Structural features of nucleosomes in interphase and metaphase chromosomes

Summary

Structural heterogeneity of nucleosomes in functional chromosomes is unknown. Here we devise the Template-, Reference- and Selection-Free (TRSF) cryo-EM pipeline to simultaneously reconstruct cryo-EM structures of protein complexes from interphase or metaphase chromosomes. The reconstructed interphase and metaphase nucleosome structures are on average indistinguishable from canonical nucleosome structures, despite DNA sequence heterogeneity, cell cycle-specific posttranslational modifications and interacting proteins. Nucleosome structures determined by a decoy-classifying method and structure variability analyses reveal the nucleosome structural variations in linker DNA, histone tails, and nucleosome core particle configurations, suggesting that the opening of linker DNA, which is correlated with H2A C-terminal tail positioning, is suppressed in chromosomes. High-resolution (3.4-3.5 Å) nucleosome structures indicate DNA sequence-independent stabilization of superhelical locations ±0~1 and ±3.5~4.5. The linker histone H1.8 preferentially binds to metaphase chromatin, from which chromatosome cryo-EM structures with H1.8 at the on-dyad position are reconstituted. This study presents structural characteristics of nucleosomes in chromosomes.

Graphical Abstract

Introduction

The nucleosome, composed of eight core histones (H2A₂, H2B₂, H3₂, and H4₂) and 144~147 bp of DNA, is the fundamental structural unit of chromosomes (Olins and Olins, 2003; Richmond et al., 1984). X-ray crystal structures of nucleosomes reconstituted in vitro with nucleosome-positioning sequences revealed that DNA wraps around the core octameric histones in a left-handed direction (Luger et al., 1997; McGinty and Tan, 2015). Many structural variants of nucleosomes can also exist in vitro (Lavelle and Prunell, 2007; Zlatanova et al., 2009) (Fig. 1A left)(Arimura et al., 2012; Bancaud et al., 2007; Furuyama et al., 2013; Kato et al., 2017; Zou et al., 2018), and cryo-electron tomography (cryo-ET) analyses suggested that the canonical nucleosome structure may not be abundant in the yeast nucleus (Cai et al., 2018a; Tan et al., 2021). However, current reconstructions of in vivo nucleosomes are lower than 20 Å resolution (Cai et al., 2018b; Chicano et al., 2019; Eltsov et al., 2018; Scheffer et al., 2011), making it difficult to assess nucleosome structural heterogeneity in chromosomes.

Figure 1. Isolation of nucleosomes from interphase and metaphase chromosomes.

A, Scopes of this study. B, The standard chromosomal nucleosome preparation. C, D, Native PAGE analysis of DNA (SYBR-safe) in nucleosome fractions isolated from sperm chromosomes formed in the Xenopus egg extract lot 1, without (C) or with (D) proteinase K treatment. Equal DNA amounts were loaded. E, SDS-PAGE analysis of nucleosomal proteins (gel code blue) isolated from sperm chromosomes in the egg extract lot 1. F, Western blot analysis of Figure 1E. G, Scatter plot of MS-detected proteins in the oligo nucleosome fractions from the extract lot 1. Right; a magnified view of the high signal region. H, MS signal intensities of representative proteins, normalized to H4. I, Linker histone-bound mono-nucleosomes detected by native PAGE and western blot. J, Protein abundance in the oligo-nucleosome fractions quantified by MS. Enrichment of intelectin-2 and glycoproteins (ovochymase and alpha2-macroglobulin family proteins) may reflect O-linked N-acetylglucosamine on chromosomes (Gagnon et al., 2015; Kelly and Hart, 1989).

Chromosome morphology dramatically changes upon entry into mitosis; diffuse interphase chromosomes are converted to rod-shaped chromatids (Fig. 1A right). Although the DNA loop formation by condensin drives large scale mitotic chromosome folding (Goloborodko et al., 2016; Hirano and Mitchison, 1994; Takahashi and Hirota, 2019), chromosomes depleted of condensin still exhibit mitosis-specific compaction (Gibcus et al., 2018; Samejima et al., 2018). Several lines of evidence indicate that nucleosomes are more densely packed in mitosis than in interphase. Nucleosomes assembled in vitro with histones purified from mitotic cells tend to aggregate more easily than those assembled with histones from interphase (Zhiteneva et al., 2017), and a cryo-ET study suggested that mitotic chromatin is more crowded than interphase chromatin (Cai et al., 2018a). Although a variety of mitosis-specific histone modifications are known (Wang and Higgins, 2013), it remains unclear if the nucleosome structure changes during mitotic chromatin compaction.

How nucleosomes pack in chromosome is also elusive. The 30 nm-fiber, which can be formed in vitro reconstituted poly-nucleosomes, (Dorigo et al., 2004; Ekundayo et al., 2017; Grigoryev et al., 2009; Robinson et al., 2006; Schalch et al., 2005; Song et al., 2014), cannot be detected in the interphase nuclei and mitotic chromosomes by cryo-EM and small angle X-ray scattering (Eltsov et al., 2008; Maeshima and Eltsov, 2008; Nishino et al., 2012). Recently, genome-wide nucleosome interaction mapping analyses indicate that short nucleosome stretches may locally cluster into ordered structures (Ohno et al., 2019; Risca et al., 2017). However, different folding patterns of such ordered nucleosome clusters can be predicted depending on the linker DNA angles and lengths (Buckwalter et al., 2017; Grigoryev, 2018; Koslover et al., 2010), while no reliable method to determine linker DNA angles in chromosomes is available.

Recent technological advances in single particle cryo-EM enabled high-resolution structure determination of proteins including endogenous proteins (Kastritis et al., 2017; Nakane et al., 2020; del Valle and Axel Innis, 2020; Yi et al., 2019), but, to the best of our knowledge, no structural comparisons of protein isolated from different cell cycle stages have been reported. Here, we report high-resolution structures of nucleosomes in interphase and metaphase chromosomes.

Results

Nucleosome isolation from interphase and metaphase chromosomes

We established a method to isolate nucleosomes from functional interphase and metaphase chromosomes while preserving intact protein-DNA structures for cryo-EM analysis (Fig. 1B). Xenopus egg extracts arrested at meiotic metaphase II by cytostatic factor (CSF) were released into interphase with Xenopus sperm nuclei (Murray, 1991). Nucleosomes assembled on sperm chromosomes, the nuclear envelope formed, and chromosomes were replicated. Mitotic chromosome compaction and spindle assembly were induced by adding fresh metaphase-arrested CSF extracts (Fig. 1B, S1A) (Shamu and Murray, 1992). Chromosomes were crosslinked with formaldehyde to preserve nucleosome structures, isolated via centrifugation, and then fragmented with micrococcal nuclease (MNase). The majority (60% - 90%) of DNA was solubilized from chromatin with MNase (Fig. S1B). The solubilized nucleosomes were separated by sucrose gradient centrifugation, yielding three fractions containing different sizes of DNA-protein complexes (Fig. 1C, S1D: mono, di, and oligo). DNA lengths isolated from these fractions were equivalent to those from mono-nucleosomes (150-180 bp) (Fig. 1C, 1D, S1E), indicating that nucleosomes may not be connected by linker DNAs but by inter-nucleosomal crosslinks in the di- and oligo- nucleosome fractions. Biochemical characteristics of these fractions were assessed by western blot and MS. The mitosis-specific histone H3 phosphorylation at Thr3 (H3T3ph) was observed in isolated metaphase nucleosomes but not in interphase nucleosomes (Fig. 1F) (Kelly et al., 2010). Known chromatin proteins, such as proliferating cell nuclear antigen (PCNA) and condensin subunits (CAPD2, SMC2, CAPG, SMC4) showed their expected cell cycle-dependent enrichment on nucleosome fractions (Fig. 1F-H, S1F, Table S1), confirming that cell cycle-specific biochemical characteristics of chromatin were preserved.

Unexpectedly, chromatosomes containing H1.8 (also known as B4, H1M, and H1Foo), the dominant linker histone variant in Xenopus eggs (Dworkin-Rastl et al., 1994; Wühr et al., 2014), were more prevalent in metaphase than in interphase (Fig. 1G-J, Table S1). On native polyacrylamide gel electrophoresis (PAGE) gels, the metaphase mono-nucleosome fraction showed an additional slower-migrating band corresponding to the chromatosome (Fig. 1C, S1D), which contains core histones and H1.8 (Fig. 1I). This H1.8 enrichment in metaphase nucleosome did not depend on crosslinking (Fig S1G-H). H1.8 was most enriched in the oligo- and di-nucleosome fractions (Fig. 1F, 1H, Table S1), consistent with the notion that H1 compacts chromatin (Renz et al., 1977).

Simultaneous cryo-EM reconstruction of multiple protein complexes enriched in the nucleosome fraction

To examine structural diversity of nucleosomes formed in chromosomes, we solved cryo-EM structures of protein complexes enriched in nucleosome fractions by devising the Template-, Reference- and Selection-Free (TRSF) pipeline (Fig. 2A). This strategy aimed to preserve conformational heterogeneity and protein diversity by avoiding commonly used cryo-EM analysis procedures that entail selection bias, such as template-based particle picking, 3D classification with references of previously reported structures, and manual 2D class selection of specific targeted protein structures. Using the TRSF pipeline, we simultaneously reconstructed distinct protein complex structures (Fig. 2A middle, S3). Guided by the MS-based protein abundance (Fig. 1J, Table S1) and sequence-based 3D structure modeling (Waterhouse et al., 2018), the structures from oligo fractions were identified as an octameric mono-nucleosome, intelectin-2, alpha2-macroglobulin family protein, and actin (Fig. S3E-H). From mono fractions, two-types of intelectin-2 structures (trimer and hexamer) and an octameric mono-nucleosome were identified (Fig. S3A-D). Although resolution of the nucleosome structures from oligo fractions were much higher than mono fractions due to the higher nucleosome concentration of oligo fractions on the cryo-EM grid (Fig S2A), nucleosome structures reconstructed from both fractions were similar (Fig 2B-E). Both interphase and metaphase nucleosome structures aligned well with the canonical octameric nucleosome crystal structure (Fig. 2C, 2E) (Luger et al., 1997). While cryo-EM maps of actin and intelectin matched published EM and crystal structures, respectively (Fujii et al., 2010; Wangkanont et al., 2016), the overall structure of the maps that we identified as alpha2-macroglobulin family protein were dissimilar to the reported crystal structure of human alpha2-macroglobulin (Fig. S4) (Marrero et al., 2012). Thus, our pipeline successfully reconstructed multiple structures including a previously unsolved structure without the need for any template-picking or manual selection, which were previously required to determine multiple structures from crude samples (Kastritis et al., 2017; Kyrilis et al., 2019; Su et al., 2021; Verbeke et al., 2018, 2020). These results were reproduced with a biological replicate (lot 2; Fig S5A-F).

Figure 2. Simultaneous cryo-EM structure reconstruction of multiple protein complexes in nucleosome fractions by the TRSF pipeline.

A, TRSF cryo-EM analysis pipeline. Shown examples of 2D classification panels and 3D references were from the metaphase oligo-nucleosome fraction. B-E, Reconstructed nucleosome structures from the interphase mono fraction (B), interphase oligo fraction (C), metaphase mono fraction (D), and metaphase oligo fraction (E) of the extract lot 1, and their superposition onto the crystal structure of the canonical nucleosome (PDB ID: 1AOI).

Averaged cryo-EM structures of interphase and metaphase nucleosomes

To assess structural variations of these chromosomal nucleosome structures, we next attempted to reconstruct averaged structures with higher resolution by repeating the cryo-EM processing with another pipeline combining Topaz, a convolutional neural network-based particle picking program (Bepler et al., 2019), and decoy classification (Fig. S6A). Unlike commonly used cryo-EM classification pipelines, our pipeline was designed to collect as many diverse nucleosome particle images as possible by removing only particles that were not classified as the nucleosome by the decoy classification (see method for a detail). Using this pipeline, structures of interphase and metaphase nucleosomes in the oligo-nucleosome fractions were reconstructed at a nominal 3.4 and 3.5 Å resolution, respectively (Fig. 3A, 3B, and S6B). Similar left-handed octameric nucleosome structures were reconstructed from the mono- and di-nucleosome fractions of interphase and metaphase chromosomes (Fig 3A-B) The high-resolution maps from oligo-nucleosome fractions allowed us to build atomic models in which the aromatic side chain of phenylalanine 98 of H2A.X-F1/F2, the predominant H2A isoform in egg (Shechter et al., 2009), could be distinguished from the leucine or isoleucine residues of other H2A isoforms (Fig. S7A). The atomic models for chromosomal nucleosomes are essentially identical to those of X-ray crystal structures of nucleosomes with homogenous nucleosome positioning sequences (Fig. 3C). This indicates that DNA sequence variability does not largely affect the histone octamer structure in functional chromosomes. Furthermore, no obvious cell cycle-dependent structural changes were observed in the histone octamer (Fig. 3D).

Figure 3. Averaged cryo-EM structures of interphase and metaphase nucleosomes.

