Histone dynamics play a critical role in SNF2h-mediated nucleosome sliding

Elucidating the mechanisms by which ATP-dependent chromatin remodeling enzymes disrupt nucleosome structure is essential to understanding how chromatin states are established and maintained. We have previously demonstrated that dynamics in the histone core are functionally important for nucleosome sliding by the ISWI-family remodeler SNF2h¹. Restraining histone octamer dynamics by site-specific disulfide crosslinks engineered in otherwise cysteine-free histones inhibited nucleosome sliding by SNF2h. Using similar approaches, others have also suggested that octamer plasticity is important in nucleosome sliding^2,3. However, a subsequent study by Chen and colleagues that reported cryo-EM structures of ISWI-family remodeler-nucleosome complexes failed to observe stable conformational rearrangements in the histone octamer⁴. The authors of this study were also unable to replicate the finding that restraining histone dynamics by single H3-H4 crosslinks (H3L82C-H4V81C, called sCX2) impacted nucleosome sliding by SNF2h¹. Initially, the authors generated sCX2 crosslinks in the background of WT Xenopus laevis histone octamers, which bear an endogenous cysteine in histone H3 (H3C110). However, the presence of the additional reactive cysteine, which can readily form disulfides under oxidizing conditions⁵, creates the possibility for multiple types of crosslinked species, complicating the interpretation of these results. Yan et al.4 also created sCX2 crosslinks in an H3C110A mutant octamer and, in this case, found that SNF2h sliding activity was impaired, as seen by Sinha et al.¹, but to a lesser extent than previously observed. Further, when the disulfide bond was apparently reduced by adding 100mM DTT to nucleosomes, sliding activity was not restored. Based on these results, Yan et al. suggested, “…that the loss of activity did not result from the disulfide bond but probably from non-specific oxidation damage to the nucleosome, which was prone to precipitate out of the solution under the condition of extensive oxidation. Prolonged treatment might have resulted in damage to the nucleosome and loss of the activity.”⁴ However, the authors did not directly test this possibility. To address this concern experimentally we first assembled nucleosomes with cysteine-free Xenopus laevis H3C110A octamers that were oxidized with CuPhe (Extended Data Figure 1A). No precipitates were observed during preparation and the oxidized octamers readily assembled into canonical nucleosomes, suggesting no gross defects. Remodeling of these oxidized nucleosomes with SNF2h after purification by glycerol gradient ultracentrifugation showed no defect compared to nucleosomes assembled from untreated H3C110A octamers (Table 1, Figure 1A).

Table 1.

Rate constant (k_obs) for nucleosome sliding of the given substrate at different concentrations of SNF2h under buffer conditions used in Sinha et al.¹

Octamer used to assemble nucleosome	kobs
[SNF2h]	1μM	50nM
H3C110A Untreated	1.6±0.1 min−1	0.03±0.01 min−1
H3C110A CuPhe Oxidized	1.8±0.1 min−1	Not Determined
H3C110A CuPhe Oxidized and then reduced	2.3±0.1 min−1	Not Determined
H3C110A sCX2 CuPhe Oxidized	0.06±0.03 min−1	Undetectable
H3C110A sCX2 CuPhe Oxidized and then reduced	2.5±0.4 min−1	0.03±0.02 min−1
H3C110A sCX2 GSSG Oxidized	0.14±.02 min−1	Not Determined
H3C110A sCX2 GSSG Oxidized and then reduced	2.2±0.6 min−1	Not Determined

Figure 1. Site-specific cysteine crosslinking, and not general oxidative damage, inhibits remodeling regardless of oxidation method.

