Conformations of linker histone H1 bound to nucleosome arrays

Abstract

Linker histone H1 plays crucial roles in nucleosome compaction and chromatin condensation. The highly basic C-terminal domain (CTD) of H1 interacts strongly with DNA, a critical aspect of its contribution to H1 function. Despite its critical roles in gene packaging in chromatin where nucleosomes are linked as an array, how H1 CTD interacts with nucleosome arrays remain poorly understood. Here we report a single-molecule FRET study of the conformation and conformational dynamics of the CTD of H1 bound to a 12-mer nucleosome array. According to our results, H1 CTDs within a nucleosome array show signs of highly heterogeneous conformations that are overall more extended and dynamic than that bound to a mono-nucleosome. This observation suggests that H1 CTD interacts randomly with two or more DNA linkers across nucleosomes in an array. This suggestion is further supported by our observation that these domains become more condensed and less dynamic as the arrays become less condensed at a lower NaCl concentration. Our results also suggest that histone H3 and H4 acetylation mimetics and tailless H3 result in H1 CTD interacting less with multiple linkers as they induce a less condensed structure of nucleosome arrays, thereby driving H1 CTD back to its own nucleosome. Our data support that H1 CTD interacts non-specifically with DNA linkers that are either local or distal and that modifications of the H3 and H4 tail domains can regulate H1-mediated chromatin condensation at both the nucleosome and nucleosome array levels.

Introduction

Genes in eukaryotes are packaged in chromatin that is composed mainly of DNA and histone proteins. Approximately 147 base-pair (bp) double-stranded (ds) DNA is first wrapped around an octameric histone protein core that comprises two copies of H2A-H2B dimers and one (H3-H4)₂ tetramer, forming a nucleosome core^1,2. Nucleosome cores are connected by DNA linkers to form nucleosome arrays in chromatin. The linker DNA between two nucleosome cores is variable and can be dynamically regulated for various genome activities^3-7. The wrapping of an approximate persistent length of DNA (~150 bp) into 1.67 helical turns around a nucleosome induces a high level of mechanical strain that is withstood by the strong electrostatic interactions between DNA and basic residues of the histone proteins^8,9. Due to this mechanical strain, the DNA near the nucleosome periphery spontaneously unwraps and re-wraps on a timescale of submilliseconds to milliseconds^10-13.

Linker histones (H1s) bind the periphery of the nucleosome core and the linker DNA and play crucial roles in the mechanisms of dynamic gene regulation^4,14-19. Linker histones have a tripartite structure with a central globular domain (GD, ~80 amino acid residues), a short unstructured N-terminal domain (NTD), and a longer unstructured (~100-120 residues) C-terminal domain (CTD)^20,21. The GD, which directs structure-specific binding of H1s to nucleosomes, interacts with the DNA helix that transits the nucleosome two-fold symmetry axis (nucleosome dyad) and bridges the DNA branches entering and exiting the nucleosome core. The CTD has been implicated in proper folding of chromatin fiber and it is rich with basic residues (~40 %) that can bind strongly to the nucleosomal and linker DNA via electrostatic interactions^22-25. The H1 CTD has the amino acid residue content emblematic of an intrinsically disordered domain and while some studies have suggested that a portion of the H1 CTD may form stable secondary structure(s) under certain conditions, many studies have shown that the CTD is disordered when H1 is free in solution^26-32. Recent studies have shown that the H1 CTD interacts non-specifically with linker DNA in nucleosomes, resulting in an ensemble of condensed conformations compared to the free protein, and that it can interact with a linker far from the nucleosome^33-36. In one of the studies, modifications of the histone H3 tail domain were shown to modulate the DNA-histone interactions near the nucleosome entry/exit regions and make the CTD conformation less compact and more dynamic³⁵.

Linker histones also play a crucial role in regulating chromatin condensation at the oligonucleosome array level, although H1 CTD interactions and conformational characteristics with nucleosome arrays remain largely unknown. Based on model systems utilizing 12-mer nucleosome arrays, a previous FRET study has suggested that the H1 CTD conformation upon binding to a nucleosome array is heterogeneous³³. Here we employed the same 12-mer nucleosome arrays and H1 with its CTD labeled with a FRET pair to study how H1 CTD interacts with a nucleosome array at a single-molecule level. Our results show that the H1 CTD adopts a mix of heterogeneous conformations that are on average more extended in nucleosome arrays than in mono-nucleosomes, suggesting that the H1 CTD interacts non-specifically with two or more DNA linkers across multiple nucleosomes. Our results also suggest that such multivalent interactions are reduced in arrays containing H3 and H4 tail acetylation mimetics or tailless H3. These results support roles for H1 CTD in regulating the overall condensation level of chromatin at a nucleosome array level that are distinct from its roles in compacting nucleosomes at a mono-nucleosome level.