A, B, Cryo-EM structures of the nucleosome core particle from interphase (A) and metaphase (B) chromosomes assembled in egg extracts lot 1. C, D, Structural comparisons between interphase nucleosomes and the canonical crystal structure of the nucleosome (PDB ID: 1AOI) (C) or between interphase and metaphase nucleosomes (D). Each histone protein chain are named separately following the definition in the nucleosome crystal structure (Luger et al., 1997). Root-mean-square deviation (R.M.S.D.) for all Cα atoms for each of the structures was calculated using PyMol and plotted (top) or mapped onto the atomic coordinates of the interphase oligo fraction nucleosome (bottom). Large values at H4 N-terminus are due to two orientations (see Fig. S9). E, Cryo-EM density of DNA around SHL 0 of nucleosome in interphase oligo fraction. F, G, H, Local resolution maps of interphase (F) and metaphase (G) nucleosomes from oligo fractions of the extract lot 1, and recombinant nucleosome (H: GUB-closed). Left, local resolution maps of the entire nucleosome. Right, local resolution maps of DNA around SHL ±3.5~4.5. Orange arrows indicate the position of phosphate group in DNA. I. Diagram of the potential factors that may affect EM density resolution. J. Graphical summary.

DNA segments at two specific superhelical locations are exceptionally ordered in chromosomal nucleosomes

Despite the heterogeneity of DNA sequences, our high-resolution metaphase and interphase nucleosome maps exhibited extraordinary homogeneity in DNA structure, such that the DNA backbone was clearly delineated (Fig. 3E). A single DNA base pair aligned at the nucleosome dyad is defined as Super Helical Location (SHL) 0 (Klug et al., 1980). Strikingly, in our high-resolution cryo-EM structures of the oligo-fraction nucleosomes, DNA segments at SHL ±3.5~4.5, as well as SHL ±0~1, exhibited higher local resolutions (Fig. 3F-G) and lower B-factor values (Fig. S7B) than those at their adjacent DNA segments did. The resolution of DNA around these segments was high enough to exhibit discrete positioning of phosphates and bases (Fig. 3E, F-G right). Two properties are likely to contribute to this high local resolution at SHL 3.5-4.5 and SHL0-1. First, the nucleotide positions of these DNA segments are strongly phased in a manner independent of DNA sequence (Fig. 3I: phasing). This was surprising since local DNA segments are known to stretch or shrink on the nucleosome to accommodate different DNA sequences or bound proteins in crystal structures (Armache et al., 2019; Chua et al., 2012; Liu et al., 2017; McGinty and Tan, 2015; Sinha et al., 2017), where different DNA conformation may be adapted depending on local bendability varied by DNA sequences (Basu et al., 2021). Second, these DNA segments must be structurally stable, since EM density may also worsen if the DNA of these regions are flexible, even if the phasing of DNA nucleotide positioning at these regions is identical among nucleosomes (Fig. 3I: Flexibility).

Similar higher local resolutions at SHL ±3.5~4.5 and SHL ±0~1 were also seen in the recombinant nucleosome with GUB nucleosome positioning DNA (Adams and Workman, 1995; An et al., 1998) (Fig. 3H, S8: “GUB-closed”). Since nucleotide positions must be phased in the homogeneous nucleosome, the result suggests that the DNA at these regions is stabilized. Supporting the high structural stability at these DNA locations, the number of DNA-histone interactions formed with basic amino acid residues on histones is particularly high in these regions (Fig. S7C-G). These data indicate that the DNA segments around SHL ±0~1 and SHL ±3.5~4.5 in chromosomal nucleosomes are strongly phased and structurally stable in a sequence-independent manner, while other regions, such as SHL ±2~3, ±5~6 and linker DNA, form relatively flexible structures and/or allow multiple DNA positioning patterns (Fig. 3J).

Qualitative identification of nucleosome structural variants

We next examined the conformational heterogeneity of nucleosome structures in chromosomes with 3D variability analysis (3DVA) in CryoSPARC (Punjani and Fleet, 2021), which models cryo-EM data into multiple series of 3D structures (motion frames), each representing a potential “trajectory” of motion (Fig. 4A, Figure360, Movie S1). Each trajectory represents a principal component of the variance present in each dataset, so each 3DVA component presented here is a representation of the major structural variations.

Figure 4. 3DVA of the nucleosomes.

A, 3DVA of the nucleosomes containing linker DNAs. Five representative structures (motion frames 3, 7, 11, 15, 19) from 3DVA component 1 class (total 20 frames) are shown. The structures of motion frames 3 and 19 were selected and overlayed for each nucleosome. EM densities for the H2A C-terminal region (orange arrows) and H1.8 (red arrows) are shown. See Movie S1 for full dataset. The identical contour level was applied for maps in each data series. B, C, NCP structural variation of GUB-open nucleosomes (B) and chromosomal nucleosomes (C).

To test if the structural variations observed in 3DVA reflect structural characteristics of nucleosomes in chromosomes rather than the fixation and freezing procedure, we determined cryo-EM structures of chromosomal nucleosomes using a modified protocol in which the same fixation and freezing procedure was applied but crosslinking was done after native nucleosomes were first released from interphase chromosomes by MNase (Fig. 4A, S8A). We called these nucleosomes “freed nucleosomes”. In addition, a more drastic structural variation including nucleosomes with wide open linker DNAs was detected when recombinant nucleosomes assembled on the GUB DNA were processed with GraFix and frozen without optimized cryoprotectant buffer (Fig. S8B). We called these structures “GUB-open nucleosomes”.

Using the 3DVA, four types of apparent structural variations were observed across different nucleosome samples (Fig. 4A, black, orange, cyan, and blue arrows, Movie S1).

Linker DNA angle variation

The linker DNA angles of interphase and metaphase nucleosomes were generally closed, relative to “freed nucleosomes”, which exhibited much larger variation in linker DNA positions (Fig. 4A, Movie S1). Quantitative assessment of this observation will be described in Figure 5.

Linker histone and H2A C-terminal tail

Extra cryo-EM density was observed in chromosomal nucleosomes, particularly when the linker DNA angle was narrowest (Fig. 4A, Movie S1, orange and red arrows). For both interphase and metaphase, the densities of the H2A C-terminal tail reached DNA at the DNA entry/exit site of the nucleosome as the DNA closed inward (Fig. 4A, Movie S1, orange arrows). In addition, as the linker DNA angle closed, an additional cryo-EM density appeared at the nucleosome dyad position, particularly in metaphase nucleosomes (Fig. 4A, Movie S1, red arrows). High resolution analysis of these extra cryo-EM densities will be described in Figure 6.

H4 N-terminal tail

Subtle structural variations around the H4 N-terminal tail were detected (Fig 4A, Movie S1, cyan arrow). Two distinct “outward” and “inward” orientations of this region have been reported. In recombinant canonical nucleosome structures, the inward orientation was clearly observed while outward orientation electron density was ambiguous (Arimura et al., 2018). The outward orientation was previously detected in nucleosomes containing histone H3 variant CENP-A and nucleosomes bound to another protein, such as Sox11 or SET8 (Ali- Ahmad et al., 2019; Arimura et al., 2019; Dodonova et al., 2020; Ho et al., 2021). Functionally, the outward orientation facilitates SET8-mediated methylation of H4K20 (Arimura et al., 2019). In the H4 N-terminal tail region of our averaged interphase and metaphase structures, cryo-EM densities were ambiguous (Fig S9B left), likely reflecting a mixture of inward and outward conformations. To clarify this, we performed a 3D classification based on the H4 tail structures, and successfully separated the outward and inward conformations (Fig S9A). This result suggests that both outward and inward H4 tail orientations are formed at comparable frequency in both interphase and metaphase nucleosomes (Fig. S9).

Major and minor structural variations within the NCP

The 3DVA analysis of GUB-open nucleosomes exhibited large structural rearrangements of the NCP (Fig. 4A blue arrows and brackets, Movie S1). To better visualize structural variations within the NCP, we employed the 3DVA analysis after subtracting the linker DNA densities from the original particles (Fig. 4B, 4C, Movie S2). This analysis revealed two different types of NCP structural variations; the major NCP structural variation, characterized by diverse NCP surface outlines seen in GUB NCPs (Fig. 4B, Movie S2, also see Fig. 7A), and the minor NCP structural variation, such as rearranged orientations of core histone α-helices and the H4 N-terminal tail observed in chromosomal NCPs (Fig. 4C, Movie S3, also see Fig. 7B). Population analysis of these structural variations will be described in Figure 7.

Linker DNA configurations of chromosomal nucleosomes

To quantitatively assess the structural variation of linker DNA configurations detected by 3DVA, we used ab initio reconstruction to generate 6-8 3D references of nucleosomes representing distinct linker DNA structural classes, and then classified all nucleosome-like particles (Fig. 5A-C, S10A-B). Linker DNAs in the majority of interphase and metaphase nucleosomes processed by our standard protocol were classified into the closed configurations, while linker DNAs in freed nucleosomes (fixed after MNase digestion of chromosomes) exhibited wider angles or could not be resolved well, likely due to their structural heterogeneity (Fig 5B). The hexasome, which lacks one H2A-H2B dimer from octameric nucleosome with partially wrapped DNA, was only observed in freed nucleosomes (Fig 5B class F).

Figure 5. The linker DNA configuration of interphase and metaphase nucleosomes.

A, 3D structure classes of nucleosomes in the oligo-nucleosome fractions of interphase chromosomes (left, extract lot 1) and metaphase chromosomes (right, extract lot 1). B, 3D structure classes of the freed nucleosome (egg extract lot2). C, Linker DNA angle definitions. D Schematic of the in silico mixing 3D classification pipeline. E-G, In silico mixing 3D classification with fixed 3D reference maps to compare the mono and oligo fractions (E), di and oligo fractions (F), and oligo fractions and their biological replicates (G). Input and output 3D maps of the in silico mixing 3D classifications were shown. Numbers and populations of particles assigned to each class are shown in Table S3. H, Graphical summary.

To quantitatively compare the linker DNA configurations between different samples, we performed in silico mixing 3D classification (Hite and MacKinnon, 2017): we combined particles from different sets of samples, and conducted a new round of classification analysis using five manually-selected 3D references with distinct linker DNA angles (Closed 1, Closed 2, Open 1, Open 2, Miscellaneous; Fig. 5D-G). Five classification replicates yielded consistent results: in all mono, di, and oligo fractions, around 60-70% of interphase or metaphase nucleosomes were assigned to two closed classes, while these proportions were reduced to 30-40% in freed mono-nucleosomes fixed after MNase digestion (Fig. 5E-G). The closed classes were also major forms in the biological replicate of oligo fractions (lot 2, Fig. 5G). These results suggest that linker DNAs are generally closed in chromosomes, but once mono-nucleosomes are released from chromosomes by MNase prior to fixation, linker DNAs can adopt more flexible open conformations, and these conformations may lead to H2A-H2B dissociation from nucleosome (Fig. 5H).

Linker histone H1.8 binds on the nucleosome dyad in metaphase chromatin

An extra dyad density reminiscent of the linker histone was observed in the metaphase nucleosome with closed linker DNA (Fig. 4A, 5A, Movie S1). Further 3D classification of Class A shown in Fig. 5A (metaphase oligo) enabled structure determination at 4.4 Å resolution (Fig. 6A, S11A-C). Since histone H1.8 preferentially cofractionated with metaphase nucleosomes (Fig. 1G-J), we built an atomic model of the H1.8-bound nucleosome and refined with our cryo-EM map. The extra EM density reasonably corresponded to our refined H1.8 atomic model (Fig. 6B). The similar structure was also reconstructed in the mono fraction in metaphase (Fig S11D). Previously, structural analyses of recombinant chromatosomes and molecular dynamics simulations had suggested that linker histones may bind to the nucleosome in multiple modes; “on-dyad”, where H1 binds at the center of the nucleosome DNA (Bednar et al., 2017; Zhou et al., 2015), and “off-dyad”, where H1 binds more than 5 bp off-center of the nucleosome DNA (Adhireksan et al., 2020; Song et al., 2014; Woods and Wereszczynski, 2020). The H1.8 position in our cryo-EM structure of the metaphase chromatosome closely corresponds to the on-dyad structure (Fig. 6C, 6D) (Bednar et al., 2017; Zhou et al., 2015).

Figure 6. Linker histone H1.8 binds on the nucleosome dyad in metaphase chromosomes.

A, Cryo-EM structure of the nucleosome class with an extra density (in red) on the dyad isolated from the oligo fraction of metaphase nucleosomes. B, Wall-eyed stereo images of the cryo-EM map superimposed on an atomic model of a H1.8-bound nucleosome (red). C, The atomic model of the H1.8-bound nucleosome superimposed on the “on-dyad” H1-bound nucleosome crystal structure (PDB ID: 5NL0). D, The cryo-EM map of the H1.8-bound nucleosome superimposed on the “off-dyad” H1 bound nucleosome cryo-EM map (emd ID: 2601). E, Reconstruction scheme of “closed” and “moderately open” linker DNA nucleosome maps. F, H2A C-tail density in the “closed” and “moderately open” linker DNA nucleosome maps. Black arrows, H2A C-tail. G, Graphical summary.

