Principal Investigator: Rajendra K. Agrawal, Ph.D.

Protein synthesis (or translation) is a fundamental process in all forms of life. An understanding of the molecular mechanisms underlying the various steps of translation is of seminal importance in a large range of fields in Biology.

We study the structure and function of ribosome, the protein synthesizing machinery of the cell and cellular organelles such as mitochondria, using cryo-EM single-particle reconstruction techniques (Frank et al., 2000; Shaikh at al., 2008). The current focus in the Agrawal laboratory is to structurally characterize the mammalian mitochondrial translation by studying molecular interactions between the mitochondrial ribosome and its ligands during all steps of protein synthesis, at 5-8 Å resolution in each case. The goal are to come to an understanding of (i) the unique features of the organellar riboosmes in general, and mitochondrial ribosomes in particular, as molecular machines, and (ii) their structural evolution from the primitive endosymbiotic bacteria, over the course of 2 billion years, to its current form within the mitochondrion of various cell types/organisms.

Cryo-EM single-particle reconstruction has evolved to become the most powerful approach to study ligand binding and conformational changes accompanying translation. Recent reviews (Agrawal et al., 2011, Agrawal and Sharma, 2012; Frank, 2012) gives only a glimpse of many results that have accumulated since 1996, when tRNA binding to the ribosome was first visualized (Agrawal et al.,1996). The EF-G mediated tRNA translocation was also first visualized by cryo-EM (Agrawal et al., 1998; 1999; Frank and Agrawal, 2000), and then main sequence of events during the translation elongation cycle were visualized (Agrawal et al., 2000). These early studies laid much of the ground work for several functional ligand-binding studies that subsequently followed during the past decade in the Agrawal laboratory and many other laboratories.  The Agrawal laboratory obtained the first cryo-EM structures of (i) the mammalian mitochondrial ribosome (Sharma et al., 2003), (ii) chloroplast ribosome (Sharma et al., 2007), and (iii) a protistan mitochondrial ribosome (Sharma et al., 2009). These studies revealed several unique features of the organellar translational machinery as reviewed recently (Agrawal and Sharma, 2012; Agrawal et al., 2011). In addition, the Agrawal laboratory continues to study specific steps of cytoplasmic translation, for example, RRF- and EF-G-dependent ribosome recyling (Agrawal et al., 2004; Barat et al., 2007; Yokoyama et al., 2012), in order to understand the underlying molecular mechanisms of the process and to compare the analogous mechanisms in the mitochondrial system. Using existing tools of cryo-EM single-particle reconstruction, and tools being developed for time-resolved cryo-EM studies (e.g., Lu et al., 2009), we are pursuing studies to unravel detailed molecular mechanism of various steps of mammalian mitochondrial translation at 5-8 Å resolution. Here are examples from some of the more recently published works:

Figure 1.  Comparison between the segmented cryo-EM maps of ribosomes from (A) bacteria (70S, E. coli), (B) chloroplasts (70S, Spinacea oleracea), (C) protistan mitochondria (50S, Leishmania tarentolae), and (D) mammalian mitochondria (55S, Bos taurus). Ribosomal Proteins of small and large ribosomal subunits are colored yellow and blue, respectively, while ribosomal RNAs are colored orange and magenta, respectively. Results revealed that mitochondrial ribosome surfaces are shielded extensively by protein masses. (E, F) Secondary structures of small and large ribosomal subunit RNAs, respectively, from bacteria (grey), mammalian mitochondria (green), and L. tarentolae mitochondria (red). Structures of mitochondrial rRNAs were derived from analyses of respective cryo-EM maps.
 

Figure 2.  A side-by-side comparison of topographies of the nascent polypeptide-exit sites in 3D cryo-electron microscopic maps of the ribosome from (A) bacteria (E. coli), (B) chloroplasts (S. oleracea), (C) protozoan (L. tarentolae) mitochondria, and (D) mammalian (B. taurus) mitochondria, suggesting that the mechanism of nascent polypeptide (red) exit in mitochondria is dramatically different from those in bacteria and chloroplasts. Pink and green are conserved rRNAs and proteins, respectively. Solid blue masses represent organelle specific protein masses.
 

Figure 3.  Structure of the mammalian mitochondrial translationa initiation factor 2 (IF2mt) as derived by single-particle cryo-EM, and location of its mitochondrial-specific insertion domain relative to the binding site of bacterial IF1.  (A) Domain alignment of E. coli IF2 and bovine IF2mt. The mature IF2mt (i.e., after deletion of the mitochondrial import sequence) starts from aa residue 78. The 37-aa insertion domain in IF2mt is highlighted in red. (B) The cryo-EM map of the E. coli 70S ribosome (30S subunit, yellow; 50S subunit, blue) in complex with IF2mt (red), initiator tRNA (green) at the P/I position (P/I stands for P-site tRNA at the initiator position). (C) Fitting of atomic models of the IF2mt and initiator tRNA into the corresponding cryo-EM densities (meshwork) extracted from the map shown in panel (B). The color codes used for various domains of IF2mt are the same as in panel A. Asterisk (*) point to the region that would correspond to domain III of IF2mt but was not modeled. (D) Binding positions of the insertion domain (red), and (E) IF1 (green), into a common binding pocket on the SSU of the ribosome.  Components of the binding pocket on SSU, the 16S ribosomal RNA helices (h18 and h44) and ribosomal protein S12, are identified.

