ac) affects the pause dynamics in the nucleosome entry region (i.e. at SHL(−6) or (−5)) by examining the dwell time of the HF state with and without H3K56ac. The HF state dwell time distribution fits well to a single exponential decay function (Fig. 2). From the fitting results, we found that H3K56ac nucleosomes show a Pol II pause duration of 276 ± 37 sec that is a 4.0-fold decrease from the pause duration of 1104 ± 211 sec in the wild-type nucleosomes (Fig. 2AB). One source of the fitting error in figure 2B is the non-acetylated histone contaminant (Fig. S1). Regardless, the total error is much smaller than the difference induced by acetylation. These results, combined with our studies on the unwrapping of DNA from the end of H3K56ac nucleosomes, support our hypothesis that H3K56ac nucleosomes accelerate transcription through the entry by slightly increasing the average DNA opening size. This effect can result in a dramatic increase in the probability of rare openings³⁰ that are large enough for Pol II passage, and subsequently, a significant increase in the success rate of escape attempts by Pol II. More detailed explanations about this mechanism are given in the discussion.

Figure 2. Acetylation at H3K56 shortens Pol II pause duration only in the entry region.

Pause durations of Pol II in the nucleosome entry region for wild-type nucleosomes (A) and H3K56ac nucleosomes (B), and at the pause location SHL(−2) for wild-type nucleosomes (C) and H3K56ac nucleosomes (D). The pause duration in the nucleosome entry region is reduced by 4.0 folds upon H3K56 acetylation (A and B) while the pause duration at SHL(−2) is not affected (C and D). The first columns of (A) and (B) start from 100 sec which is the measurement start time. The errors indicate fitting errors.

It is reasonable to postulate that H3K56ac would not affect the pause dynamics after Pol II passes through the entry region because H3K56 interacts with DNA only in the entry region. We analyzed the durations of the next pause at SHL(−2) which is represented by the MF state (Fig. S5). The total numbers of nucleosomes that we analyzed are 641 and 862 for the wild-type and H3K56ac nucleosomes, respectively. We selected nucleosomes showing all three FRET states (HF, MF, and LF states) starting from the HF state to evaluate the dwell time of the MF state. Some FRET trajectories start in the MF state because not all Pol II pauses in the HF state. We discarded the trajectories starting in the MF state because we could not pinpoint the onset of the MF state (46 % and 36 % of the total number of the nucleosomes monitored for wild-type and acetylated nucleosomes). These nucleosomes represent likely the ones with no pauses in the entry region and filtering out these will not affect the analysis of pause dynamics. In many cases, the HF state skips the MF pause to go directly to the LF state, which we also excluded from the analysis (45% and 56% for wild-type and acetylated nucleosomes). Finally, we also discarded the data that showed arrested nucleosomes as they would remain in a stable FRET state until the fluorophores photobleach (5% and 2% for wild-type and acetylated nucleosomes). When we analyzed the data, no notable difference was found in the pause durations at SHL(−2) between the wild-type and H3K56ac nucleosomes (Fig. 2CD). The pause durations in both cases are very short at around 5 sec, which is consistent with the weak swift pauses reported at SHL(−2)¹⁹.

