The combination of cryo-FIB and in situ cryo-electron tomography enables ultrastructural analysis of disease-related protein aggregates

Article image: 
Alexander Rigort,1,2 Qiang Guo,1 Felix J.B. Bäuerlein,1 Rubén Fernández-Busnadiego1
1. Department of Molecular Structural Biology, Max Planck Institute of Biochemistry, 82152 Martinsried, Germany.
2. Thermo Fisher Scientific, FEI Deutschland GmbH, Fraunhofer Straße 11, 82152 Planegg, Germany
Alex Rigort obtained his PhD in cell biology from the University of Bonn and then joined the Max Planck Institute of Biochemistry in Munich for a postdoctoral position in cryo-electron microscopy. 
Following a short postdoc in Israel at the Ben-Gurion University in Beer-Sheva, he returned to the Max Planck Institute where he developed and used cryo-focused ion beam instrumentation for applications in electron tomography. He then moved on to the scientific instruments manufacturing industry where he worked for Zeiss and since 2015 for the Materials & Structural Analysis Division at Thermo Fisher Scientific.
A pathological hallmark of many neurodegenerative diseases is the accumulation of protein aggregates in inclusion bodies. This aggregation plays a crucial role in mediating neurotoxic effects. To gain insight into the structure of protein aggregates inside cells advanced cryo-electron tomography methods were applied by the researchers. Sample thinning by cryo-focused ion beam microscopy plays a key role in this approach, as it enables the preparation of thin lamellas from vitrified cells containing protein aggregates. Imaging aggregate structure by cryo-tomography at molecular resolution and in three dimensions within pristinely preserved cellular environments sheds new light into the cellular mechanisms of neurodegeneration.
R.F.-B. acknowledges funding from the European Commission FP7 GA ERC-2012-SyG_318987–ToPAG (Toxic Protein Aggregation in Neurodegeneration).
Corresponding author
Further reading
Link to workflow video:
An aging population comes along with intrinsic social and economic burdens, like the increased occurrence of neurodegenerative disorders such as Huntington's disease (HD), amyotrophic lateral sclerosis (ALS or Lou Gehrig's disease), and Parkinson's disease. Although it is well established that protein aggregation is a hallmark of many of these neurodegenerative disorders, the molecular mechanism(s) underlying aggregation have eluded neuro and cell biologists for decades. Does aggregation ultimately lead to cellular cytotoxicity and cell death? And what can be done to inhibit, or even halt, the process?
Promisingly, recent progress in cryo-electron microscopy (cryo-EM) have addressed many of the challenges faced in high-resolution imaging of biological samples. These advances represent a quantum leap in our understanding of the structural basis of fundamental biological processes.1-3 Cryo-EM is also instrumental in investigating the basis of neurodegenerative disorders at the structural level. The work described here combines cryo-focused ion beam microscopy (cryo-FIB) with cryo-electron tomography (cryo-ET) to examine the structure of large protein aggregates in situ, that is, within their native cellular context. Sample thinning by cryo-FIB is key for such an approach and allows for unperturbed views of proteins within their functional cellular environments.4 This article will introduce the technological advances and the overall workflow enabling these studies and give an overview of two significant recent examples, which have been described in detail elsewhere.5, 6
Researchers seeking to investigate the molecular structure of biological samples with high-resolution transmission electron microscopy (TEM) face a number of challenges rooted in the nature of biological materials. Generally, biological samples are composed of lighter elements that generate little contrast in a TEM. Techniques that work well to enhance contrast at lower resolution, such as selective staining with heavier elements, introduce distortions and artifacts at the molecular level. Longer exposures and higher currents that might improve contrast are not possible because the materials are easily damaged by high-energy electrons. Additionally, most biological materials exist naturally in a hydrated state, in which the presence of water is necessary to maintain the native structure; however, liquid water is incompatible with the vacuum environment of the microscope. Preparations that “fix” the sample by substituting non-volatile materials for water distort structure at the molecular level.