(A) Treatment of H3C110A histone octamer with CuPhe or GSSG does not affect nucleosome sliding by SNF2h. Left. Native gel sliding assay with saturating concentrations of SNF2h (1μM) with or without saturating ATP and 15nM cy3-labeled nucleosomes using the indicated histone octamer. Time points were quenched with excess ADP and plasmid DNA and resolved on a 6% (29:1 bis) acrylamide gel. Higher migrating species are more centrally-positioned nucleosomes. Right. Quantification of the data shown at the left plotted as the fraction of end-positioned nucleosomes over time. The experiment was performed 3 times independently with similar results. (B) Left. Native gel remodeling assay as in A. Right. Quantification of the data shown at the left plotted as the fraction of end-positioned nucleosomes over time with the plot zoomed in to the first 10 minutes of the reaction to better evaluate fits. This experiment was performed 3 times independently with similar results. (C) Mean observed rate constants (k_obs) from 3 independent experiments. Error bars reflect the standard error of the mean (SEM). Uncropped images for panels A-B are available as source data online and the numerical values plotted in the graphs are provided in Supplementary Table 1. In all figure panels “Reduced” nucleosomes refer to nucleosomes that were assembled from octamers oxidized using the indicated method and then subjected to reducing conditions before nucleosome assembly.

To test whether the oxidation method may impact the results in the context of H3C110A sCX2 octamer, we generated H3C110A sCX2 crosslinked nucleosomes using CuPhe or oxidized glutathione (GSSG), a gentler oxidizing agent that does not generate free radicals and has been used previously to investigate nucleosome dynamics⁶. Oxidation with CuPhe went to near-completion as assessed by a non-reducing SDS-PAGE gel, while GSSG oxidation was less efficient (Extended Data Figure 1A). We then reduced a portion of each H3C110A sCX2 crosslinked octamer by dialysis into buffer containing 100mM DTT (Extended Data Figure 1B). H3C110A sCX2 octamers showed no visible precipitates upon oxidation or after reduction and, like cysteine-free octamers, readily assembled into nucleosome core particles that were purified by glycerol gradient ultracentrifugation. H3C110A sCX2 nucleosomes assembled from octamers oxidized with CuPhe slowed SNF2h nucleosome sliding ~40-fold compared to reduced controls under saturating concentrations of SNF2h, while nucleosomes assembled from octamers oxidized by GSSG slowed sliding ~15-fold (Figures 1B and C). This quantitative difference can be attributed to the different crosslinking efficiencies of the two oxidation protocols and cannot be explained by nucleosome disassembly (Supplemental Figures 1 and 2). Reduction of either type of oxidized octamer fully restored SNF2h-dependent sliding activity of nucleosomes assembled from their respective octamers (Figures 1B and C). To further test whether the oxidation-dependent impact on SNF2h-dependent sCX2 nucleosome remodeling is due to the formation of a disulfide bond, we also tested remodeling on nucleosomes assembled using oxidized H3C110A octamers containing single cysteines from the sCX2 cysteine pair that were purified by ultracentrifugation (Extended Data Figure 2A). In this side-by-side experiment, SNF2h remodeled CuPhe oxidized H3C110A sCX2 nucleosomes ~30-fold slower than untreated H3C110A nucleosomes in line with our earlier observations, while single cysteine-containing nucleosomes were remodeled only ~2-fold slower. Oxidation-dependent remodeling defects were also observed with sub-saturating concentrations of enzyme (Extended Data Figure 2B). These results indicate that non-specific oxidative damage to the octamer cannot explain the remodeling defects.

In order to determine whether any differences in the remodeling assay protocols could explain the discrepancies between the two studies, we repeated the remodeling reaction under the conditions used by Yan et al. ⁴. While the remodeling reactions were faster overall (Extended Data Figure 2C), even under these conditions a disulfide bond between the sCX2 cysteine pair causes a>6-fold reduction in reaction rate (Extended Data Figure 2C). Due to the faster remodeling reaction, nucleosomes are already ~60 % remodeled at the first time point we are able to capture (0.3 min). As a result, the rate constant reported here for non-oxidized nucleosomes should be considered an underestimate and the 6-fold defect should be considered a lower limit.

These results (summarized in Table 1), (i) rule out non-specific octamer oxidation causing inhibition of sliding; (ii) reproduce the inhibitory effects seen by Sinha et al. when the H3L82C-H4V81C disulfide bond is formed in an H3C110A nucleosome; and (iii) show that this defect is reversible when the oxidized octamer is reduced prior to nucleosome assembly.