To assess the structural correlation between extra density for the H2A C-terminal tail in nucleosomes and closed linker DNAs (Fig. 4A, Movie S1), we generated 3D maps of interphase nucleosomes with closed and open linker DNAs at comparable resolutions (4.51 Å and 4.52 Å) (Fig. 6E-F). In the nucleosome structure with closed linker DNAs, the cryo-EM density corresponding to the H2A C-terminal tail was attached to the DNA, while the corresponding EM density was weaker and detached from the DNA in the nucleosome structure with open linker DNAs (Fig. 6F, black arrows), suggesting a role for the H2A C-terminal tail in influencing the linker DNA angle (Fig. 6G).

Structural variations of the NCP in chromosomes

3DVA analysis revealed major and minor NCP structural variations (Fig. 4B-C). The major structural variants are characterized by jagged NCP outlines with shifted positioning of the H3-H4 tetramer (Fig. 7A, Movie S2). To examine if the major NCP variants exist in chromosomal nucleosomes, we employed in silico mixing 3D classification for each 3DVA component using four reference 3D maps; one representing the canonical nucleosome structure (generated from composite particles of interphase and metaphase), and three representing the major nucleosome structural variants (picked from 3DVA motion frames of GUB-open NCP, which clearly exhibited major NCP rearrangements; see Methods and Fig S12A-B) (Fig. 7A, C). While more than 60% of NCPs were assigned to the major NCP structural variants in “freed nucleosomes”, only 30~40 % of the interphase and metaphase NCPs from oligo nucleosome fractions were assigned to these variants (Fig. 7C). The results suggest that formation of the major NCP structural variants is suppressed at least in clustered oligo nucleosomes.

Figure 7. Structural variation of chromosomal NCP.

A, Types of major NCP structural variations identified in 3DVA of GUB-open nucleosomes. B, Types of minor NCP structural variations identified in 3DVA of chromosomal nucleosomes. C, D, In silico mixing 3D classification of major NCP variants (C) and minor NCP variants (D). Numbers and populations of particles assigned to each class are shown in Table S3. E, Surface hydrophobicity of the chromosomal NCP structural variants. Hydrophobicity indexes were mapped on the EM maps of the representative motion frames of 3DVA component 1 of chromosomal NCP. F, Electrostatic potential of the chromosomal NCP structural variants. Surface electrostatic potentials for the atomic models for the representative motion frames of 3DVA component 3 of chromosomal NCPs are shown. The identical contour level was applied for all EM maps.

The minor NCP structural variants are characterized by rotation of H3 α1 and reorientation of the H4 N-terminal tail (Fig. 7B, magenta square), opening and closing of two H3 α-helices at the dyad (violet square), and opening and closing of the H2A-H2B acidic patch (blue square). To quantitatively analyze these minor NCP structural variants, we first performed in silico mixing 3D classification to isolate particles assigned to the canonical nucleosome class from major NCP variants (Fig S12C). We then conducted a second round of in silico mixing 3D classification against these selected particles using minor structural variants seen in chromosomal nucleosomes by 3DVA as references (Fig. 4C). This classification analysis revealed that minor NCP structural variants were prevalent in both interphase and metaphase chromosomes (Fig 7D). In the metaphase nucleosome, the amount of minor NCP structural variants was slightly increased compare to interphase (Fig 7D). In some of these variants, buried hydrophobic amino acid residues were exposed to the NCP surface (Fig. 7E), and the electrostatic potential at the H2A-H2B acidic patch was redistributed (Fig 7F), possibly affecting interaction between the NCP to other proteins and nucleosomes. These results suggest that minor NCP structural rearrangements can occur in chromosomes even in the context where the linker DNA opening and formation of major NCP structural variants are suppressed.

Discussion

Characteristics of nucleosome structures in interphase and metaphase chromosomes

Here we present sub-nanometer resolution structures of nucleosomes from interphase nuclei and metaphase chromosomes. On average, reconstructed nucleosome-like structures isolated from both interphase and metaphase chromosomes are identical to canonical left-handed octameric nucleosomes assembled on defined nucleosome positioning sequences (Fig. 2, 3) (Chua et al., 2016; Luger et al., 1997). Although it is still possible that minor populations of nucleosomes form alternative structures, we propose that our reconstructions represent the major forms in mitotic and interphase chromosomes for the following reasons. First, among the structures of five most abundant protein complexes (nucleosome, chromatosome, alpha2-macroguloblin, intelectin-2, actin filament) reconstructed by the TRSF pipeline, the only reconstructed nucleosome-like structure was the octameric left-handed nucleosome (Fig. 2, S3, S5). Second, around 60-90% of chromosomal DNA was solubilized after MNase digestion (Fig. S1B), suggesting that the solubilized nucleosomes analyzed by cryo-EM did not represent minor fractions of chromatin. Third, structures of “freed nucleosomes”, which were crosslinked after release from chromatin by MNase, were more heterogenous than those of chromosomal nucleosomes and even included the hexasome (Fig. 5), demonstrating that our procedures enabled preservation and detection of dynamic histone and DNA movements. Altogether, our analysis revealed that the major nucleosome structure in chromatin is the octameric left-handed nucleosome, and that global histone architectures are generally homogeneous.

One of the key structural features that we found in chromosomal nucleosomes is the prevalence of nucleosomes with closed linker DNA angles, which are comparable to those from compacted recombinant poly-nucleosome cryo-EM structure composed of “tetra-nucleosomal repeating units” (Fig S10C) (Song et al., 2014). It has been proposed that nucleosomes in chromatin form “nucleosome motifs”, comprised of specific oligo-nucleosome arrangements (Krietenstein and Rando, 2020; Ohno et al., 2019; Ricci et al., 2015; Risca et al., 2017). It is possible that observed nucleosomes with closed linker DNA angles reflect such nucleosome motifs.

3DVA showed that closed linker DNA angles are highly correlated with the visibility of the H2A C-terminal tail and H1.8 (Fig. 4A, 6A, 6F, Movie S1). H1.8 density was only visible on nucleosomes with the narrowest linker DNA angle, in agreement with the previous observation that H1 binding constricts the linker DNA angle (Bednar et al., 2017). Since the C-terminal tail of H2A is highly divergent among H2A variants and is a target of various modifications (Corujo and Buschbeck, 2018), it is tempting to speculate that linker DNA conformation is regulated by these H2A variants and modifications. Consistent with this idea, a recent cryo-EM study showed that the linker DNA is open in the nucleosomes reconstituted with the H2A variant H2A.Z.2.2, which has a short C-terminal tail (Zhou et al., 2020).

Local resolution analysis suggests that DNA segments at SHL −1 to +1 and SHL ± 4 are structurally rigid and uniformly positioned despite highly heterogeneous DNA sequences (Fig 3F-I). This is consistent with a single molecule force-measurement study showing that the dyad position (SHL 0) and ±40 bp from the dyad (SHL ± 4) is resistant to mechanical DNA unzipping (Hall et al., 2009), and RNA polymerase II halts around SHL ±5 and SHL ±1 on the nucleosome during transcription in vivo and in vitro (Kujirai et al., 2018; Weber et al., 2014). Since the H2A.X-F residues 15-44, interact with DNA at SHL ±4, are highly conserved among H2A variants across different species, uniform positioning of those two DNA segments may be a common feature of nucleosomes (Fig. S7G). The ATPase module of many nucleosome remodeling factors binds to SHL ±2.5 or ±6, suggesting that these factors are optimized to target the nucleosome regions that easily allow DNA repositioning (Ayala et al., 2018; Eustermann et al., 2018; Farnung et al., 2017; Han et al., 2020; He et al., 2020; Liu et al., 2017; Willhoft et al., 2018; Yan et al., 2019; Ye et al., 2019).

Implications for models of chromatin compaction during the cell cycle

Hirano hypothesized that condensin drives chromatin compaction and suppresses DNA unwinding by converting (−) crossings of linker DNAs to (+) crossings, which is inhibited in interphase by linker histones, and this inhibition released by the phosphorylation of linker histones in M phase (Hirano, 2014). However, our cryo-EM analysis did not reveal mitotic conversion of linker DNAs. The majority of interphase and metaphase nucleosomes contained closed (−) crossing linker DNAs (Fig. 5, S10), while histone H1.8 preferentially associates with the metaphase nucleosome at the on-dyad position and stabilizes (−) crossing (Fig. 5A, 6A). Our MS and cryo-EM results suggest that H1.8 in interphase may not strongly bind to nucleosome even though H1.8 accumulates in interphase nuclei (Maresca et al., 2005). This preferential association of H1.8 to mitotic nucleosomes may contribute to local chromatin compaction, since linker histone dependent chromatin compaction was reported in vitro and in vivo (Gibbs and Kriwacki, 2018; Gibson et al., 2019; Turner et al., 2018; Willcockson et al., 2021; Yusufova et al., 2021). Interestingly, our recent study suggests that H1.8 suppresses chromosome binding of condensins and DNA topoisomerase II to make chromosomes shorter and limit chromosome individualization (Choppakatla et al., 2021). Thus, stable H1.8 binding to the nucleosome is likely to be enhanced during mitosis not only for local chromatin compaction but also for a long-range DNA folding and topological organization.

Since most histone chaperones and nucleosome remodelers dissociate from mitotic chromatin (Fig S1F) (Funabiki et al., 2017; Jenness et al., 2018), one might speculate that more NCP structural variants exist in interphase than in metaphase. However, our data suggest that the formation of major NCP structural variants are suppressed in both interphase and metaphase, and the population of minor NCP structural variants was even slightly higher in metaphase than in interphase (Fig. 7). In the minor NCP structural variants, many interspaces were newly generated on the NCP surface (Fig 7E, Movie S3), mimicking a situation where nucleosome binding of CENP-C and Swi6 increased solvent accessibility of buried histone residues (Falk et al., 2015; Sanulli et al., 2019). Changes in the hydrophobicity and electrostatic potential on the nucleosome surface would potentially affect specificity and affinity of nucleosome-binding targets (Fig. 7E-F) (Skrajna et al., 2020). Exposure of hydrophobic surfaces may enhance weak, transient multivalent nucleosome-nucleosome interactions (Sanulli et al., 2019). Changes in charge distribution at the acidic patch may affect binding targets, including the H4 N-terminal tail, which is also important for nucleosome-nucleosome interactions (Chen et al., 2017). Thus, the slight population-level changes of these minor structural variants during the cell cycle may be functionally related to global changes in nucleosome-interacting proteins and the chromatin compaction mechanism. Future studies are needed to understand the mechanistic basis and functional significance of these structural variations of the nucleosome.

Limitations of the Study

We note four limitations in this study.

First, preservation of nucleosome structure variations, especially linker DNA angles, is likely affected by fixation conditions. Indeed, linker DNA conformations of the recombinant nucleosomes containing GUB DNA significantly differ between two sample preparation protocols (Fig S8B, S8K). In our standard protocol, in which nucleosomes were crosslinked by formaldehyde and frozen with optimized cryoprotectant buffer, a majority of nucleosomes had closed linker DNA conformations. In contrast, the proportion of open linker DNA conformations increased when samples were crosslinked by GraFix without magnesium and frozen on the cryo-EM grid without the cryoprotectant buffer (GUB-open nucleosomes). Even with our standard protocol, notable structural differences can be seen between lot 1 and lot 2; apparent enrichment of H4 outward orientation (Fig S9C) and nucleosome minor structural variants (Fig 7D) in metaphase seen in lot 1 was less obvious in lot 2. We noticed that spindles and nuclei clustered more frequently in lot 2 than in lot 1 (Fig S1A, S8C), potentially leading to reduced penetration of crosslinkers to nucleosomes in lot 2. In addition, MNase lots were different between these biological replicates, possibly causing subtle differences in linker DNA lengths (Fig. 1D, S8E), which might have affected H1.8 stabilization on the structure.

Second, our structural preservation may be affected by the cryo-EM grid freezing process. For example, protein adsorption to the air-water interface can cause sample denaturation (D’Imprima et al., 2019), leading to grid-to-grid variability in structure preservation. Although cryoprotectant buffer was optimized to prevent sample denaturation (Fig S2B), we cannot rule out a possibility that fragile nucleosome structural variants were selectively denatured or altered during the freezing step. However, to minimize this issue, we excluded problematic grids based on signatures such as poor reconstruction of actin filaments and prevalence of naked DNA fragments (Fig. S5G).

Third, we were able to reconstruct high resolution (< 4 Å) nucleosome structures only with the oligo-nucleosome fraction, raising the possibility that some features (i.e. DNA structural stabilization at SHL −1 to +1 and SHL ± 4; suppression of major NCP structure variant formation; two distinct orientations of the H4 N-terminal tail; association of H2A C-terminal tail to DNA) are specific to subpopulations of chromatin with clustered nucleosomes.

Fourth, we could not determine the structures of less abundant or weakly associated nucleosome-protein complexes. For example, while substantial amount of H1.8 were also detected interphase nucleosome fractions (Fig 1H: roughly 30~50% of metaphase), we could not extract the H1.8 bound nucleosome structure in cryo-EM analysis (Fig. 6). It is possible that H1.8 in interphase associates with nucleosomes weakly or at various positions, making it difficult to reconstruct the H1.8 bound nucleosome structure.