Materials and Methods

Nucleosome array reconstitution

Plasmids containing a 12-mer 601 nucleosome array DNA sequence with a 207 nucleosome repeat length were purified from an E. coli cell stock as previously described³³. Briefly, E. coli cells containing the plasmid were grown overnight in LB medium, followed by cell lysis and DNA extraction with ethanol precipitation. The extracted DNA was digested with restriction endonucleases (HindIII and XbaI) to leave sticky ends. Short fragments of ds-DNA with complementary sequence to these sticky ends were purchased (Integrated DNA Technologies, Coralville IA). One of the fragments has a biotin labeled at the 5’ end for surface immobilization. These fragments were ligated to the digested plasmid DNA and purified with a centrifugal filter device to remove shorter digestion products. The final array DNA was confirmed with an agarose gel (see the “Naked DNA” lane on a gel shown in Fig. S1).

Nucleosome arrays were assembled by gradual and continuous salt dialysis of histone/DNA mixture at an optimum ratio against a buffer containing (10 mM Tris-HCl, 1mM EDTA, 1mM DTT,) (pH 7.5) from 1.5 M to 50 mM NaCl for 26 hours, followed by an additional 3 hour dialysis against the same buffer with 5 mM NaCl. To determine the ratio of histone/DNA, the array DNA was titrated with Xenopus laevis histone octamer (unmodified or modified) purchased from the Histone Source (Colorado State University) at varying histone/DNA ratios to test the histone saturation level at each ratio. The shift in the array DNA molar mass at each ratio was used to gauge histone saturation. As the ratio increases, the molar mass of the array increases and stops increasing at the optimum ratio. This ratio was selected as the optimum histone/DNA ratio for histone saturation of the array DNA (Fig. S1). To test for proper saturation, the assembled arrays were digested with a restriction endonuclease EcoRV, as the arrays contain a target digestion site in every linker DNA. Thus, properly saturated array will result in 12 nucleosomes while any unsaturated portion will result in a naked DNA of 207 bp. The digestion product is analyzed with native PAGE to confirm proper array assembly (Fig. S2).

Linker histone H1 purification and labeling

Wildtype Xenopus laevis H1.0 (referred to as H1 in this study) G101C/K195C double mutant was expressed in E. coli BL21 (DE3) and purified as described previously⁴⁶. Briefly, cells were grown in 50 ml LB media with 100 μg/ml ampicillin overnight at 37 °C. One ml of the culture was transferred to a liter of LB media (100 μg/ml ampicillin) and cultured at 37 °C until the OD reached 0.4 – 0.6. The culture was induced with 0.4 mM IPTG and incubated for another 3 – 4 hours at 37 °C, cells pelleted by centrifugation at 4,400g, resuspended into 20 ml of lysis buffer (10 mM Tris-HCl, 1 mM EDTA, 0.1 % TritonX-100, 0.1 M PMSF, 1 mg lysozyme) with a protease inhibitor cocktail (Roche) followed by incubation for 30 - 40 minutes at room temperature. The suspension was adjusted to 1 M NaCl and cells lysed with a Branson sonicator at 20 % amplitude for 60 seconds on and off cycles (3 – 5 times). Lysates were centrifuged at 20,000g for 30 minutes at 4 °C, and clear supernatants were subjected to purification by ion-exchange chromatography using Biorex-70 resin (Bio-Rad). The concentration of H1 was determined by quantitative comparison with an H1 standard, whose concentration had been determined by amino acid analysis as described earlier²⁹. H1 G101C K195C was labeled with a ~5-fold molar excess of a 50/50 mix of maleimido-Cy3 and maleimido-Cy5 (GE Healthcare) for 30 minutes at room temperature in the dark. The protein concentration of Cy3 and Cy5 labeled H1 was determined using quantitative SDS-PAGE gels by comparison with an H1 standard. The efficiency of labeling was determined by comparison to total fluorescence of dye standards and known labeling controls and is 90%.

Surface preparation for single-molecule measurements

We employed a supported lipid bilayer on a microscope slide to avoid non-specific binding of H1 as previously described^35,37-39. Briefly, a sub-monolayer of biotinylated poly-ethyleneglycol silane (biotin-PEG-silane) with molecular weight of 3400 g/mol (Laysan Bio, Arab AL) was first grafted on the surface of a clean microscope quartz slide by incubating a dry slide in an anhydrous acetonitrile solution of 20 pM bio-PEG-silane for 3 min followed by multiple washes with Nanopure^™ water. Upon drying with 99.99 % pure nitrogen gas, the slides are stored in a vacuum chamber. Right before using the slide, five fluidic channels of ~3 mm width separated by ~2 mm wide 0.1 mm thick double-sided tape pieces are constructed by sandwiching the channel layer with a cover glass. The volume of each channel is approximately 6 μL.