Figure 4.  Segmented cryo-EM maps of the bacterial 70S ribosome post-termination complex (PoTC) in complex with ttRRF and in complex with ttRRF and EF-G. (A) The PoTC•ttRRF complex. The 70S ribosome (yellow, 30S subunit; and blue, 50S subunit), with ttRRF (red); (B) The PoTC•ttRRF•EF-G•GDP•FA complex, shown in a matching view as in panel A. Densities for both ttRRF (red) and ecEF-G (orange) are readily identifiable. Position occupied by domain II of ttRRF in panels A is now occupied by domain V of EF-G. In both complexes density corresponding to mRNA (dark blue) can also be seen. (C) Molecular interactions between the ribosome and ttRRF in the PoTC•ttRRF complex. (D) Molecular interactions between ttRRF and EF-G within the PoTC•ttRRF•EF-G•GDP•FA complex. Thumbnails to the lower left of panel C and lower right of panel D depict overall orientations of the ribosome in those panels.

Principal Investigators: Terrence Wagenknecht, Ph.D. (skeletal muscle), Zheng Liu, Ph.D. (cardiac muscle)

Introduction: RyRs function as intracellular calcium release channels that are particularly prevalent in muscle where they are associated with the sarcoplasmic reticulum (SR). RyRs are homotetramers constructed from a subunit of molecular mass 565 kDa, making these receptors the largest known ion channels (Samso and Wagenknecht, 1998; Fill and Copello, 2002; Wagenknecht and Samso, 2002). The Wagenknecht and Liu laboratories have been engaged in structural studies of isolated RyRs from skeletal and heart muscle by cryoelectron microscopy and three-dimensional reconstruction (Liu et al., 2004; Radermacher et al., 1992). 3D reconstructions have shown that about 80% of the mass of the RyR is located in the cytoplasm as a complex assembly of ~10 distinct domains (Radermacher et al., 1994). Besides their function in releasing calcium, which is the immediate stimulus for muscle contraction, RyRs play a key role in excitation-contraction (E-C) coupling. E-C coupling refers to the process by which neuron-induced depolarization of the muscle plasma membrane (sarcolemma) and its invaginations known as transverse (t)-tubules leads to release of Ca2+ from the SR via the RyRs. Perhaps the most important of the sarcolemma/t-tubule components is an L-type calcium channel known as the dihydropyridine receptor, which serves as the voltage sensor in e-c coupling (Rios and Pizarro, 1991;Jones, 2002). Interactions between the dihydropyridine receptor and the cytoplasmic region of the RyR are thought to mediate communication between the plasma membranes/transverse tubules and the transmembrane, ion-conducting region of the RyR in the SR. The figure below shows schematically the relationship of RyRs to DHPRs in skeletal muscle as well as some of the other protein components of the triad junction (CaM, calmodulin; FKBP, FK506-binding protein)

Figure 1.  Triad junction schematic showing major components

Single-particle cryo-EM of Ryanodine Receptors: The RyR is a ligand-gated channel. Physiologic ligand activators of the skeletal RyR include the dihydropyridine receptor (the voltage-sensor), Ca2+ at micromolar levels, and millimolar ATP. Inhibitors include Ca2+ and Mg2+ at millimolar concentrations. In a pioneering study that employed time-resolved cryo-microscopy, Unwin demonstrated that the acetylcholine receptor, also a ligand-gated ion channel, could be imaged in two of its functional states, a closed and an open state (Unwin, 1995). Time-resolved techniques were required because when the acetylcholine receptor is induced to open by physiologic agonists (in this case acetylcholine), the open state is short-lived, lasting only 50-100 ms before assuming a desensitized (inactivated) state. Efforts to “lock” RyRs in functional states using various chemical treatments for structural analysis have produced tentative models, but their physiologic significance is not known (see Figure below and Samso et al. 2009).  We are developing new time-resolved cryo-EM methods to RyR (Lu et al., 2009; Shaikh et al. 2009) that are sufficiently efficient, in terms of quantity of sample required and distribution of the specimen on the grid, so as to be applicable to characterizing discrete functional states of RyRs (e.g., closed, open, inactivated or adapted) " (Schiefer et al., 1995; Laver and Lamb, 1998; Valdivia et al., 1995;  Gyorke and Fill, 1993) ; Lamb et al., 2000; Fill et al., 2000; Wang et al. 2004).