The resolution attainable in a TEM improves for thinner samples. Imaging cells by cryo-ET requires samples that are thin enough to allow electrons with energies of 300 keV to be transmitted — typically not much more than 300 nm. Most cells exceed these dimensions and are too thick to study intact in cryo-transmission electron microscopy (cryo-TEM), and must consequently be thinned before they can be imaged. The development of the ultramicrotome was arguably the most important advancement in enabling the preparation of sufficiently thin specimens. However, its application under cryo conditions has not proven very useful. Mechanical cryo-sectioning has a low throughput, and specimens suffer from artifacts such as an unpredictable degree of compression that restricted the usefulness of this preparation technique and limited the obtainable image quality. Finally, simply locating the structure of interest can be difficult in the vast complexity of the natural cellular environment. To meet these challenges, a robust multimodal workflow that combines sample localization, preparation and imaging steps (Fig. 1) is needed. 

Figure 1.
Cryo-ET workflow. (A) TEM grids containing biological samples such as cells are rapidly vitrified to avoid damage caused by crystallization, (B) Cells containing fluorescently labeled structures are identified in a light microscope, (C) Labeled cells are relocated in a FIB/SEM, in which a thin lamella for TEM analysis is created by milling away material above and below the feature of interest, (D) The lamella is transferred to a TEM, where a tomographic image series is acquired, (E) With the help of computational algorithms, repeating structures are identified (shown here: two different functional states of 26S proteasome) and high-resolution 3D models are generated by a process termed sub-tomogram averaging.
Cryo-TEM, in which the sample is made vacuum-compatible by rapid freezing, addresses many of these limitations. The key to its success is a freezing process so fast that liquid water forms a non-crystalline vitreous ice, thus avoiding the damage caused by the formation of crystals at slower freezing rates. Before a frozen cell can be examined by tomography, the target site within the cell first must be localized and then a thin lamella of the target site must be prepared. 
The localization of the feature of interest, such as a fluorescently labeled aggregate (see Fig. 2A and Fig. 3B), is performed by cryo-correlative fluorescence microscopy. The vitrified sample, along with the coordinates of the target region, is transferred to the focused ion beam/scanning electron microscope (FIB/SEM). Here, the superimposition of the cryo-light microscopy (cryo-LM) image with the scanning electron image acquired by the SEM/FIB permits reliable navigation to the features of interest. 

Figure 2.
(A) Overlay of live cell imaging and cryo-SEM images of primary neurons grown on an EM grid and containing a HttEx1 inclusion (green). (B and C) Side view of (A) imaged by FIB-induced secondary electrons before (B) and after (C) lamella preparation. Green circles indicate the coordinates of the inclusion obtained from correlation with light microscopy. (D) TEM low magnification image of the lamella shown in (C). (E) Tomographic slice from an HttEx1 inclusion interacting with cellular membranes in a HeLa cell. (F) 3D rendering of (E) showing HttEx1 fibrils (cyan/blue), vesicles embedded in the fibrillar network (white), ER membranes (red), and ER-bound ribosomes (green). (G) Magnified tomographic slice showing a vesicle containing a membrane-bound ribosome (white arrowhead) and contacted by HttEx1 fibrils (red arrowheads). This vesicle might have formed by the rupture of ER membranes upon interactions with HttEx1 fibrils. (H) 3D rendering of (G). Scale bars: (A) 50 µm (B,C) 5 µm, (D) 1 µm, (E,F) 250 nm, (G,H) 100 nm. Modified from5 with permission.
Following localization, the focused ion beam is used to prepare a thin, electron-transparent lamella by removing material above and below the target region. The SEM provides visual control over the FIB milling process, permitting the creation of cryo-lamellas as thin as 100–200 nm (Fig. 2B,C). Cryo-FIB preparation allows precise control of the location of the lamella and avoids the destructive effects of mechanical sectioning. After the milling step, thin cryo-lamellas are transferred to the cryo-TEM, where the actual tomographic image acquisition takes place.
