Beyond Protein Structure Determination with MicroED

Abstract

Microcrystal electron diffraction (MicroED) was developed at the Janelia Research Campus as a new modality in electron cryomicroscopy (cryoEM), with the term MicroED first coined in 2013. Since then, MicroED has not only made important contributions for pushing the resolution limits of cryoEM protein structure characterization but also of peptides, small-organic and inorganic molecules, and natural-products that have resisted structure determination by other methods. This review showcases important recent developments in MicroED, highlighting the importance of the technique in fields of studies beyond protein structure determination where this MicroED is beginning to have paradigm shifting roles.

Introduction

MicroED^1,2 has been pushing the limits of cryoEM in determining structures of macromolecular assemblies, peptides, and chemical compounds^1,3–5. Prior the technological advancements in detectors^6–9 and software^10–13 that made the “cryoEM resolution revolution¹⁴” possible, structure determination of biological assemblies, peptides, and chemical compounds has been dominated X-ray crystallography. To date, there are close to 150,000 depositions in the Protein Data Bank (PDB), comprised of roughly 90% X-ray, 8% NMR, and 2% EM structures. Although the first atomic resolution structure by cryoEM was already reported in 2005¹⁵, the highest growth in TEM structure deposition has occurred in the last five years.

The field of cryoEM includes at least four major techniques: cryo-electron tomography (cryo-ET)^16,17, single-particle-analysis (SPA)^18–20, 2-dimenstional (2D) electron crystallography^21,22, and MicroED^1,2 (Figure 1). All of these cryoEM techniques exploit the advantage that electrons interact orders of magnitude more strongly with materials than X-rays, allowing the use of samples that are not tractable by other methods²³. While CryoET and SPA use imaging, the crystallographic cryoEM methods of 2D electron crystallography and MicroED can take advantage of electron diffraction. 2D-electron crystallography is typically used for structure determination of 2D arrays, which traditionally have been of membrane proteins that are crystallized within the native environment of the lipid bilayer²². In contrast, MicroED uses 3D crystals and the data is collected by continuous rotation to yield structures of a wide range of samples including soluble and membrane proteins, peptides, small organic and inorganic molecules, semi-conductors, and natural-products⁵.^24,25,21,26

Figure 1. The four major modalities of cryoEM.

From left to right. A model of the HeLa cell nuclear periphery (Reprint from reference 24 with permission from AAAS) by tomography. Model of COVID-19 spike protein by SPA (Reprint from reference 25 with permission from AAAS). The structure of bacteriorhodopsin determined by electron crystallography (Reprint from reference 21) and the structure of catalase determined by MicroED (Reprint from reference 26).

During MicroED experiments, crystals are harvested and prepared in a number of ways for embedment on EM grids (discussed below)¹ (Figure 2). Biological samples, which are more sensitive to radiation, are typically vitrified to protect from radiation damage and to withstand the high-vacuum within the electron microscope^1,27. Once well-diffracting crystals are detected, MicroED datasets are collected by exposure of the sample to an electron beam in diffraction mode during continuous rotation of the stage² (Figure 2). MicroED data are then collected on a fast camera as a movie, where each frame contains a diffraction pattern representing a wedge of the reciprocal space²⁷. Because continuous rotation for MicroED is analogous to the rotation method in X-ray crystallography, the data collected can be directly processed by existing standard software such as Mosflm²⁸, XDS²⁹, DIALS³⁰, SHELX³¹ and HKL2000³² (Figure 2).

Figure 2. MicroED workflow.

From left to right. Crystals are grown, harvested, and placed directly on EM grids. Manual or automated screening to assess for electron diffraction of the crystals. When a well diffracting crystal is identified, a diffraction dataset is collected by continuous rotation. The diffraction dataset is then processed using standard crystallography software leading to model building and refinement.

After data-processing, the phases are determined and structures are built using the density maps¹. Given its wide-application and ability to extract structural information from nanocrystals, often a billionth the volume of those needed for X-ray crystallography, there are a growing number of structures determined by MicroED. Since its inception in 2013, there are close to 100 PDB entries produced by MicroED, with the highest growth in the just the past two years. To date, several laboratories have published MicroED studies and the number of practitioners is growing. Still, there are several challenges that lay ahead for MicroED including additional methodologies for sample preparation and technological developments for data collection that currently limit widespread usage of the technique. Below we discuss the most recent developments and strategies to expand the use of the MicroED.