Despite these limitations, our analysis revealed several structural features of nucleosomes in chromosomes in a reproducible manner. Nucleosome structures containing diverse genomic DNA sequences are remarkably similar between interphase and metaphase; DNA linkers tend to exist in the closed conformation, yet minor NCP structural variations are commonly found. Future improvements to this strategy may capture high-resolution structures of nucleosome-nucleosome interaction and other protein-DNA complexes while they are functioning on chromosomes.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to the Lead Contact, Hironori Funabiki (funabih@rockefeller.edu)

Materials Availability

The plasmid (pUC18-GUB_176x4) generated in this study is available from the lead contact upon request.

Data and Code Availability

Cryo-EM raw data are available in the EMPIAR. EM maps are available in the EMDR. Atomic models are available in the PDB. IDs for each data are listed in star method table, Table 1 and Table 2. This paper does not report any original code. Any additional information required to reanalyze the data reported in this work paper is available from the Lead Contact upon request.

Table 1.

Data collection, model refinement and validation for simultaneous TRSF reconstruction of cryo-EM structures of multiple protein complexes. Related to Figure 2

Micrograph name	Interphase_oligo_lot1				Metaphase_oligo_lot 1				Interphase_oligo _lot2	metaphase_oligo _lot2
Deta collection and prosessing
Microscope	FEI Talos Arctica				FEI Talos Arctica		FEI Talos Arctica		FEI Talos Arctica	FEI Talos Arctica
Magnification	28,000				28,000		28,000		28,000	28,000
Voltage (kV)	200				200		200		200	200
Camera	Gatan K2 Summit				Gatan K2 Summit		Gatan K2 Summit		Gatan K2 Summit	Gatan K2 Summit
Electron exposure (e−/Å)	33.1				38.3		35.3		34.2	34.2
Defocus range (μm)	−1 ~ −3				−1 ~ −3		−1 ~ −3		−1 ~ −3	−1 ~ −3
Pixel size (Å)	1.5				1.5		1.5		1.5	1.5
Micrographs	1,325				2,676				1,156	1,044
EMPIAR ID of raw micrographs	EMPIAR-10692				EMPIAR-10691				EMPIAR-10747	EMPIAR-10746
EM 3D maps	Octameric nucleosome	Intelectin- 2	F-actin	alpha2- macroglobulin	Octameric nucleosome	Intelectin- 2	F-actin	alpha2- macroglobulin	Octameric nucleosome	Octameric nucleosome
Initial particle images (no.)	812,976				1,317,710				320,524	384,740
High resolution group particles(no.)	66,426				166,673				35,700	55,801
Low resolution group particles(no.)	746,550				1,151,037				284,824	328,939
particle for 3D clasification (no.)	812,976				1,317,710				320,524	384,740
Particle images (no.)	62,014	89,728	10,922	9,721	116,723	150,300	38,321	49,363	40,640	40,822
Symmetry imposed	C1	C3	C1	C1	C1	C1	C1	D2	C1	C1
Map resolution (FSC=0.143)	4.2	4.5	7.6	8.07	4.2	4.7	4.5	5.5	4.42	4.32
EM data resource ID	EMD-22793	-	-	EMD-22794	EMD-22795	-	-	EMD-22796	EMD-23820	EMD-23822

Table 2.

Data collection, model refinement and validation for averaged nucleosome structure analysis. Related to Figure 3, 6

Micrograph name	Interphase mono lot 1	Interphase di lot 1	Interphase oligo lot 1	Metaphase mono lot 1	Metaphase di lot 1	Metaphase oligo lot 1		Interphase, 'Freed' (Crosslinked after MNase) lot 2		Interphase oligo lot 2	Metaphase oligo lot 2		GUB-closed nucleosome	GUB-open nucleosome
Deta collection and prosessing
Microscope	FEI Titan Krios	FEI Titan Krios	FEI Talos Arctica	FEI Titan Krios	FEI Titan Krios	FEI Talos Arctica		FEI Talos Arctica		FEI Talos Arctica	FEI Talos Arctica		FEI Talos Arctica	FEI Talos Arctica
Magnification	25,000	20,000	28,000	25,000	20,000	28,000		28,000		28,000	28,000		28,000	28,000
Voltage (kV)	300	300	200	300	300	200		200		200	200		200	200
Camera	Gatan K2 Summit	Gatan K2 Summit	Gatan K2 Summit	Gatan K2 Summit	Gatan K2 Summit	Gatan K2 Summit		Gatan K2 Summit		Gatan K2 Summit	Gatan K2 Summit		Gatan K2 Summit	Gatan K2 Summit
Electron exposure (e−/Å)	45.0	55.5	33.1	44.7	54.0	38.3	35.3	37.3	31.1	34.2	34.2
Defocus range (μm)	−1 ~ −3	−1 ~ −3	−1 ~ −3	−1 ~ −3	−1 ~ −3	−1 ~ −3		−1 ~ −3		−1 ~ −3	−1 ~ −3		−1 ~ −3	−1 ~ −3
Pixel size (Å)	1.33	1.6	1.5	1.33	1.6	1.5		1.5		1.5	1.5		1.5	1.5
Micrographs	1,622	1,270	1,325	1,637	1,347	2,676		2,049		1,375	1,044
EMPIAR ID of raw micrographs	EMPIAR-10742	EMPIAR-10744	EMPIAR-10692	EMPIAR-10743	EMPIAR-10745	EMPIAR-10691		EMPIAR-10750		EMPIAR-10747	EMPIAR-10746		EMPIAR-10749	EMPIAR-10748
EM 3D map. Name	Nucleosome	Nucleosome	Nucleosome	Nucleosome	Nucleosome	Nucleosome	Nucleosome + H1.8	"Free" nucleosome		Nucleosome	Nucleosome	Nucleosome + H1.8	GUB-closed nucleosome	GUB-open nucleosome
Initial particle images (no.)	147,921	213,608	571,558	116,669	167,734	898,553		259,060		178,949	194,837		203,052	199,125
Particles for nucleosome ab-initio reconstruction (no.)	13,434	21,650	85,994	11,325	9,515	154,208		25,123		61,931	50,002		86,720	75,171
Particles for 'decoy' ab-initio reconstruction (no.)	134,487	100,946	485,564	105,344	102,722	744,345		133,700		117,018	144,835		116,332	123,954
particle for 3D clasification (no.)	147,921	122,596	571,558	116,669	112,237	898,553		158,823		178,949	194,837		203,052	199,125
Particle images (no.)	13,798	15,653	89,191	13,350	6,027	155,413	27,383	20,835		52,163	52,941	6,399	76,391	62,823
Symmetry imposed	C1	C1	C1	C1	C1	C1	C1	C1		C1	C1	C1	C1	C1
Map resolution (FSC=0.143)	5.12	4.74	3.38	5.64	8.1	3.5	4.42	5.44		3.54	3.77	5.6	3.77	4.52
CTF rifinement	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	-	✓ (Relion 3.1)		✓ (Relion 3.1)	✓ (Relion 3.1)	-	✓ (Relion 3.1)	✓ (Relion 3.1)
Particle polishing	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	✓ (Relion 3.1)	-	✓ (Relion 3.1)		✓ (Relion 3.1)	✓ (Relion 3.1)	-	✓ (Relion 3.1)	✓ (Relion 3.1)
Postprocess Pixel size	1.3	1.66	1.47	1.3	1.66	1.47	-	1.47		1.47	1.47	-	1.47	1.47
EM data resource ID	EMD-22797	EMD-22798	EMD-22790	EMD-22800	EMD-22801	EMD-22791	EMD-22792	EMD-23826		EMD-23819	EMD-23821	-	EMD-23823	EMD-23824
Atomic coordinate name	-	-	Nucleosome	-	-	Nucleosome	Nucleosome + H1.8	-	-	-	-	-	-	-
Pixel size			1.47			1.47	1.47
Model resolution (Å)			3.38			3.5	4.4
Initial model used (code)			6DZT, 5NL0			6DZT, 5NL0	5NL0
Map sharpening			RELION 3.1 (Postprocess)			RELION 3.1 (Postprocess)	CryoSPARC, bsoft (blocres/blocafilt)
Map sharpening B factor (Å2)			−60.3386			−30.7959	-
Model composition
Non-hydrogen atoms			12,169			12,408	13,710
Protein residues			755			760	842
Nucleotide			302			312	344
Model vs. Data
CC (mask)			0.81			0.8	0.73
CC (box)			0.74			0.73	0.83
CC (peak)			0.7			0.69	0.72
CC (volume)			0.79			0.78	0.76
R.m.s. deviations
Bond lengths (Å)			0.008			0.008	0.008
BoBond angles (°)			0.691			0.753	0.867
Validation
MolProbity score			2.42			2.56	2.96
Clashscore			6.86			9.71	30.6
Ramachandran plot
Favored (%)			94.99			95.3	93.93
Allowed (%)			5.01			4.57	6.07
Disallowed (%)			0			0.13	0
Protein data bank ID			7KBD			7KBE	7KBF

Experimental Model and Subject Details

Xenopus laevis sperm and eggs were collected by the method described previously (Murray, 1991). Sperm were collected from the matured (7.5-9 cm) male Xenopus laevis (Nasco, LM00715). Eggs were collected from the matured (9+ cm) female Xenopus laevis (Nasco, LM00535). Frogs were housed in the water tanks with temperature controlled at 16-20 °C at the Rockefeller University Comparative Bioscience Center. Prior to ovulation, 11 frogs were transferred at the certified satellite animal facility with temperature control in the Funabiki lab. Animal husbandry and protocol (#20031) approved by the Institutional Animal Care and Use Committee at the Rockefeller University were followed.

Method details

Xenopus egg extracts and chromatin formation

Cytostatic factor (CSF) metaphase-arrested Xenopus laevis egg extracts were prepared with the method described previously (Murray, 1991). To prevent spontaneous cycling of egg extracts, 0.1 mg/ml cycloheximide was added to the CSF extract. For interphase chromosome preparation, Xenopus laevis sperm nuclei (final concentration 2000/μl) were added to 8 ml of CSF extracts, which were incubated for 90 min at 20 °C after adding 0.3 mM CaCl₂, which releases CSF extracts into interphase. To monitor spindle assembly, Alexa594-labeled-bovine brain tubulin (final concentration 19 nM) was added to the extract during the incubation. For metaphase sperm chromosome preparation, cyclin B Δ90 (final concentration 24 μg/ml) and 4 ml fresh CSF extract was added to 8 ml of the extract containing interphase sperm nuclei prepared with the method described above. The extracts were incubated for 60 min at 20 °C, during which each tube was gently mixed every 10 min.

Nucleosome isolation from Xenopus egg extracts chromosomes

To crosslink the Xenopus egg extracts chromosomes, nine times volume of ice-cold buffer XL (80 mM PIPES-KOH [pH 6.8], 15 mM NaCl, 60 mM KCl, 30 % glycerol, 1 mM EGTA, 1 mM MgCl₂, 10 mM β-glycerophosphate, 10 mM sodium butyrate, 2.67 % formaldehyde) was added to the interphase or metaphase extract with chromosomes, which was further incubated for 60 min on ice. These fixed samples were layered on 3 ml of buffer SC (80 mM HEPES-KOH [pH 7.4], 15 mM NaCl, 60 mM KCl, 1.17 M sucrose, 50 mM glycine, 0.15 mM spermidine, 0.5 mM spermine, 1.25x cOmplete EDTA-free Protease Inhibitor Cocktail (Roche), 10 mM β-glycerophosphate, 10 mM sodium butyrate, 1 mM EGTA, 1 mM MgCl₂) in 14 ml centrifuge tubes (Falcon, #352059) and spun at 3,300 (2,647 rcf) rpm at 4 °C for 40 min using JS 5.3 rotor in Avanti J-26S (Beckman Coulter). Chromatin was collected from the bottom of the centrifuge tube. Collected chromatin was further spun through buffer SC as described above. Chromatin was collected from the bottom of the centrifuge tube, and then this step was repeated once again. Only at the 1^st round of centrifugation over sucrose cushion, chromatin trapped at the boundary between the extract and buffer SC was also collected and applied for 2^nd round centrifugation over sucrose cushion. The isolated interphase nuclei and metaphase chromosomes were associated with white aggregations and spindle microtubules, respectively. To remove them, chromosomes were gently pipetted with a 2 ml pipette. These chromosomes were pelleted by centrifugation at 5,492 rpm (2,500 rcf) using SX241.5 rotor in Allegron X-30R (Beckman Coulter). The chromosome pellets were resuspended with 200 μL of buffer SC. To digest chromatin, 1.5 U/μL (0.1 U/μL for the egg extract lot 2) of MNase (Worthington Biochemical Corporation) and CaCl₂ were added to 7.4 mM, and the mixture was incubated at 4 °C for 6 h. MNase reaction was stopped by adding 900 μL MNase stop buffer (15 mM HEPES-KOH [pH 7.4], 150 mM KCl, 5 mM EGTA, 10 mM β-glycerophosphate, 10 mM sodium butyrate, 5 mM DTT). The soluble fractions released by MNase were isolated by taking supernatants after centrifugation at 13,894 rpm (16,000 rcf) at 4 °C for 30 min using SX241.5 rotor in Allegron X-30R (Beckman Coulter). The supernatants were collected and layered onto the 10-22 % linear sucrose gradient solution with buffer SG (15 mM HEPES-KOH [pH 7.4], 50 mM KCl, 10-22 % sucrose, 10 μg/ml leupeptin, 10 μg/ml pepstatin, 10 μg/ml chymostatin, 10 mM sodium butyrate, 10 mM β-glycerophosphate, 1 mM EGTA) and spun at 32000 rpm (max 124,436 rcf) and 4 °C for 13 h using SW55Ti rotor in Optima L80 (Beckman Coulter). The samples were fractionated from the top of the sucrose gradient.