H3K56ac has no effect on elongation rate

To investigate whether H3K56ac alters the transcription elongation rate in the nucleosome, we measured the transition times between two major pauses, SHL(−6) or (−5) (HF) to SHL(−1) (LF). Nucleosomes showing a single step transition from the HF to the LF state were filtered out because we cannot distinguish fast elongation from fluorophore photobleaching. The transition time distribution represents the elongation time distribution and follows a single exponential decay function (Fig. 3). We removed the events that are too fast to be precisely measured (0–1 sec columns in figure 3). Sample FRET trajectories showing various elongation times are shown in figure S6. The elongation rates in both wild-type and H3K56ac nucleosomes are the same within error (0.96 ± 0.04 and 0.89 ± 0.04 sec, respectively), indicating that H3K56ac has no effect on the elongation rate (Fig. 3). As the HF state is a pause at SHL(−6) or (−5), we can estimate the elongation rate in each case. The elongation rate through the wild-type nucleosome is 52 ± 3 and 42 ± 2 nt/s when we assume that the HF state is a pause at SHL(−6) and (−5), respectively. As SHL(−5) is the dominant pause location between the two, we conclude that the elongation rate is most likely 42 ± 2 nt/s. The rate through the H3K56 acetylated nucleosome is 45 ± 2 nt/s. In other words, wild-type nucleosomes take 24 ± 1 ms to transcribe one nucleotide while H3K56ac nucleosomes need 22 ± 1 ms. These values are consistent with previously published results^{31, 48} and essentially the same within error, indicating that H3K56ac does not alter the transcription elongation rate through the nucleosome.

Figure 3. Transcription elongation rate through the nucleosome is unaffected by H3K56ac.

Elongation times of Pol II from the nucleosome entry region to the final pause SHL(−1) for wild-type nucleosomes (A) and H3K56ac nucleosomes (B). H3K56ac did not induce any difference in the elongation rate. The first columns of the histograms start at 1 sec because we discarded single-step elongation events that belong to the histogram column 0 – 1 sec. The errors indicate fitting errors.

DNA rewrapping events are rare in both wild-type and acetylated nucleosomes

Next, we examined whether nucleosomal DNA rewraps as Pol II passes through the pause locations. Based on the nature of the nucleosome structure, we postulated that nucleosomal DNA would not rewrap spontaneously in the absence of other factors such as histone chaperone^49–53. Pol II has also been suggested to function as histone chaperone^{17, 54}. DNA rewrapping after Pol II passage can be identified by an increasing FRET transition from the LF or MF state to the MF or HF state. We counted these nucleosomes with and without H3K56ac. Our data indicates that rewrapping events occurred in 3.6 ± 1.9 % (23 out of 641, the error is the standard error of binomial distributions) of the wild-type nucleosomes and 4.4 ± 2.0 % (38 out of 862) of the H3K56ac nucleosomes. These values are identical within error, indicating no effect of H3K56ac on the efficiency of nucleosome reassembly that is already low in the wild-type nucleosome. These values also confirm that spontaneous nucleosome reassembly during Pol II translocation in real-time is very inefficient although this conclusion does not rule out that nucleosomes may reassemble by the aid of another factor (e.g. histone chaperone) after Pol II runs off completely.

Nap1 has no effect on the nucleosome dynamics during transcription elongation by Pol II

We investigated the roles of the histone chaperone Nap1 in regulating the nucleosome dynamics during Pol II transcription. Nap1 competes against DNA for histone binding. Therefore, unlike H3K56ac that facilitates spontaneous nucleosome dynamics, Nap1 may play a direct enzymatic role in regulating DNA-histone interactions during transcription. First, we explored whether Nap1 affects the pause dynamics in the nucleosome entry region (Fig. 4A). The result indicates that the entry pause duration in the presence of Nap1 is 1277 ± 381 sec which is the same within error as the pause duration in the absence of Nap1, 1104 ± 211 sec (Fig. 2A). We also analyzed the transcription elongation rate in the presence of Nap1 (Fig. 4B). By measuring the transition times between two major pauses (HF to LF), we found that the elongation rate is 1.18 ± 0.19 sec which is the same within error as the rate in the absence of Nap1 (Fig. 3A). Thus, the results indicate that Nap1 has no effect on either the pause dynamics or the transcription elongation rate.

Figure 4. Transcription elongation kinetics through the nucleosome is unaffected by Nap1.

Distributions of the pause durations of Pol II in the nucleosome entry region (A) and of the elongation rates (B) in the presence of 600 nM Nap1. The first columns start at non-zero values because the measurement in (A) started at 300 sec and the single time-point transitions in (B) were discarded. The errors indicate fitting errors.