Lipid vesicles made of 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-550] (PEG550 PE) were prepared by sonicating a lipid suspension (2 mg/mL) in a buffer containing 10 mM Tris-HCl (p.H. 7.8) and 100 mM NaCl for multiple 1 min/1 min sonication/cooling cycles until the suspension becomes transparent. The solution was then extruded through a polycarbonate membrane filter of 100 nm pore size to produce lipid vesicles of a narrow distribution of sizes. The vesicle solution (2 mg/mL) is aliquoted (100 μL) and stored at −80 °C after flash freezing. An aliquot is thawed and diluted by 2-fold, and sonicated for 1 min. A flow channel is injected with 20 – 50 μL of the lipid solution and incubated for 45 min in a cold room (4 °C) for bilayer deposition. The flow channel is flushed with clean buffer.

Single-molecule FRET experiments and data analysis

We collected single-molecule fluorescence resonance energy transfer (smFRET) data following previously published protocols^37,38. Following is a brief description of the process. A fluidic channel passivated with a lipid bilayer is flushed with a buffer containing Tris-HCl (pH 8) and 100 mM NaCl. The clean channel is injected with a streptavidin solution in PBS (0.1 mg/mL) and incubated for 10 min. Free streptavidin is flushed with the PBS buffer twice. The channel is then injected with a nucleosome array solution (5 pM) and incubated for 10 min, followed by 2x wash with an imaging buffer (10 mM Tris-HCl pH 8, 1 mM EDTA, 2 mM Trolox, 0.8 mM PCD, 8 mM PCA, 0.1 mM BSA, and 100 mM NaCl). Next, the channel is incubated with a FRET-labeled H1 solution (mixture of fluorophore-labeled and unlabeled H1 at a molar ratio of 1:11) at a total concentration of 60 pM in a buffer containing 10 mM Tris-HCl pH 8, 1 mM EDTA, 2 mM Trolox, 0.8 mM PCD, 8 mM PCA, 0.1 mM BSA, and 5, 20, or 50 mM NaCl for 10 min, followed by 2x wash with the same buffer. To determine the optimum labeled:unlabeled H1 ratio, ratios of 1:12, 1:11, 1:10, and 1:9 were tested. The ratio 1:11 yields the maximum number of spatially resolved single H1 molecules exhibiting FRET.

The surface of the fluidic channel is illuminated with an evanescent field of a 532 nm laser (CrystaLaser, Reno NV) via a total internal reflection geometry and the fluorescence images with a 100 ms signal integration time were continuously recorded with an EMCCD camera (Ixon Ultra 897 from Andor Technology, Belfast Ireland) every 100 ms, producing an image stack or a movie. Movie recording is stopped when the majority of Cy3 fluorescence spots are photobleached, which typically takes 2 – 3 minutes. Each movie contains ~100 strong fluorescent spots. Due to the labeling scheme, only two thirds of the strong Cy3 fluorescent spots contain a FRET pair. Also, due to the stochastic nature of the labeled:unlabeled H1 ratio in each array, many FRET spots contain two or even three Cy3 fluorophores, which are filtered out from further analysis. From each movie, ~30 H1 molecules emitting FRET signals from a single Cy3-Cy5 pair are identified. Out of these, only those with a signal-to-noise ratio higher than 3 are selected to minimize measurement errors. In the end, less than 20 H1 molecules per movie are typically analyzed further. Each FRET pair is analyzed to produce a time trajectory of the fluorescence intensities from Cy3 (FRET donor) and Cy5 (FRET acceptor) at every 100 ms time point, and their FRET efficiency. The FRET efficiency is approximated with the Cy5 intensity divided by the total intensity from both Cy3 and Cy5 (i.e., FRET efficiency $\approx {\mathrm{I}}_{\text{Cy}5}∕({\mathrm{I}}_{\text{Cy}3}+{\mathrm{I}}_{\text{Cy}5})$, where $\mathrm{I}$ is the fluorescence intensity and the subscript marks the fluorophore). The FRET efficiencies while both fluorophores are active before photobleaching are combined together to construct a histogram of their distribution or a swarm plot.