To prepare “freed nucleosome”, the interphase nucleosomes that was crosslinked after MNase, chromosome-containig Xenopus egg extracts were diluted with the buffer XL without formaldehyde and incubated for 60 min on ice. The soluble chromatin fraction after stopping the MNase reaction using the same method as described above was concentrated to 60 μL using Amicon Ultra centrifugal filter 100K (Millipore Sigma; 100K filter was chosen to avoid concentration of MNase). The concentrated nucleosomes were crosslinked by adding 540 μL buffer XL and incubated on ice for 1 h. The crosslinking reaction was stopped by diluted with 1.8 ml buffer Q (15 mM HEPES-KOH [pH 7.4], 15 mM NaCl, 60 mM KCl, 1 mM EGTA, 1 mM MgCl₂, 10 mM β-glycerophosphate, 10 mM sodium butyrate, 500 mM glycine). Quenched sample was layered onto the 10-22 % linear sucrose gradient solution with buffer SG and spun at 27,000 rpm (max 129,657 rcf) at 4 °C for 20 h using SW40Ti rotor in Optima L80 (Beckman Coulter). The centrifuged samples were fractionated from the top of the sucrose gradient.

Native and SDS-PAGE analysis of nucleosomes

For the native PAGE for nucleosome (Fig 1C, S8D, S8G), nucleosome fractions containing 100 ng DNA were to load onto a 0.5x TBE 6 % native PAGE gel. For the native PAGE for nucleosomal DNA (Fig 1D, S8E, S8H), nucleosome fractions containing 100 ng DNA were mixed with 1 μL of 10 mg/ml RNaseA (Thermo Scientific) and incubated at 55 °C for 30 min. To deproteinize and reverse-crosslink DNA, RNaseA treated samples were then mixed with 1 μL of 19 mg/ml Proteinase K solution (Roche) and incubated at 55 °C for overnight. Samples were loaded to 0.5x TBE 6 % native PAGE. Native PAGE gels were stained by SYBR-safe or SYTO-60 to detect DNA. For SDS-PAGE analysis (Figure 1E, S8F), nucleosome fractions containing 200 ng DNA were loaded to a 4-20 % gradient gel.

Western blotting of native PAGE to detect H1.8-bound nucleosome

Fifteen μL of the interphase-mono and metaphase-mono fractions just after sucrose gradient was applied onto 6 % native PAGE with x0.5 TBE, and DNA was stained by SYBR-safe (Invitrogen). After acquiring image, the native PAGE gel was soaked in transfer buffer (25 mM Tris, 192 mM glycine and 20% methanol) and blotted onto the nitrocellulose membrane (GE) using TE42 Tank Blotting Units (Hoefer) at 15 V, 4 °C for 4 h. Histone H4 was detected by 2.4 μg/ml of mouse monoclonal antibody (CMA400) provided by H. Kimura (Hayashi-Takanaka et al., 2015). H1.8 was detected by 1μg/ml of rabbit polyclonal antibody against full-length Xenopus laevis H1.8 (Jenness et al., 2018). As secondary antibodies, IRDye 680LT goat anti-rabbit IgG (Li-Cor 926-68021; 1:10,000) and IR Dye 800CW goat anti-mouse IgG (Li-Cor 926-32210; 1:15,000) were used. The images were taken with Odyssey Infrared Imaging System (Li-Cor).

Nucleosome dialysis and concertation for Cryo-EM and MS

Nucleosome contained sucrose gradient fractions were collected and dialyzed against 600 ml dialysis buffer EM (10 mM HEPES-KOH [pH 7.4] 30 mM KCl, 1 mM EGTA, 0.3 μg/ml leupeptin, 0.3 μg/ml pepstatin, 0.3 μg/ml chymostatin, 1 mM sodium butyrate, 1 mM β-glycerophosphate) using Tube-o-dialyzer 15 kDa (G-Biosciences) at 4 °C. The samples were further dialyzed twice against the fresh dialysis buffer. The samples were concentrated using Amicon Ultra centrifugal filters 10K (Millipore Sigma). Absorbance at wavelength 260 nm were 2.705, 1.342, 3.756, 2.564, 2.06, 5.244, 1.389, 2.479, and 2.915 for interphase mono nucleosome (lot 1), interphase di nucleosome (lot 1), interphase oligo nucleosome (lot 1), metaphase mono nucleosome, metaphase di nucleosome (lot 1), metaphase oligo nucleosome (lot 1), interphase oligo nucleosome (lot 2), metaphase oligo nucleosome (lot 2), and freed nucleosome (lot 2) respectively. The isolated nucleosomes were stored at 4 °C.

Mass spectrometry

The samples containing 800 ng DNA were incubated at 95 °C for 30 min to reverse crosslink. DNA amounts were estimated by the 260 nm absorbance. Samples were applied to SDS-PAGE (4-20% gradient gel) (Bio-Rad). The Gel was stained by Coomassie Brilliant Blue G-250 (Thermo Fisher). Gel bands were excised by scalpel and cut into pieces of approximately 2 mm x 2 mm. Destaining, in-gel digestion and extraction was performed as described (Shevchenko et al., 2007). Briefly, bands were excised with scalpel and de-stained using 30 % acetonitrile, 100 mM ammonium bicarbonate in water. Gel pieces were dehydrated using 100 % acetonitrile. Disulfide bonds were reduced with dithiothreitol, and cysteines were alkylated using iodoacetamide. Proteins were digested by hydrating the gel pieces in a solution of sequencing grade trypsin and endopeptidase LysC in 50 mM ammonium bicarbonate. Digestion proceeded overnight at 37 °C. Peptides were extracted into 70 % acetonitrile, 5 % formic acid for three iterations. Extracted peptides were purified using in-house constructed micropurification C18 tips. Purified peptides were analyzed by LC-MS/MS using a Dionex3000 HPLC equipped with a NCS3500RS nano- and microflow pump coupled to a Q-Exactive HF mass spectrometer (Thermo Scientific). Peptides were separated by reversed phase chromatography (solvent A: 0.1% formic acid in water, solvent B: 80% acetonitrile, 0.1% formic acid in water) across a 70-min gradient. Spectra were recorded in positive ion data-dependent acquisition mode, fragmenting the 20 most abundant ions within each duty cycle. MS1 spectra were recorded with a resolution of 60 k and AGC target of 3e6. MS2 spectra were recorded with a resolution of 30 k and ACC target of 2e5. Spectra were queried against a Xenopus laevis database (Wühr et al., 2014) (3413 sequences) concatenated with common contaminants using MASCOT through Proteome Discoverer v.1.4 (Thermo Scientific). Oxidation of M and acetylation of protein N-termini were set as variable modifications and carbamidomethylation of C was set as static modification. A false discovery rate of 1% was applied. A large number of identified peptides map to multiple highly similar proteins. iBAQ values of these peaks were summed before drawing the figures (Schwanhüusser et al., 2011).

Western blotting

Nucleosome fractions containing 200 ng DNA were applied for SDS-PAGE with 4-20 % gradient SDS-PAGE gel (Bio-rad). Proteins were transferred to the nitrocellulose membrane (GE) from the SDS-PAGE gel using TE42 Tank Blotting Units (Hoefer) at 15 V, 4 °C for 4 h. As primally antibodies, 1 μg/ml of Mouse monoclonal H3T3ph antibody 16B2 (Kelly et al., 2010) and 1 μg/ml of mouse monoclonal PCNA antibody (Santa Cruz SC-56) were used. For H4 and H1.8 detection, antibodies described above were used as primally antibodies. As secondary antibodies, IRDye 680LT goat anti-rabbit IgG (Li-Cor 926-68021; 1:10,000) and IR Dye 800CW goat anti-mouse IgG (Li-Cor 926-32210; 1:15,000) were used. The images were taken with Odyssey Infrared Imaging System (Li-Cor).

Histone preparation for recombinant nucleosome

All histones were purified with the method described previously (Zierhut et al., 2014). Briefly, bacterial expressed X. laevis H2A, H2B, H3.2, and H4 were purified from inclusion bodies. Histidine-tagged histones (H2A, H3.2, and H4) or untagged H2B expressed in bacteria were resolubilized from the inclusion bodies by incubation with the 6 M guanidine HCl. For Histidine-tagged histones, the resolubilized histidine-tagged histones were purified with Ni affinity chromatography using Ni-NTA beads (Qiagen). For untagged H2B, resolubilized histones were used for H2A-H2B dimer formation without further purification. To reconstitute H3–H4 tetramer and H2A–H2B dimer, denatured histones were mixed at a concentration of ~45 μM and dialyzed to refold histones with removing the guanidine. Histidine-tags were removed by TEV protease treatment, and H3–H4 tetramer and H2A–H2B dimer were isolated with gel-filtration chromatography on a HiLoad 16/60 Superdex 75 pg column (GE Healthcare). Fractions containing (H3–H4 tetramer and H2A–H2B dimer) were concentrated and stored at −80 °C.

DNA preparation for recombinant nucleosome

Four tandem 176 bp GUB DNA sequences (An et al., 1998) were cloned into pUC18 vector (pUC18-GUB_176x4). The E.coli DH5a that possessing the plasmid was cultured in 1.5x TBG-M9-YE medium with 25 μg/ml carbenicillin at 37 °C for overnight. The plasmid was purified using plasmid plus giga kits (Qiagen) following standard protocol. The 176 bp GUB DNA was cut out from the purified plasmid with EcoRV (New England Biolabs). The plasmid backbone was removed by polyethylene glycol precipitation using PEG-6000. The 176 bp GUB DNA was recovered from the supernatant of the polyethylene glycol precipitation fraction and further purified with gel-filtration chromatography on a HiLoad 16/60 Superdex 75 pg column (GE Healthcare) using TE buffer (10 mM Tris-HCl [pH 7.5] and 0.1 mM EDTA). DNA containing fractions were collected and stored at −20 °C.

The DNA sequence of the 176 bp GUB DNA is the following. 5’- ATCCC TCTAG ACGGA GGACA GTCCT CCGGT TACCT TCGAA CCACG TGGCC GTCTA GATGC TGACT CATTG TCGAC ACGCG TAGAT CTGCT AGCAT CGATC CATGG ACTAG TCTCG AGTTT AAAGA TATCC AGCTG CCCGG GAGGC CTTCG CGAAA TATTG GTACC CCATG GAAGA T-3’

Recombinant nucleosome reconstitution

Approximately 1.6 μM of purified 176 bp GUB DNA was mixed with 3.2 μM H3-H4 reconstituted histone dimers and 3.5 μM H2A-H2B reconstituted dimers with the salt dialysis method described previously (Zierhut et al., 2014). The sample mixture was transferred into a dialysis cassette and placed into a high salt buffer (10 mM Tris-HCl [pH 7.5], 1 mM EDTA, 2M NaCl, 5 mM β-mercaptoethanol, and 0.01% Triton X-100). Using a peristaltic pump, the high salt buffer was exchanged with low salt buffer (10 mM Tris-HCl [pH 7.5], 1 mM EDTA, 50 mM NaCl, 5 mM β-mercaptoethanol, 0.01% Triton X-100) at roughly 2 ml/min for overnight at 4 °C. In preparation for cryo-EM image collection, the dialysis cassette containing the sample was then placed in a buffer containing 10 mM HEPES-HCl (pH 7.4) and 30 mM KCl, and dialyzed for 48 h at 4 °C. The sample was then incubated at 55 °C for 2 h and centrifuged at 4 °C, 5000 rpm for 5 min in 5415D centrifuge (Eppendorf) to remove aggregates.

Recombinant nucleosome with open linker DNA conformation (GUB-open) purification

The reconstituted nucleosomes were purified and fixed using the GraFix method. Briefly, a sucrose gradient was created using 2.3 ml of a 20 % sucrose solution containing 15 mM HEPES [pH 7.4] and 4 % paraformaldehyde aqueous solution (EM Grade) (Electron Microscopy Sciences) and 2 ml of a 10% sucrose solution containing 15 mM HEPES [pH 7.4]. Reconstituted nucleosome samples were overlaid onto the sucrose gradient and spun at 32000 rpm for 20 h at 4 °C using SW55Ti rotor in Optima L80 (Beckman Coulter). Fractions containing the fixed samples were collected and dialyzed in 15 mM HEPES (pH 7.4) for 48 h at 4 °C with one fresh exchange of buffer halfway through, using a Tube-O-DIALYZER (50 kDa) (G-Biosciences). Samples were concentrated using an Amicon Ultra 100K centrifugation filter (Millipore) to final concentration of 247 ng/μl DNA.