Next, we examined whether Nap1 facilitates nucleosome reassembly during Pol II progression. We counted the nucleosomes with DNA rewrapping, which can be identified by an increasing FRET transition from the LF or MF state to the MF or HF state. Our data shows that rewrapping events occurred in 2.3 ± 1.5 % (6 out of 260) of the nucleosomes in the presence of Nap1. This result indicates that Nap1 does not promote nucleosome reassembly during Pol II progression in real time although this conclusion does not rule out a possibility that Nap1 can mediate nucleosome reassembly after Pol II runs off completely.

Three-color smFRET confirms inefficient nucleosome reassembly during transcription

We looked further into the recovery of the nucleosome during transcription by Pol II in real time with three-color smFRET. Nucleosome recovery has been suggested to proceed through two mechanisms: one involves dissociation and reassociation of one H2A-H2B dimer while the other involves displacement of the entire histone octamer^{17, 49–53}. As the 601 sequence is an extremely strong nucleosome positioning sequence⁴⁵, the first mechanism is more likely applicable. We employed a three-color smFRET system to monitor the DNA dynamics and histone H2A-H2B dynamics simultaneously (Fig. 5A). The nucleosomal DNA was labeled with a FRET pair (Cy3 and Cy5) at the same locations as in figure 1 to report DNA unwrapping and rewrapping. A T112C mutant of H2B was labeled with Cy5.5 to introduce a second FRET acceptor to report histone H2A-H2B dissociation and association. According to the relative intensities from the three fluorophores, we can identify the nucleosomes fully intact, DNA unwrapped, and/or one or both H2A-H2B displaced (see Fig. S7 for further details). In order to avoid premature photobleaching of Cy5.5 before Pol II passage, we imaged the fluorescence only for 500 ms every 7.5 seconds during a 3-minute time period (see further details in Materials and Methods). Our imaging scheme enabled us to monitor the nucleosome dynamics without premature photobleaching of the majority of the fluorophores. After Pol II passage, we counted the population of intact nucleosomes and intermediates (Figs. 5B and S7). Our data indicates that most 601 nucleosomes are left with DNA unwrapped and H2A-H2B displaced upon Pol II passage. Only 5.5 ± 2.3 % (47 out of 857) of the nucleosomes are kept intact or reassembled, confirming the result from the two-color smFRET measurements based on DNA re-wrapping (3.6 ± 1.9 %). We also found some partially reassembled nucleosomes that show the intensity trajectories of hexasome-like intermediates formed by one dimer reassociation after both dimers were displaced (Figs. 5B and S7). These particles account for 2.5 ± 1.6 % (21 out of 857) of the nucleosomes. Therefore, the efficiency of at least one H2A-H2B dimer retained or re-associated is 8.0 ± 2.8 %. Overall, our results indicate that most nucleosomes are left disassembled with unwrapped DNA and both H2A-H2B dimers displaced during transcription. It is noteworthy that the majority (>93 %) of H2A-H2B dimers displaced from the nucleosome still remains in the elongation complex “locally” as they show Cy5.5 within the complex that does not FRET with Cy3 or Cy5 (Fig. S7). This may be explained by the re-positioning of the dimer along the DNA. H2A-H2B binds DNA in the absence of the tetramer. Alternatively, it was proposed that electrostatic interactions between negatively-charged surfaces on Pol II can retain histones in the elongation complex^{17, 54}.

Figure 5. A three-color smFRET experimental system reports both DNA and H2A-H2B dynamics during and after transcription by Pol II.

(A) Nucleosomal DNA is labeled with a FRET pair, Cy3 and Cy5, at the +112^th and +34^th nucleotides from the entry site, to report DNA wrapping and unwrapping during transcription. H2B T112C is labeled with Cy5.5 maleimide to introduce a second acceptor fluorophore that reports histone H2A-H2B dissociation and reassociation. (B) Schematics show the fate of nucleosomes after transcription. See figure S7 for intensity changes of the three fluorophores induced by DNA and H2A-H2B dynamics during transcription. The vast majority of the nucleosomes are either fully or partially disassembled after transcription.