Results

The CTD of H1 bound to a nucleosome array adopts a mixture of heterogeneous conformations that are overall more extended than that bound to a mono-nucleosome

We constructed a single-molecule fluorescence resonance energy transfer (smFRET) experimental system where we can monitor the CTD conformations of a single H1 that is bound within a nucleosome array (Fig. 1A). We used a 12-mer nucleosome array based on 12 repeats of the Widom 601 nucleosome positioning sequence with a nucleosome repeat length (NRL) of 207 bp. We chose 207 bp NRL because this spacing provides physiologically relevant linker DNA for stable H1 binding at nucleosomes regardless of their stacking structures, which allows reproducible higher-order chromatin condensation^40,41. One end of the array DNA was digested with a restriction endonuclease that leaves a sticky end for ligating a short biotinylated DNA fragment with the complementary sticky end. Nucleosome arrays were reconstituted via gradient salt dialysis at an optimum histone:DNA ratio determined by titration (see details in Materials and Methods). The reconstituted arrays were confirmed by restriction endonuclease digestion assays to show proper saturation with histones. The nucleosome arrays were immobilized via biotin-streptavidin-biotin conjugation onto a microscope slide surface that was passivated by a lipid bilayer (Fig. 1A). We labeled H1 G101C/K195C with Cy3 and Cy5, placing a FRET pair near the two ends of the CTD to monitor its overall condensation level as previously reported^33,35. Prior to binding to nucleosome arrays, we mixed labeled and unlabeled H1 at a ratio of 1:11 to ensure that each array gets about one fluorescently-labeled H1 CTD. The ratio was initially calculated approximately according to the labeling efficiency of H1 and then eventually determined by titration to maximize the number of single fluorescently labeled H1s observable in our smFRET setup (Fig. 1B).

Figure 1.

Experimental setup and examples of experimental data. (A) Experimental setup with 12-mer nucleosome arrays immobilized on a microscope slide surface that is passivated with a lipid bilayer. Near the ends (G101C/K195C) of the C-terminal domain (CTD) of H1 are labeled with a FRET pair Cy3 and Cy5 which report its overall condensation level. (B) A typical fluorescence image of the H1 molecules immobilized on a microscope slide. (C) Examples of smFRET time trajectories from single H1 molecule bound to a nucleosome array, showing a dynamic (upper) and a static (lower) CTD conformation. (D) A FRET histogram from 330 H1 molecules reports the overall condensation level of CTD at 50 mM NaCl.

Fluorescence signals from arrays containing a single labeled H1 were confirmed by single-step photobleaching events and were analyzed to obtain time trajectories of FRET efficiencies at a temporal resolution of 100 ms, the time resolution of FRET signal acquisition (Fig. 1C). No fluorescence spots were detected in the absence of nucleosome arrays, confirming that the signals are from array-bound H1 molecules. The vast majority of the H1 molecules observed (310 out of 330 molecules) show dynamically changing FRET efficiencies among various FRET values (Fig. 1C upper), which is a strong indication of a mixture of heterogeneous CTD conformations. The rates of FRET changes under this condition (50 mM NaCl) are 0.50 and 0.30 /s, for up and down transitions, respectively, according to HMM analyses (Figure S3). These rapid transitions correspond to one FRET-up event per every 2.0 sec and one FRET-down event per 3.3 sec (see the full data in Figure S3). The FRET trajectories from a total of 330 H1 molecules were combined to construct a FRET histogram showing the distribution of the FRET efficiencies from the heterogeneous conformational mixture (Fig. 1D). Even the small number of H1 molecules showing static FRET (Fig. 1C lower) display a wide distribution of FRET values (Fig. S4), further supporting the conformational heterogeneity. These observations suggest each H1 rapidly explores a wide distribution of CTD-DNA interactions within a nucleosome array. Each count in the FRET histogram (Fig. 1D) is a single occurrence of FRET with a 100 ms duration. Therefore, each count in the histogram represents the result of an independent observation event of the CTD conformation for 100 ms, and their average value should represent the ensemble average of the heterogeneous H1 CTD conformations. The distribution shows an average FRET value of 0.45. This value is far lower than 0.63 which is the average smFRET efficiency reported from H1 molecules bound to a mono-nucleosome³⁵. This result agrees well with the previously reported ensemble FRET value (0.4 – 0.5) from H1 molecules labeled at the same FRET locations mixed with nucleosome arrays under the same salt concentration (50 mM NaCl)³³. These results combined indicate that H1 CTD adopts an ensemble of rapidly interchanging and heterogeneous conformations that are on average more extended when H1 binds nucleosome arrays than mono-nucleosomes. Considering that a H1-bound nucleosome array adopts a condensed structure at 50 mM NaCl⁴², we suggest that the overall more extended conformations of H1 CTD are due to interactions with multiple DNA linkers across nucleosomes in an array. Such interactions would help reduce inter-nucleosomal distance, thereby contributing to chromatin condensation at a nucleosome array level.

