Electron diffraction and crystal imaging methods

cryo-em structure refinement and validation

protein design and scaffolding for cryo-EM imaging methods

structural biology of protein condensates in plants

electron diffraction and crystal imaging methods

Our research aims to expand the utility of electron diffraction and crystal imaging methods in collaboration with DOE/national facilities. Our teams work to facilitate data collection pipelines for BER user communities, while establishing methods to systematically probe microcrystals grown from protein targets that play a critical role in the microbial or plant projects.

The structural elucidation of macromolecules that either cannot form crystals, or form nano-scale crystals is of growing interest. Nanocrystallography has been facilitated by kilometer-long high-powered x-ray lasers that yield atomic structures from slurries of thousands to billions of crystals. On the laboratory scale, electron diffraction methods have determined protein structures from single similarly sized crystals. As such, electron diffraction excels at determining structures from certain types of crystalline macromolecules but faces major challenges in its application to DOE-relevant targets.

UCLA-DOE scientists are actively developing approaches to overcome these limitations and speed throughput of crystal screening and data collection. UCLA-DOE scientists are likewise developing new hybrid imaging-diffraction methods to obtain structures from ever smaller regions of crystals, or from partially disordered crystals by imaging.

In partnership with various DOE national laboratories, UCLA-DOE scientists are expanding on existing tools for 4DSTEM to determine accurate protein structures from 4DSTEM tilt series and are complementing electron diffraction measurements with tomography tilt series of crystals to provide information from partially ordered crystal domains. In conjunction, UCLA-DOE scientists are expanding the capabilities of existing software for fragment-based molecular replacement of data that otherwise yields no viable solutions.

We are working to establish electron microscopy and diffraction as enabling capabilities for the ab initio determination of protein structures from nanocrystals.

Despite the recent revolution in technologies driving the elucidation of small and large molecular structures, the atomic structures of many categories of molecules remain out of reach. Proteins that are too small for direct electron imaging via single particle methods and that fail to form large crystals remain a challenge. In addition, the determination of ligands or other small molecular partners bound to small proteins remains out of reach by conventional EM methods.

Our teams are enabling the use of electron microscopy and diffraction methods to interrogate nanocrystals formed by these molecules and complexes via three modalities: i) the application of electron nanodiffraction combined with energy filtering and serial automated processing, ii) the implementation and robust application of fragment-based molecular replacement for ab initio atomic structure determination from electron diffraction data, and iii) the use of tomographic reconstruction to obtain structural details from semi-crystalline molecular nanoarrays.

UCLA-DOE scientists have recently yielded successful solutions for a 1.6 Å MicroED dataset of Proteinase K from distant homologues and various small peptides. The approach is implemented using the CCP4 ARCIMBOLDO program that substitutes the atomicity constraint underlying direct methods with stereochemical constraints. UCLA-DOE scientists have established protocols to grow crystals in bacteria and extract them in large quantities. By both negative stain EM and cryoEM imaging UCLA-DOE scientists have demonstrated that the lattice structure of protein crystals remains intact after purification. Fourier transforms of crystal images show that the lattice remains ordered out to at least 50 Å resolution; ongoing improvements aim for atomic resolution detail.

Left: Image of a protein crystal from a tilt series (acquired at PNNL).
Middle: Two-dimensional class averages obtained from overlapping windows within the image on left, showing intensity variations consistent with protein lattice features.
Right: Fragment-based phasing of electron diffraction structures; two example structures are shown, both of short peptides. The approach is expected to generalize to small and large proteins including microbe-associated enzymes and plant miniproteins.

cryo-EM structure refinement and validation

We continue to develop web-based tools for cryo-EM structure refinement and validation, which will include servers for assessment of atomic charges and instrument parameter calibration.

The recent surge in cryo-EM research activities brings new challenges to the fore. Several key gaps have arisen related to best practices for atomic structure refinement and validation by cryo-EM and electron diffraction. One of these concerns the need for accurate scattering factors for electron diffraction studies; current practice relies on form factors optimized for x-ray scattering, which are defective in modeling scattering from charged amino acid residues.