DISCUSSIONS

Our single-molecule FRET experimental systems enabled us to monitor the DNA motions and H2A-H2B dynamics in the nucleosome during transcription elongation in real-time in a time-resolved manner⁴². Based on the measurements, we investigated the roles of histone-DNA interactions in regulating the Pol II transcription elongation kinetics and examined the nucleosome structural changes. It has been suggested that Pol II does not actively dissociate nucleosomal DNA from the histone core due to the act of transcription^{20, 31}. These papers reported the frequency of transcription pauses in the presence and absence of a nucleosome based on a dual-trap optical tweezers setup where the transcription template is stretched through Pol II. According to the measurements, the pause frequency is increased significantly in the presence of a nucleosome, which cannot be explained if Pol II can actively break through a DNA-histone interface. Instead, Pol II gets stalled when it encounters a roadblock such as a strong DNA-histone interface^55–60. Fluctuations in DNA-histone interactions that occasionally induce a large histone-DNA gap would open a path for a paused Pol II to resume. Our results suggest that spontaneous opening of histone-DNA interfaces heavily impacts the efficiency of transcription by facilitating resumption of a paused Pol II. In the first set of experiments, we addressed the role of H3K56 acetylation in the progression of Pol II through the nucleosome. It has been established that H3K56 acetylation facilitates spontaneous DNA opening in the nucleosome entry region where K56 is located^{26, 30}. Consistent with this, we found that H3K56 acetylation dramatically shortens the pause duration in the entry region of the nucleosome by 4.0 folds. As H3K56 acetylation removes a positive charge on the αN helix and consequently weakens histone interaction with DNA termini, it will increase the frequency of spontaneous DNA opening and the size of opening on average. The frequency of opening increases by 50 % according to our previous report²⁶. This is a considerable impact although it cannot account for the 4.0-fold increase we observed here. This remarkable increase can be explained by the following mechanism as we proposed³⁰. The 50 % increase in the opening frequency is for openings with all possible sizes while the opening required for Pol II escape should be very large and rare. These rare large opening events make up the tailing edge of an approximate gaussian distribution as we modeled (Fig. S8)³⁰. This tailing edge decays dramatically as the opening size gets larger. Conversely, the tailing edge rises dramatically by shifting the entire distribution to the larger side. When the entire gaussian distribution shifts toward the larger side upon H3K56 acetylation, therefore, the frequency of tailing edge events (i.e. rare large opening events) will increase dramatically (Fig. S8). The remarkable 4.0-fold increase in the escape rate of paused Pol II strongly supports that paused Pol II requires rare large DNA opening ahead of it in order to escape. The frequency of such rare events can be significantly increased by a small increase in the average opening. Without invoking any unknown or mysterious source, this mechanism can explain how spontaneous nucleosome dynamics can play a major role in regulating the efficiency of nucleosomal transcription. This effect is significant because the overall efficiency of nucleosomal transcription is determined mainly by pause durations that last for tens to hundreds of seconds while the elongation rate is very fast at tens of milliseconds per nucleotide.

Previous reports suggest that nucleosomes reassemble upon Pol II passage in moderately transcribed genes^{17, 49–53}. Our results indicate that only <5 % of the nucleosomes reassemble during Pol II progression in real-time. The three-color single molecule FRET measurements report the same level of nucleosome reassembly and 8.0 ± 2.8 % histone H2A-H2B (one or both) recovery during transcription. These results indicate that most nucleosomes do not reassemble during transcription in real time. One major difference between our experimental system and the previously reported ones is that our system inherently filters out nucleosomal changes unrelated to Pol II progression. Any change in the nucleosome after Pol II run-off is not reported in our system because the nucleosome would have already left the Pol II. Therefore, our results indicate that Pol II mediates nucleosome reassembly very inefficiently during transcription. This conclusion does not rule out nucleosome reassembly after Pol II runs off completely, most likely utilizing a plethora of chromatin assembly factors and ATP-dependent chromatin remodeling complexes. According to our three-color measurements, the majority (>93%, 857 out of 915 nucleosomes) of the H2A-H2B dimers displaced during Pol II progression are still observed within the elongation complex. Therefore, it is feasible that Pol II or an unwrapped portion of the nucleosomal DNA harbors H2A-H2B during elongation and helps reassemble the nucleosome after Pol II passes. Further investigations remain in order to test whether extended DNA at the exit side of the nucleosome would make Pol II stay longer with the nucleosome and facilitate nucleosome reassembly. Of note, extended DNA may induce technical problems such as heterogeneity in nucleosome positioning which will require a major redesign of the experimental system.

