Nuclear processes depend on the organization of chromatin, whose basic units are cylinder-shaped complexes called nucleosomes. A subset of mammalian nucleosomes in situ (inside cells) resembles the canonical structure determined in vitro 25 years ago. Nucleosome structure in situ is otherwise poorly understood. Using cryo-ET and 3-D classification analysis of budding yeast cells, here we find that canonical nucleosomes account for less than 10% of total nucleosomes expected in situ. In a strain in which H2A-GFP is the sole source of histone H2A, class averages that resemble canonical nucleosomes both with and without GFP densities are found ex vivo (in nuclear lysates), but not in situ. These data suggest that the budding yeast intranuclear environment favors multiple non-canonical nucleosome conformations. Using the structural observations here and the results of previous genomics and biochemical studies, we propose a model in which the average budding yeast nucleosome’s DNA is partially detached in situ.
This important paper exploits new cryo-EM tomography tools to examine the state of chromatin in situ. The experimental work is meticulously performed, with a vast amount of data collected, with convincing interpretation of possible nucleosomal structures in yeast. The majority of these nucleosomes appear not to possess stable octameric configurations. Although it cannot be excluded that there could be an unknown bias in class detection, we agree that the authors have gone as far as is possible, given constraints in current technology and analysis, to support their model. Despite the provocative nature of this report, it is our hope that such work will spark debate, and further the development of exciting new tools that can interrogate the dynamic nature of the nucleosome, and its associated functions in situ.
Eukaryotic chromosomes are polymers of DNA-protein complexes called nucleosomes. An octamer of proteins, consisting of a heterotetramer of histones H3 and H4 and two heterodimers of histones H2A and H2B, resides at the nucleosome’s center (Luger et al., 1997). Canonical nucleosomes resemble 10 nm wide, 6 nm thick cylinders and have 145 – 147 base pairs of DNA bent in 1.65 left-handed superhelical gyres around the histone octamer (Zhou et al., 2019; Zlatanova et al., 2009). In contrast, non-canonical nucleosomes have either partially detached DNA, partially detached histones, fewer than 8 histones, or a combination of these features (Zlatanova et al., 2009). Both X-ray crystallography and single-particle cryo-EM have shown that reconstituted nucleosomes, either alone or within a complex, are largely canonical in vitro (Zhou et al., 2019). Nucleosome structures in situ inside cells remain mysterious.
In the context of chromatin, nucleosomes are not discrete particles because sequential nucleosomes are connected by short stretches of linker DNA. Variation in linker DNA structure is a source of chromatin conformational heterogeneity (Collepardo-Guevara and Schlick, 2014). Recent cryo-EM studies show that nucleosomes can deviate from the canonical form in vitro, primarily in the structure of DNA near the entry/exit site (Bilokapic et al., 2018; Fukushima et al., 2022; Sato et al., 2021; Zhou et al., 2021). In addition to DNA structural variability, nucleosomes in vitro have small changes in histone conformations (Bilokapic et al., 2018). Larger-scale variations of DNA and histone structure are not compatible with high-resolution analysis and may have been missed in single-particle cryo-EM studies.
Molecular-resolution (2 – 4 nm) studies of unique objects like cells may be obtained by cryo-electron tomography (cryo-ET), a form of cryo-electron microscopy (cryo-EM) that generates 3-D reconstructions called cryotomograms. These studies reveal life-like snapshots of macromolecular complexes because the samples are prepared and then imaged in an unfixed, unstained, frozen-hydrated state. Most eukaryotic cells are too thick for cryo-ET, so thinner frozen-hydrated samples are made by cutting by cryomicrotomy or thinning by cryo-focused ion beam (cryo-FIB) milling (Ng and Gan, 2020; Strunk et al., 2012). These two approaches respectively produce cryosections and plank-like samples called cryolamellae. Subvolumes called subtomograms contain independent copies of the cells’ macromolecular complexes. These subtomograms can be further studied by averaging, which increases the signal-to-noise ratio, and classification, which facilitates the analysis of heterogeneity. Large macromolecular complexes such as ribosomes and proteasomes have been identified in situ by 3-D classification followed by comparison of the class averages to known structures – an approach called purification in silico (Beck and Baumeister, 2016). Using the purification in silico approach, we previously showed that canonical nucleosomes exist in cryotomograms of yeast cell lysates ex vivo and in a HeLa cell cryolamella (Cai et al., 2018a; Cai et al., 2018b; Cai et al., 2018c). Herein, we use the term ex vivo to describe nucleosomes from lysates instead of the term in vitro, which is more commonly used to describe either reconstituted or purified mononucleosomes. However, our 3-D structural analysis did not generate canonical nucleosome structures from cryosectioned fission yeast cells (Cai et al., 2018b). The discrepancy between nucleosome class averages ex vivo and in situ could have either technical or biological origins. As a biological explanation for the absence of canonical nucleosome class averages in situ, we hypothesized that yeast nucleosomes are either conformationally or constitutionally heterogeneous.
In this work, we test this heterogenous-nucleosome hypothesis by using cryo-ET to image both wild-type cells and strains that have nucleosomes bearing GFP as a density tag. We use the budding yeast S. cerevisiae, herein called yeast, because it has only two copies of each histone gene and because it is more amenable to gene editing. Our work compares the chromatin in both lysates and thin cellular cryo-EM samples prepared primarily by cryo-FIB milling. To obtain more information about nucleosomes in situ, we create one strain in which H2A-GFP is the sole source of H2A, meaning that the nucleosomes are expected to project one or two extra densities from their surface. Canonical nucleosomes are abundant in nuclear lysates and, in the H2A-GFP-expressing strain’s nuclear lysates, have extra densities consistent with GFP. In contrast, canonical nucleosomes account for less than 10% of the expected number of total nucleosomes in wild-type cell cryolamellae. Furthermore, neither canonical nucleosomes nor nucleosome-like particles with extra protruding densities were detected in the H2A-GFP-expressing strain. These findings suggest that the yeast intracellular environment disfavors the canonical nucleosome conformation.
Canonical nucleosomes are abundant in wild-type yeast lysates
The crystal structure of the reconstituted yeast nucleosome (White et al., 2001) shows a canonical structure that is largely indistinguishable from the first published one (Luger et al., 1997). To describe the various views of the nucleosome, herein we use the compact nomenclature introduced by (Zhou et al., 2019): the disc view is along the superhelical axis, and the gyre view is along the pseudo-dyad axis (Figure 1A). The side view, which was not defined by Zhou et al, is orthogonal to both the disc and gyre views.
The original RELION subtomogram-analysis workflow (Bharat and Scheres, 2016) involves 2-D classification, followed by 3-D classification (Figure S1A), whereas in the alternative approach, 2-D classification is bypassed and subtomograms are subjected directly to 3-D classification (Figure S1B); for brevity, we term this alternative method “direct 3-D classification”. Direct 3-D classification is limited by computer hardware (Kimanius et al., 2016), but can detect more canonical nucleosomes (Cai et al., 2018a); see the Methods for more details. In this study, the original workflow is used on a subset of samples to show example 2-D class averages for comparison with other studies. However, the conclusions in this paper are drawn from direct 3-D classification.
Our previous subtomogram analysis of nuclear lysates (Cai et al., 2018c) revealed that nucleosomes from the wild-type strain YEF473A (Bi and Pringle, 1996) adopt the canonical structure ex vivo. We repeated this experiment on the strain BY4741 (Brachmann et al., 1998), which serves as the wild-type and parent strain for the histone-GFP tagging mutants described later. In this experiment, yeast nuclei are isolated, lysed, then deposited on an EM grid. Cryotomograms of BY4741 nuclear lysates reveal the crowded nucleosome-like particles and other cellular debris (Figure S2). Next, we template matched for nucleosome-like particles using a nucleosome-sized featureless cylinder as a reference. Direct 3-D classification (Figure S3, A and B) followed by 3-D refinement produced an 18 Å-resolution subtomogram average of a BY4741 canonical nucleosome class (Figure S3C). The subtomogram average of wild-type yeast nucleosomes in nuclear lysates has longer linker DNA (Figure 1B) than the crystal structure, which was reconstituted using a 146-bp DNA fragment. Because the nucleosome-repeat length of budding yeast chromatin is ∼168 bp (Brogaard et al., 2012), this extra length of DNA may come from an ordered portion of the ∼22 bp linker between adjacent nucleosomes.
Canonical nucleosomes are rare in wild-type cells in situ
Cryo-FIB milling is a compression-free method to thin frozen-hydrated cells (Hayles et al., 2007; Mahamid et al., 2015; Marko et al., 2006; Medeiros et al., 2018; Rigort et al., 2010; Villa et al., 2013). This technique uses a beam of gallium ions to thin a cell under cryogenic conditions, producing a frozen-hydrated plank-like cellular sample called a cryolamella. Canonical nucleosomes are detectable in a HeLa cell cryolamella (Cai et al., 2018a), meaning that cryo-FIB milling does not grossly perturb canonical nucleosomes in situ. We prepared cryolamellae of wild-type BY4741 cells and then collected defocus-phase-contrast tilt series. Cryotomograms of yeast cryolamellae showed that nuclei were packed with nucleosome-like particles (Figure S4). Two-dimensional class averages of template-matched nucleosome-like particles reveal densities that have the approximate size and shape of nucleosomes (Figure S5A). We then subjected the particles belonging to the most nucleosome-like 2-D class averages to 3-D classification, following the original RELION classification workflow. However, none of these 3-D class averages resemble canonical nucleosomes (Figure S5B). While many of the nucleosome-like class averages have dimensions similar to the canonical nucleosome, none of them have densities that resemble the distinctive 1.65 left-handed gyres of DNA. To rule out the possibility that canonical nucleosomes were missed during 2-D classification, we performed direct 3-D classification using 100 classes. None of the resultant 100 class averages resemble a canonical nucleosome (Figure S6).
To determine if our template-matching and classification workflow missed the canonical nucleosomes, we re-did both template matching and 3-D classification of BY4741 wild-type yeast cryolamellae with an intentionally biased reference. Instead of a featureless cylinder, we used the yeast nucleosome crystal structure (White et al., 2001) as the reference (Figure S7A). If canonical nucleosomes are abundant in yeast, they should be detected as a class average that resembles a low-resolution nucleosome crystal structure. No canonical nucleosome class averages were seen in this control experiment (Figure S7B and C).