Figure 2.  Open and closed states of the ryanodine receptor

RyRs are known to interact directly with several proteins in skeletal muscle to form a signal-transducing complex known as the triad junction. One of our goals is to use the isolated RyR as an "assembly platform" to prepare complexes containing each of these ligands. Progress has been made in our lab and in others to determine the binding locations of ligands such as, FK506-binding protein (Wagenknecht et al. 1997; Samso et al. 2006), calmodulin  (Wagenknecht et al., 1997; Samso and Wagenknecht, 2002; Huang et al., 2012), and CLIC2, (Meng et al. 2009). Of particular interest is the interaction of the RyR with the dihydropyridine receptor (Samso et al.,1999), which is thought to represent the core of the signal-transducing apparatus of e-c coupling. Sequence-specific antibodies represent another class of ligand that is being used to probe the structure of the RyR by means of in vitro assembly and 3D cryo-microscopy.

We have developed a cryo-EM based technique to locate, with nanometer-precision, specific amino-acid residues in the 3D structure of RyRs by site-specific insertion of green fluorescent protein (GFP) into RyR’s sequence followed by expression in cultured cells, affinity purification of RyRs, and cryo-EM (Liu et al. 2002, 2004).  The inserted GFP typically appears as roughly spherical region of excess density attached to the surface of one of the domains comprising RyR.  More recently, dual insertions of the GFP-related fluorescent proteins, CFP and YFP, have allowed the construction of experimental systems for investigating fluorescence resonance energy transfer between CFP and YFP in cultured cells expressing the recombinant RyRs (Liu et al. 2010; Wang et al. 2011). The CFP/YFP pairs allow dynamic FRET experiments that have detected interdomain movements of the CFP- and YFP-labeled domains when the RyR is induced to switch from the closed to the open states.  The figure below shows a composite of several localization studies by the GFP approach and direct mapping of RyR-specific ligands.

Figure 3.  Site-specific mapping by 3D cryo-EM. The green spheres show location of a phosphorylation site that was labeled with a GFP inserted after amino acid Tyr-2801 in the sequence or heart RyR.  A monoclonal antibody ligand that is specific for an epitope nearby to Tyr-2801 is shown in orange, and the location of bound FK506-binding protein is shown in red (Liu et al. 2010). The three views are: left, top view from the cytoplasmic surface, which in situ would face the transverse tubule; middle, side view; right, bottom view showing the surface that would face the SR lumen. The numerals 1–11 on the cytoplasmic assembly indicate the distinguishable structure domains, according to our earlier nomenclature.  Each ligand is present four times due to the homotetrameric nature of RyRs. The length of an edge of RyR in its square-shaped orientations is 280 Å.

Cryo-electron tomography of triads: Work in other laboratories has described the isolation of membrane fractions from skeletal muscles that contain functionally intact triads (e.g., Ikemoto et al., 1994).  Low resolution microscopy on these preparations shows the presence of intact triads (i.e. sarcoplasmic reticulum-derived vesicles joined to transverse-tubule-derived vesicles with RyRs visible in the intermembrane space in the junctional regions) (Kim et al., 1990), and cryo-EM of frozen-hydrated triads yields images in which the bridging RyRs are directly visible. We determined the first tomograms from frozen-hydrated isolated triad junctions which are illustrated in the figures below (Wagenknecht et al., 2002; Renken et al., 2009).  The reconstructions resolved some of the individual ryanodine receptors and revealed their relationship to the dense mat of luminal density attributed to calsequestrin that underlies the RyR-enriched regions of the SR. Frozen-hydrated triads should be a readily amenable to the application of high-precision immuno-EM to identify protein components that are resolved in the tomograms. Currently we working to determine cryo-tomograms from thin sections of frozen-hydrated, structurally intact muscle which are being prepared by a new technique, focused ion beam milling (FIB), that is being developed in-house (Marko et al., 2007)

Figure 4.  Selected examples of frozen-hydrated isolated TC vesicles and triads. A) Selected examples of TC vesicles.  In middle panel, a segment of a contaminating collagen fiber is present above the TC vesicle. The dense material in the lumens of the vesicles is attributed to calsequestrin. In some preparations, exemplified by rightmost panel, some of the vesicles contained an apparently organized form of calsequestrin. Arrowheads indicate locations of cytoplasmic regions of ryanodine receptors (feet). (B) Three isolated triads.  The TC vesicles and TT are indicated in the panels. Ryanodine receptors are sometimes resolved within the junctional gaps (middle panel). [from Wagenknecht et al., 2002].