Protein Structure Determination by MicroED

MicroED was first developed and demonstrated on proteins^1,26 and later applied to a cohort of diverse samples. Formation of large crystals continue to be the most challenging and time-consuming step for X-ray crystallography and neutron diffraction, especially for membrane proteins and protein complexes³³. The small crystals that are typically formed by membrane proteins and protein complexes can often diffract electrons using very low exposures to minimize radiation damage (0.01 e–/Ų)¹. One of the earliest membrane protein structures determined by MicroED was the Ca²⁺ ATPase (PDB 3J7T/U)³⁴ (Figure 3A). The Ca²⁺ ATPase structure illustrated the utility of MicroED for generating Coulomb potential maps to detail information about the charged-states of amino-acid sidechains, cofactors, metals, and ligands³⁴. Since Ca²⁺ ATPase, there have been several important structures determined by MicroED including the non-selective sodium-potassium (NaK) channel (PDB 6CPV)³⁵ and the complex of the transforming growth factor beta paired type II (TGF-βm:TβRII) (PBD 5TY4)³⁶ (Figure 3A–B). The MicroED structure of NaK was similar to the previously determined X-ray structure of NaK³⁷ but like the Coulomb maps for Ca²⁺ ATPase, the structure of NaK by MicroED allowed generation of density maps to unambiguously place Na⁺ within the ion channel and to capture a new state³⁵. The heterodimeric complex between TGF-βm and TβRII plays essential roles in the adaptive immune response and maintenance of the extracellular matrix³⁸. Unlike Ca²⁺ ATPase and NaK, which form nanocrystals, the structure of TGF-βm:TβRII were obtained from fragmentation of large, imperfect crystals (discussed below)³⁶, expanding the application MicroED studies to include much larger crystals. Compared with the parent crystals, this approach led to better data by MicroED, and ultimately atomic-resolution structures³⁶.

Figure 3. Representative MicroED structures of membrane proteins, protein complexes, small-molecules, and natural-products.

(A) MicroED structures of the membrane proteins rendered as cyan and green ribbon for Ca²⁺ ATPase (PDB 3J7T/U)³⁴ and NaK (PDB 6CPV)³⁵, repectively. (B) MircoED structure of the TGF-pm:TpRII protein-protein complex (PBD 5TY4)³⁶. Protein rendered as pink and magenta ribbon for TGF-pm and TpRII, respectively. (C) MircoED structure of the HIV-GAG-bevirimat rendered as blue ribbon (PDB 6N3U)³⁹. (D) A galley of small-molecules with structures determined by MircoED. (E) Examples of natural-products with structures determined by MircoED.

MicroED in Drug Discovery

MicroED already made important contributions to drug discovery by determining structures of protein-drug complexes and supra-resolution of small-molecules and natural-products, often directly from powders without crystallization. The MicroED structure of HIV-GAG, which plays important roles in the life-cycle of HIV, was solved in complex with the antiviral drug, bevirimat (PDB 6N3U)³⁹ (Figure 3C). The HIV-GAG-bevirimat complex provided important information about the antiviral drug mechanism and was the first demonstration of drug discovery using MicroED, laying the foundation for its therapeutic development.

MicroED was originally intended for studying protein assemblies^1,26, however, it was rapidly recognized that this technique is a powerful tool for the characterization of small-molecules and natural-products. In 2016, the structure of the sodium channel blocker carbamezapine was determined to ~1 Å resolution⁴⁰. In 2018, a method for small molecule sample preparation using powder to structure was described for carbamezapine³³ and later expanded to several small organic molecules⁴. In addition, the structure of MBBF4⁴¹, a methylene blue derivative with wide medical applications including as photo-activatable antimicrobial agent⁴², was also solved. Since then, several structures of small-molecules were reported by MicroED. These MicroED structures include Grippostad⁴¹, an antiviral drug for treatment of the common cold and the flu⁴³, and a recent example of the non-fulleren acceptor (NFA) semi-conductive material ITIC-Th (Figure 3D).

