Combining per-particle cryo-ET and cryo-EM single particle analysis to elucidate heterogeneous DNA-protein organization

Abstract

Cryo-electron microscopy single particle analysis (cryo-EM SPA) and cryo-electron tomography (cryo-ET) have historically been employed as distinct approaches for investigating molecular structures of disparate sample types, focusing on highly purified biological macromolecules and in situ cellular contexts, respectively. However, these techniques offer inherently complementary structural insights that, when combined, provide a more comprehensive understanding of complex biological systems. For example, if both techniques are applied to the same purified biological macromolecules, cryo-ET has the ability to resolve highly flexible yet strong signal features on individual target molecule which will not be preserved in the high-resolution cryo-EM SPA results. In this review, we highlight recent achievements utilizing such application to unveil new insights into the chromatin assembly and activities of DNA-protein assemblies. This convergence of cryo-EM SPA and cryo-ET holds great promise for elucidating new structural aspects of these essential molecular processes.

Graphical Abstract

Complementary structural information provided by per-particle cryo-ET analysis and cryo-EM single particle analysis.

Cryo-EM single particle analysis generates a high-resolution map of the structurally conserved region among particles while cryo-ET characterizes, in 3D, the heterogeneous features on a per-particle basis. The joining of these techniques reveals heterogeneous organization about a highly resolved core structure.

In recent years, the field of structural biology has been revolutionized by cryogenic electron microscopy (cryo-EM), driven by advancements of its hardware, software, and sample preparation techniques [1]. Two prominent methods for structural elucidation, cryo-EM single particle analysis (SPA) and cryo-electron tomography (cryo-ET), have significantly contributed to this revolution. Cryo-EM SPA is the preferred choice for solving structures of highly purified biological macromolecules [1]. This technique relies on gathering 2D projections of thousands of randomly oriented molecules, which are then computationally aligned to generate high-resolution 3D density maps. While powerful, a caveat of this method is that only highly reoccurring and stable features are retained while unique and flexible features are often left unresolved. On the other hand, cryo-ET has established itself as the premier technique for resolving macromolecular structures within their native cellular context [2]. Unlike SPA, cryo-ET collects 2D projections of the same region of interest while tilting the sample, which are then used to calculate a 3D volume (tomogram) of that region with its unique and flexible features preserved [2].

While canonically applied to different type of samples, cryo-EM SPA and cryo-ET offer highly complementary information. Applying both methods on the same type of sample could allow for the unresolved flexible features lost in cryo-EM SPA to be visualized through cryo-ET and the high-resolution cryo-EM SPA result to supplement the relatively lower resolution but already 3D particles seen in cryo-ET. Therefore, a hybrid approach of cryo-EM SPA and cryo-ET will provide a more comprehensive structural view of biological macromolecules, especially those featuring highly flexible yet strong signal features, such as long DNA strains associated with protein molecules. Below we use the recent studies of chromatin organization and DNA-protein assemblies to highlight the power of this emerging usage of cryo-EM structural analysis.