Recombinant nucleosomes with closed linker DNA conformation (GUB-closed) purification

Reconstituted nucleosome with GUB DNA was dialyzed against the XB-CSF buffer (10 mM HEPES-KOH [pH 7.4], 100 mM KCl, 1 mM EGTA, 1 mM MgCl₂, 0.1 mM CaCl₂, 1.7 % sucrose) and concentrated to 60 μL using Amicon Ultra centrifugal filters 100K (Millipore Sigma). The concentrated nucleosome was crosslinked by diluted with 540 μL buffer XL and incubated on ice for 1 h. The crosslinking reaction was stopped by diluting with 1.8 ml buffer Q. Quenched sample was layered onto the 10-22 % linear sucrose gradient solution with buffer SG and spun at 27000 rpm (max 129,657 rcf) and 4 °C for 20 h using SW40Ti rotor in Optima L80 (Beckman Coulter). The centrifuged samples were fractionated from the top of the sucrose gradient. Nucleosome-containing fractions were collected and dialyzed against 600 ml dialysis buffer EM using Tube-o-dialyzer 15 kDa (G-Biosciences) at 4 °C. The samples were concentrated using Amicon Ultra centrifugal filters 10K (Millipore Sigma) to a final concentration of 448 ng/μl DNA.

Cryoprotectant buffer optimization

Unlike previously reported recombinant nucleosome with 601 DNA (Chua et al., 2016), nucleosomes from Xenopus egg extract chromosomes were thoroughly denatured during cryo-EM grid freezing without cryoprotectant (Fig S2B left). In those girds, many free DNAs that may be released from broken nucleosomes are observed. This suggest that chromosomal nucleosomes with native DNA sequence are fragile than nucleosome with strong positioning sequence. To reduce the structural denaturization during cryo-EM grid freezing, cryoprotectant buffer were screened and optimized to the buffer containing 1% trehalose and 0.1% 1,6,-hexanediol. Under this condition, although a small amount of free DNAs are still observed, most of nucleosome-like structure remained intact (Figure S2B right).

Cryo-EM grid preparations and data collection

For the nucleosomes purified from Xenopus egg extracts and GUB-closed nucleosomes, 2 μl of sample was mixed with 0.5 μl of cryoprotectant buffer (10 mM HEPES-KOH [pH 7.4], 30 mM KCl, 1 mM EGTA, 0.3 μg/ml Leupeptin, 0.3 μg/ml Pepstatin, 0.3 μg/ml Chymostatin, 1 mM Sodium Butyrate, 1 mM β-glycerophosphate, 5% trehalose, 0.5% 1,6,-hexanediol) just before loading the sample on the cryo-EM grid. 2.5 μl of the sample was frozen onto a glow discharged Quantifoil Gold R 1.2/1.3 300 mesh grid (Quantifoil). For GUB-open nucleosome, 2.5 μl of the samples was frozen onto a glow discharged Quantifoil Gold R 1.2/1.3 300 mesh grid (Quantifoil) without adding cryoprotectant buffer. The samples were frozen under 100% humidity, 20 sec incubation, and 5 sec blotting time using the Vitrobot Mark IV (FEI). Grids were imaged on a Talos Arctica (FEI) installed with a K2 Camera (GATAN) and a field emission gun operating at 200 kV or Titan Krios (FEI) installed with a K2 Camera (GATAN) and a 300 kV field emission gun. All data were collected as 50 frames/movie in super resolution mode. The conditions for the data collection were listed in Table 1 and 2.

TRSF cryo-EM structure reconstruction

Briefly, particles were picked from micrographs (Fig. S2A) using the CryoSPARC blob picker and were then split into two groups based on the resolution and effective classes assigned (ECA) value of the 2D classification (Fig. 2A: 2D classification) (Punjani et al., 2017). Multiple 3D maps were generated from both groups by CryoSPARC ab initio reconstruction (Punjani et al., 2017), resulting in two types of maps: 3D maps derived from particles with high-resolution features, and “decoy” maps derived from the particles that were noisy, low-resolution, or ambiguous. All picked particles and 3D maps were then used for multireference 3D classification by CryoSPARC heterogenous refinement (Fig 2A: decoy classification, Fig S3A-B, S3E-F). Compared to other workflows, our pipeline uniquely combines three main strategies that allowed for unbiased assessment of conformational diversity and in silico purification of our extremely heterogeneous sample.

First, we avoided using template-based particle picking, which can miss particles that do not resemble the template. Instead, particles were picked by CryoSPARC blob picker, which does not require a template map (Fig 2A: blob pick). 2D class averaging revealed diverse particle views and protein heterogeneity, confirming the success of this particle picking approach (Fig 2A: 2D classification).

Second, when reference maps for 3D classification were generated, we avoided introducing a human selection bias. In a common classification pipeline, one might subjectively select 2D classes that are similar to the expected structure of a target protein, and then generate a 3D initial map using these selected 2D classes (Bell et al., 2016; Vargas et al., 2014). Alternatively, one might use previously-determined cryo-EM or crystal structures as references for 3D classification (Bell et al., 2016; Grigorieff, 2016; de la Rosa-Trevín et al., 2016; Punjani et al., 2017; Scheres, 2012). While these common methods may work for structural analysis of homogenous samples, they may exclude un-recognizable structures in heterogenous samples. In contrast, employing CryoSPARC’s ab initio reconstruction (Punjani et al., 2017), our pipeline successfully generated multiple independent 3D structures simultaneously from heterogeneous particles that were selected based on the resolution and ECA of the 2D classes, rather than recognizable visual features. This allowed us to remove biases commonly generated by manual particle selection and use of externally derived 3D references.

Finally, both 3D reference generation and 3D classification were performed using all picked particles (Fig 2A: decoy classification). Common pipelines use multiple copies of a single 3D map for 3D classification (Lyumkis et al., 2013; Scheres et al., 2007). While these pipelines are suited for classifying minor structural variations of the target protein, they may miss rare or radically altered structures. Instead, we used twenty-four distinct 3D references, all of which were data-derived and thus encompassed the innate heterogeneity of the sample. These include ten “target” classes to resolve heterogeneous high-resolution particles and fourteen “decoy” classes to filter out noise and low-quality protein particles (Fig 2A: decoy classification, Fig S3A-B, S3E-F). This decoy classification approach is reminiscent of “random-phase” 3D classification (Gong et al., 2016) and multireference 3D classification with structurally dissimilar cryo-EM maps as decoys (Nguyen et al., 2019). However, using data-derived decoys instead is useful to tailor 3D references to the specific noise present in the dataset without introducing any external biases (Lilic et al., 2020).

Specifically, both for the interphase oligo fraction and metaphase oligo fraction, data were processed with the identical procedure. Movie frames were motion corrected and dose weighted with a binning factor of 2 using MOTIONCOR2 (Zheng et al., 2017) with RELION3.0 (Scheres, 2012). Particles were picked by blob picker (minimum particle diameter = 50 Å, maximum particle diameter = 300 Å, minimum separation diameters = 100 Å) and extracted (extraction box size = 320 pixels, Fourier crop to 160 pixels) using CryoSPARC v2.15 (Punjani et al., 2017). Extracted particles were applied for 2D classification with 400 classes using CryoSPARC v2.15. Using 2D classification results, particles were split to the lower resolution group and the higher resolution group. The 2D classes whose resolution were higher than 28 Å (did not considered for the samples with egg extract lot 2) and effective classes assigned (ECA) value better than 2 (1.6 for the samples with egg extract lot 2) were assigned as the higher resolution group. 2D classes containing obvious ice signals, carbon signals, and strong small dots were manually removed from the higher resolution group and reassigned as the lower resolution group. 3D initial models were generated with ab initio reconstruction of CryoSPARC v2.15 (10 classes for the higher resolution group, 14 classes for the lower resolution group, maximum resolution = 20 Å, class similarity = 0). 3D classifications were performed using all eighteen 3D classes generated by Ab-initio reconstruction and all picked particles using the heterogeneous refinement of CryoSPARC v2.15 (Refinement box size = 160 voxels). After the 3D classification, particles for each reasonable structure class were re-extracted with the original pixel size with optimized box sizes for each particle. Nucleosome classes, intelectin classes, actin classes, and alpha2-macroglobulin classes were re-extracted with 256, 100, 320, 320 box sizes, respectively. New 3D references were prepared with the re-extracted particles by ab initio reconstruction and further refined with homogenous refinement or non-uniform refinement using CryoSPARC v2.15. For the interphase actin and intelectin classes, multiple models were generated by ab initio reconstruction (Class similarity = 0) and further “decoy” classifications were performed. For the mono fractions, the same pipeline was used with several modifications: micrographs were denoise with Topaz denoise to improve blob pick accuracy, maximum particle diameter of blob pick = 200 Å, initial extraction box size = 256 pixels (Fourier crop to 128 pixels), higher resolution group threshold; < 20Å resolution, < 2.0 ECA. Final extraction box size = 200 pixels. Chimera and ChimaraX were used for 3D map visualization (Pettersen et al., 2004).

Actin cryo-EM structure analyses as a quality control of the EM grids

For the EM grids of oligo nucleosome fractions that contain decent amount of actin filament, we used the actin filament cryo-EM structure as an internal control for assessing the sample denaturization and grid qualities. Actin particles were picked by Topaz v0.2.3 using actin particles separated from the simultaneous cryo-EM structures reconstruction as training models (Bepler et al., 2019). For each EM grids for oligo fractions of the chromosomal nucleosomes, 7500 particles were randomly selected and extracted (box size = 320 pixels). Using identical input 3D map, maps were refined for each EM grids separately with homogenous refinement of CryoSPARC v2.15 (without helical refinement). Grids in which the filament actin resolution surpassed 5 Å were qualified and used for this study.

Cryo-EM image processing for the averaged nucleosome structure determination

While our TRSF reconstruction pipeline (Fig. 2) excelled at reducing particle selection bias, the particle picking process inevitably missed many nucleosome particles on the micrographs: a 100 Å minimum inter-particle distance was needed to prevent redundant picking of large proteins by CryoSPARC blob picker, but this excluded many clustered nucleosome particles.

Unlike the TRSF pipeline, this Topaz and decoy classification-based pipeline was exclusively focused on analyzing nucleosome-like structures. However, we expect that structural diversity within these nucleosome-like particles was still preserved for three reasons. First, although Topaz initially required manual selection of particles for training, it failed to remove even lectin particles (Fig. S6A: decoy classification), suggesting that the selection was not stringent enough to exclude nucleosome structural variants. Second, 3D decoy classification was performed with all picked particles since manual particle curation could propagate bias. Third, only one nucleosome 3D map was used for decoy classification to include all nucleosome-like particles within one reconstruction, while multiple decoy classes were used to exclude unrelated particles (Fig. S6A: decoy classification). As we will describe later, both octasome and hexasome particles were assigned to the nucleosome class, suggesting that the output maps of this pipeline retain structural diversity of the nucleosome in chromosomes.

Specifically, All Cryo-EM images of interphase mono-, interphase di-, interphase oligo-, freed-, metaphase mono-, metaphase di-, metaphase oligo-, and in vitro reconstituted nucleosome were processed with the same procedure (Fig S6A). Movie frames were motion-corrected and dose-weighted with a binning factor of 2 using MOTIONCOR2 (Zheng et al., 2017)with RELION3.0 or 3.1 (Scheres, 2012). Particles were picked by Topaz v0.2.3 with 500~2000 manually picked nucleosome-like particles as training models (Bepler et al., 2019). Picked particles were extracted using CryoSPARC v2.15 (extraction box size = 200 or 256 pixels) (Punjani et al., 2017). Extracted particles were applied for 2D classification with 200 classes using CryoSPARC v2.15. Using 2D classification result, particles were split to the nucleosome-like groups and the non-nucleosome-like groups. For the interphase di- and metaphase di- data, intelectin 2D classes were removed, and remaining particles were applied for second round of 2D classification, because abundant intelectin-2 particles inhibited the formation of nucleosome-like 2D classes. Four of 3D initial models were generated for both groups with ab initio reconstruction of CryoSPARC v2.15 (Class similarity = 0). One nucleosome like model was selected and used as a given model of 3D classification with all four of the “decoy” classes generated from non-nucleosome-like group. After the first round 3D classification, the particles that were assigned to the “decoy” classes were removed, and remained particles were applied for the second round 3D classification with the same setting with the first round. These steps were repeated until 90~95 % of particles were classified as a nucleosome-like class. Repeat times of this step were described in Table S2. Using RELION 3.1 beta, CTF refinement, Bayesian polishing, and postprocessing were performed. Local resolution was calculated with the Blocres and Blocfilt in the Bsoft package (Cardone et al., 2013). Chimera and ChimaraX were used for 3D map visualization (Pettersen et al., 2004).