Our previous study of a HeLa cell (Cai et al., 2018a), which detected canonical nucleosomes, used a Volta phase plate (VPP). VPP data has more low-resolution contrast than defocus phase-contrast data. To test if canonical nucleosomes in yeast cryolamellae are detectable in VPP data, we recorded VPP tilt series and reconstructed tomograms of BY4741 cell cryolamellae (Figure S8). Subtomogram analysis of the VPP tomograms by 2-D classification (Figure S9A), followed by 3-D classification did not reveal a canonical nucleosome class average in BY4741 (Figure S9B). When we performed direct 3-D classification using 100 classes, we detected one class average that resembles a canonical nucleosome (Figure S10A, Movie S1). A second round of classification revealed two types of class averages, one of which resembles a slightly elongated nucleosome (Figure S10B, Movie S2). Refinement of these two types of nucleosomes produced 24 Å-resolution averages that differ in the amount of linker DNA visible (Figures 1C and S10C). The class average that has more ordered linker DNA vaguely resembles the chromatosome, a form of the nucleosome that has linker DNA crossing at the entry-exit site and in contact with a linker histone (Bednar et al., 2017; Zhou et al., 2015). However, the DNA does not appear to cross at the DNA entry-exit site, and, at the present resolution, it is not possible to determine if the linker histone is present. Unlike plunge-frozen complexes, which interact with the air-water interface and have biased orientations (Noble et al., 2018), complexes in situ are not subject to such biases. Visualization of the angular distribution of the canonical nucleosomes shows that the disc views are undersampled and likely missed by the classification analysis (Figure S10D), which we also observed in our analysis of HeLa nucleosomes in situ (Cai et al., 2018a). This missing hemisphere of views results in roughly half of the canonical nucleosomes going undetected. In summary, the use of VPP and relatively thin (≤ 160 nm) cryolamellae made it possible to detect two canonical nucleosome classes that differ in the amount of ordered linker DNA, similar to what we observed in a HeLa cell (Cai et al., 2018a).
Following our previous work (Cai et al., 2018a), we performed a negative control by analyzing a tomogram of a region in the cytoplasm (Figure S11), which does not have any nucleosomes. We performed template matching using the same reference as for our analysis of nuclei and then direct 3-D classification into 100 classes. None of the resultant class averages resemble a canonical nucleosome (Figure S12), confirming that our analysis was not biased by the cylindrical reference.
Our classification detected only 769 canonical nucleosomes. If we account for the undersampling of disc views (Figure S10D), we estimate there are ∼1,500 canonical nucleosomes detected in the 5 tomograms. In comparison, we estimate that in the single HeLa cell cryolamella cryotomogram (Cai et al., 2018a), there were more than 2,000 nucleosomes in a nucleus volume ∼1/6th of the total analyzed here in BY4741. The percentage of HeLa nucleosomes that are non-canonical in situ is unknown and will require further study. To visualize the distribution of canonical nucleosomes, we remapped the two class averages back into their positions in the original tomogram that contains the largest number of canonical nucleosomes (Figure 2). The remapped model shows that the canonical nucleosomes are scattered throughout the sampled nuclear volume. There are no large clusters of canonical nucleosomes like what we saw near the nuclear envelope of a HeLa cell. Using experimentally determined values for nucleosome number and chromatin volume (Oberbeckmann et al., 2019; Uchida et al., 2011), the tomograms we analyzed are expected to hold 25,000 nucleosomes. Therefore, the nucleosomes (both canonical and non-canonical) should pack with an order-of-magnitude higher density than visualized in our remapped model (Figure 2B). In summary, our calculations suggest that the vast majority (> 90%) of the nucleosomes in BY4741 yeast are non-canonical.
Histone GFP-tagging and visualization ex vivo
Our previous cryo-ET analysis of nucleosomes in a HeLa cell revealed that subtomogram 3-D classification is sensitive to features much smaller than the nucleosomes, as evidenced by the separation of canonical-nucleosome class averages that differ by ∼10 bp of linker DNA near the dyad (Cai et al., 2018a). Furthermore, studies of flagella (Oda and Kikkawa, 2013) and pilus machines (Chang et al., 2016) showed that subtomogram averages of complexes in situ can reveal either the presence or absence of protein densities as small as fluorescent proteins. These observations led us to attempt to use a GFP tag to facilitate nucleosome identification in situ. Our strategy is to compare subtomogram averages of nucleosome-like particles in strains that express only wild-type histones versus those that express GFP-tagged histones. Note that this tagging strategy did not work as intended because we could not detect tagged nucleosomes 3-D classes in situ. However, this negative result provided an important clue about the nature of nucleosomes inside yeast cells (see below).
Histones can accept a genetically encoded GFP tag at either the N- or C-terminus. An N-terminal GFP tag is not expected to be visible in subtomogram averages because it would be separated from the histone’s globular domain by the long, flexible N-terminal “tail”. Therefore, we fused GFP to the histone C-terminus, which does not have a flexible tail (Figure S13, A and B) and we further confined the GFP by eliminating the peptide linker that is included in popular GFP-tagging modules (Figure S13C). S. cerevisiae has two copies of each histone gene (Figure S13D), which are arranged as gene pairs. The H2A and H2B genes are arranged as gene pairs HTA1-HTB1 and HTA2-HTB2 (Hereford et al., 1979). To maximize our chances of detecting nucleosome class averages that have an extra density, we first sought to create strains in which a histone-GFP fusion is the sole source of one class of histones (Figure S13, E and F). We deleted the entire HTA2-HTB2 gene pair to prevent its amplification as circular DNA molecules by the flanking retrotransposon elements (Libuda and Winston, 2006); the resulting strain is called LGY0012. Next, we inserted the GFP gene at the 3’ end of HTA1, without a linker, to generate LGY0016, making H2A-GFP the sole source of H2A. We confirmed LGY0016’s genotype by PCR analysis (Figure S14, A and B), Sanger sequencing, and immunoblots using anti-H2A or anti-GFP antibodies (Figure S14C). Accordingly, the LGY0016 nuclei showed bright fluorescence (Figure S14D). We also constructed the strain LGY0015, which expresses both a H2B-GFP fusion without a linker peptide (Figure S14, E – H) and an untagged copy of H2B. We were unable to create strains that have either H2B-GFP, H3-GFP, or H4-GFP as the sole H2B, H3, and H4 sources, respectively (see H3- and H4-tagging experiments below). Consistent with this low tolerance for a GFP-tagged histone as the sole source of a histone type, the LGY0016 doubling time is ∼50% longer than for wild-type BY4741 (130 minutes versus 85 minutes) in rich media.
We next performed cryo-ET of the nuclear lysates of LGY0016 and LGY0015 cells (Figures S15 and S16). The 2-D class averages of LGY0016 and LGY0015 nuclear lysates resemble those seen in single-particle cryo-EM studies of reconstituted nucleosomes (Chua et al., 2016), though with lower-resolution features (Figure 3A, S17A). Note that 2-D classification uses a circular mask, meaning that it will not bias the shape of the resultant class averages to resemble, for example, the double-lined motifs seen in nucleosome side and gyre views. To increase the number of detected nucleosomes, we used direct 3-D classification into 40 classes. We obtained canonical nucleosome 3-D class averages this way in the lysates of both strains (Figure S18 and S19; Movies S3 and S4). These class averages have the unmistakable structural motifs of canonical nucleosomes, such as a 10-nm diameter, 6-nm thick cylindrical shape, and the left-handed path of the DNA densities. There is more DNA than the crystal structure’s 1.65 gyres because lysate chromatin samples have linker DNA. All these properties are consistent with the subtomogram analysis of nucleosomes from nuclear lysates of wild-type strains BY4741 (Figure 1B) and YEF473A (Cai et al., 2018c).
Subsequent classification rounds revealed nucleosome class averages that have an extra density projecting from one or both faces (Figures S18B and S19B). One example of each class (zero, one, or two extra densities) was refined to ∼25 Å resolution (Figure 3B, S17B). The position of the extra density is consistent with the H2A C-terminus being closer to the DNA entry-exit point (Figure 3C) and the H2B C-terminus being far from the DNA entry-exit point (Figure S17C). In principle, LGY0015 cells can assemble nucleosomes that have two copies of H2B-GFP. The absence of LGY0015 nucleosome classes with two densities suggest that they are either unstable or too rare to detect by 3-D classification. We focused our in situ cryo-ET analysis on LGY0016 cells because the GFP tags are easier to recognize on nucleosomes from nuclear lysates of this strain.
Canonical nucleosome classes are not detected in LGY0016 cells in situ
In an attempt to detect more yeast nucleosomes in situ, we did three types of imaging experiments on LGY0016 cells. We performed cryo-ET of cell cryolamellae with and without the VPP (Figures S20 and S21) and we also imaged cryosections with the VPP (Figure S22). The cryolamellae benefit from the absence of compression artifacts while the cryosections benefit from being thinner on average. We performed template matching and then subjected the hits directly to 3-D classification, which was needed to detect canonical nucleosomes in BY4741 above. The 3-D class averages from all three samples resembled neither canonical nucleosomes nor cylindrical bodies with one or more protruding densities that is expected from the lysate samples (Figures 4, S23, S24, Movie S5). Some class averages have two linear motifs that resemble the double DNA gyres opposite the DNA entry-exit point, but these DNA-like densities do not go 1.65 times around the center of mass as expected of canonical nucleosomes (Luger et al., 1997). Small variations in the classification parameters such as mask size, class number did not reveal any canonical nucleosome classes in LGY0016 cell cryotomograms. We therefore conclude that canonical nucleosomes are rare inside budding yeast nuclei and that non-canonical nucleosomes are the vast majority. At present, we do not know what the non-canonical nucleosome structures are, meaning that we cannot even determine if one non-canonical structure is the majority. Until we know the non-canonical nucleosomes’ structures, we will use the term non-canonical to describe all the nucleosomes that do not have the canonical (crystal) structure.