Nucleosome arrangements inside cells

Cryo-EM SPA studies of nucleosomes have provided invaluable insights into DNA organizations of these subunits of chromatin [3–5*]. However, these studies are typically limited to a single nucleosome core particle or an array of a few particles. Interestingly, Arimura et al. isolated native interphase as well as metaphase nucleosomes from Xenopus egg extracts and used cryo-EM SPA to solve the structure of each nucleosome [5*]. The authors found that the structure of these native nucleosomes was, on average, indistinguishable from previous structures [5*]. However, such focus of SPA on stable and reoccurring features limits its ability to capture the full heterogeneity of chromatin that is necessary to accommodate dynamic biological processes such as DNA replication, repair, and transcription. In contrast, cryo-ET has the unique capability of visualizing individual nucleosomes in 3D [6–9]. Furthermore, cryo-ET can tolerate thicker sample thicknesses of about 400 nm [10]. These attributes have allowed cryo-ET to elucidate chromatin organization and map the position and orientation of nucleosomes in situ within the crowded nuclear environment [6–9]. There the nucleosomes were found to have an irregular distribution throughout the nucleus. In addition, variations of DNA packaging around each nucleosome were revealed [9,11]. The distance between DNA gyres per nucleosome spanned 2-4 nm compared to what was resolved by X-ray crystallography and cryo-EM SPA structures which is about 3 nm [11–13]. The entry and exit of DNA wrapped around nucleosomes were variable and deviated from the canonical 1.65 superhelical turns [9]. It is important to note that the characterization of chromatin organization in situ relied on a template matching approach to identify nucleosomes [6–9]. Therefore, the heterogeneity is further exacerbated because nucleosomes that are too conformationally heterogeneous or variably bound by different proteins may be missed in these analyses. This characterization of chromatin and individual nucleosomes in situ highlights the variability unable to be captured by cryo-EM SPA alone. Notably, these studies do not capture a critical aspect of chromatin organization, namely the interconnectivity. For example, nucleosomes positioned next to one another in 3D may be far apart along the linear DNA (i.e., topologically associating domains). Therefore, visualization of the interconnectivity will reveal insights into the nucleosome spacing along DNA, boundaries between chromatin regions, the trajectory of chromatin, and any patterns associated with these features. Cryo-ET is well positioned to visualize highly flexible yet strong signal features such as DNA, but these analyses have been limited. A primary reason for not assigning the interconnectivity was mostly because the linker DNA could not be unambiguously visualized due to its dense packing in situ.

DNA arrangement in isolated chromatin

To investigate linker DNA conformation within the context of chromatin organization, cryo-ET has been employed to image reconstituted and native chromatin with different components. Therefore, a more direct relationship between chromatin organization and different sample conditions may be observed. Furthermore, this approach allows for the sample freezing in thinner ice, enhancing contrast for 3D visualization and quantitative conformational analysis of linker DNA.

Reconstituted Chromatin

The assembly of reconstituted chromatin has allowed researchers to dissect linker DNA organization as a function of various histones. For example, it was observed that centromere protein A (CENP-A), a histone H3 variant, reduces the angle between entry and exit linker DNA strands, resulting in greater compaction of nucleosome arrays compared to canonical histone H3 [14]. Additionally, histone H1 was found to facilitate the invasion of free entry DNA into one of the inner nucleosome core particles of a nucleosome array, with the rate of condensate formation of nucleosome arrays increasing in correlation with histone H1 concentration [15].

While linker DNA was characterized under various conditions, these investigations utilized the Widom 601 nucleosome positioning sequence to reconstitute the chromatin [14–16]. It is worth noting that this sequence, although valuable for biochemical studies due to its reliable nucleosome positioning, may not fully represent biological conditions as it can adopt non-natural configurations and bias the analysis [17–19**].

Chromosomes

To capture a more natural chromatin organization, Beel et al. inspected purified mitotic chromosomes by cryo-ET [20**]. This approach, distinct from previous studies inspecting native chromatin in densely packed in situ conditions, allowed for partial chromatin decondensation within the physiological range. As a result, the authors were able to clearly visualize the connectivity between nucleosomes (Figure 1A) [20**]. It was observed that the DNA linker lengths covered a broad range (10-110 bp) (Figure 1B) and the angle between entry and exit linker per nucleosome also varied between 40 and 80 degrees (Figure 1C). These key observations highlight that the canonical 146bp nucleosome core particle represents an average and that nucleosomes are variably spaced apart and over- or under-wrapped. Similarly, Jentik et al. characterized purified human chromatin that was decondensed by fragmentation via micrococcal nuclease (MNase) [19**]. Additionally, the authors of this study utilized deep-learning-based regression models to denoise the tomograms to facilitate segmentation. Through this approach, linker DNA and even histone tails were visualized and allowed characterization of chromatin organization as a function of Mg²⁺ and PAD4 dependent citrullination. Interestingly, non-uniform nucleosome stacking was abundant and condensed by the addition of Mg²⁺ and unfolded by PAD4 mediated citrullination (Figure 1D and 1E). Altogether, 3D analysis by cryo-ET of purified chromatin has revealed DNA organization between nucleosomes at a per-particle level, and these analyses have propelled our understanding of chromatin organization.

Figure 1.