For the 3D classification to analyze linker DNA structure variety shown in Fig 5, S8K, and S10A-B, six to eight 3D references were generated with ab initio reconstruction of CryoSPARC v2.15 using purified nucleosome-like particles listed in Table S2 (Class similarity = 0.9). 3D classifications was performed using all newly generated 3D references and purified particles using the heterogeneous refinement of CryoSPARC v2.15.

For the atomic model building, initial atomic models of linker DNA containing nucleosomes were built by combining linker DNA moiety of gH1.0 binding nucleosome (PDB ID: 5NL0) and nucleosome moiety of ScFv binding nucleosome (PDB ID: 6dzt) and by replacing amino acids sequence to that of Xenopus laevis core histones using Coot (Bednar et al., 2017; Emsley and Cowtan, 2004; Zhou et al., 2019). The initial atomic model was docked onto the cryo-EM map postprocessed by RELION 3.1 beta and refined using Coot and PHENIX (Liebschner et al., 2019).

3D Variability Analysis (3DVA)

For the 3D structure dynamics analysis of nucleosome containing linker DNA shown in Fig 4A, the purified nucleosome particles used for 3D classification analysis (listed in Table S2) were used for the 3DVA in CryoSPARC v2.15 (Punjani and Fleet, 2021). Numbers of 3DVA components and filter resolution were set by user. After several round of tryout, numbers of 3DVA components were manually optimized to 2, which can explain most of the representative protein movements of chromosomal nucleosome without making redundant component. For the interphase nucleosomes crosslinked after MNase, filter resolution was set to 12 Å, whereas for other samples, filter resolution was set to 8 Å. All output 3D maps were loaded with Chimera (Pettersen et al., 2004) and movies were created from the maps.

For the 3DVA of NCPs shown in Fig 4B, C, linker DNA densities of each nucleosome were removed from original particles listed in Table S2 using particle subtraction in cryoSPARC v2.15.

For GUB-open nucleosomes, the new initial 3D map was generated with linker DNA subtracted particles by ab initio reconstruction of cryoSPARC v2.15. The 3D map was further refined with homogeneous refinement of cryoSPARC v2.15. Particle stack and mask generated by NU-refinement were used for 3DVA of cryoSPARC v2.15 (number of modes = 2, Filter resolution = 8Å).

For chromosomal NCPs, 100,000 particles were selected from interphase or metaphase nucleosome particles whose linker DNAs had been subtracted. Particles were mixed, and new initial 3D map was generated by ab initio reconstruction of cryoSPARC v2.15. Particle stack and mask generated by homogeneous refinement were used for 3DVA of cryoSPARC v2.15 (number of modes = 4, Filter resolution = 8Å).

To calculate the surface hydrophobicity and electrostatic potential and global R.M.S.D., an atomic model was generated for each NCP structural variant. Using PHENIX package, the atomic model of interphase oligo nucleosome was docked on to the 3D maps of NCP structural variants generated by 3DVA. Docked atomic models were refined with rigid body refinement in PHENIX. Global R.M.S.D. of Cα and P atoms were calculated by PyMol. To show the surface hydrophobicity (Fig. 7E), hydropathy indexes (Kyte and Doolittle, 1982) were assigned to each amino acid residue and mapped on the original EM maps by Chimera. To show the electrostatic potential (Fig. 7F), surface electrostatic potentials were calculated by APBS and PDB2PQR (Baker et al., 2001; Dolinsky et al., 2004) using refined atomic models for each NCP structural variant, and the colored surface electrostatic potentials were drawn by Chimera.

In silico mixing 3D classification for assessing linker DNA angle

For the in silico mixing 3D classification for analyzing linker DNA structure variation shown in Fig 5 and S6, five different linker DNA angle 3D maps were selected as the fixed templates for all samples (“closed 1” : metaphase oligo class A (Fig 5A), “closed 2” : metaphase oligo class D (Fig 5A), “open 1” : interphase oligo class E (Fig 5A), “open 2” : interphase crosslink after MNase class E (Fig 5B), and “misc.” : metaphase oligo class F (Fig 5A)). As decoy maps, 4 noise classes (shown in Fig. S6A: decoy classification) were used. Because of the original pixel size discrepancy, in silico mixing 3D classification for mono nucleosome fractions (original image pixel size = 1.33 Å/pixel), di nucleosome fractions (original image pixel size =1.6 Å/pixel), and others (original image pixel size = 1.5 Å/pixel) were performed separately. For the mono nucleosome fractions, from the purified particle set of each interphase and metaphase nucleosome sample (listed in Table S2), 17000 particles were selected randomly and extracted (original image pixel size= 1.33Å/pixel, extraction box size = 300 pixels, Fourier crop to 150 pixels). Particles of mono nucleosome fractions were mixed with 17,000 particles randomly selected from the purified particle set (listed in Table S2) of interphase oligo nucleosome (egg extract lot1), metaphase oigo nucleosome (egg extract lot1), and freed nucleosomes (original image pixel size = 1.5Å/pixel, extraction box size = 266 pixels, Fourier crop to 150 pixels). For the di nucleosome fractions, 12,000 particles of interphase or metaphase nucleosomes were extracted (original image pixel size= 1.6 Å/pixel, extraction box size = 180 pixels, Fourier crop to 150 pixels). Particles of di nucleosome fractions were mixed with the 12,000 particles randomly selected from the purified particle set (listed in Table S2) of interphase oligo nucleosomes (egg extract lot1), metaphase oligo nucleosomes (egg extract lot1), and freed nucleosomes (original image pixel size = 1.5 Å/pixel, extraction box size = 192 pixels, Fourier crop to 150 pixels). For oligo nucleosome fractions, 25,000 particles of interphase nucleosomes (egg extract lot1), metaphase nucleosomes (egg extract lot1), interphase nucleosomes (egg extract lot2), metaphase nucleosomes (egg extract lot2), and freed nucleosomes were selected from the purified particle set (listed in Table S2), extracted (image pixel size = 1.5 Å/pixel, extraction box size = 200 pixels), and mixed. After mixing these particles, 3D classification was performed five times using all five 3D maps and merged particle set using heterogeneous refinement of CryoSPARC v2.15 (refinement box size =200, initial resolution = 32 Å). The particles assigned to each class were counted (Table S3) and plotted.

In silico mixing 3D classification for assessing H4 N-terminal tail orientations

The analysis pipeline was depicted in Fig. S9A. Since each nucleosome particle has two H4 N-terminal tails and structural discrepancy between these tails deterred the 3D classification, one out the two H4 N-terminal tails was mathematically subtracted before in silico mixing 3D classification. In addition, to use the structural information of both side of the H4 tails, particles were symmetrically doubled prior to the H4 tail density subtraction. From each interphase oligo (egg extract lot1), metaphase oligo (egg extract lot1), interphase oligo (egg extract lot2), metaphase oligo (egg extract lot2), and recombinant nucleosome (GUB-closed), 50,000 particles were selected were mixed. Using mixed particles, 3D map of the nucleosome was refined applying C2 symmetry with homogeneous refinement of CryoSPARC v2.15. The particles were then symmetrically doubled using symmetry expansion of CryoSPARC v2.15. Density around the one side of the H4 N-terminal tail and both side of the linker DNA were subtracted from each particle using particle substruction of CryoSPARC v2.15.

To prepare 3D templates for the in silico mixing 3D classification, 3D map was refined using density subtracted particles with homogeneous refinement of CryoSPARC v2.15 (C1 symmetry). The H4 N-terminal tail density was masked, and the H4 N-terminal tail structural variants were generated by of CryoSPARC v2.15 (simple mode, number of modes = 2, Filter resolution = 6Å, number of frames = 20). The 3D maps representing ‘inward’ or ‘outward’ H4 tail orientation were selected as 3D templates for the in silico mixing 3D classification.

In silico mixing 3D classifications were performed using twelve of 3D templates (4 of the ‘inward’ H4 tail maps, 4 of the ‘outward’ H4 tail maps, and 4 of decoy maps) and 500,000 of the density subtracted particles with heterogeneous refinement of CryoSPARC v2.15 (C1 symmetry, refinement box size =200, initial resolution = 16 Å, number of initial random assignment iterations = 0). The in silico mixing 3D classification was performed five times. The particles assigned to each class were counted (Table S3) and plotted.

After the in silico mixing 3D classification, particles assigned to the 4 of ‘inward’ H4 tail maps and the 4 of the ‘outward’ H4 tail maps ware mixed, respectively. 3D maps of the nucleosomes with ‘inward’ and ‘outward’ H4 tail were refined with homogeneous refinement of CryoSPARC v2.15 (C1 symmetry). For both of the ‘inward’ and ‘outward’ map, particles were split back to the original image group. Using these particle sets, 3D maps of the nucleosomes were refined with homogeneous refinement of CryoSPARC v2.15 (C1 symmetry).

In silico mixing 3D classification for assessing NCP structural variation

The analysis pipeline for the in silico mixing 3D classification of major NCP variation shown in Fig. 7C was depicted in Fig. S12A. Three 3D maps (frame 4, 11, and 18) for each component were selected from 3DVA of GUB-open NCP, which had generated 21 motion frames per each component. In the 3DVA, largest structural differences were found between motion frames 1 and 21, and they were structurally most deviated from the averaged GUB NCP structure, while the structure of the motion frame 11 was most similar to that of the averaged GUB NCP. We chose frames 4 and 18, instead of frames 1-3 and 19-21, for reference maps because frames 1 and 21 contain many noisy densities (e.g., gaps in helices and densities outside of the NCP). As a 3D reference of canonical nucleosome structure in chromosome, a refined 3D map generated from 100,000 each of linker DNA subtracted interphase oligo nucleosome, metaphase oligo nucleosome particle was used. As decoy maps, 4 noise classes shown in Fig. S6A were used. As input particles for in silico mixing 3D classification,25,000 each of linker DNA subtracted interphase oligo nucleosome (egg extract lot 1), metaphase oligo nucleosome particle (egg extract lot 1), interphase oligo nucleosome (egg extract lot 2), metaphase oligo nucleosome particle (egg extract lot 2), and freed nucleosome was were randomly selected and mixed (Total 125,000 particles). Using these 3D reference maps and particles, 3D classification was performed five times using heterogeneous refinement of CryoSPARC v2.15 (refinement box size =200, initial resolution = 20 Å, number of initial random assignment iterations = 0). To confirm the accuracy of 3D classification, output 3D maps were compared to the input maps of the in silico mixing 3D classification, and the parameters of the in silico mixing 3D classification were optimized for output 3D maps to preserve the feature of input maps (Fig S12B). For each 3DVA component, the in silico mixing 3D classification was performed five times. The particles assigned to each class were counted (Table S3) and plotted.

The analysis pipeline for the in silico mixing 3D classification of minor NCP variation shown in Fig. 7D was depicted in Fig. S12C. Prior to the in silico mixing 3D classification using minor NCP variant maps, the particles of major NCP variant were removed by performing the in silico mixing 3D classification using 1 canonical NCP map, 6 major NCP variant maps (GUB-open, frame 4, 11, and 18 from component 1 and 2), and 4 decoy maps as input 3D maps (Fig. S12C). As input particles same 175,000 mixed particles for assessing the major NCP variation were used. Using these 3D reference maps and particles, 3D classification was performed using heterogeneous refinement of CryoSPARC v2.15 (refinement box size =200, initial resolution = 20 Å, number of initial random assignment iterations = 0). Using the 73,682 particles assigned to the canonical nucleosome, in silico mixing 3D classification of minor NCP variation were performed. As input maps for in silico mixing 3D classification, three 3D maps (frame 4, 11, and 18) of each component generated by 3DVA of chromosomal NCP were selected. The 3DVA of chromosomal NCP (using metaphase and interphase oligo nucleosome particles) generated 21 motion frames per each component. Frames 11 of each component are most similar to the canonical nucleosome structure in chromosome, while we chose frames 4 and 18 of components 1-4 to represent NCP structural variants, as the rationale explained above for the GUB nucleosome in silico mixing 3D classification. As decoy maps, 4 noise classes shown in Fig. S6A were used. Using these 3D reference maps and particles, 3D classification was performed five times using heterogeneous refinement of CryoSPARC v2.15 (refinement box size =200, initial resolution = 13 Å, number of initial random assignment iterations = 0). To confirm the accuracy of 3D classification, output 3D maps were compared to the input maps of the in silico mixing 3D classification, and the parameters of the in silico mixing 3D classification were optimized for output 3D maps to preserve the feature of input maps (Fig S12D). For each 3DVA component, the in silico mixing 3D classification was performed five times. The particles assigned to each class were counted (Table S3) and plotted.

Reconstruction of “closed” and “moderately open” linker DNA nucleosome maps (Fig. 6F)

Using the interphase nucleosome particles shown in Table S2, twenty classes of 3D maps were generated by 3DVA of cryoSPARC v2.15 (cluster mode, number of modes = 2, Filter resolution = 8Å, number of frames = 20). To generate the “closed” and “open” linker DNA nucleosome maps at the comparable resolution, five “closed” linker DNA classes (20,738 particles) and five “moderately open” linker DNA classes (19,120 particles) were selected, and initial 3D maps were generated by ab initio reconstruction of cryoSPARC v2.15. Initial 3D models were further refined with homogeneous refinement in cryoSPARC v2.15.