To note, Yan et al. tested reversibility in a different manner than described above. Specifically, while their oxidation reaction was carried out in the context of an octamer, their reduction reaction was carried out in the context of a nucleosome. In our experience, reducing buried disulfides in the context of a fully assembled nucleosome is difficult, likely because the disulfide crosslink is less solvent accessible in a nucleosome compared to an octamer. When we reduced crosslinked H3C110A sCX2 nucleosomes following the same protocol as Yan et al. (treatment of crosslinked nucleosomes with 100mM DTT at 37°C for 1hr), we found that SNF2h sliding activity was not restored, consistent with their report (Extended Data Figure 3A). Although Yan et al. showed these conditions were sufficient to completely break the disulfide bond when assayed by non-reducing SDS-PAGE, we reasoned that if the sample was not buffer exchanged to remove excess DTT prior to denaturation in SDS-PAGE loading buffer, once-buried disulfides could become exposed upon denaturation and rapidly reduced by the residual DTT. This would cause an overestimation of the degree of reduction by SDS-PAGE. To test this possibility, we reduced H3C110A sCX2 crosslinked nucleosomes with 100mM DTT for 1hr at 37°C and either directly added the sample to SDS loading buffer or quenched the residual DTT with 5-fold excess N-ethyl maleimide before SDS treatment. While in the unquenched sample the disulfide bond is completely reversed, quenching the sample prior to SDS treatment causes retention of the disulfide bond (Extended Data Figure 3C). This retention cannot be explained by re-oxidation of the disulfide bond upon quenching as N-ethyl maleimide would also react with cysteine thiols, blocking new disulfide bond formation. In addition, oxidation of the cysteines within octamers in the absence of catalysts such as CuPhe or GSSG is extremely slow (≥48 hours)¹. It is thus possible that the apparent inability by Yan et al. to reverse the nucleosome sliding defects associated with octamer crosslinking may have been due to an inability to completely reverse crosslinking in fully assembled nucleosomes, and that the reduction they observe occurs while processing the samples for analysis by SDS-PAGE. The results underscore the importance of performing both oxidation and reduction reactions in the context of the histone octamer when attempting to generate or reverse crosslinks to investigate the role of histone dynamics.

In their Reply⁷, Chen and colleagues now propose that the disulfide bond may damage the integrity of the histone octamer, an argument that is different from their original hypothesis of non-specific oxidative damage. To support this claim, the authors provide data suggesting that the sCX2 disulfide crosslinked histone octamer has reduced stability. However, they do not provide evidence for reduced stability in the context of a nucleosome, which is the relevant substrate. We contend that the data we present here tests if the sCX2 disulfide bond impacts nucleosome stability. Specifically, when comparing unremodeled nucleosomes assembled from oxidized and reduced sCX2 octamers, after incubation in reaction buffer with SNF2h for more than an hour, we observe similar levels of intact nucleosomes, suggesting no stability differences between the two samples (Figure 1B, and Supplementary Figure 2A, see “No ATP” lanes).

Importantly, our results do not invalidate the other conclusions drawn in the article by Yan et al.⁴ nor do they cast doubt on the quality of the structural data presented. Cryo-EM structures represent only the fraction of states that can be reconstructed to high resolution. Lowly-populated or highly dynamic states would likely be lost during the averaging required for high resolution analysis. Consistent with this possibility, a recent cryo-EM study captured multiple stable ISWI-nucleosome states⁸. While the highest resolution structures showed no clear evidence of conformational dynamics, lower resolution structures showed local reduction in the cryo-EM density that overlaps both with dynamic regions captured by NMR¹ and with the crosslinked residues tested here and in Sinha et al¹. These results are consistent with SNF2h promoting the formation of an ensemble of highly dynamic conformations that cannot be directly visualized by cryo-EM averaging. Consequently, the absence of a stable deformed histone octamer in the cryo-EM density reported by Yan et al. is still fully compatible with the conclusion that dynamic histone deformation is important for nucleosome sliding.