Ribosome control for sample, data, and analysis pathologies
We may have missed canonical nucleosomes if there were pathologies with either our cryolamellae, data, or workflow, which would result in grossly misclassified and misaligned particles. To test this hypothesis, we performed template matching, 3-D classification and alignment on cytoplasmic ribosomes. Because ribosomes are so large (> 3 megadaltons) and have been studied extensively in situ, any of the pathologies listed above would result in either the absence of ribosome class averages or extremely low resolution. Using our control tomogram of the cytoplasm, which is densely packed with ribosomes, we obtained a subtomogram average at between ∼33 Å resolution based on the Fourier shell correlation (FSC) = 0.5 cutoff criterion (or 28 Å using the FSC = 0.143 criterion), from 1,150 particles (Figure S25). As a comparison, we simulated densities with the yeast ribosome crystal structure between 15 Å and 30 Å resolution and found that our average has density features consistent with this resolution range. The resolutions of our nucleosome and ribosome averages (∼24 Å and 33 Å) are comparable to recent in situ subtomogram averages using similar numbers of particles (∼500 to 1,500) (Laughlin et al., 2022; van den Hoek et al., 2022). Therefore, our cryolamellae, data, and workflow do not show evidence of pathologies.
GFP tagging of H3 and H4 also does not reveal nucleosomes
In cryolamellae of LGY0016 cells, the absence of nucleosome-like class averages that have an extra density bump suggested that previously unappreciated properties of the H2A-H2B heterodimer make this histone pair a poor candidate for the GFP fusion tag strategy (see Discussion). We therefore revisited GFP tagging of H3 and H4, which are found in all known and speculated forms of nucleosomes and non-canonical nucleosomes (Zlatanova et al., 2009). Because all attempts to make H3- or H4-GFP “sole source” strains failed, we tested strains that had one wild-type copy and one GFP-tagged copy of one of these histones. We tagged members of the HHF1-HHT1 gene pair, which encode histone H4 (HHF1) and H3 (HHT1). HHF1-HHT1 is flanked by the same transposon elements as HTA2-HTB2, so gene amplification here would not result in the wild-type copy outnumbering the tagged copy. The genotypes and phenotypes were verified by confirmation PCRs, western blots, Sanger sequencing (not shown), and fluorescence confocal microscopy (Figures S26 and S27). We created H3-GFP strains without a linker (LGY0007) and with linker sequences RIPGLIN (LGY0002) and GGSGGS (LGY0070); this latter linker was introduced previously (Verzijlbergen et al., 2010). For H4, we could only obtain a strain with a H4-GFP-expressing strain by using the flexible GGSGGS linker (LGY0071; Figure S27). Flexible linkers are not expected to facilitate tag-based identification because the tag can occupy a much larger volume and get blurred out as a result but was the only one tolerated in our H4 tagging attempts.
We next prepared nuclear lysates of each of these strains, performed cryo-ET, and subjected them to direct 3-D classification analysis (Figures S28 – S31). Like the other strains, the lysates of each of these four strains had large numbers of canonical nucleosomes (Figures S28A, S29A, S30A, S31A). Of the canonical nucleosomes, we were only able to detect class averages that have the extra density in the LGY0007 nuclear lysates (Figure S28); the nuclear lysates of the other strains do not have nucleosome class averages with a GFP density. Finally, to test if nucleosomes, canonical or non-canonical, with the GFP density bump, could be detected in situ, we performed cryo-ET of LGY0007 cell cryolamellae with the VPP (Figure S32). We then performed template matching and subjected the hits to direct 3-D classification analysis. Compared to the BY4741 (wild type) cell cryolamella VPP class averages (Figure S10A), the 3-D class averages in LGY0007 cell cryolamellae resemble neither canonical nucleosomes nor cylindrical bodies with a protruding density (Figure S33). Therefore, H3- and H4-GFP fusions cannot be used to detect non-canonical nucleosomes and much more work needs to be done to identify non-canonical nucleosomes.
Cryo-EM of cells has made the structural analysis of chromatin organization in situ feasible. The earliest in situ studies were done with projection cryo-EM images and 2-D Fourier analysis of cryosectioned cells, which revealed that long-range order is absent in chromatin in situ (Eltsov et al., 2008; McDowall et al., 1986). Cryo-ET studies later showed that short-range order is present in isolated chicken erythrocyte nuclei (Scheffer et al., 2011), but absent in picoplankton and budding yeast (Chen et al., 2016; Gan et al., 2013). These early cryo-ET studies revealed little about nucleosome structure in situ because, as Eltsov et al observed, the nucleosomes in those samples appeared like smooth ellipsoids (Eltsov et al., 2018). Recent cryo-ET studies using state-of-the-art electron-counting cameras and energy filters have revealed that in cryotomographic slices of cell nuclei, there are densities that resemble nucleosome side and gyre views in which gyre-like features are resolved (Cai et al., 2018a; Cai et al., 2018b; Eltsov et al., 2018). This higher-fidelity data made it possible to use 3-D classification to detect canonical nucleosomes in nuclear envelope-associated chromatin in a HeLa cell (Cai et al., 2018a).
Our use of 3-D classification detected canonical nucleosomes in wild-type yeast nuclear lysates and canonical nucleosomes both with and without GFP tags in nuclear lysates of some strains that bear histone-GFP fusions. We also detected a canonical nucleosome class average in wild-type cell cryolamellae imaged with a VPP. The in situ detections suggest that there are only ∼1,500 canonical nucleosomes (taking account of the undersampling of nucleosomes in the disc view) out of the 25,000 nucleosomes expected of the total sampled nuclear volume. Note that the percentage of canonical nucleosomes in lysates cannot be accurately estimated because we cannot determine how many nucleosomes in total are in each field of view. The estimates from the cryolamellae are more reliable because the expected numbers of nucleosomes (canonical or not) can be estimated from genomics, biochemical, and X-ray tomography data (see Methods). When we analyzed LGY0016 cell cryolamellae, in which H2A-GFP is the sole source of H2A, we did not detect canonical nucleosome classes or any cylindrical nucleosome-sized structures that have extra density bumps. Likewise, analysis of cryolamellae of LGY0007 did not reveal either a canonical nucleosome-like class average or any cylindrical nucleosome-sized density with an extra density bump, even though H3-GFP should be incorporated into the H3-H4 tetramer, the central component of the nucleosome.
The absence of either a canonical nucleosome-like class average that has an extra density bump, or a cylinder with an extra density bump in the LGY0016 strain suggests that the H2A-H2B heterodimer is mobile in situ. By “mobility”, we are not implying that H2A-H2B is dissociated. We mean that H2A-H2B is attached to the rest of the nucleosome and can have small differences in orientation. The H2A/H2B heterodimers are in contact with the H3/H4 heterotetramer because crosslinking and single-particle fluorescence imaging experiments showed that approximately 80% of H2B – and presumably H2A to which it stably dimerizes – is bound to chromatin in situ (Mohan et al., 2018; Ranjan et al., 2020). H2A-H2B mobility was hinted at by the observations from X-ray crystallography that yeast nucleosomes lack the hydrogen bonds that stabilize the two H2A-H2B heterodimers in metazoan nucleosomes (White et al., 2001) and that interfaces between H2A-H2B and H3 are moderately exposed to small-molecule probes (Marr et al., 2021). Furthermore, recent simulations show that the H2A/H2B heterodimer may adopt numerous orientations and small displacements, with either fully or partially unwrapped DNA, while remaining in contact with H3-H4 (Ishida and Kono, 2022). We also cannot rule out the possibility that expression of H2A-GFP makes nucleosomes less ordered in situ.
Like LGY0016 (H2A-GFP sole source), none of the LGY0007 (H3 plus H3-GFP) in situ nucleosome-like class averages had an extra density consistent with the GFP density tag (only seen in lysates). Furthermore, there were no canonical nucleosome class averages – either with or without the GFP density – seen among the class averages of LGY0007 cell cryolamella nucleosome-like particles. These observations suggest that the expression of H3-GFP makes nucleosomes less ordered in situ. Altogether, our experiments show that in budding yeast, canonical nucleosomes are rare in situ and that the expression of GFP-tagged histones leads to further perturbations of nucleosome structure in situ.
Our data is consistent with a model in which yeast canonical nucleosomes are abundant ex vivo (Figure 5A). They adopt multiple non-canonical conformations in situ (Figure 5, B and C) and rarely adopt the canonical one (Figure 5D). Note that the blurring in these panels implies the heterogeneity in the positions of the DNA and proteins, not their actual motion. In addition to H2A-H2B and DNA conformational heterogeneity, diverse nucleosome-associated proteins that bind to multiple positions (Figure 5C) would further increase the heterogeneity. A high abundance of non-canonical nucleosomes means that more of the genome would be accessible, consistent with yeast having high levels of transcription. Nucleosome heterogeneity is also consistent with the absence of chromatin long-range order in situ because crystalline oligonucleosome arrays can only form if sequential nucleosomes adopt nearly identical conformations (Ekundayo et al., 2017; Robinson et al., 2006; Song et al., 2014). Further investigation is needed to identify the biochemical and biophysical factors responsible for the abundance of non-canonical nucleosomes in yeast and to determine their diverse structures (Zlatanova et al., 2009).
In combination with our previous work on a HeLa cell (Cai et al., 2018a), we show that globular complexes as small as canonical nucleosomes (∼200 kilodaltons) can be detected by 3-D classification of subtomograms from cryolamellae that were imaged with a VPP. Given the low throughput of cryo-FIB milling, the low yield of cryolamellae thinner than 160 nm, and the technical difficulties of VPP imaging, it is not yet feasible to systematically determine the conditions needed to detect sub-200 kDa complexes by purification in silico. More depositions of eukaryote cryolamellae cryo-ET datasets in the public database EMPIAR (Iudin et al., 2016) may enable a thorough search of data-collection and processing-parameter space.