H1 CTD assumes more condensed conformations at lower NaCl concentrations

In order to explore whether the proximity of distal linkers influences H1 CTD structure, we repeated the smFRET measurements at lower levels of NaCl at 5 mM and 20 mM (Fig. 2). No fluorescence spots were detected in the absence of nucleosome arrays, confirming that the signals are from array-bound H1 molecules at these NaCl concentrations. Nucleosome arrays in the presence of H1 in vitro are condensed to mimic a fibril structure at 50 mM NaCl while they adopt a loose zig-zag structure at 5 mM NaCl and a contacting zig-zag structure at ~20 mM NaCl^43-45. These conclusions are based on experiments with hexa-nucleosome or longer arrays and various experimental techniques including ultracentrifugation and electron microscopy. Of note, recent cryo-EM structures of short tetra-nucleosome arrays show inter-nucleosomal stacking even at a very low salt level⁴⁰. However, inter-nucleosomal stacking in a tetra-nucleosome array is likely facilitated by the relatively freely moving nucleosomes in a short array, not accurately capturing inter-nucleosomal interactions in a longer array where nucleosome motions and array conformations are restricted. Thus, our 12-mer nucleosome arrays at 5 and 20 mM NaCl likely have more distantly spaced nucleosomes than at 50 mM NaCl, raising the possibility that the H1 CTD interacts less with DNA from neighboring nucleosomes at these lower NaCl concentrations. Accordingly, we observed less extended H1 CTD conformations for the less condensed arrays, as indicated by the greater number of high-FRET events observed at the lower NaCl levels (Fig. 2). We examined any effect of [NaCl] variations on the H1-CTD conformation at the mono-nucleosome level to observe no significant FRET differences among the cases of 5, 20, and 50 mM NaCl (Fig. S5). Furthermore, freely diffusing H1 molecules in the absence of nucleosome arrays do not show any significant FRET variations in the range of 5 – 50 mM NaCl (Fig. S6). These results confirm that the differences in the array-bound H1 molecules induced by [NaCl] variations are due to the changes in the conformation of the arrays and their interactions with the H1-CTD. The differences are significant according to the student’s t-test (two-sided) while the average FRET value increases by a relatively small margin, demonstrating the distinct benefit of sub-population analyses with our single-molecule measurements. The changes in the population density of the higher FRET conformations (FRET > 0.45 = average FRET) are 23 % from 50 mM to 5 mM and 21 % from 50 mM to 20 mM, respectively. The numerators of these values are the differences in the high FRET populations between two cases and the denominators are the entire FRET populations of the 5 mM and 20 mM cases, respectively. These results support that H1 CTD interacts with multiple DNA linkers across nucleosomes within a nucleosome array and that such interactions are less prevalent at 5 and 20 mM NaCl. This inhibition is likely due to the weak charge screening effect of the lower NaCl concentrations, resulting in hindered inter-nucleosomal packing in nucleosome arrays^46,47. Our observations suggest that a proper ionic strength is required for H1 to contribute to chromatin condensation at a nucleosome array level.

Figure 2.

Changes in H1 CTD conformations induced by lower NaCl concentrations. (A) (upper) Three smFRET histograms reveal a larger higher-FRET population at a lower NaCl concentration. (lower) Three histograms of increased high-FRET (FRET > 0.45 = average FRET of the 50 mM case) events at a lower NaCl level show considerable population increases. The numbers of H1 molecules observed at 5, 20, and 50 mM NaCl are 174, 197, and 330 molecules, respectively. (B) Swarm plots of the smFRET events shown in A show statistically significant differences according to two-sided student’s t-tests (***: p<0.001). The table shows the average FRET values of the three cases.

More extended CTD conformations correlate with an increased population displaying dynamic FRET