Figure 5.  Surface-rendered representation of triad. SR- and TT-derived vesicles in green and red,  respectively. Yellow spheres in lumen of SR represent calsequestrin and continuous yellowslabs near junctional surfaces of SR represent the condensed calsequestrin.  Blue structures correspond to feet/RyRs. Magenta particles in lumen of TT are of unknown identity.[from Wagenknecht et al, 2002]

Figure 6.  Partial model for the triad junction ‘‘couplon” based on cryo-electron tomography. Top: view from abovethe SR membrane (light blue) showing RyR (yellow), calsequestrin (dark blue), secondary SR membrane densities (green), and cytoplasmic TT-associated densities (pink).  Bottom: Side view of the couplon model showing expected location of the TT membrane (black) and tethers (dashed lines) inferred from the class averages.[from Renken et al., 2009]

Principal Investigators: Michael Koonce, Ph.D. (Microtuble motors); Haixin Sui, Ph.D. (Microtubule complexes) Bruce McEwen (Microtubules at the kinetochore)

Microtubules and microtubule-based complexes are essential in all types of cells.  Molecular motors on microtubules are responsible for movement of cells and within cells, from mitosis to muscle action, as well for transport of materials within cells, and they form part of the cell’s cytoskeletal scaffold.  Members of our group are studying three aspects of microtubules: (1) structure and function of microtubule-based motors, (2) assembly, maintenance, and regulation of microtubule-based complexes, and (3) interaction of microtubules with the kinetochore during mitosis.

Microtubule motors: The laboratory of Mike Koonce focuses on microtubule-based motors, the dynein and kinesin isoforms that participate in organelle transport and cell division.  Our work utilizes combinations of gene knockouts, live cell imaging, and electron microscopy to investigate cell processes that are disrupted as a consequence of motor malfunction.  Characterization of dynein has produced the first detailed structural models of how the motor may work and has revealed that dynein plays an important role in organizing the interphase microtubule network.  Further analysis of kinesin motors has identified new functions in microtubule organization and in connecting the cell centrosome with the nucleus.  The work is performed in Dictyostelium discoideum, an outstanding model organism for cell motility studies.

Microtubule complexes: The Haixin Sui laboratory studies functional mechanisms of cytoskeleton assemblies in cells and organelles using structural imaging methods.  Through cryo-EM and image processing, the lab has achieved high-quality structural maps for microtubule singlets of multiple structural types.  We have also obtained the molecular architecture of microtubule doublets.  These results have provided structural insight into their assembly and mechanical properties.

Microtubules (singlets): The microtubules are critically important in maintaining shapes of cells and organelles (cilia), and in ensuring correct dell division. They also serve as the railways for motor-based intracellular transportation.  All these functions induce mechanical stress on microtubule structure and require a combination of rigidity and flexibility.  The high-quality maps of various types of microtubules have been achieved at a resolution better than 9A, which enabled us to construct reliable structural models.  Comparison of these structures offers the following insight into the lateral deformation and bending property of microtubules.  (1) The loop-loop interaction offers flexibility for microtubule lateral deformation.  (2) The molecular configuration of protofilament lateral interaction limits the degree of lateral deformation.  (3) Local structures surrounding the inter-protofilament interaction stabilize the loop interactions and contribute to the rigidity of microtubules.  (4) The seam is not significant in the mechanical properties of microtubules (Sui and Downing, 2010).

Figure 1.  Structures of various types of microtubules: (a) Density maps of different structural types of microtubules containing protofilament numbers of 11 (pink), 12 (orange), 13 (yellow), 14 (green), 15 (blue), 16 (purple). These density maps were reconstructed from cryo-electron micrographs of microtubule mixtures obtained by polymerization of purified tubulin. The background image represents the typical morphology of the frozen-hydrated microtubule mixture with microtubules of various diameters. (b) The density maps clearly revealed secondary structural features and enabled constructing structural models of microtubules with good accuracy.

Microtubule doublets: Microtubule doublets are usually found in arrays of nine doublets and serve as the major framework in eukaryotic cilia or flagella.  This framework contributes to the elastic property required for the mechano-sensory function of primary cilia and the beating function of motile cilia.  Both microtubule singlets and doublets are mainly composed of tubulin, but the doublets also incorporate a number of other structural-protein components.  Using cryo-electron tomography and image averaging, we obtained a 3D structural map of intact microtubule doublets at a resolution better than 3 nm.  The map disclosed an 8% lateral deformation of the complete microtubule (A-tubule) in the doublet.  The deformation suggests a flexibility of inter-protofilament interaction, which allows the microtubules to undergo a certain degree of lateral deformation. The map also revealed protein densities bound to the inside of the microtubule. These proteins, named intra-luminal microtubule-associated proteins (iMAPs; Amos 2010), are believed to play important roles in stabilizing or modifying inter-protofilament interactions and contribute to the lateral deformation of microtubules (Sui and Downing, 2006; Downing and Sui 2007).

Figure 2.  Microtubule doublet: In the left panel, the receding gray-scale image is one of the images in one tomographic tilt series. The image just below is a vertical section through the 3-D reconstruction generated from this series. The density in gold color is the resultant 3D map of microtubule doublets with the tubulin structural model fitted into it. In the middle and right panels, the density map reveals an 8% lateral distortion of the complete microtubule (A-tubule). Protein densities other than tubulins, colored yellow, are clearly revealed inside the microtubule doublets. Inside the A-tubule, intraluminal microtubule- associated proteins (iMAPs) are visualized on the intraluminal wall, where the inter-protofilament interactions locate for integration of the microtubule structure.