Our in situ cryo-ET-based model adds to the body of work on non-canonical nucleosomes. ChIP-seq analysis has detected sub-nucleosomes in situ, though the abundance was unknown (Rhee et al., 2014). MNase-seq has also detected nucleosomes in a partially unwrapped and partially disassembled state in situ (Ramachandran et al., 2017). A recent Hi-C variant “Hi-CO” presented evidence that < 147 bp of yeast nucleosomal DNA is protected from MNase attack (Ohno et al., 2019), which is consistent with the partial detachment of core DNA. Molecular dynamics advances in all-atom and coarse-grained simulations are now showing that nucleosomes are far more dynamic than previously appreciated (Armeev et al., 2021; Brandani et al., 2021; Farr et al., 2021; Huertas et al., 2021; Ishida and Kono, 2021). The DNA-unwrapping associated with nucleosome breathing was shown to disfavor ordered helical oligonucleosome structures (Farr et al., 2021). Nucleosome breathing is also evident in high-throughput atomic-force microscopy experiments, which revealed that only ∼30% of nucleosomes are fully wrapped, and that approximately half of the nucleosomes have an opening angle (measured between the entry/exit DNA arms) 60° larger than the fully wrapped one (Konrad et al., 2021). Nucleosomes that have non-canonical nucleosome properties, such as lower stability or exposure of internal surfaces, have been reported in fission yeast (Koyama et al., 2017; Sanulli et al., 2019), which may explain why we did not observe canonical nucleosomes in cryosections in those cells either (Cai et al., 2018b). Some nucleosomes in human and fly cell cryosections appear “gaping”, i.e., with the inter-DNA-gyre distance slightly larger than ∼2.7 nm (Eltsov et al., 2018). Partial DNA detachment has been seen in complexes between nucleosomes and remodelers and transcription factors (Eustermann et al., 2018; Farnung et al., 2017; Liu et al., 2017; Sundaramoorthy et al., 2018; Sundaramoorthy et al., 2017; Willhoft et al., 2018), methyltransferases (Bilokapic and Halic, 2019; Jang et al., 2019), or transcription-related complexes (Dodonova et al., 2020; Kujirai et al., 2018). Larger amounts (up to ∼25 bp) of DNA detachment have been seen very rarely in cryo-EM structures (Bilokapic et al., 2018; Zhou et al., 2021). A recent study of cryosectioned fly embryos has also presented evidence of nucleosome-like structures such as hemisomes and three-gyre structures (Fatmaoui et al., 2022). The prevalence and functional consequences of non-canonical nucleosomes in situ remain to be studied in other organisms.
We now consider the alternative hypothesis that canonical nucleosomes are the dominant form in situ and that we have missed them as a result of our image processing. Many of the 3-D class averages appear to have multiple gyre-like densities, which may arise from nucleosome stacking. However, classes with multiple gyre-like densities are also found in the class averages from the cytoplasm, which does not have nucleosomes (Figure S12). These are “junk” classes that result from the averaging of different particle species into the same class. Importantly, stacked canonical nucleosomes are so conspicuous that they could not have been missed in tomographic slices. For instance, two stacked canonical nucleosomes have twice the mass of a single canonical nucleosome and would have dimensions 12 nm by 10 nm (Figure S34). Multiple stacked canonical nucleosomes would appear like 10-nm-thick filaments.
Another alternative hypothesis is that the lack of disc views, corresponding to nucleosomes whose “face” is parallel to the cryolamella surface, resulted in canonical nucleosomes going undetected. This hypothesis would be supported only if the orientations of canonical nucleosomes were biased. However, biased orientation in cryo-EM experiments is usually a result of interactions with the air-water interface (Noble et al., 2018), which does not exist for nucleosomes inside the nucleus of a cell. Another hypothesis is that the missing views resulted in elongated class averages because of the missing wedge or missing cone in Fourier space. This artifact would only affect the distribution of canonical versus non-canonical nucleosomes if canonical nucleosomes have a preferred orientation, which we have just argued against due to the nature of our cellular samples. The class averages from non-disc-view nucleosomes do not have a missing-wedge/cone artifact because they include nucleosomes whose superhelical axes are in the X-Y plane; when averaged together, the Fourier transforms of these nucleosomes “fill in” Fourier space because they sample numerous rotations about their super-helical axes. Missing-wedge-free reconstructions are exemplified by plunge-frozen actin filaments, which lie parallel to the EM grid. As illustrated in a recent study (Merino et al., 2018), reconstructions of filamentous actin are free of missing-wedge distortions even though the axial views are completely absent.
Some of the particles in the tomographic slices resemble donuts. Because nucleosome disc views also resemble donuts, it is possible that we completely missed these particles in either our template matching or classification analyses. As stated above, even if we missed all the disc views, our conclusions would not change because the orientations of canonical nucleosomes inside of cellular samples are not biased.
Furthermore, the donut-like particles cannot be canonical nucleosomes because they are too wide (> 12 nm). A subset of them were detected by template matching and classification and separated into their own class, which resembles a cylinder with a rounded cap (Movie S5, row 5 column 1).
Another hypothesis for the low numbers of detected canonical nucleosomes is that the nucleoplasm is too crowded, making the image processing infeasible. However, crowding is an unlikely technical limitation because we were able to detect canonical nucleosome class averages in our most-crowded nuclear lysates, which are so crowded that most nucleosomes are butted against others (Figures S15 and S16). Crowding may instead have biological contributions to the different subtomogram-analysis outcomes in cell nuclei and nuclear lysates. For example, the crowding from other nuclear constituents (proteins, RNAs, polysaccharides, etc.) may contribute to in situ nucleosome structure, but is lost during nucleus isolation.
Limitations of the study
Because naked linker DNA cannot yet be seen in situ in yeast, the connectivity between the nucleosomes (both canonical and non-canonical) could not be followed. The subtomogram analysis was unable to resolve the structures of the non-canonical nucleosomes. Because of the low resolution, we could not assess the variability in the structure and composition of the canonical nucleosomes. All of these limitations could potentially be addressed by increasing the data quality and the numbers of classes and subtomograms per class, which require advances at all stages of structural cell biology. These advances include improvements in cryo-FIB milling throughput and reproducibility (Buckley et al., 2020; Tacke et al., 2021; Zachs et al., 2020), cryo-EM cameras, laser phase plates (Turnbaugh et al., 2021), template-matching / segmentation software (Bepler et al., 2020; Moebel et al., 2021), and subtilt refinement software such as emClarity (Himes and Zhang, 2018), EMAN2 (Chen et al., 2019), Warp/M (Tegunov et al., 2021) and RELION 4 (Zivanov et al., 2022). Because structural cell biology is still a nascent field, the optimum combination of sample prep, imaging, and data analysis will require thorough exploration of the parameter space at each step. The nucleosome VPP subtomogram averages presented here were limited to ∼24 Å, even though the data was recorded close to focus as suggested by a higher-resolution study of purified ribosomes (Khoshouei et al., 2017). Limiting factors include lower particle numbers, alignment inaccuracy, VPP charging, and the lack of contrast transfer function (CTF) modeling. In single-particle cryo-EM analysis, high-resolution analysis (3 Å or better) requires accurate modeling and then compensation for the CTF. A key first step of CTF modeling is the visualization of Thon rings in Fourier power spectra. We plotted Fourier power spectra in both 2-D with CTFFIND (Rohou and Grigorieff, 2015) and as a 1-D rotational average with IMOD (Mastronarde, 1997), but could not see Thon rings (Figure S35), meaning that CTF compensation is not feasible. A study of VPP single-particle cryo-EM analysis recommends intentionally underfocussing to 500 nm to increase the number of Thon rings, making CTF modeling easier (Danev et al., 2017). However, the samples used in that study were much thinner (plunge-frozen proteasomes) and the dose used per image was much higher (∼40 electrons / Å2). Because the subtomogram analysis of smaller complexes inside cryolamellae faces multiple challenges, more investigation is needed to determine the optimum parameters for increased resolution.
Another challenge is the positive identification of the non-canonical nucleosome species. The approaches attempted here (the use of a GFP tag) failed. An alternative approach would be to remap the various averages and show that the linker DNA of sequential non-canonical nucleosomes point to each other, as we have previously done for HeLa chromatin in situ (Cai et al., 2018a). Unfortunately, this approach is not suitable for yeast because only the canonical class averages have linker-DNA densities – none of the other class averages have densities that resemble linker DNA. Because non-canonical yeast nucleosomes in situ are unknown structures, we would have to identify individual instances using a “GFP of cryo-EM”, in the form of a compact fusion protein-like structure that can be directly visualized in tomographic slices without classification and subtomogram averaging. Several attempts have been made in recent decades (Diestra et al., 2009; Mercogliano and DeRosier, 2007; Nishino et al., 2007; Wang et al., 2011). More work is needed to test the suitability of such tags as “GFPs of cryo-EM” in situ.
We thank Kerry Bloom, Vu Nguyen, and Carl Wu for advice on fluorescent-protein tagging of histones; Bill Rice and Ed Eng for help with cryo-EM data collection; John Heumann for discussions about PEET and for implementing a parallelized duplicate-removal routine; Rado Danev, Kliment Verba, and Shenping Wu for advice on the VPP. The Quadro P6000 used in this work was kindly donated by the NVIDIA Corporation. ZYT, SC, JKC, and LG were supported by a Singapore Ministry of Education Tier 1 grant R-154-000-A49-114 and Tier 2 grant MOE2019-T2-1-140. AJN was supported by a grant from the NIH National Institute of General Medical Sciences (NIGMS) (F32GM128303). Some of this work was performed at the Simons Electron Microscopy Center and National Resource for Automated Molecular Microscopy located at the New York Structural Biology Center, supported by grants from the Simons Foundation (SF349247), NYSTAR, and the NIH National Institute of General Medical Sciences (GM103310) with additional support from NIH S10 RR029300.
Declaration of interests
The authors declare no conflicts of interest.
Materials and methods
Yeast Strain and Growth Conditions
All yeast strains were streaked on yeast extract peptone dextrose (YPD) agar plates (2% w/v peptone, 1% w/v yeast extract, 2% w/v glucose for liquid media; additional 2% w/v agar for agar plates) at 30°C and cultured in YPD in conical flasks shaking at 200 – 250 RPM at 30°C. All yeast strains are of mating type a. Modifications or lack thereof to all histone genes were authenticated by PCR and Sanger sequencing (Bio Basic Asia Pacific Pte Ltd, Singapore) in all yeast strains.