Most H1 molecules (314 out of 330 molecules or 95.2 ± 1.2 % of the total H1 CTD observed at 50 mM NaCl) show dynamically changing smFRET signals at our observation time resolution of 100 ms (Fig. 1C upper) while a small number of H1 molecules (16 out of 330 molecules or 4.8 ± 1.2 %) show a static FRET level (Fig. 1C lower). We counted static FRET trajectories in two steps. First, we classify that a FRET trajectory is potentially static when there are no visually identified FRET changes accompanying a change in the Cy3 fluorescence intensity anti-correlated with a concomitant and opposite change in the Cy5 intensity until one of the fluorophores is photobleached. After the classification, for each potentially static FRET trajectory, we examine at every time point whether a FRET efficiency change is equal or larger than the 2σ (σ = standard deviation) of the FRET signal with anti-correlated Cy3 and Cy5 intensities. The standard deviation is calculated from the FRET efficiencies at all time points where the trajectory shows “static” FRET before photobleaching. If the trajectory passes this automated test, we count the trajectory as static. This scheme ensures, with 95% confidence, that no apparent FRET transitions occur in each of these trajectories. Our observations of mostly dynamic FRET trajectories indicate that the CTD of an H1 molecule bound to a nucleosome array has dynamically changing conformations. When the NaCl concentration is decreased to 20 mM and 5 mM, the population showing a static FRET state increases to 10 ± 2 % (19 out of 197 molecules) and 12 ± 2 % (20 out of 173 molecules). The errors are according to the binomial distribution statistics. At a lower NaCl concentration, the arrays are less condensed, which would restrict the H1 CTD to more local interactions, resulting in it being less extended overall (i.e., higher average FRET). Conversely, a larger dynamic FRET population at a higher NaCl concentration suggests that more distal CTD-DNA interactions (i.e., lower average FRET) are correlated with more dynamic conformations of H1 CTD. This suggestion is further supported by the larger population of H1 CTDs within arrays showing dynamic FRET (95.2 ± 1.2 % at 50 mM NaCl) than that bound to a mono-nucleosome (52.2 ± 2.7 % at 50 mM NaCl) as was previously reported³⁵. These results support that the H1 CTD in nucleosome arrays becomes more dynamic as it extends to interact with distal regions of DNA across nucleosomes that become more accessible at higher salt conditions. On the contrary, H1 CTD becomes more static at a lower salt as it becomes more condensed by interacting more with local DNA.

H3 acetylation mimetics, tailless H3, and H4 acetylation mimetic affect H1 CTD conformations in a salt-dependent manner

We explored how the changes in the nucleosome and nucleosome arrays induced by H3 acetylation or removal of the H3 tail would affect H1 CTD binding and conformations in nucleosome arrays. Histone H3 tail acetylation weakens H3 tail-DNA interactions and is typically associated with gene activation and active transcription^48-50. To test the effects of H3 acetylation, we employed an acetylation mimetic of H3, H3 6KQ, where six acetylation sites (K4, K9, K14, K18, K23, and K27) are mutated to glutamine^35,51. Tailless H3 (gH3) has also been previously employed to study the roles for H3 N-terminal tail in regulating DNA-histone interactions and H1 CTD conformations^35,51. When H1 binds a nucleosome array containing H3 6KQ or gH3 at 50 mM NaCl, its CTD shows a higher average FRET and less dynamic FRET (Figs. 3 and 4). At the nucleosome array level, these changes are consistent with less H3 tail bridging interactions between nucleosomes upon these H3 modifications⁵², which results in less condensed nucleosome arrays, thereby inducing more local than distal CTD-DNA interactions. At the nucleosome level, these modifications should make a wider region of entry/exit DNA available for H1-CTD interactions due to weakened nucleosome wrapping. Such effects should result in more extended and dynamic H1 CTD conformations overall (i.e., lower and more dynamic FRET) as was previously reported³⁵. Our results likely reflect contributions from both nucleosome structural changes and array condensation. According to the results, the dominant effects of these H3 modifications on H1-CTD conformations at 50 mM NaCl arise from weakened nucleosome array condensation.

Figure 3.

Changes in H1 CTD conformations induced by histone H3 and H4 modifications. (A) (upper) Three sets of four smFRET histograms reveal a change in the higher-FRET population induced by H3 6KQ, gH3, and H4 5KQ modifications. (lower) The histograms of increased high-FRET (FRET > 0.45 = average FRET of the unmodified arrays at 50 mM NaCl) events show considerable population increases. The numbers of H1 molecules observed at 5, 20, and 50 mM NaCl are 132, 135, and 168 molecules for H3 6KQ, 128, 82, and 103 molecules for gH3, and 162, 133, and 133 molecules for H4 5KQ. (B) Swarm plots of the smFRET events shown in (A) show statistically significant differences according to two-sided student’s t-tests (***: p<0.001). (C) Average FRET efficiency values of the cases shown in (A).

Figure 4.

Population densities of H1 molecules showing dynamic smFRET trajectories bound to nucleosome arrays containing unmodified (WT), H3 6KQ, gH3, and H4 5KQ modified histone. The table shows the average values of the dynamic FRET population densities. The numbers in the parentheses are the number of dynamic molecules out of total number of molecules analyzed in each case. The standard deviations are according to the binomial distribution statistics.