Optimized electron scattering equations will provide critical improvements in high resolution modeling.  We are developing and providing web-based access to new algorithms for treating electron scattering.  These advances will aid research progress toward an atomic level understanding of molecular systems.  On the broadest scale, our team is developing tools that will enhance the technical quality of EM research across national laboratories and the community at large.

Single particle cryogenic electron microscopy (cryoEM) and continuous-rotation electron diffraction (MicroED) are techniques of high utility for structural elucidation of macromolecular assemblies. The two methods rely on data collected using transmission electron microscopes to recapitulate 3D maps of electrostatic potential (ESP).

These representations of macromolecules are derived from interactions between a relatively coherent and parallel, high-energy (100-300keV) incident electron beam and a macromolecule under interrogation. The UCLA-DOE institute recognizes an important opportunity in facilitating rigorous analysis and refinement of cryoEM structures and is developing publicly accessible tools to facilitate this goal. Our enabling capabilities team is developing a server to aid in atomic charge assignment for cryoEM maps. In addition, we are implementing new tools to assess and correct cryoEM and X-ray map distortions due to inaccuracies in experiment geometry during data collection.

Examples of calculations performed based on parametrized electron scattering factors (right). The energy-dependent elastic scattering cross section for Carbon is shown (left) alongside specific elastic cross sections depicted for 100,200 and 300keV (middle).

Protein design and scaffolding for cryo-EM imaging methods

We are applying new Darpin-selection and cryo-EM scaffolding methods on protein structural targets identified in the Microbes and Plants projects. This novel method will produce near-atomic resolution reconstruction of target proteins.

One of the most pressing limitations in the field of cryo-EM is the challenge of determining structures of smaller proteins or nucleic acids.  Small molecules less than approximately 50 kDa pose a challenge due to complex issues of signal-to-noise, leaving a large fraction of cellular components outside the reach of this powerful technique.  With DOE funding, our team has developed the leading technique to overcome this size barrier, with relatively minor improvements remaining before application to far-reaching problems.  We will apply our protein engineering tools to design specialized scaffolds for determining the structures of key DOE-BER proteins in the sub 50kD size range by cryo-EM.  Targets for these first-of-kind structure determination studies will be based on collaborations internally (with the Microbes team) and with multiple DOE-BER supported laboratories. Success in pilot projects will lay the groundwork for future pipelines for more routine determination of atomic structures by cryo-EM, especially where flexibility presents challenges for crystallography, even for smaller proteins.

In 2018 our team developed the first workable scaffold for near-atomic imaging small proteins by cryo-EM, with subsequent studies and a review following.  The crux of our solution is the use of our designed protein cages as the scaffold core, to which we fuse a Darpin (designed ankyrin repeat protein) as a modular binding domain (see figure).  Darpins have been developed as a facile platform for generating and selecting mutant versions (in loops connecting helical repeats) that bind essentially any protein of interest using various laboratory evolution methods (phage/yeast/bacterial display).  Besides adding the required size and favorable symmetry, the system overcomes the two hurdles that have prevented prior success in scaffold development: rigidity and modularity.  We chose Darpins over other immunoglobulin-type systems (i.e., antibodies) because their helical nature provides a mechanism for relatively rigid attachment to the designed cage core through a continuous/shared helical segment.  The resolution demonstrated so far is 3.8 Å for the bound target.  The next steps in developing this system are to improve rigidification and to demonstrate useful imaging applications (the proof of principle study imaged GFP as the cargo).  While applications of biomedical interest are the domain of other agencies (NIH), the aim of the present work is to develop use cases for proteins and enzymes in the bioenergy space to advance DOE science.

Our proof of principle work relied on imaging test cargo proteins (e.g., GFP) for which cognate Darpin sequences had already been published. There are many [>100] known Darpin sequences selected for binding diverse targets , but our applications to DOE-mission-relevant targets requires development of an in-house pipeline for selecting Darpins for binding specific targets. In that effort, so far we have created a Darpin library (with variable loops) and piloted a bacterial-display syste

To reach higher resolution, we will use next-generation scaffolds we have recently developed (unpublished) in which the protruding Darpin appendages are rigidified by bridging interactions.

Using newer scaffolds under development, we hope to break the 3 A resolution barrier for the first time.  If this is possible it will set another landmark for imaging small proteins by EM, with far-reaching impact.