Bacterial Growth Conditions
All plasmids were provided in DH5-Alpha Escherichia coli. The bacteria were streaked on LB agar plates with ampicillin (40 g/L LB Broth with agar (Miller), 100 μg/mL ampicillin) at 37°C and cultured in LB liquid medium with ampicillin (25 g/L LB Broth (Miller), 100 μg/mL ampicillin) in vent cap tubes shaking at 200 – 250 RPM at 37°C.
Plasmid extraction and linearization
The plasmid pFA6a-GFP(S65T)-His3MX6 (Longtine et al., 1998) was a gift from John Pringle (Addgene 41598; Addgene, Watertown, MA) and pFA6a-link-yoTagRFP-T-Kan (Lee et al., 2013) was a gift from Wendell Lim & Kurt Thorn (Addgene 44906); both were given in the form of bacterial stabs. Note that pFA6a-link-yoTagRFP-T-Kan is the source of the KanR marker used to delete HTB2-HTA2. Five mL of bacteria were cultured overnight, then plasmids were extracted with the QIAprep Spin Miniprep Kit (QIAGEN, Hilden, Germany) following the manufacturer’s instructions.
Extracted plasmids were linearized by double digestion with a reaction containing 1 µg of plasmid DNA, 5 µL of 10× rCutSmart™ Buffer (New England BioLabs, Ipswich, MA), 10 units each of SalI and EcoRV restriction enzymes (New England BioLabs) topped up to 50 µL with nuclease-free water. The reaction mixture was heated at 37°C for 15 minutes for the digestion reaction, then 80°C for 20 minutes to inactivate the enzymes.
The strain details are shown in Table S1. Primers were from IDT (Integrated DNA Technologies, Inc., Singapore) and listed in Table S2. Q5 PCR Master Mix (New England BioLabs) was used for PCR reactions. BY4741 (Brachmann et al., 1998) served as the wild-type strain. Tagging and deletion cassettes were amplified from linearized pFA6a plasmids (Lee et al., 2013; Longtine et al., 1998) with a PCR reaction containing 1 ng of template DNA and 0.5 μM of each primer, using a PCR program of 98°C for 30 seconds, 30 cycles of 98°C for 5 seconds, 60°C for 10 seconds and 72°C for 1.5 minutes, then 72°C for 5 minutes.
Cells were transformed using the lithium acetate / PEG4000 method reported in (Nishimura and Kanemaki, 2014). Overnight cell culture in YPD was diluted to 0.1 OD600 in 25 mL of YPD and grown to 0.3 OD600. Ten mL of cells were collected, centrifuged at 1,600 × g at 25°C for 3 minutes, and the supernatant was removed. The cells were then washed twice with 10 mL sterile water with centrifugation at 1,600 × g at 25°C for 3 minutes. The cells were resuspended in 1 mL sterile water, transferred to a new 1.5 mL collection tube, centrifuged at 17,900 × g at 25°C for 1 minute, washed in 1 mL of TE/LiAc (10 mM tris, 1 mM ethylenediaminetetraacetic acid (EDTA), 100 mM lithium acetate), with centrifugation at 17,900 × g for 1 minute, and resuspended in 50 μL of the same buffer. Fifty μL of cell suspension was transferred to a new 1.5 mL collection tube containing 5 μL of 10 mg/mL salmon sperm DNA (Sigma-Aldrich, Burlington, MA) plus 5 μL of PCR-amplified cassette DNA. 360 µL of TE/LiAc/PEG (10 mM tris, 1 mM EDTA, 100 mM lithium acetate, 40% w/v Polyethylene glycol 4000) was added and incubated with shaking at 25°C for 30 minutes. Forty µL of dimethyl sulfoxide was added, the suspension was incubated at 42°C for 15 minutes in a water bath, then cooled on ice for 2 minutes. The suspension was centrifuged at 13,000 x g for 1 minute, the liquid was removed, and the cells were resuspended in 300 μL of 1× TE buffer, pH 8.0 (10 mM tris, 1 mM EDTA). Two hundred μL of this cell suspension was plated on a selection plate and incubated for several days at 30°C. Histidine selection plates were created with 6.7 g/L yeast nitrogen base without amino acids (Sigma-Aldrich), 1.92 g/L yeast synthetic drop-out medium supplements without histidine (Sigma-Aldrich), 2% w/v glucose and 2% w/v agar. G418 selection plates were created by adding G418 to the molten agar (YPD for G418 single selection, histidine auxotrophy medium for double selection) to a concentration of 200 mg/L before it was poured.
Transformants were verified by PCR and Sanger sequencing. Genomic DNA was extracted with the DNeasy Blood & Tissue Kit (QIAGEN) following the manufacturer’s instructions. Confirmation PCR was performed with 1 µg of template genomic DNA and 0.5 μM of each primer, using a PCR program of 94°C for 2 minutes, 30 cycles of 94°C for 1 minute, 62°C for 1 minute and 72°C for 3.5 minutes, then 72°C for 5 minutes. PCR products were purified with the QIAquick PCR Purification Kit (QIAGEN) following the manufacturer’s instructions, with water used for the final elution before they were sent for Sanger sequencing.
DNA gels and immunoblots
PCR products were electrophoresed in 2% agarose in Tris-acetate-EDTA and visualized with FloroSafe DNA Stain (Axil Scientific Pte Ltd, Singapore). The electrophoresis was performed at 100 Volts for 60 – 80 minutes before visualization with a G:Box (Syngene).
To generate protein samples for immunoblot analysis, ∼20 OD600 units of yeast cells were pelleted and stored at −80°C for at least 1 hour. Cells were then resuspended in 200 µL of ice-cold 20% trichloroacetic acid (TCA) and vortexed with glass beads at 4°C for 1 minute 4 times. For each sample, 500 µL of ice-cold 5% TCA was added, mixed with the pellet, then transferred to a new 1.5 mL collection tube. Another 500 µL of ice-cold 5% TCA was mixed with each pellet and transferred to the same 1.5 mL collection tube as before, so that each collection tube had 1 mL total volume. The tubes were then left on ice for 10 minutes and centrifuged at 15,000 × g at 4°C for 20 minutes. The TCA was aspirated from the tube, then the pellet was resuspended in 212 µL of Laemmli sample buffer with 2-mercaptoethanol added (Bio-Rad, Hercules, CA). To neutralize the residual TCA, 26 µL of 1 M Tris, pH 8 was added. This mixture was heated at 95°C for 5 minutes, centrifuged at 25°C at 15,000 × g for 10 minutes, then 5 µL of the supernatant was subjected to SDS-PAGE.
The primary and secondary antibodies used for immunoblots are shown in Table S3. Proteins were electrophoresed in 10% Mini-PROTEAN® TGX™ Precast Protein Gels (Bio-Rad). Either Precision Plus Protein™ WesternC™ Standards (Bio-Rad) or Invitrogen™ MagicMark™ XP Western Protein Standard (Thermo Fisher Scientific, TFS, Waltham, MA) served as the ladder. The proteins were electrophoresed at 100 Volts for 1 hour at 25°C. The gels were transferred onto Immun-Blot® PVDF Membranes (Bio-Rad) at 4°C in transfer buffer (3.02 g/L tris, 14.4 g/L glycine, 20% methanol). The transfer was performed at 100 Volts for 30 minutes. The membranes were then blocked with 2% BSA in 1× Tris-Buffered Saline, 0.1% Tween® 20 Detergent (TBST) for 1 hour. This was followed by 50 µg/mL avidin in TBST for 30 minutes, then a wash with TBST for 20 minutes if Precision Plus Protein™ WesternC™ Standards ladder was used. All antibody dilution factors are reported in Table S3. The membranes were probed with primary antibodies at the stated dilutions in 2% BSA in TBST for 1 hour at 25°C. Membranes were washed with TBST for 20 minutes three times at 25°C. The membranes were then probed with secondary antibodies in 2% BSA in TBST for 1 hour, with additional Precision Protein™ StrepTactin-HRP Conjugate (Bio-Rad) at 1:10,000 dilution if Precision Plus Protein™ WesternC™ Standards ladder was used. Finally, the membranes were washed with TBST for 10 minutes 3 times, then treated with a 50:50 mixture of Clarity Western Peroxide Reagent and Clarity Western Luminol/Enhancer Reagent (Bio-Rad) for 5 minutes before visualization by chemiluminescence on an ImageQuant LAS 4000 (Cytiva, Marlborough, MA).
Cells were grown to log phase (OD600 = 0.1 – 1.0), of which 2 OD600 units of cells were collected, pelleted at 5,000 × g for 1 minute, then resuspended in 1 mL of YPD. Four µL of cell culture was then applied to a glass slide and pressed against a coverslip. The cells were imaged live at 23°C with an Olympus FV3000 Confocal Laser Scanning Microscope (Olympus, Tokyo, Japan) equipped with a 1.35 NA 60× oil-immersion objective lens. GFP fluorescence was acquired using the 488 nm laser-line, with a DIC image recorded in parallel with the fluorescence image. Images were captured as Z-stacks thick enough to sample the GFP signals through all the nuclei in each stage position. Additional details regarding data collection are shown in Table S4.
Preparation of nuclear lysates
Yeast nuclei were prepared with reagents from the Yeast nuclei isolation kit (Abcam 206997, Cambridge, United Kingdom), unless noted otherwise. Yeast cells (30 mL, OD600 ∼1) were pelleted at 3,000 × g for 5 minutes at 25°C. The pellet was washed twice with 1 mL water (3,000 × g, 1 minute). The pellet was then resuspended in 1 mL Buffer A (pre-warmed to 30°C) containing 10 mM dithiothreitol. The suspension was incubated in a 30°C water bath for 10 minutes. Cells were then pelleted at 1,500 × g for 5 minutes and then resuspended in 1 mL Buffer B (prewarmed to 30°C) containing lysis enzyme cocktail (1:100 dilution). The suspension was incubated in a 30°C shaker for 15 minutes for cell-wall digestion and then pelleted at 1,500 × g for 5 minutes at 4°C. The pellet was resuspended in 1 mL pre-chilled Buffer C with protease inhibitor cocktail (1:1,000 dilution). The cells were lysed by 15 up-and-down strokes with a pre-chilled glass Dounce homogenizer on ice. The lysate was incubated with shaking for 30 minutes at 25°C. The cell debris was pelleted at 500 × g for 5 minutes at 4°C. The supernatant was then transferred to a new tube. The nuclei were pelleted at 20,000 × g for 10 minutes at 4°C. The nuclear pellet was resuspended in 10 – 20 µL pre-chilled lysis buffer (50 mM EDTA and 1:1,000 protease inhibitor cocktail dilution) and incubated for 15 minutes on ice.