At 5 and 20 mM NaCl, the average FRET difference between WT arrays and arrays with modified H3 becomes smaller and even reversed. In contrast to arrays in 50 mM NaCl, H1 CTDs became more condensed (exhibiting higher FRET values) when arrays containing modified H3s were in low salt conditions (Fig. 3). A decrease in the dynamic FRET population is also observed upon H3 modifications (Fig. 4). We suggest that these changes are due to the following reasons. At these lower salt concentrations, nucleosome arrays adopt a less condensed structure with fewer nucleosome-nucleosome interactions (Fig. 5). Therefore, we expect that the effects of these modifications would resemble the effects observed at the mono-nucleosome level, wherein such modifications result in more extended CTD conformations overall (i.e., lower FRET)³⁵.

Figure 5.

Summary of results and a schematic model for H1 CTD conformation and dynamics in nucleosome arrays. (A) A summary chart constructed with the average FRET values listed in Fig. 3C. (B) A schematic model to summarize the conformational changes and dynamics induced by NaCl and histone modifications. Nucleosome array condensation induces extended and dynamic H1 CTD.

Somewhat surprisingly, we observed a decreased population of H1 CTDs exhibiting dynamic FRET at lower NaCl concentrations with H3 6KQ and gH3 arrays (e.g., 80 ± 3 and 83 ± 3 % respectively for H3 6KQ and gH3 arrays at 5 mM NaCl) compared to WT arrays (e.g., 88 ± 2 % at 5 mM NaCl) (Fig. 4). In contrast, in mono-nucleosomes the same H3 modifications result in more dynamic H1 CTD conformations compared to WT nucleosomes^35,53. However, even the H1 CTDs in the H3-modified mono-nucleosomes are still less dynamic (77 ± 5 and 69.1 ± 3.0 % respectively for H3 6KQ and gH3) than those in the H3-modified arrays (80 ± 3 and 83 ± 3 % respectively for H3 6KQ and gH3 arrays at 5 mM NaCl), confirming the effect of these modifications on making H1 CTD more dynamic regardless³⁵.

We assume that the dynamic FRET populations of CTD in mono-nucleosomes would only decrease at a lower NaCl level, as a lower ionic strength would only attenuate the DNA-histone dynamics in a nucleosome. Therefore, a shift from distal to local DNA-CTD interactions would only result in overall a decreasing dynamic FRET population. Of note, no significant changes are observed in the dynamic population density across varying salt levels for either H3 6KQ or gH3 modified arrays, although we are tempted to speculate that two conflicting effects of these modifications on regulating array condensation and nucleosome compaction may have opposing influences as the ionic environment varies. Regardless, our observations of CTD conformations in nucleosome arrays suggest that H3 acetylation regulates how H1 CTD interacts with DNA in nucleosome arrays and eventually how H1 condenses chromatin structure at a nucleosome array level.

Histone H4 tail lysine residues are implicated in intra- and inter-nucleosomal histone-DNA and histone-histone interactions to stabilize higher-order chromatin structures^1,54,55. Acetylation of these residues have been reported to weaken these effects in model nucleosome arrays in both inter- and intra-array contexts^56-58. We employed an H4 tail acetylation mimic H4 5KQ which contains lysine to glutamine mutations at five tail lysine residues (K5, K9, K12, K16, and K20). The residue K20 was included for its recently reported role in compacting the H4 tail⁵⁹. The effects of H4 5KQ mimetic on the conformations of H1 CTD are overall similar to those of H3 6KQ and gH3 (Fig. 3). At 50 mM NaCl where nucleosome arrays are condensed, the average FRET increases and the dynamic FRET population decreases, supporting more local DNA-CTD interactions upon H4 tail acetylation likely due to less condensation of nucleosome arrays. At 20 mM and 5 mM NaCl, the average FRET decreases and the dynamic FRET population also decreases, similarly to the changes induced by H3 6KQ and gH3. One significant change of note in FRET dynamics is a decrease in the dynamic FRET population as the NaCl level is lowered to 5 mM from 50 mM. This observation suggests that H4 5KQ induces more local DNA-CTD interactions at 5 mM NaCl than at 50 or 20 mM NaCl. This change may indicate that H4 5KQ effect on relaxing inter-nucleosomal interactions is stronger at 5 mM than at 50 or 20 mM NaCl as is anticipated from their electrostatic nature. Overall, our results suggest roles for H4 tail acetylation in regulating gene packaging via H1 CTD interactions with nucleosomes in an array context.