Three-dimensional reconstruction of single (i.e., isolated, freestanding) macromolecules from low-dose electron micrographs is a technique pioneered in Albany (Frank et al. 1978; 2006; Radermacher et al. 1987; Frank, 1996; 2006).  Briefly, molecules are flash-frozen in a thin (<1000 Å) layer of water, which becomes vitrified (Grassucci et al., 2007).  The single-particle projections visible in the electron micrograph show the molecule in many different views.  In the single-particle reconstruction approach, the relative angles between these views are determined, either by setting up a defined data collection geometry (as in the random-conical reconstruction approach; see Radermacher et al. 1987) or by making use of intrinsic mathematical relationships among different projections of the same object (common lines, or angular reconstitution approach; van Heel, 1987; Penczek et al. 1996). Based on this initial estimate of angles, a first, coarse 3D reconstruction is computed.

Next, this 3D density map is refined by an iterative technique that consists of the following steps: (i) compute a set of equispaced projections from the density map (the reference), covering the entire half-sphere of orientations; (ii) compare each of the experimental images with each of the computed (predicted) projections, to find the angle of best match; (iii) with the new angle assignments, compute a new, refined 3D map; (iv) compare the new set of angles with the previous set of angles. If the discrepancy is below a certain threshold, stop, else go to (i) for a new iteration. This approach of refinement is termed 3D projection matching (Penczek et al. 1994).

After refinement, the density maps are corrected for the effects of the contrast transfer function, which causes the Fourier components sitting in different zones of the Fourier transform to have alternate phases. Another correction concerns the falloff of the Fourier amplitudes, which is caused by instabilities of the stage and charging. This correction could be guided by experimental average amplitude curves obtained by low-angle X-ray scattering (Gabashvili et al. 2000).

The single-particle reconstruction approach makes use of the assumption that all molecules visible on the micrograph have the same structure and conformation. If this assumption is not met, then classification methods must be used to extract the homogeneous sub-populations. The most straightforward way to achieve this is by using two or more different 3D references that represent the different structures or conformations (Valle et al., 2002; Gao et al., 2004). Each particle image is then compared with the set of the references to find the one that it most closely resembles (e.g., Barat et al., 2007). This method, which is also referred to as “supervised classification”, obviously fails if the alternative reference structures are not available.  In such cases, methods of “unsupervised classification”, for example, the maximum likelihood approach (e.g., Scheres, 2010), can be useful in extracting unique conformational states.

Density maps representing molecules at 8-15 Å resolution can be interpreted by docking atomic coordinates of known components (Rossmann, 2000; Rossmann et al., 2005). For resolutions of 12 Å or better, 3D motif search based on locally normalized cross-correlation can be employed (Rath et al., 2003). Further molecular interpretation can be carried out by making use of various software packages such as “O” (Jones et al., 1991), Modeller (Eswar et al., 2006), SSE hunter (Baker et al., 2007), and molecular dynamics based flexible fitting (MDFF) approach (Trabuco et al., 2009). Refined coordinates of the fitted structures can then be further analyzed for various interactions and biological interpretations

All operations that are required for single-particle reconstruction are contained in the package SPIDER (Frank et al. 1981, 1996, 2006; Shaikh et al. 2008), which was developed in Albany. SPIDER is maintained and further developed under an NIH grant to ArDean Leith of our 3D-EM Group, and has been freely distributed to hundreds of labs under the GPLv2 license. The SPIDER website.

Single-particle reconstruction has been applied to a large number of macromolecular complexes, and has enabled the study of molecular machines engaged in many functions pivotal for the maintenance of the cell, such as DNA repair, transcription, splicing, translation, protein folding, and protein decay.  In-house projects at the Wadsworth Center benefiting from this technology are centered around the structure and function of the ribosome.

One of our primary tools is TEM cryo-electron tomography, a technique that provides high-resolution 3-D images of biological material in its native state - without structural disruption caused by chemical fixation, dehydration, or staining.  Specimens are examined in the vitreously frozen state, after use of rapid freezing methods that avoid formation of ice crystals that could damage the structure. 

In tomography, a 3D reconstruction is generated by back-projecting a series of images that are recorded in the TEM as the specimen is tilted over a wide range (Frank et al. 2006).  For vitreously frozen samples, a “low-dose” imaging protocol is used that avoids unnecessary electron irradiation of the sample, which is easily damaged by even a moderate electron dose.  We were the first US lab to present results using this technique (Rath et al. 1997; Marko et al. 1999).

Thin films of suspensions of isolated organelles, e.g. mitochondria or small cells such as bacteria, are routinely prepared by plunge-freezing directly on TEM grids.  Cells can be cultured on TEM grids, and if thin enough, they can also be plunge-frozen.  Tissue samples or cell suspensions are high-pressure frozen, which avoids formation of crystalline ice in depths up to about 100 μm. 

Frozen samples that are over about 500 nm in thickness need to be thinned before imaging in the TEM.  At first, this was done by cryo-ultramicrotomy, and we published the first tomograms of vitreously frozen tissue (Hsieh et al. 2002).  We subsequently improved our cryo-ultramicrotomy approach (Hsieh et al. 2006), but we found that mounting the sections on grids is problematic, limiting the usefulness of cryo-ultramicrotomy for TEM tomography.  In addition, we were not able to overcome all the artifacts that are generated in the sections by the action of the knife, the most recalcitrant artifact being compression in the direction of cutting.