Among the disassembled nucleosomes detected during transcription, some were identified as hexasomes. Structural and biochemical studies have established that hexasomes can survive transcription^{16, 49–52}. While our system allows us to detect hexasome generation during transcription, it is not designed to monitor any further changes of the hexasomes. Future studies can be designed to investigate the transcription pathways through hexasomes.

Histone chaperones mediate assembly and disassembly of the nucleosome and have been hypothesized that they may affect the kinetics of nucleosomal transcription^{7–11, 51}. In order to investigate the roles of a generic histone chaperone Nap1 in regulating nucleosome dynamics during transcription, we carried out smFRET measurements in the presence of Nap1. Our results indicate that Nap1 has no effects on the pause dynamics, the elongation rate, and the nucleosome reassembly efficiency during transcription. This result is consistent with ensemble biochemical studies that indicated Nap1 on its own did not affect Pol II progression through the nucleosome⁶¹. Instead, Nap1 only had an effect in the presence of the ATP-dependent chromatin remodeling complex RSC.

It might not be surprising that Nap1 failed to have an effect. Structural studies of Pol II on the nucleosome and its intermediates shed some light on possible explanations. In the paused state at SHL(−6), nucleosomal DNA is completely wrapped around the histone core²⁹. Nap1 cannot bind H2A-H2B in fully wrapped nucleosome³⁹, and therefore cannot impact on the SHL(−6) pause dynamics. Even in the paused state at SHL(−5) where only ~15 base pairs of DNA gets unwrapped²⁹, the interactions between DNA and H2A-H2B are mostly preserved^{29, 62, 63}. Moreover, the nucleosome and histone surfaces around the H2A-H2B region are sterically hindered by Pol II in the SHL(−5) pause, which will inhibit the action of Nap1. The next paused state at SHL(−2), albeit rare to form, may be shortened by Nap1 because a large surface area of H2A-H2B is exposed that can be a good target for Nap1 action. Unfortunately, we could not measure the pause duration at SHL(−2) in the presence of Nap1 because of the too small number of pauses observed at this location. This result may indicate that Nap1 does significantly shorten the duration of SHL(−2) pauses, resulting in extremely rare pauses observable. Regardless, the effect of Nap1 at this pause location should be minimal because it is already a very weak pause location.

We did not observe any change in the number of reassembled nucleosomes during Pol II progression in the presence of Nap1 despite its widely accepted function of nucleosome assembly^{36, 39–41}. As is mentioned above, our measurements record nucleosomal changes only during Pol II progression because the nucleosome will dissociate from Pol II after Pol II runs off. Therefore, our result indicate that Nap1 does not mediate nucleosome reassembly during transcription in real time although it does not rule out that Nap1 may promote nucleosome reassembly after Pol II passes through the nucleosome.

In conclusion, our results suggest that Pol II progression through a nucleosome is enabled by spontaneous opening of DNA-histone interfaces and that Pol II or histone chaperone does not mediate nucleosome disassembly or reassembly very efficiently during transcription in real time. The role of H3K56ac in facilitating transcription elongation is to shorten the pause duration in the entry region of the nucleosome, which is likely by accelerating the spontaneous opening of the DNA-histone interface in this region.