In summary, we replicate our earlier findings that restraining histone dynamics with disulfide crosslinks interferes with SNF2h-mediated nucleosome sliding. We further show that this finding is robust to the method of crosslinking and cannot be explained by non-specific oxidative damage to the nucleosome. The inability to replicate this finding by Yan et al.⁴ might be due to incomplete reduction of the disulfide bond in the context of the nucleosome. Overall, we contend that our results support an important role for histone octamer plasticity during SNF2h-mediated nucleosome sliding.

Methods

A detailed description of the methods is available in Supplementary Note 1.

Extended Data

Extended Data Fig. 1. Preparation of crosslinked and reduced histone octamers.

SDS-PAGE gels of Xenopus laevis of histones used to prepare all nucleosomes in this study. (A) Wild type (WT) Histone H2A-H2B dimer as well as H3C110A histone octamer and H3C110A sCX2 (H3 L82C, H4 V81C) histone octamer oxidized using copper phenanthroline (CuPhe) or oxidized glutathione (GSSG). (B) The same samples used in (A) treated with 100mM DTT in order to reduce the disulfide bond. Uncropped images are available as source data files. SDS-PAGE analysis was only performed once for all samples.

Extended Data Fig. 2. Remodeling of oxidized nucleosomes is slowed specifically due to disulfide bond formation and is robust to remodeling conditions.

(A) Left. Native gel remodeling assay with saturating SNF2h (1μM), saturating ATP, and 15nM cy3-nucleosomes as in Figure 1. Middle. Quantification of the experiment at the left including a plot of all time points; and for ease of comparison a plot of the first 15 minutes of the reaction normalized to the best-fit parameters for Y0 and plateau. This experiment was performed 3 times with similar results. Right. Mean observed rate constants (kobs) from 3 independent experiments. Error bars reflect the standard error of the mean (SEM). (B) Left. Native gel remodeling assay with sub-saturating SNF2h (50nM), saturating ATP, and 15nM cy3-nucleosomes as in Figure 1. Middle. Quantification of the experiment on the left. This experiment was performed 3 times with similar results. Right. Mean and SEM of the observed rate constants (kobs) from 3 independent experiments. The asterisk denotes that the rate constant for the oxidized reaction condition was too slow to reliably quantify with the time points taken. (C) Left. Native gel remodeling assay under the conditions of Yan et al. 2 using 50nM SNF2h, saturating ATP, and 15nM cy3-nucleosomes. Remodeling overall is substantially faster likely because of the different conditions used (higher temperature, lower salt concentration, the absence of .02% (v/v) NP40, and the presence of 0.1mg/mL BSA). Middle. Quantification of the gel on the left along with the indicated observed rate constants. This experiment was performed once. Right. Time courses shown in the middle panel normalized to the best fit parameters for Y0 and Plateau of the exponential decay and zoomed in to the first 10 minutes of the reaction to better evaluate the fits. Uncropped images for panels A-C are available in Source Data and values obtained in quantifications are in Supplementary Table 1.

Extended Data Fig. 3. Disulfide reduction is impaired in the context of the nucleosome.

(A) Native gel remodeling assay with saturating SNF2h (1μM), saturating ATP, and 15nM cy3-nucleosomes as in Figure 1. Nucleosomes containing the oxidized sCX2 bonds were generated by oxidizing the H3C110A sCX2 octamer using CuPhe, and then assembling nucleosomes. Treatment of these nucleosomes with excess DTT as in Yan et al. fails to reverse the remodeling defect. (B) Scheme for the samples run in C. Nucleosomes treated with DTT were either directly added to non-reducing SDS-PAGE loading buffer or quenched with 500mM N-Ethyl Maleimide freshly dissolved in DMSO (final [DMSO]≈10%(v/v)). Additionally, a condition where N-Ethyl Maleimide and DTT were added simultaneously is included to evaluate the efficacy of the quench. (C) SDS-PAGE of samples treated as in B. Samples with reducing agent quenched prior to running on the gel are near-completely oxidized. The experiments shown here were performed once.

Data availability

All relevant data are included in the paper and its supplementary information files. Uncropped images of gels and blots are provided as source data and numerical values plotted in the graphs are available in Supplementary Table 1.