MicroED has also proven its usefulness for the structure determination of several natural-products that have previously been challenging or, in some cases, impossible to determine by other techniques. Unlike their synthetic small-molecules counterparts, biosynthesized natural-products are typically larger, structurally dynamic, obtained in small amounts, and are difficult to crystallize, posing considerable challenges for X-ray studies. Even when natural-products do crystallize, these crystals are often too small and are not useful for X-ray diffraction^44,45. Brucine is an alkaloid toxin currently being tested for its anticancer properties⁴⁶ (Figure 3E). The MicroED structure of brucine at 0.9 Å resolution allowed for definitive assignment of its two chiral centers, key for understanding its toxicity and anticancer properties⁴ (Figure 3). Brucine, while large compared to small-molecules, is relatively small compared to amino-acid derived natural-products called ribosomally synthesized and post-translationally modified peptides (RiPPs), including 3-thiaGlu⁴⁴ and thiostreptin⁴⁷ (Figure 3E). Glutamylated thiols, similar to the peptide modification of 3-thiaGlu, have been shown to block jasmonate and ethylene signaling pathways⁴⁸. Thiostreptin is an antibiotic currently used in veterinary medicine⁴⁷ (Figure 3E). When efforts failed by X-ray crystallography, MicroED readily provided a 0.9 Å resolution of the 3-thiaGlu peptide (PDB 6PO6)⁴⁴. While thiostreptin has been studied by NMR⁴⁴ and X-ray crystallography⁴⁹ previously, the ease of its characterization speaks to the robustness of MicroED for structural characterization of large, flexible natural-products (Figure 3E). Like the difficulties encountered for 3-thiaGlu, the structures of 3-substituted oxindole derivatives (Figure 3E), which contains a new stereocenter at the γ carbon installed by an enzyme through directed-evolution, was only solved with the application of MicroED⁴⁵. These recent studies demonstrate strategies for determining the absolute configuration in small molecules based on an internal marker^44,45.

Strategies for Preparing Large Crystals for MicroED Experiments

Electrons interact much more strongly with material than X-ray²³. This phenomenon, however, results in high absorption and, thus, electrons can only penetrate very thin materials. Thick crystals that are >500nm must be thinned before MicroED data can be collected³⁶. There are two strategies for trimming large crystals to thicknesses suitable for MicroED including mechanical fragmentation (typically by sonication, vigorous pipetting, or vortexing)³⁶ and milling with a focused ion beam (FIB)^50–52. Fragmentation has been successful for determining protein structures from large crystals of lysozyme, TGF-βm TβRII, xylanase, thaumatin, trypsin, proteinase K, thermolysin, and a segment of the protein tau³⁶. Moving forward, the most current and promising technique for trimming large crystals for MicroED is FIB milling^50–52. During FIB milling, a crystal is repeatedly exposed a gallium beam to trim away the surrounding materials and generate lamellas with controllable thicknesses. As proof of principle, the structure of several proteins, including lysosome and proteinase K, have been determined by FIB milling and MicroED^50–54 (Figure 4). FIB milling crystals is a relatively slow process but even then about 10 crystals can be prepared per day.

Figure 4. FIB milling of crystals for MicroED.

(A) Image of select proteinase K crystals at high magnification before milling. The arrow indicates the crystal that was milled in (C). (B) FIB image of select crystal from (A) after milling the top of the crystal. (C) FIB image of crystal after milling and cleaning both the top and bottom of the crystal leaving a lamella indicated by an arrow. (Reprint from reference 52).

Out Running Radiation Damage and Structural Dynamics Using Fast Cameras

Radiation damage in structural studies continue to be a major challenge leading to poor processing statistics and map quality^55,56. When electrons penetrate the sample they deposit energy and this energy deteriorates the sample, referred to as radiation damage. Radiation damage can be categorized into two forms: global and site-specific damage. Global radiation damage typically results in the disruption of the crystal lattice and can be detected during data-processing when decreases in overall diffraction intensities and increases in B-factors are observed^57,58. On the other hand, site-specific radiation damage is not uniform, is not typically detected during data-processing, and observable only during examination of the real-space map⁵⁹. The degree of radiation damage depends, among other things, on the content of the sample, the surrounding solution, and is proportional to the amount of energy used during diffraction studies. For MicroED, site-specific radiation damage has been illustrated to occur on specific amino-acids including cysteines, glutamate, and aspartic acids⁶⁰. To curb the effects of radiation damage, samples are often vitrified for cryoEM studies⁶¹. However, even the combination of vitrification with exposure to extremely low doses of electrons (0.01 e–/Å2/s) during MicroED experiments can still lead to observable radiation damage⁶⁰.