We were able to overcome the limitations of cryo-ultramicrotomy by instead using a focused-ion-beam (FIB) to thin vitreous specimens, a technique first developed in our lab (Marko et al. 2006, 2007), this technique is described under the Cryo-focused-ion-beam (FIB) preparation of vitreous specimens for cryo-TEM section.

Resolution in TEM cryo-tomography is ultimately limited by the damage to the specimen due to electron irradiation.   Resolution can be increased if more tilt images can be recorded without increasing the electron dose (McEwen et al. 2002).  This is facilitated by increasing the contrast in the TEM, without increasing the electron dose.  Zero-loss energy filtering and TEM phase-plate imaging (both implemented on our JEM-4000FX and JEM-3200FSC/PP TEMs) accomplish this task.  We were the first US lab to implement TEM phase-plate imaging (Marko et al. 2005, 2011), and the technique is described under the Phase-plate imaging for cryo-TEM section.

If there are multiple, identical, examples of a structure in a tomogram, averaging them considerably improves the resolution.  As have other labs, we developed a motif-search algorithm (Ramos) that can detect the "signature" of known macromolecular complexes inside tomograms (Rath et al. 2003).  A novel implementation of the Ramos algorithm was used to generate a partial map of the calcium-release unit from multiple electron tomograms of frozen-hydrated triad junctions isolated from skeletal muscle (Renken et al. 2009).

Principal Investigator: Michael Marko

Most whole cells in culture are too thick to be imaged at high resolution in the cryo-TEM, and frozen pieces of tissue also need to be thinned before imaging. Cryo-ultramicrotomy can be used to produce thin slices of tissue or cell pellets, but the mechanical cutting process causes artifacts, and the slices do not lie flat on the TEM grid.  We developed the new technique of cryo-FIB milling to overcome the drawbacks of cryo-ultramicrotomy (Marko et al. 2005; 2006; 2007a,b; 2008; 2010)

The FIB-SEM instrument typically consists of an SEM column (vertical) and a focused ion beam (FIB) column (at an angle of about 50 degrees). The FIB is used for milling away material, while the SEM is used to view the results non-destructively.  The ion used most frequently is gallium (Ga+).

Figure 1.  Effect of ion-beam-milling of frozen tissue with a 30 keV Ga+ beam (plots made with www.srim.org). When the ion beam is directed at a glancing angle to the sample, most of the ions are stopped within 10-15 nm of the sample surface so this material is sputtered away.  Some ions beyond this range are implanted in the upper 10 nm of the new surface.  With glancing-angle milling, density differences within the sample have little effect on the smoothness of the surface. However, if the ion beam is directed normal to the sample, the sputtering depth is much greater.

Figure 2.  Cells can be cultured on TEM grids, and then high-pressure frozen on the grids.  Alternatively, small cells (such as bacteria) that don’t exceed a few micrometers in thickness can be plunge-frozen on the grid.  Before freezing, the cells can be photographed by light microscopy and cells of interest (e.g. based on a fluorescent label) and be identified.  After freezing, the grids are cut in half and preselected cells that are on the cut edge can be thinned with the ion beam to produce a suitable TEM sample.

Figure 3.  Cryo-electron tomograms of a cryo-FIB milled E. coli cell (a–h). (a) 1.8-nm-thick tomographic slice at cut surface (frost particles at arrows) with well-preserved membranes and ‘bridging’ structure between inner and outer membranes (arrowhead and in inset). (b) Slice 304 nm below that in a; inset region (indicated by arrowhead) with three putative ribosomes (arrows) and small vesicle (arrowhead) attached to inner membrane. (c) Cross-section along line indicated in b; surface frost particle at arrow.  Membrane contours at bottom appear incomplete because of limited angular range of tilt series. (d) Detail from 205-nm-deep slice. (e) Density profile of bacterial envelope, scanned in the direction of the arrow and averaged along length of bars in d.  (f) Detail near end of cell, with intermediate dense layer (arrow) and three ca. 50-nm diameter vesicles (arrowheads) also seen at top left in h.  (g) Surface-rendered model of bacterial envelope.  (h) Model showing vesicle cluster. (i,j) High-pressure frozen, microtomed cell. (i) Projection image of 160-nm-thick microtome section.  (j) A 1.8-nm-thick tomographic slice showing compression in (vertical) cutting direction that caused narrowed space between inner and outer membranes (arrow), as well as disruption of cell wall structure (arrowheads) due to plastic deformation of section. Deformation was correlated with crevasses (horizontal ripples) formed during sectioning. Scale bars, 250 nm (a–c,g,i and j); 30 nm (insets in a and b, and in d and f); 50 nm (h). [from Marko et al., 2007]

Figure 4.  Tissue pieces or cell suspensions are high-pressure frozen in 3-mm diameter specimen carriers of our own design.  These carriers fit into a TEM cryo-transfer stage.  A diamond trimming knife, in the chamber of a cryo-ultramicrotome, is used to cut the carrier in half, exposing the tissue. The tissue is then shaped to provide a 20-μm-thick slab spanning the 1000-μm area of tissue, and including part of the aluminum carrier metal on both sides for mechanical support and electrical conductivity.