Nuclei lysates (3 µL) were added to a glow-discharged CF-4/2-2C-T grid (Protochips, Morrisville, NC). The grid was plunge-frozen using a Vitrobot Mark IV (blot time: 1 second, blot force: 1, humidity: 100%, temperature: 4°C).
Preparation of cryosections
Self-pressurized freezing was done based on a modified version of a published protocol (Yakovlev and Downing, 2011). Yeast cells (30 mL, OD600 = 0.2 – 0.6) were pelleted and resuspended in a dextran stock (40 kilodalton, 60% w/v, in YPD) to a final concentration of 30%. Cells were then loaded into a copper tube (0.45 / 0.3 mm outer / inner diameters). Both ends of the tube were sealed with flat-jaw pliers. The tube was held horizontally and dropped into the liquid-ethane cryogen. The tube’s ends were removed under liquid nitrogen with a tube-cut tool (Engineering Office M. Wohlwend, Sennwald, Switzerland).
Gold colloid solution (10-nm diameter, 5 µL at 5.7 ×1012 particles/mL in 0.1 mg/mL BSA) was applied to a continuous-carbon grid (10-nm thick carbon) and then air-dried overnight. Cryosections were controlled by a custom joystick-based micromanipulator (MN-151S, Narishige Co., Ltd., Tokyo, Japan) (Ladinsky et al., 2006; Ng et al., 2020). Seventy nm-thick frozen-hydrated sections were cut at −150°C in a Leica UC7/FC7 cryo-ultramicrotome (Leica Microsystems, Vienna, Austria). The EM grid was positioned underneath the ribbon using a Leica micromanipulator (Studer et al., 2014). The cryosection ribbon (∼3 mm long) was then attached to the grid by operating the Crion (Leica Microsystems) in “charge” mode for ∼30 seconds (Pierson et al., 2010).
Preparation of cryolamellae
Cells were plunge-frozen and then cryo-FIB milled using the method of Medeiros and Bock (Medeiros et al., 2018) as follows. Immediately before plunge-freezing, mid-log phase (OD600 ∼0.6) yeast cells were pelleted at 4,000 × g for 5 minutes. They were then resuspended in YPD media containing 3% (v/v) dimethyl sulfoxide as cryo-protectant to a final OD600 of approximately 2.5. Four µL of the cells were subsequently deposited onto Quantifoil R2/4 200 mesh copper grids (Quantifoil Micro Tools GmbH, Jena, Germany), which were then manually blotted from the back with Whatman® Grade 1 filter paper for approximately 3 – 5 seconds. The grids were then plunged into a 63/37 propane/ethane mixture (Tivol et al., 2008) using a Vitrobot Mark IV (humidity: 100%, temperature: 4°C). Cryo-FIB milling was performed on a Helios NanoLab 600 DualBeam (Thermo Fisher Scientific, TFS, Waltham, MA) equipped with a Quorum PolarPrep 2000 transfer system (Quorum Technologies, Laughton, United Kingdom). Plunge-frozen yeast samples were coated with a layer of organometallic platinum using the in-chamber gas injection system and the cold-deposition method (Hayles et al., 2007). Cryolamellae were then generated as follows: bulk material was first removed using the FIB at 30 kV 2.8 nA, followed by successive thinning of the cryolamellae at lower currents of 0.28 nA and 48 pA.
Cryo-ET data collection and reconstruction
All cryo-ET data were collected on Titan Krioses (TFS). Tilt series were collected with either TFS Tomo4, Leginon (Suloway et al., 2009), SerialEM (Mastronarde, 2003), or PACE-tomo (Eisenstein et al., 2022). Images were recorded either on a Falcon II (TFS) in integration mode or as movie frames on a K2 or K3 summit camera (Gatan, Pleasanton, CA) in super-resolution mode. The pixel sizes for the K2 and K3 data were chosen so that when binned to the same level, they closely match the ∼7 Å pixel size in our previous in situ study of HeLa chromatin (Cai et al., 2018a). Smaller pixel sizes were chosen for Falcon II data because this camera has lower detective quantum efficiency than the K-series cameras (Ruskin et al., 2013). Movies were aligned with either MotionCor2 (Zheng et al., 2017) or IMOD alignframes (Mastronarde, 1997). Prior to starting data collection of cryolamellae, the stage was pre-tilted to either −10° or −15° to account for the milling angle. Additional data-collection details are shown in Table S5.
Volta Phase Plate (VPP) imaging was done using the protocol of Fukuda and colleagues (Fukuda et al., 2015). The VPP position and condenser stigmator were adjusted such that the Ronchigram (the projected pattern visible in the camera) was free of elongated features, indicating no astigmatism, and is flat rather than grainy, indicating that the VPP is on-plane. Notably, the VPP needs to be heated, or else the phase shift will rapidly exceed π/2 radians (Danev et al., 2014). Two different VPP assemblies were used, which we will refer to as the NYSBC-2019 and NUS-2020 ones. These VPPs required different heater power settings to maintain a stable phase shift. The NYSBC-2019 and NUS-2020 VPP heaters were operated at ∼100 mW and ∼370 mW, respectively.
Cryotomograms were reconstructed using IMOD’s eTomo workflow (Mastronarde, 1997). Tilt series of lysates and cryosections were aligned using the gold beads as fiducials while those of cryolamellae were aligned with patches as fiducials. Only the tilt series that exhibited the minimal amount of drift and sample warping were analyzed. To further improve the tilt series alignment around the chromatin, fiducials (beads or patches) were chosen on the chromatin regions. For cryolamella patch tracking, tilt series were binned to a pixel size of 6.8 Å, the “Break contours into pieces w/ overlap” was set to 10, and the low-frequency rolloff sigma and cutoff radius were set to 0.03 and 0.1 pixel−1, respectively. A boundary model was created so that only the chromatin was enclosed. Fiducial patches were manually deleted if they overlapped with debris like ice crystals or if they mistracked. We did not detect a correlation between the alignment residual (Table S6) and our ability to detect canonical nucleosomes.
CTF estimation and phase flipping were done on the defocus phase contrast datasets using IMOD’s ctfplotter and ctfphaseflip programs. Prior to reconstruction, the cryosection and cryolamellae tilt series were binned to a final pixel size of 6.8 Å using the eTomo antialiasing option. Two tomogram versions were reconstructed for each tilt series. For visualization purposes, the tilt series were low-pass filtered to attenuate spatial frequencies beyond 25 Å to 30 Å resolution, prior to tomogram reconstruction. For classification analysis, the tilt series were low-pass filtered with a Gaussian rolloff starting at 15 Å resolution for lysates and 20 Å for cryosections and cryolamellae. In the classification jobs, the resolution of the data was further limited to 25 Å or 20 Å (see next section for details). More details of the datasets analyzed in this paper are shown in Table S6.
Nucleosome template matching, classification, and subtomogram averaging
A featureless round-edged 10 nm-diameter, 6 nm-thick cylindrical template was created using the Bsoft program beditimg (Heymann and Belnap, 2007). A cubic search grid with a 12-nm spacing was created with the PEET program gridInit (Heumann, 2016). Regions that had high-contrast artifacts from surface contaminants and ice crystals were excluded. Template matching was done using PEET (Heumann, 2016; Heumann et al., 2011; Nicastro et al., 2006), with a duplicate-removal cutoff distance of 6 nm. To accelerate the runs, no orientation search was done around the cylindrical axis and the resolution was attenuated starting at 70 Å on account of the smooth appearance of the template. Candidate hit lists of different cross-correlation cutoffs were generated using the PEET program createAlignedModel, then visualized together with the tomograms in 3dmod. The cross-correlation cutoff that eliminated spurious densities (primarily empty nucleoplasm) was chosen. The final numbers of subtomograms analyzed for each sample are in Table S7 and in the figures and figure legends.
Classification and subtomogram analysis were done with RELION (Kimanius et al., 2016; Scheres, 2012), following the workflows in Figure S1. In the published workflow (Figure S1A) (Bharat and Scheres, 2016), each subtomogram is first averaged by projecting the entire volume (16 nm along the Z axis), which introduces contributions from densities above and below the candidate nucleosome. To minimize the influence of other densities, pseudo-projections were created by averaging ∼12 nm along the Z axis, using the ot_relion_project.py script. 2-D classification using mask diameters ranging from 120 Å to 140 Å produced clear nucleosome-like classes, though the smaller masks included fewer adjacent densities. Note that in RELION, only circular masks are available for 2-D classification, meaning that it is not possible for the pseudo-projected densities to appear cylinder-like due to truncation by the mask. Densities that belonged to the most nucleosome-like classes were exported for 3-D classification, split into 30 classes. The resolution cutoffs were 25 Å for 2-D and 20 Å for 3-D classification. To eliminate the influence of adjacent densities during 3-D classification, a smooth cylindrical mask with a cosine-shaped edge was applied. The mask was created using beditimg and relion_mask_create. Because the GFP densities protrude from the nucleosome surface, we used a 9 nm-tall cylindrical mask for the analysis of nucleosomes with GFP tags, such as LGY0016 nucleosomes. The percentage of subtomograms belonging to each class was extracted with the script count_particles.awk from Guillaume Gaullier (Gaullier, 2021).
As observed in our previous study (Cai et al., 2018a), some canonical nucleosomes were lost in the 2-D classification process. We therefore used the alternative workflow in which the template matching hits were directly classified in 3-D (Figure S1B). To accommodate the increased diversity of complexes (some of which would have been removed had 2-D classification been done), we used 40 classes for BY4741, LGY0015, and LGY0016 nuclear lysates and 100 classes for cell cryolamellae and nuclear lysates of strains in which H3 or H4 were GFP tagged, without prior 2-D classification. These jobs crashed frequently because the RELION memory usage scales up with the number of classes (Kimanius et al., 2016). We were able to eliminate the crash problem by using higher-memory GPUs and by decreasing the number of translational search steps from 5 to 3 or 4. The canonical nucleosome classes were subjected to “Gold-standard” 3-D refinement (Henderson et al., 2012). No map sharpening was applied. Subtomogram class-average volumes were visualized with UCSF Chimera (Pettersen et al., 2004).