Success in individual projects will advance our atomic understanding in specific areas related to cellulosic formation/degradation.  This will impact the progress of internal institute projects (in the Microbe and Plant areas) as well as projects in collaborating BER funded laboratories (CCRC and UCSB).

Designed scaffolds developed with DOE funding for imaging small proteins by cryo-EM. (A) Diagram highlighting the protein construction methods used to create rigid and modular scaffolds for displaying diverse proteins for imaging (Liu et al. 2018). (B) First results on a test target scaffold protein (GFP), reaching a new benchmark of 3.8 Å for a small protein (Liu et al. 2019).

structural biology of protein condensates in plants

A new and rapidly expanding area of molecular and cellular biology is liquid-liquid phase separation (LLPS), also known as condensates or membraneless organelles. In a wide variety of normal metabolic states, as well as under stress conditions, cells undergo phase separation in which certain cellular components, often protein-RNA complexes, are found at high concentrations in subcellular bodies. Stress conditions giving rise to LLPS include temperature stress, chemical stress, dehydration stress, and viral infection. The structural basis of LLPS has just started to come to light through applications of crystallography, cryo-EM, and computational methods. LLPS is driven by multivalent interactions among proteins and nucleic acids, in some cases in the nucleus and in other cases in cytoplasm. LLPS could possibly be driven by protein-glycan interactions in cases where glycan polymers exist. The protein-protein interactions are generally between low-complexity domains (LCDs). The LCDs form pairs of interacting beta sheets, with both similarities and differences with amyloid-like organization. One type of adhesive segment found in LCDs is called a LARKS, an acronym for Low-complexity Amyloid-Like Reversible Kinked Segments. These have been mapped in genomes by a computer algorithm that has been applied to yeast, microbes, and other model organisms. UCLA-DOE scientists will apply novel predictive algorithms to identify LLPS-like structural motifs in the genomes of terrestrial plants, including those important for biomass production (plant LLPS/amyloid server).

Based on these predictions, UCLA-DOE scientists, in collaboration with our Plant Team, will interrogate structures of interacting low-complexity domains by cryoEM and of their adhesive segments by crystallography. These studies will lead the way to deeper understanding of plant metabolism and ways of regulating it. SAURs, ARFs, NAC-domain, and MYB-domain proteins are classes of proteins that may form LLPS as part of their regulation, which are also essential for plant growth and development. Expanded networks are expected to reveal novel insights into the regulation of growth, development, and cell wall biosynthesis. The role of phase separation dynamics in regulating processes relevant for feedstock and biofuels production will be a critical extension of our understanding of transcriptional regulatory networks and structures.

Plant LLPS/amyloids are anticipated to exist and perform a myriad of roles in cell biology. The set of LLPS/amyloid-forming proteins in plants is not fully defined but likely intersects with the set of plant prion proteins and condensate-associated proteins, both of which might adopt amyloid-like structures. A single pair of mating beta strands can facilitate amyloid formation, while kinked sheets might facilitate reversible amyloid folds. LARKS, adhesive segments, often found in low-complexity regions of proteins are among the types of structures anticipated to drive amyloid-like phase separation in biological systems. An abundance of LARKS in a protein low-complexity domain suggests its capacity to undergo liquid-liquid phase separation.

A web server LARKSdb has been constructed to permit users anywhere to predict the locations of LARKS in six-residue windows within protein sequences of interest. LARKSdb is a service of the UCLA-DOE Institute found at LARKSdb. A similar server predicts six-residue windows likely associated with pathogenic amyloid folds (ZIPPERDB). However, despite success in predicting pathogenic and low complexity amyloid forming segments (zipperDB, larksDB) no structure-based prediction method has been proven to confidently predict functional LLPS/amyloid-forming sequences in plants.

We hypothesize that plant LLPS/amyloid structures are distinct from those of pathogenic amyloids and instead resemble functional amyloids/prions. The structural motifs observed in functional amyloid fibrils to date can therefore be used to predict the propensity of plant sequences to form functional amyloids with a high degree of accuracy. Importantly, structures of post-translationally modified amyloids can serve as scaffolds for glycan-interacting motifs, or other modifications.