TEMs for cryo-EM studies are typically equipped with highly-sensitive direct-electron detectors designed for optimal imaging^9,14,62. These highly sensitive cameras, however, have not been used extensively for MicroED because of concerns of damage to the sensors. As such, MicroED data are typically collected on indirect-electron detectors such as the complementary metal oxide semiconductor (CMOS)-based CetaD and TVIPS TemCam-F416 cameras. This strategy, however, limits the availability of MicroED because most TEMs are typically outfitted with top-of-the-line direct-electron detectors and not CMOS cameras. To expand the use the TEMs, the Falcon III direct-electron camera was tested for MicroED data collection⁶³. These studies demonstrate that MicroED data collected at lower electron exposure, to avoid camera damage of Falcon III, lead to greater mean completeness relative to CMOS detectors. As proof of principle, examination of maps of proteinase K from data collected on the Falcon III camera preserved the disulfide bonds (which are highly susceptible to radiation damage)⁶³ (Figure 5A). A similar approach was used recently with a Gatan K2 direct-electron detector in counting mode with an exposure 25 times less than for the Falcon III detector and a seemingly damage-free structure of Proteinase K has been determined.

Figure 5. MicroED radiation damage and experimental phasing.

(A) Disulfide bonds of the proteinase K structures determined from data collected on Falcon III and CetaD. The density around the two disulfide bonds indicates increasing radiation damage as an effect of increasing dose. The positive difference density around Cβ of Cys283 in the CetaD data (black arrow) indicates a partially dislocated S atom. The 2mFo – DFc densities (blue meshes) are contoured at 1.5σ (Reprint of reference 63). (B) Fourier difference maps between the damaged and undamaged structure of the GSNQNNF peptide using the phases at 2.5-Å resolution, contoured at 3s (Top left). Maps of the experimental phases of the peptide extended to 1.4 Å (Top right). Maps of an intermediate model-building step (Bottom left) to generate the final peptide structure at 1.4 Å (Bottom right). Reprint from reference 64.

Outrunning radiation damage using the direct electron detector Falcon III has been instrumental in determining the structures of samples that are highly susceptible to damage^44,45 and we believe that many more such examples would be forthcoming. Radiation damage was recently used to establish a pipeline for phasing MicroED data⁶⁴ (Figure 5B). A low-damage followed by a high-damage data sets were taken from the very same crystal. A difference Patterson was calculated and allowed for generation of initial phases. Following cycles of model building and refinement, the structure of a protein peptide was determined (Figure 5B). This study demonstrates the ability to extract meaningful phase information using radiation damage in MicroED.

Concluding Remarks

MicroED is proving to be an important new tool in structural biology not only in determining structures of proteins but also of peptides, small organic and inorganic molecules, and natural-products. Continuous rotation MicroED is paradigm shifting because it has proven to be a robust, fast, and efficient method for determining the structures of small-molecules, natural-products, and semi-conductor materials without crystallization and directly from mixtures⁴. To our knowledge no other structural biology method is capable of determining atomic resolution structures directly from mixtures, making MicroED a useful and powerful tool for an array of problems that are yet to be explored. Moving forward, the application of FIB-milling and fast-cameras will certainly expand MicroED for structure determination of varying types samples with a wide-range of crystal sizes and facilitate time resolved studies. These advancements could ultimately be applied to establish automate pipelines that mirror those for X-ray crystallography to facilitate obtaining MicroED structures. Early examples of automation in MicroED has already reported in which several hundred complete data sets could be collected automatically overnight to demonstrate similar throughput as synchrotrons⁶⁵. The application of these advancements would make MicroED more accessible for widespread use. ^66,67