Biased-reference classification control
A 14 Å-resolution density map was simulated from the yeast nucleosome crystal structure PDB 1ID3 (White et al., 2001) using the Bsoft program bgex. To account for the artifacts associated with defocus phase contrast and CTF correction, a 6 µm underfocus was applied and then “corrected” for using the Bsoft program bctf. The map was also subjected to the 20 Å-resolution low-pass filter that was used on the tilt series using the IMOD program mtffilter. Template matching was done using this reference, including data to higher resolution (28 Å) than that for the less-biased search above and including a search around all Euler angles. The hits were subjected to 2-D classification to remove obvious non-nucleosomal densities. Next, the hits were 3-D classified using the simulated map as an initial alignment reference. To maximize the model bias, the template was only low-pass filtered to 20 Å-resolution instead of the recommended 60 Å (Bharat and Scheres, 2016).
Estimation of nucleosomes sampled per cell cryolamella
First, the average concentration of nucleosomes in the chromatin was estimated. The absolute number of nucleosomes per cell determined from genomics is 60,000 (Oberbeckmann et al., 2019). Soft X-ray tomography measurements revealed that the average G1 nucleus volume is 2 µm3, of which 20% is nucleolus (Uchida et al., 2011). Accordingly, chromatin (the nuclear volume not taken by the nucleolus), which contains the vast majority of nucleosomes, occupies ∼1.6 µm3. These two experimental values give an average nucleosome concentration of 37,500 per µm3. Next, the nuclear volume sampled by subtomogram analysis was determined by first drawing one closed contour around the chromatin using 3dmod; this closed contour encloses only the portion of the tomogram that was analyzed by template matching. The volume of this closed contour, which is one-voxel thick, was extracted using the command: imodinfo -F model.mod This command outputs the quantity “Cylinder volume”, in cubic pixels (voxels). The total tomographic volume sampled (Table S8) was obtained by multiplying Cylinder volume, the voxel volume (0.31 nm3 for 0.68 nm pixel size), and the number of tomographic slices that contain chromatin. The BY4741 VPP tomograms summed to 0.67 µm3, which yields ∼25,000 nucleosomes. For the HeLa cell in (Cai et al., 2018a) (EMPIAR-10179), the nucleus volume analyzed was 0.12 µm3.
Simulations of nucleosome tomographic slices
Atomic models of the nucleosome (PDB 1KX5) (Davey et al., 2002) were manually positioned in UCSF Chimera and edited to remove the N-terminal tails. A 3-D density map was calculated with the Bsoft program bgex. There is no software that simulates Volta contrast, so we approximated the Volta-induced phase shift by setting the amplitude contrast to 100% and the defocus to zero in the Bsoft program bctf, followed by “correction”, also done with bctf. A tilt series was calculated from the simulated map using the IMOD program xyzproj. The tilt series was then aligned by cross-correlation and then backprojected using IMOD’s eTomo workflow. Parameters such as pixel size, tilt range, and tilt angle were kept as close to the experimental ones as possible. Tomographic slices were made in IMOD slicer at the same thickness as for the real cryotomograms.
Ribosome subtomogram analysis control
Ribosomes were analyzed using the same software packages as for nucleosomes, as detailed in the workflow in Figure S25. Candidate ribosomes were template matched in a single VPP tomogram (from Figure S11), using a 25-nm-diameter sphere as a reference. The candidate ribosomes were filtered by cross correlation so that obvious false positives (vacuum) were excluded, leaving 3,816 hits. These particles were subjected to direct 3-D classification with k = 10, resulting in six non-empty classes. Particles belonging to the two ribosome classes (1,150 total) were pooled and “gold-standard” refined, yielding a density map at 28 Å resolution (FSC = 0.143) / 33 Å resolution (FSC = 0.5). Density maps were simulated at 15 Å, 20 Å, and 30 Å resolution using the Bsoft program bgex and the yeast ribosome crystal structure (Ben-Shem et al., 2011). To approximate the use of the VPP, the amplitude contrast was set to 90% and the defocus to −1 µm.
Materials, data, and code availability
All S. cerevisiae strains generated in this study are available upon request. A subtomogram average of a BY4741 canonical nucleosome ex vivo, a double-GFP tagged LGY0016 nucleosome ex vivo, and the two BY4741 canonical nucleosome classes in situ have been deposited at EMDB as entry EMD-31086. All raw cryo-ET data, reconstructed tomograms and BY4741 cryolamellae VPP in situ class averages have been deposited in EMPIAR under entry EMPIAR-10678. All auxiliary scripts have been deposited at GitHub (https://github.com/anaphaze/ot-tools) and are publicly available as of the date of publication. Any additional information required to reanalyze the data reported in this paper is available upon request.
Movie S1. Direct 3-D classification of BY4741 (Wild type) nucleosome-like particles in VPP tomograms of cell cryolamellae, round 1
The progress of 30 rounds of 3-D classification is shown. There are 100 classes, initialized with a smooth nucleosome-sized cylindrical reference. The final iteration (30) is also shown in Figure. S10, with only the most nucleosome-like classes shaded.
Movie S2. Direct 3-D classification of BY4741 (Wild type) nucleosome-like particles in VPP tomograms of cell cryolamellae, round 2
The most nucleosome-like classes from round 1 were selected and subjected to a second round of classification, with 4 classes, again initialized with a smooth nucleosome-sized cylindrical reference.
Movie S3. Direct 3-D classification of LGY0016 (H2A-GFP) lysate nucleosomes, round 1
The progress of 30 rounds of 3-D classification is shown. There are 40 classes, initialized with a smooth nucleosome-sized cylindrical reference. The final iteration (30) is also shown in Figure. S18, with only the most nucleosome-like classes shaded.
Movie S4. Direct 3-D classification of LGY0016 (H2A-GFP) lysate nucleosomes, round 2
The most nucleosome-like classes from round 1 were selected and subjected to a second round of classification, with 10 classes, again initialized with a smooth nucleosome-sized cylindrical reference. Two “junk” classes were removed.
Movie S5. Direct 3-D classification of LGY0016 (H2A-GFP) nucleosome-like particles in VPP tomograms of cell cryolamellae, round 1
3-D classification analysis of LGY0016 (H2A-GFP) nucleosome-like particles from cryolamellae imaged with a VPP. The initial reference is a featureless cylinder. There are 100 classes.
- Histone dynamics mediate DNA unwrapping and sliding in nucleosomesNat Commun 12
- Cryo-Electron Tomography: Can it Reveal the Molecular Sociology of Cells in Atomic Detail?Trends Cell Biol 26:825–837
- Structure and Dynamics of a 197 bp Nucleosome in Complex with Linker Histone H1Mol Cell 66:384–397
- The structure of the eukaryotic ribosome at 3.0 A resolutionScience 334:1524–1529
- Topaz-Denoise: general deep denoising models for cryoEM and cryoETNat Commun 11
- Resolving macromolecular structures from electron cryo-tomography data using subtomogram averaging in RELIONNat Protoc 11:2054–2065
- ZDS1 and ZDS2, genes whose products may regulate Cdc42p in Saccharomyces cerevisiaeMol Cell Biol 16:5264–5275
- Nucleosome and ubiquitin position Set2 to methylate H3K36Nat Commun 10
- Histone octamer rearranges to adapt to DNA unwrappingNat Struct Mol Biol 25:101–108
- Designer deletion strains derived from Saccharomyces cerevisiae S288C: a useful set of strains and plasmids for PCR-mediated gene disruption and other applicationsYeast 14:115–132
- The kinetic landscape of nucleosome assembly: A coarse-grained molecular dynamics studyPLoS Comput Biol 17
- A map of nucleosome positions in yeast at base-pair resolutionNature 486:496–501
- Automated cryo-lamella preparation for high-throughput in-situ structural biologyJ Struct Biol 210
- The in situ structures of mono-, di-, and trinucleosomes in human heterochromatinMol Biol Cell 29:2450–2457
- Cryo-ET reveals the macromolecular reorganization of S. pombe mitotic chromosomes in vivoProc Natl Acad Sci U S A 115:10977–10982
- Natural chromatin is heterogeneous and self-associates in vitroMol Biol Cell 29:1652–1663
- Architecture of the type IVa pilus machineScience 351
- Budding yeast chromatin is dispersed in a crowded nucleoplasm in vivoMol Biol Cell 27:3357–3368
- A complete data processing workflow for cryo-ET and subtomogram averagingNat Methods 16:1161–1168
- 3.9 A structure of the nucleosome core particle determined by phase-plate cryo-EMNucleic Acids Res 44:8013–8019
- Chromatin fiber polymorphism triggered by variations of DNA linker lengthsProc Natl Acad Sci U S A 111:8061–8066
- Volta potential phase plate for in-focus phase contrast transmission electron microscopyProc Natl Acad Sci U S A 111:15635–15640
- Using the Volta phase plate with defocus for cryo-EM single particle analysisElife 6
- Solvent mediated interactions in the structure of the nucleosome core particle at 1.9 a resolutionJ Mol Biol 319:1097–1113
- Visualization of proteins in intact cells with a clonable tag for electron microscopyJ Struct Biol 165:157–168
- Nucleosome-bound SOX2 and SOX11 structures elucidate pioneer factor functionNature 580:669–672
- Parallel cryo electron tomography on in situ lamellaebioRxiv
- Capturing Structural Heterogeneity in Chromatin FibersJ Mol Biol 429:3031–3042
- Nucleosome conformational variability in solution and in interphase nuclei evidenced by cryo-electron microscopy of vitreous sectionsNucleic Acids Res 46:9189–9200
- Analysis of cryo-electron microscopy images does not support the existence of 30-nm chromatin fibers in mitotic chromosomes in situProceedings of the National Academy of Sciences 105:19732–19737
- Structural basis for ATP-dependent chromatin remodelling by the INO80 complexNature 556:386–390