Characterization of Linker DNA and chromatin organization.

(A) A 4-nm slab of a tomogram displayed with inverted contrast. Selected fibers with nucleosome core particle map (8140-EMD [13]) manually docked and linker DNA traced. Scale bars, 10 nm. (B) Cropped densities depicted in mesh representation, with nucleosome core particle map (8140-EMD [13]) manually docked and linker DNA length indicated. (C) Example of DNA entering and exiting nucleosomes. Highlighting the variable angles between DNA entering and exiting nucleosomes. (D) Example subtomogram of fragmented native chromatin treated with 0.75 mM Mg²⁺(left) with centroid, axis, plane (CAP) modeling of nucleosomes and linker DNA (right). Scale bar, 30 nm. (E) Same as (D) but treated with PAD4. Scale bar, 30 nm. Panels A-C adapted from Beel et al. [20**]; panels D-E adapted from Jentik et al. [19**].

Conformational landscape of purified macromolecular assemblies

The scope of cryo-ET imaging purified DNA-protein assemblies has been expanded beyond the above-mentioned chromatin to isolated DNA containing macromolecules. These complexes offer a particular intriguing subject for cryo-ET analysis, as they may exist in unique states that are challenging to be resolved by cryo-EM SPA.

For example, DNA and RNA origami structures have been analyzed by cryo-ET to elucidate their conformational dynamics and inform their optimization [21–23]. Researchers were able to assess distinct DNA origami structures as a function of gold nanoparticle size [21]. The authors of this study found that 5 nm gold nanoparticles facilitated formation of two different structures while 10 nm gold nanoparticles biased formation into one structure. As these confirmations were not homogeneous, each unique 3D structure was only able to be determined by cryo-ET.

Similarly, cryo-ET was utilized to characterize the assembly process of Mediator Preinitiation Complex (Med-PIC) along DNA and its regulation of transient promoter-enhancer communication using S. cerevisiae as a model system [24**]. Gorbea et al. assembled highly purified transcription factors and prepared the same sample for both cryo-EM SPA and cryo-ET. There, the core protein components were highly resolved using cryo-EM SPA but the DNA was largely missing from the averaged electron density (Figure 2, upper left row). Therefore, the interplay between DNA and the core complex was not understood and its relationship throughout transcription activation remained to be determined. However, cryo-ET and its unique ability to visualize individual molecules in 3D was able to clearly visualize the DNA and trace its organization about the core complex (Figure 2, lower left row). To elucidate the interplay between DNA and the core complex, subtomogram averaging was performed in which each individual particle in a 3D volume (subtomogram) was aligned in 3D and averaged to generate higher-resolution density map of the core protein structure. This procedure allowed the subtomogram average to be fit back into each individually aligned subtomogram to systematically trace DNA and study DNA-protein interaction sites and configurations on a per-particle basis for statistical analysis (Figure 2, lower left row).

Figure 2.

Hybrid application of Cryo-ET and cryo-EM SPA.

Cryo-EM SPA generates a high-resolution map while cryo-ET characterizes the flexible DNA on a per-particle basis. The joining of these techniques reveals flexible DNA organization about a highly resolved core complex which can be studied in the presence of additional transcription factors. Adapted from Gorbea et al. [24**]

The authors of this study observed that Med-PIC is predominantly assembled at the upstream activating sequence (enhancers in higher eukaryotes) in contact with Mediator (Figure 2, upper right panel). Since this configuration is not immediately poised for transcription, this prompted the addition of another general transcription factor to the purified system to generate a more complete Med-PIC (Med-PIC⁺). The presence of the additional general transcription factor dramatically altered the organization in which Med-PIC⁺ is primarily assembled at the core promoter in a form amenable to transcription (Figure 2, lower right panel). Furthermore, cryo-ET visualized simultaneous dual DNA contact of both promoters on each PIC suggesting a role in bidirectional transcription. This highlights the utility of cryo-ET per-particle analysis since SPA alone would not have been able to distinguish between simultaneous dual DNA contact of each promoter or individual DNA contact at one promoter. Therefore, the combination of cryo-EM SPA and cryo-ET per-particle analysis facilitated a more complete description of the assembly process of Med-PIC along DNA as it becomes poised for transcription.