Figure 5.  Preparation of high-pressure frozen tissue for the cryo-TEM. Step 1 (upper left): An area about 100 μm wide is milled at the top and bottom of the edge-on orientation of the 20-μm thick tissue slab, reducing the thickness to about 5 μm.  The side view is shown at the middle left.  Step 2: the 5x100 μm “lamella” is further thinned to about 1 μm in thickness, in two areas about 40 μm wide (lower left).  Step 3: Each 1x40 μm lamella is thinned to 100-200 nm, forming the final TEM specimen, about 10 μm wide and 50 μm long (upper right).  A magnified edge-on view is shown at the middle right, and a magnified portion of the TEM lamella is shown at the lower right.  Note that because the milling was at a glancing angle, no detail of the specimen can be seen in the SEM.

Figure 6.  Cryo-TEM image of a FIB-thinned lamella of skeletal muscle tissue. Z-lines, sarcoplasmic reticulum vesicles (SR) and t-tubules (TT) are indicated.  This is the first-ever TEM image of muscle tissue that is free of any artifacts due to chemical fixation, dehydration, staining, or mechanical sectioning (Marko et al. 2010).

Principal Investigator: Michael Marko

A phase plate increases the contrast of vitreously frozen macromolecules and cell components by two to five times, providing 3D reconstructions with higher resolution, and aiding the detection of weak features (Danev and Nagayama, 2001; Danev et al. 2009; Marko and Danev, 2012).

Phase-plate TEM is particularly beneficial for cryo-tomography (Marko et al. 2005, 2007, 2010) because the increase in image contrast allows each tilt image to be recorded at a lower electron dose.  Thus, more tilt images can be collected at the same total dose, and the tomographic resolution is increased.  Low-contrast details are made more visible, and image interpretation is simpler because of the lack of CTF oscillation.

We are working on making phase-plate imaging more available to the cryo-EM community (Marko et al. 2007, 2010, 2011).  This includes implementation of automatic centering of the phase plate (Marko and Leith 2012) and improved phase plates (Striemer et al. 2012). 

Figure 1.  Phase plates are placed in the back focal plane of the objective lens (the objective aperture position), where they are normally designed to produce a 90-degree phase shift between unscatteredelectrons and electrons that are scattered by the specimen.  In the case of the Zernike phase plate, the unscattered electrons (and those scattered at very low angles) pass through the central hole in the phase plate, and are not phase-shifted. [Figure from Danev and Nagayama, 2001]

Figure 2.  Phase-plate construction: SEM of a thin-film window in a silicon TEM grid (provided by TEMWindows, Rochester, NY).  A 700-nm-diameter hole (arrow) has been FIB-milled in the center.  Inset: Higher-magnification SEM of the phase-plate central hole.  Subsequently, the silicon grid, containing 16-25 individual phase-plate windows, is coated on both sides with thin metal or another conductive film.

Figure 3.  Details of phase-plate imaging:  Contrast-transfer function curves for a 300 kV TEM (Cs=5.0 mm), set for nearly parallel illumination (0.01 mrad convergence angle).  (A) Zernike phase plate (red) with 50-nm underfocus and 90-degrees phase shift, compared with a typical value of 1 μm underfocus with no phase plate (blue).  With the phase plate, there is a band of high, uniform transfer, from very low resolution up to 0.4 nm.  (B) Phase plate as in (A), but non-phase-plate curve shown at Scherzer underfocus (110 nm). Note the absence of low-frequency transfer without the phase plate.  (C) Compromise use of a phase plate with 60-degree phase shift, used with 90 nm underfocus.  This pushes the (theoretical) resolution with phase plate to 0.34 nm at the expense of some loss of transfer at lower resolution; nevertheless, the transfer at 2 nm and below is still far higher than without a phase plate.  Scattering in the carbon film slightly reduces maximum transfer with the phase plate (as shown), and may also limit the practical resolution for cryo-TEM to about 0.5 nm due to the effects on the envelope function.

Figure 4.  Example of phase-plate imaging: Vitreously frozen section of liver tissue, 200 nm thick.  Comparison of in-focus phase-plate imaging and conventional high-underfocus (20 μm) imaging without a phase plate.  Imaging was done at 400 kV with zero-loss energy filtering.  (A) Near-focus image without phase plate, (C) near-focus image with phase plate, (D) conventional high-underfocus image. Note the improved visibility of ribosomes on the endoplasmic reticulum at (*).  The insets in C and D show traces across mitochondrial membranes within the small boxes at the upper right of each image.  The rotationally averaged power spectra are shown in B: IF corresponds to A, UF corresponds to D, and PP corresponds to A.  Scale bar = 100 nm  From Marko et al., 2011.