Discussions

Our smFRET experimental system to study H1 CTD conformations in nucleosome arrays provided two distinct benefits. One is that we can filter out any signal from aggregated H1 or arrays by excluding FRET trajectories showing multiple fluorophores. Histones are naturally prone to aggregation in aqueous solution and their characterization is often complicated by such aggregation. The other is that we can investigate sub-populations of H1 CTD conformations and their dynamics to pursue deeper understanding of how H1 CTD interacts with DNA in nucleosome arrays. Our data strongly support that H1 CTD adopts more extended and dynamic conformations when bound to a nucleosome array than to a mono-nucleosome. This conclusion is based on the decreased average smFRET efficiency showing more dynamic changes, which is consistent with previous reports with ensemble and single-molecule FRET measurements^33,35. In the previous ensemble study, an H1 CTD construct labeled at the same FRET locations was employed to measure the ensemble FRET efficiencies at varying NaCl concentrations with the same 12-mer nucleosome arrays. The efficiency of intra-CTD FRET was 0.42 at 50 mM NaCl, which agrees well with our result of 0.45. With the same H1 CTD FRET-labeling scheme, an smFRET investigation when H1 CTD binds a mono-nucleosome was also carried out³⁵. According to the results, the FRET efficiency distributes widely, covering the entire range of FRET, revealing a heterogeneous ensemble of CTD condensation levels and conformations, which agrees well with our current observations with nucleosome arrays. On average, however, the FRET efficiency is distinctly lower (0.42 vs 0.63) when CTD is bound to a nucleosome array than to a mono-nucleosome. These results strongly support that H1 CTD is overall more extended when it binds nucleosome arrays than mono-nucleosomes. This can be due to the different DNA linker conformations in an array than those in a nucleosome and/or to the CTD interacting with two or more DNA linkers from multiple nucleosomes. It seems unlikely that variation in linker conformations in a condensed nucleosome array would account for the observed wide distribution in FRET efficiencies, suggesting additional contributions likely provided by diverse CTD interactions to juxtaposed linker DNAs of neighboring nucleosomes (Fig. 5B). Therefore, we suggest that the lower average FRET in an array is due to such more distal CTD-DNA interactions which add another layer in regulating the chromatin structure by H1 at a nucleosome array level.

We also observed mostly dynamic H1 CTD conformations at a time resolution of 100 ms. According to our previous report on H1 CTD FRET conformations in mono-nucleosomes, H3 6KQ and gH3 induce lower and more dynamic FRET, indicating a more extended H1 CTD that is also more dynamic. Some of the anticipated changes induced by H3 6KQ and gH3 in the nucleosome structure are correlated with more open and dynamic conformations near the edge of the nucleosome³⁵. These changes would provide an additional region of DNA that H1 CTD can bind, thereby extending the CTD and inducing lower FRET on average. According to the increased dynamic FRET population with these changes, such extended conformations of the CTD are not as stable and exhibit dynamically interconverting conformations at the mono-nucleosome level. On the basis of these results, we suggest that the more extended H1 CTD in arrays is due to its interactions with neighboring linker DNA and is correlated with more dynamic conformational changes. Therefore, we conclude that the mostly dynamic FRET population of H1 CTD bound with nucleosome arrays correlates well with the more extended CTD conformations due to these distal DNA-CTD interactions. This conclusion agrees well with the more condensed and less dynamic H1 CTDs observed at the lower salt concentrations (5 and 20 mM), as H1 CTD interacts more with local than distal linker DNA in a looser or a less condensed array.

The changes in the average FRET efficiency induced upon H4 modification (H4 5KQ) at all NaCl concentrations are similar to those observed with H3 6KQ and especially gH3. This is likely because arrays adopt a less condensed conformation at these NaCl concentrations where an H1 CTD would interact more with its local linker DNA. Based on these results, we suggest that the effects of these modifications on DNA-CTD interactions in condensed nucleosome arrays are exerted more strongly via modulating the nucleosome array condensation than via regulating the nucleosome termini conformations. Combined with a previous report³⁵, our results support that H3 and H4 tail acetylation affects H1 CTD interactions with DNA to regulate gene compaction at both mono-nucleosome and nucleosome-array levels with their relative contributions depending on the condensation level of chromatin.

The majority of H1 CTD shows fluctuating FRET signals indicating dynamic conformational changes with some statistically significant shifting in the population density upon NaCl concentration variations or histone modifications. The changes in the unmodified arrays induced by decreasing NaCl levels or histone modifications are in line with previous results, suggesting that distal DNA-CTD interactions are more dynamic than local DNA-CTD interactions with or without H3/H4 modifications. Unlike H3 modifications, H4 5KQ results in a significantly decreasing trend in the dynamic FRET population as the NaCl level is lowered, suggesting that H4 acetylation has a stronger impact than H3 modifications on the nucleosome array structure at a lower NaCl level.