Challenges and opportunities for per-particle 3D analysis by cryo-ET

Although there have been significant advances in per-particle 3D analysis by cryo-ET, challenges remain that limit its widespread use.

Data collection speed and quality

A large number of particles is necessary to confidently sample the conformational landscape. However, cryo-ET suffers from lower throughput than cryo-EM SPA. This is mostly due to the time required for accurate tracking during the imaging of the sample from a wide range of angles. Several recent method developments have significantly improved cryo-ET data collection speed, including fast-incremental single exposure methods [25] or parallel/montage imaging of adjacent regions during tilt series collection [26–28]. However, several qualifications about the local geometry of the sample and precision of the electron microscope stage must be met. Therefore, data collection schemes and technology need to be further optimized to make it more widely applicable.

Image restoration and segmentation

For high-resolution structural study, biological samples in general can only tolerate a total electron dosage of ~40-80e/A² which must be distributed amongst several 2D projections for cryo-ET. This reduces the contrast and number of frames per tilt angle for accurate motion correction. This problem is further exacerbated at higher tilt angles in which the sample thickens. A solution to compensate for the sample thickening for enough image contrast is to increase the dose as the tilt angle increases. However, this causes one to reach the tolerated dose limit much quicker. Collections schemes that reduce the number of 2D projections (e.g., smaller tilt range and larger tilt increments) help stay beneath the tolerated dose limit but are at the expense of increased missing information in the Fourier space. Recent methods have improved data quality post data collection [29**–31*]. For example, deep learning models have been shown to dramatically improve the contrast to aid in per-particle characterization and these methods have been applied to chromatin both purified and in situ [19**,29**–32]. Furthermore, utilizing high-resolution maps obtained by single particle analysis may help deep learning models to facilitate subtomogram averaging and per-particle characterization for other biological samples. It will be exciting to see how these perform on more complex macromolecules.

Current segmentation methods exhibit robustness when it comes to segmenting large, highly contrasting, and repetitive features like membranes, ribosomes, or nucleosomes [33,34]. However, these methods have yet to demonstrate reliability in tracing flexible DNA strands in various context [20**,24**]. Consequently, manual tracing remains the predominant approach, despite being exceedingly laborious and susceptible to user error and bias. Current deep learning models assist in assigning unmodeled density but are restricted to higher-resolution density maps [32,35]. Therefore, the development of new computational approaches capable of assigning unmodeled density during per-particle characterization would greatly benefit these research directions.

Targeting

The development of cryo-ET along with techniques to identify molecules of interest such as cryo-correlative super-resolution light and electron microscopy yields exciting opportunities to explore fundamental biology [36–41]. For example, the overall organization of native chromosomes can be further dissected by relating the local organization to particular proteins of interest (e.g., CENP-A, PAD4) identified by super-resolution cryo-fluorescence microscopy. Similarly, the sub-chromosomal regions such as heterochromatin, euchromatin, or the centromere can be locally analyzed. Additionally, identification of post-translational modifications on particular proteins (e.g., histone citrullination) and their effect on local chromatin organization present another avenue of exploration.

Conclusion:

Cryo-ET, especially on purified samples, is at the forefront to resolve highly flexible yet strong signal features on a per-particle basis. Although this review uses long DNA strands as the example of such features, this approach has the potential for a variety of samples and questions. For example, the core of the nuclear pore complex has been resolved to high resolution while the disordered linker nucleoporins remain to be observed [42–44]. How the disordered nucleoporins are organized about the nuclear pore complex may reveal critical insight into its mechanism and presents as an attractive avenue for this method. Similarly, one may characterize the conformational space of a system as a function of time and visualize conformational transitions and transient states (i.e., time-resolved per-particle characterization). The use of purified samples is ideal because time-resolved cryo-EM requires precise conditions to establish the kinetics of the system [45]. Therefore, sample preparation for time-resolved cryo-EM aligns well with that for cryo-ET per-particle characterization. The dual utilization of cryo-EM SPA and cryo-ET presents an exciting new approach to deepen our structural understanding of these complex molecular processes.