Cryo-EM image processing for the H1.8-bound nucleosome structure determination

41,416 particles assigned to the class that has the extra density of the metaphase oligo fraction were applied for homogeneous refinement (Fig S11A). Refined particles were further purified with the heterogeneous refinement using an H1-bound class and an H1-unbound class as decoys. The purified particles were further refined and split to four of 3D classes using the ab initio reconstruction of (4 classes, class similarity = 0.9) and heterogeneous refinement of CryoSPARC v2.15. Two classes that had reasonable extra density were selected and refined with homogeneous refinement. The final resolution was determined as 4.42Å with the gold stand FSC threshold (FSC = 0.143).

For the atomic model building, H1 segment of the gH1.0-bound nucleosome structure (PDB ID: 5NL0) was replaced with the H1.8 structure model built using SWISS-MODEL: homology modeling tool (Waterhouse et al., 2018). The initial atomic model was docked onto the cryo-EM map locally filtered by Bsoft (Blocres and Blocfilt) and refined using Coot and PHENIX (Cardone et al., 2013; Emsley and Cowtan, 2004; Liebschner et al., 2019). Chimera and ChimaraX were used for 3D map visualization (Pettersen et al., 2004).

Quantification and Statistical Analysis

In silico mixing 3D classification analyses were performed five times for each analysis (Fig 5E, 5F, 5G, 7C, 7D, and S9C). The relative populations of the particles assigned to each class were counted for each round of classification analysis. Average and standard deviation of populations of five independent in silico mixing 3D classification analyses were calculated using Excel (Microsoft). Number of the particles for each class used for each analysis and their classification results are reported in Table S3.

Supplementary Material

1

Figure360. Introduction video for 3DVA. Related to Figure 4A.

3

Movie S1. The 3DVA of the interphase oligo (lot 1), metaphase oligo (lot 1), freed (lot 2), and GUB-open nucleosome. Related to Figure 4, 5, 6, 7.

4

Movie S2. The 3DVA of GUB-open nucleosomes. Related to Figure 4, 7.

5

Movie S3. The 3DVA of chromosomal nucleosomes. Related to Figure 4, 7.

6

Table S1. Mass spectrometry analysis results. Related to Figure 1

Sheet 1. Raw data of the Mass spectrometry analysis of the mono, di, and oligo fractions of interphase and metaphase chromosome.

Sheet 2. Modified data with summing the iBAQ values of peptides that separately mapped to highly similar protein isoforms.

7

Table S3. Number of particles in each 3D class after in silico mixing 3D classification. Related to Star methods.

8

Figure S1-S12, Table S2

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-histone H4 mouse monoclonal antibody	(Hayashi-Takanaka et al., 2015)	CMA400
Anti-histone H1.8 (Xenopus laevis) mouse monoclonal antibody	(Jenness et al., 2018)	RU1974
Anti-histone H3T3ph mouse monoclonal antibody	(Kelly et al., 2010)	16B2
IRDye 680LT goat anti-rabbit IgG	LI-COR	926-68021
IR Dye 800CW goat anti-mouse IgG	LI-COR	926-32210
Biological samples
Xenopus laevis sperm nuclei	Nasco Isolated from male Xenopus laevis	LM00715 (male Xenopus laevis))
Xenopus laevis egg	Nasco Laid by female Xenopus laevis	LM00535 (female Xenopus laevis)
Alexa594-labeled-tubulin	(Hyman et al., 1991) Isolated from bovine brain	N/A
Chemicals, peptides, and recombinant proteins
Chorionic gonadotropin human	Sigma	CG10-1VL
Pregnant Mare Serum Gonadotropin (PMSG)	Prospec	HOR-272
Cycloheximide	Sigma	C-7698
CaCl2	Sigma Aldrich	C7902-500G
Cyclin B Δ90	(Glotzer et al., 1991)	N/A
Cysteine	Sigma	C7352-1KG
PIPES	Alfa Aesar	A16090
NaCl	Millipore Sigma	SX0420-5
Glycerol	Alfa Aesar	36646
EGTA	Sigma Aldrich	E4378-250G
MgCl2	Acros organic	197530010
β-glycerophosphate	Acros organic	410991000
Sodium butyrate	Aldrich	303410-100G
Formaldehyde	Fisher bioreagents	BP531-25
HEPES	Akron biotech	AK1069-1000
Sucrose	Fisher chemical	S5-3
KCl	Fisher chemical	P217-3
Spermidine	Sigma	S-2501
Spermine	Sigma	S1141-5G
cOmplete EDTA-free Protease Inhibitor Cocktail	Roche	11873580001
MNase	Worthington Biochemical Corporation	LS004798
Leupeptin	Millipore Sigma	El8
Pepstatin	Sigma	P5318-25MG
Chymostatin	Millipore Sigma	El6
RNaseA	Thermo Scientific	EN0531
Proteinase K solution	Roche	03115828001
SYBR-safe	Invitrogen	S33102
4-20 % gradient SDS-PAGE gel	BioRad	4561096
SYTO-60	Invitrogen	S11342
Tris	Sigma	T1503-5KG
Glycine	Sigma	G7126-5KG
Amicon Ultra centrifugal filter 100K	Millipore Sigma	UFC510024
Methanol	Fisher chemical	A452SK-4
Tube-o-dialyzer 15 kDa	G-Biosciences	786-618
Amicon Ultra centrifugal filters 10K	Millipore Sigma	UFC501024
Coomassie Brilliant Blue G-250	Calbiochem	3340
Xenopus laevis H2A, H2B, H3.2, H4	(Zierhut et al., 2014)	N/A
Ni-NTA beads	Qiagen	30210
TEV protease (S219V)	(Tropea et al., 2009)	N/A
Carbenicillin	Alfa Aesar	J61949
EcoRV (New England Biolabs)	New England Biolabs	R0195S
PEG-6000	Fluka	81253
Triton X-100	BioRad	161-0407
Methanol	Fisher chemical	A452SK-4
Paraformaldehyde aqueous solution	Electron Microscopy Sciences	157-8
Tube-O-DIALYZER (100 kDa) (G-Biosciences)	G-Biosciences	786-619
Trehalose	Sigma	T0167-100G
1,6,-hexanediol	Alfa Aesar	A12439
Deposited data
Recombinant poly-nucleosome (emd2601) (Song et al., 2014)	EMDR (Song et al., 2014)	emd2601
Recombinant Xenopus laevis canonical nucleosome	PDB (Luger et al., 1997)	1AOI
Recombinant gH1.0 binding nucleosome	PDB (Bednar et al., 2017)	5NL0
Recombinant ScFv binding canonical nucleosome	PDB (Zhou et al., 2019)	6DZT
Cryo-EM map of nucleosome in interphase oligo fraction (lot 1, TRSF)	This paper; EMDR	EMD-22793
Cryo-EM map of alpha-2-macroglobulin family protein in interphase oligo fraction (lot 1, TRSF)	This paper; EMDR	EMD-22794
Cryo-EM map of nucleosome in metaphase oligo fraction (lot 1, TRSF)	This paper; EMDR	EMD-22795
Cryo-EM map of alpha-2-macroglobulin family protein in metaphase oligo fraction (lot 1, TRSF)	This paper; EMDR	EMD-22796
Cryo-EM map of nucleosome in interphase mono fraction (lot 1, averaged)	This paper; EMDR	EMD-22797
Cryo-EM map of nucleosome in interphase di fraction (lot 1, averaged)	This paper; EMDR	EMD-22798
Cryo-EM map of nucleosome in interphase oligo fraction (lot 1, averaged)	This paper; EMDR	EMD-22790
Cryo-EM map of nucleosome in metaphase mono fraction (lot 1, averaged)	This paper; EMDR	EMD-22800
Cryo-EM map of nucleosome in metaphase di fraction (lot 1, averaged)	This paper; EMDR	EMD-22801
Cryo-EM map of nucleosome in metaphase oligo fraction (lot 1, averaged)	This paper; EMDR	EMD-22791
Cryo-EM map of freed nucleosome (crosslinked after MNase, averaged)	This paper; EMDR	EMD-23826
Cryo-EM map of nucleosome of interphase oligo fraction (lot 2, averaged)	This paper; EMDR	EMD-23819
Cryo-EM map of nucleosome of metaphase oligo fraction (lot 2, averaged)	This paper; EMDR	EMD-23821
Cryo-EM map of GUB-closed nucleosome (averaged)	This paper; EMDR	EMD-23823
Cryo-EM map of GUB-open nucleosome (averaged)	This paper; EMDR	EMD-23824
Cryo-EM map of H1-bound nucleosome of metaphase oligo fraction (lot 1)	This paper; EMDR	EMD-22792
Atomic coordinates of nucleosome of interphase oligo fraction (lot 1, averaged)	This paper; PDB	7KBD
Atomic coordinates of nucleosome of metaphase oligo fraction (lot 1, averaged)	This paper; PDB	7KBE
Atomic coordinates of H1 bound nucleosome of metaphase oligo fraction (lot 1)	This paper; PDB	7KBF
Raw micrographs of interphase mono fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10742
Raw micrographs of interphase di fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10744
Raw micrographs of interphase oligo fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10692
Raw micrographs of metaphase mono fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10743
Raw micrographs of metaphase di fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10745
Raw micrographs of metaphase oligo fraction (egg extract lot 1)	This paper; EMPIAR	EMPIAR-10691
Raw micrographs of ‘freed’fraction (Crosslinked after Mnase, interphase)	This paper; EMPIAR	EMPIAR-10750
Raw micrographs of interphase oligo fraction (egg extract lot 2)	This paper; EMPIAR	EMPIAR-10747
Raw micrographs of metaphase oligo fraction (egg extract lot 2)	This paper; EMPIAR	EMPIAR-10746
Raw micrographs of GUB-closed nucleosome	This paper; EMPIAR	EMPIAR-10749
Raw micrographs of GUB-open nucleosome	This paper; EMPIAR	EMPIAR-10748
Recombinant DNA
pUC18-GUB_176x4	This paper	N/A
Software and algorithms
MASCOT through Proteome Discoverer v.1.4	Thermo Scientific	N/A
MOTIONCOR2	(Zheng et al., 2017)	https://emcore.ucsf.edu/ucsf-software
RELION3	(Scheres, 2012)	https://github.com/3dem/relion
CryoSPARC v2.15	(Punjani et al., 2017)	https://cryosparc.com
Chimera	(Pettersen et al., 2004)	https://www.cgl.ucsf.edu/chimera/
ChimaraX	(Pettersen et al., 2004)	https://www.rbvi.ucsf.edu/chimerax/
Topaz v0.2.3	(Bepler et al., 2019)	https://github.com/tbepler/topaz
Bsoft	(Cardone et al., 2013)	https://lsbr.niams.nih.gov/bsoft/
Coot	(Emsley and Cowtan, 2004)	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PHENIX	(Liebschner et al., 2019)	https://phenix-online.org/documentation/reference/refinement.html
APBS	(Baker et al., 2001)	https://www.cgl.ucsf.edu/chimera/docs/ContributedSoftware/apbs/apbs.html
PDB2PQR	(Dolinsky et al., 2004)	https://www.cgl.ucsf.edu/chimera/docs/ContributedSoftware/apbs/pdb2pqr.html
SWISS-MODEL	(Waterhouse et al., 2018)	https://swissmodel.expasy.org
RING 2.0 webserver	(Piovesan et al., 2016)	http://old.protein.bio.unipd.it/ring/
Pyem	(Asarnow et al., 2019)	https://github.com/asarnow/pyem
ImageJ	(Schneider et al., 2012)	https://imagej.nih.gov/ij/
Python	Python Software Foundation	https://www.python.org
Jupyter Notebook	Project Jupyter	https://jupyter.org
Microsoft Excel	Microsoft	https://www.microsoft.com/en-us/microsoft-365/excel
Other
Xenopus laevis MS database	(Wühr et al., 2014)	https://scholar.princeton.edu/wuehr/protein-concentrations-xenopus-egg
Centrifuge rotor	Beckman Coulter	SX241.5
Ultra centrifuge rotor	Beckman Coulter	SW55Ti
Ultra centrifuge rotor	Beckman Coulter	SW40Ti
Centrifuge rotor	Beckman Coulter	JS 5.3
Centrifuge	Beckman Coulter	Allegron X-30R
Ultra centrifuge	Beckman Coulter	Optima L80
Centrifuge	Beckman Coulter	Avanti J-26S
Imaging System	Li-cor	Odyssey Infrared Imaging System
LC-MS/MS system	Thermo Scientific	Dionex3000 HPLC, NCS3500RS nano- and microflow pump, Q-Exactive HF mass spectrometer
Plunge freezer	FEI	Vitrobot Mark IV
Plasma Cleaner	Gatan	Solarus II
Cryo-EM	FEI	Talos Arctica TEM
Cryo-EM	FEI	Titan Krios TEM
Cryo-EM detector	GATAN	K2 Camera