- Nucleosome-Chd1 structure and implications for chromatin remodellingNature 550:539–542
- Nucleosome plasticity is a critical element of chromatin liquid-liquid phase separation and multivalent nucleosome interactionsNat Commun 12
- Cryo-electron tomography and deep learning denoising reveal native chromatin landscapes of interphase nucleibioRxiv
- Electron cryotomography of vitrified cells with a Volta phase plateJ Struct Biol 190:143–154
- Structural and biochemical analyses of the nucleosome containing Komagataella pastoris histonesJ Biochem 172:79–88
- Chromatin in a marine picoeukaryote is a disordered assemblage of nucleosomesChromosoma 122:377–386
- A technique for improved focused ion beam milling of cryo-prepared life science specimensJ Microsc 226:263–269
- Outcome of the first electron microscopy validation task force meetingStructure 20:205–214
- Isolation of yeast histone genes H2A and H2BCell 18:1261–1271
- PEET. Vol
- Clustering and variance maps for cryo-electron tomography using wedge-masked differencesJ Struct Biol 175:288–299
- Bsoft: image processing and molecular modeling for electron microscopyJ Struct Biol 157:3–18
- emClarity: software for high-resolution cryo-electron tomography and subtomogram averagingNat Methods 15:955–961
- Histone tails cooperate to control the breathing of genomic nucleosomesPLoS Comput Biol 17
- Torsional stress can regulate the unwrapping of two outer half superhelical turns of nucleosomal DNAProc Natl Acad Sci U S A 118
- Free Energy Landscape of H2A-H2B Displacement From NucleosomeJ Mol Biol 434
- EMPIAR: a public archive for raw electron microscopy image dataNat Methods 13:387–388
- Structural basis of recognition and destabilization of the histone H2B ubiquitinated nucleosome by the DOT1L histone H3 Lys79 methyltransferaseGenes Dev 33:620–625
- Subtomogram analysis using the Volta phase plateJ Struct Biol 197:94–101
- Accelerated cryo-EM structure determination with parallelisation using GPUs in RELION-2Elife 5
- High-throughput AFM analysis reveals unwrapping pathways of H3 and CENP-A nucleosomesNanoscale 13:5435–5447
- In vitro reconstitution and biochemical analyses of the Schizosaccharomyces pombe nucleosomeBiochem Biophys Res Commun 482:896–901
- Structural basis of the nucleosome transition during RNA polymerase II passageScience 362:595–598
- Vitreous cryo-sectioning of cells facilitated by a micromanipulatorJ Microsc 224:129–134
- Architecture and self-assembly of the jumbo bacteriophage nuclear shellNature
- Improved blue, green, and red fluorescent protein tagging vectors for S. cerevisiaePLoS One 8
- Amplification of histone genes by circular chromosome formation in Saccharomyces cerevisiaeNature 443:1003–1007
- Mechanism of chromatin remodelling revealed by the Snf2-nucleosome structureNature 544:440–445
- Additional modules for versatile and economical PCR-based gene deletion and modification in Saccharomyces cerevisiaeYeast 14:953–961
- Crystal structure of the nucleosome core particle at 2.8 A resolutionNature 389:251–260
- A focused ion beam milling and lift-out approach for site-specific preparation of frozen-hydrated lamellas from multicellular organismsJ Struct Biol 192:262–269
- Focused ion beam milling of vitreous water: prospects for an alternative to cryo-ultramicrotomy of frozen-hydrated biological samplesJ Microsc 222:42–47
- Global histone protein surface accessibility in yeast indicates a uniformly loosely packed genome with canonical nucleosomesEpigenetics Chromatin 14
- Dual-axis tomography: an approach with alignment methods that preserve resolutionJ Struct Biol 120:343–352
- SerialEM: A program for automated tilt series acquisition on Tecnai microscopes using prediction of specimen positionMicroscopy and Microanalysis 9
- Cryo-electron microscopy of vitrified chromosomes in situEMBO J 5:1395–1402
- Robust workflow and instrumentation for cryo-focused ion beam milling of samples for electron cryotomographyUltramicroscopy 190:1–11
- Concatenated metallothionein as a clonable gold label for electron microscopyJ Struct Biol 160:70–82
- Structural transitions of F-actin upon ATP hydrolysis at near-atomic resolution revealed by cryo-EMNat Struct Mol Biol 25:528–537
- Deep learning improves macromolecule identification in 3D cellular cryo-electron tomogramsNat Methods 18:1386–1394
- VivosX, a disulfide crosslinking method to capture site-specific, protein-protein interactions in yeast and human cellsElife 7
- Investigating eukaryotic cells with cryo-ETMol Biol Cell 31:87–100
- Serial Cryomicrotomy of Saccharomyces cerevisiae for Serial Electron CryotomographyBio-Protocol 10
- The molecular architecture of axonemes revealed by cryoelectron tomographyScience 313:944–948
- Rapid Depletion of Budding Yeast Proteins via the Fusion of an Auxin-Inducible Degron (AID)Curr Protoc Cell Biol 64
- A genetically encoded metallothionein tag enabling efficient protein detection by electron microscopyJ Electron Microsc (Tokyo 56:93–101
- Routine single particle CryoEM sample and grid characterization by tomographyElife 7
- Absolute nucleosome occupancy map for the Saccharomyces cerevisiae genomeGenome Res 29:1996–2009
- Novel structural labeling method using cryo-electron tomography and biotin-streptavidin systemJ Struct Biol 183:305–311
- Sub-nucleosomal Genome Structure Reveals Distinct Nucleosome Folding MotifsCell 176:520–534
- UCSF Chimera--a visualization system for exploratory research and analysisJ Comput Chem 25:1605–1612
- Improving the technique of vitreous cryo-sectioning for cryo-electron tomography: electrostatic charging for section attachment and implementation of an anti-contamination glove boxJ Struct Biol 169:219–225
- Transcription and Remodeling Produce Asymmetrically Unwrapped Nucleosomal IntermediatesMol Cell 68:1038–1053
- Live-cell single particle imaging reveals the role of RNA polymerase II in histone H2A.Z evictionElife 9
- Subnucleosomal structures and nucleosome asymmetry across a genomeCell 159:1377–1388
- Micromachining tools and correlative approaches for cellular cryo-electron tomographyJ Struct Biol 172:169–179
- EM measurements define the dimensions of the “30-nm” chromatin fiber: evidence for a compact, interdigitated structureProc Natl Acad Sci U S A 103:6506–6511
- CTFFIND4: Fast and accurate defocus estimation from electron micrographsJ Struct Biol 192:216–221
- Quantitative characterization of electron detectors for transmission electron microscopyJ Struct Biol 184:385–393
- HP1 reshapes nucleosome core to promote phase separation of heterochromatinNature 575:390–394
- Cryo-EM structure of the nucleosome core particle containing Giardia lamblia histonesNucleic Acids Res 49:8934–8946
- Evidence for short-range helical order in the 30-nm chromatin fibers of erythrocyte nucleiProceedings of the National Academy of Sciences 108:16992–16997
- RELION: implementation of a Bayesian approach to cryo-EM structure determinationJ Struct Biol 180:519–530
- Cryo-EM study of the chromatin fiber reveals a double helix twisted by tetranucleosomal unitsScience 344:376–380
- Thinning of large mammalian cells for cryo-TEM characterization by cryo-FIB millingJ Microsc 247:220–227
- A new tool based on two micromanipulators facilitates the handling of ultrathin cryosection ribbonsJ Struct Biol 185:125–128
- Fully automated, sequential tilt-series acquisition with LeginonJ Struct Biol 167:11–18
- Structure of the chromatin remodelling enzyme Chd1 bound to a ubiquitinylated nucleosomeElife 7
- Structural reorganization of the chromatin remodeling enzyme Chd1 upon engagement with nucleosomesElife 6
- A streamlined workflow for automated cryo focused ion beam millingJ Struct Biol 213
- Multi-particle cryo-EM refinement with M visualizes ribosome-antibiotic complex at 3.5 A in cellsNat Methods 18:186–193
- An improved cryogen for plunge freezingMicrosc Microanal 14:375–379
- High-power near-concentric Fabry-Perot cavity for phase contrast electron microscopyRev Sci Instrum 92
- Quantitative analysis of yeast internal architecture using soft X-ray tomographyYeast 28:227–236
- In situ architecture of the ciliary base reveals the stepwise assembly of intraflagellar transport trainsScience 377:543–548
- Recombination-induced tag exchange to track old and new proteinsProc Natl Acad Sci U S A 107:64–68
- Opening windows into the cell: focused-ion-beam milling for cryo-electron tomographyCurr Opin Struct Biol 23:771–777
- A ferritin-based label for cellular electron cryotomographyStructure 19:147–154
- Structure of the yeast nucleosome core particle reveals fundamental changes in internucleosome interactionsEMBO J 20:5207–5218
- Structure and dynamics of the yeast SWR1-nucleosome complexScience 362
- Freezing in sealed capillaries for preparation of frozen hydratedsectionsJ Microsc 244:235–247
- Fully automated, sequential focused ion beam milling for cryo-electron tomographyElife 9
- MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopyNat Methods 14:331–332
- Structural Mechanisms of Nucleosome Recognition by Linker HistonesMol Cell 59:628–638
- Nucleosome structure and dynamics are coming of ageNat Struct Mol Biol 26:3–13
- Structural basis of nucleosome dynamics modulation by histone variants H2A.B and H2A.Z.2.2EMBO J 40
- A Bayesian approach to single-particle electron cryo-tomography in RELION-4.0Elife 11
- The nucleosome family: dynamic and growingStructure 17:160–171
- The structure of the eukaryotic ribosome at 3.0 A resolutionScience 334:1524–1529
- Advances in Single-Particle Electron Cryomicroscopy Structure Determination applied to Sub-tomogram AveragingStructure 23:1743–1753
- Natural chromatin is heterogeneous and self-associates in vitroMol Biol Cell 29:1652–1663
- Implementation of a cryo-electron tomography tilt-scheme optimized for high resolution subtomogram averagingJ Struct Biol 197:191–198
- Structure of the yeast nucleosome core particle reveals fundamental changes in internucleosome interactionsEMBO J 20:5207–5218
- The molecular structure of green fluorescent proteinNat Biotechnol 14:1246–1251