Figure 6.  Example of phase-plate imaging:  Cryo-TEM of microtuble doublets from Chlamydomonas flagella. Top: conventional underfocus imaging.  Bottom: In-focus phase-plate imaging.  Note the clarity of broken-off radial spokes, seen as a ”fringe” below the microtubule doublet. Each microtubule is about 25 nm in diameter. Image recorded at 200 keV.  Specimen courtesy of Haixin Sui.

Principal Investigators: Rajendra K. Agrawal, Ph.D., Terrence Wagenknecht, Ph.D.

The goal of structural biology is to understand physiological function from three-dimensional organization.  Electron microscopy (EM) is one general technique to obtain a 3D structure. Whereas the end product of conventional EM is a static structure, it would be informative to observe a structure in a sequence of states in order to undertand how it works.  Our goal is to develop methods to initiate reactions involving macromolecules, trap them in intermediate states, and visualize them using electron microscopy.  We have implemented three modes to prepare time-resolved samples for EM but current development is concentrated on the first mode: a Microfluidic Mixer-Sprayer

Microfluidic Mixer-Sprayer: We have developed a novel integrated microfluidic device that allows small volumes of two reactants to be mixed in less than 1 ms, react for a time set by the device, and then sprayed onto a dry EM grid as it is being plunged into cryogen (Lu et al. 2009, 2010; Barnard et al 2009).  Several devices have been constructed that allow reaction times ranging from 10 – 500 ms to be investigated. The current configurations require about 10 microliters of each macromolecular reactant per grid at a concentration in the micromolar range.

Figure 1.  Micromixer development. (a) Configuration of the micromixer, with detailed dimensions labeled. SEM images on the bottom show the mixing channel as fabricated using DRIE etching technology (left), and the details of half of the butterfly mixing unit (right).  (b) Fluorescence microscopy images of mixing fluorescein and Fluoro Sphere Red in the fabricated mixer at various flow rates.  (c) Experimental and simulated mixing indices showing essentially complete mixing at flow rates above 4 μL/sec. (d) Activity of malate dehydrogenase is preserved at various flow rates.

Figure 2.  Experimental setup with microspray generation shown: (a) Diagram of the experimental system, including syringe pump for liquid transport, monolithic device for mixing and spraying, and cryo-EM grid for droplet collection. (b) Photo of the entire experimental system setup.  (c) Close-up of the device with fittings, which is mounted on a plastic holder.  A tweezers (connected to the plunger) with an EM grid placed at the exit of the spray nozzle.  (d) Fine microspray (blue arrows) generated by the device with the liquid flow rate at 6.0 μL/s, and with nitrogen gas pressure at 50 psi (set by nitrogen tank regulator).

Figure 3.  Monolithic microfluidic devices for mixing and spraying (a) Detailed schematic of monolithic device (micromixer/sprayer) design showing (1) device design layout, (2) micromixer configuration, (3) SEM image of the mixer, (4) external atomization nozzle design, (5) SEM image of the nozzle, (6) device top view, and (7) device bottom view. (b) The droplet size distributions for two of the devices with the external atomization nozzles, showing the droplet size mainly in the range of 10-30 μm.

Figure 4.  Cryo-EM of ribosomes and ferritin that were mixed and sprayed by the monlithic device #2. Upper left: low magnification micrograph showing the edge of a droplet that has spread sufficiently thin for image. Right: High magnification view of a hole containing thin ice and showing presence of both ferritin (dark structures) and ribosomes (lighter structures). Lower left: Surface representation of a three-dimensional reconstruction of the ribosome determined from images such as that shown in B. Resolution is 18.9Å. [Adapted from Lu et al. J. Struct. Biol., 2009]

Figure 5.  Classification of ~4400 partially re-associated 70S images (out of 17,000 images) using the 3-D maximum likelihood (ML3D) method. (a) Comparison between two volumes (#1 [yellow mesh] and #2 [blue surface]) showing absence of bridges (B1b, B8, and BX) in volume #2. (b) Comparison between volumes #1 [yellow mesh] and #3 [white surface], showing absence of bridges (B8 and BX) in volume #3 and presence of additional bridge B1a in volume #3.

Flash photolysis: There exist chemical reagents that become able to react only upon irradiation with light. We have mounted a light pipe leading from a xenon lamp, trained on a point in the path of the grid as it is cryofixed (Shaikh et al. 2009). The two reactants of interests -- one in a unreactive, "caged" form -- are both on the grid at the start of the experiment. The grid is blotted (before the reaction has started) and then flashed. The length of time between flashing and cryo-fixation can be adjusted by the user.

Concurrent with these sample-preparation methods are computational techniques to distinguish between what are likely to be heterogeneous states appearing in images. We have several in-house projects to test the above methods, including the ryanodine receptor with calcium, the ribosome release complex with release factor-1, and various ribosomal complexes with GTP (Shaikh et al 2010).