The images in the tomographic series are acquired by tilting the sample in known increments. Individual projection images are then combined computationally in a procedure known as back-projection, which creates the 3D tomographic volume. From the 3D tomographic volume, higher resolution structures of particles (i.e. proteins or protein complexes) can be obtained by averaging out noise in a process termed sub-tomogram averaging. Sub-tomogram averaging is a method in which repeating structures within a tomogram are computationally extracted from the data set and treated as individual particles.7 The procedure is conceptually similar to single particle analysis, yielding higher resolution structures of the macromolecules detected in situ
Cryo techniques alone do not address the lack of contrast in biological samples. Contrast can be improved by zero-loss energy filtering, which removes inelastically scattered electrons — those possessing poor localization and coherence and therefore not contributing to high-resolution information — before they reach the detector.
In cryo-ET, phase contrast is usually induced by a slight negative defocusing of the objective lens. Contrast arises from an interference effect that modulates both low and high frequencies. It is mathematically described by the contrast transfer function, an oscillating function in spatial frequencies that has regions of inverted contrast, effectively limiting resolution in tomograms. Increasing defocus increases apparent contrast but decreases resolution. A newly developed hardware device, the Volta phase plate,8 offers a remedy here, as it is capable of enhancing (phase) contrast without the need of defocusing, especially for lower spatial frequencies. Last but not least, recent breakthroughs in detector technologies have made a major contribution to improved sensitivity and readout speeds, allowing researchers to obtain higher resolution data from cryo-tomograms. This progress was made possible by newly developed direct electron detectors. In contrast to the charge-coupled device (CCD)–based detectors, direct detectors avoid blurring and distortion due to the generation and transfer of photons. Their higher sensitivity in terms of detective quantum efficiency and higher frame-rates of image acquisition make fast recordings possible and enable procedures to correct for beam-induced specimen motion and drift in post-acquisition image processing.
The huge potential of cryo-ET — in particular its capability to perform structural studies in situ (inside vitrified cells) — can be illustrated by two recent studies investigating the structural basis of protein aggregate-associated neurodegenerative diseases. For a number of diseases, including Huntington’s disease, it was known for years that protein-aggregation results in cytotoxicity and the formation of large intracellular aggregates known as inclusions.9 However, a detailed understanding of the structure of these aggregates and their interactions with other cellular components was missing. 
HD is caused by the expansion of a CAG repeat in the gene coding for the huntingtin protein. The CAG codon is translated into the amino acid glutamine (Q), and inclusion bodies containing fragments of polyQ-expanded huntingtin are found in the brain of HD patients. In the first study,5 the structure of inclusions, formed by polyQ-expanded huntingtin exon 1 (HttEx1), was investigated by cryo-ET inside vitrified primary neuronal mouse and HeLa cells. Cryo-correlative microscopy was used to identify fluorescently labeled HttEx1 inclusions inside cells (Fig. 2A). Cryo-FIB was then employed to create 150–250 nm thin lamellas capturing parts of the inclusion (Fig. 2B, C, D). 
Phase plate tomograms from these lamellas showed that the inclusions are composed of a network of radially arranged amyloid-like fibers (Fig. 2E, F). These data revealed that the fibrils forming polyQ inclusions can distort and may even rupture organellar membranes (Fig. 2G, H). It was observed that the fibrils interact with cellular endomembranes, particularly of the endoplasmic reticulum (ER), deforming ER membranes and altering ER organization and dynamics. These findings suggest that aberrant interactions between fibrils and cellular organelles contribute to the deleterious cellular effects of protein aggregation.
The second study6 investigated a different type of protein aggregates related to frontotemporal dementia (FTD) and ALS associated with mutations in the C9orf72 gene. The root cause for these aggregates is a pathologic expansion of a GGGGCC sequence within C9orf72, which gives rise to five different di-peptide repeat proteins that aggregate in the brain of patients. Amongst these proteins, poly-GA is the most abundant. Poly-GA aggregates are common in the brain of C9orf72 ALS/FTD patients, and poly-GA expression in cultured neurons leads to the formation of large aggregates. Poly-GA aggregates were analyzed by in situ cryo-ET within neurons (Fig. 3) in a similar way as described for the first study.

Figure 3.
 (A) Cryo-LM overview of an EM grid containing poly-GA-transfected primary neurons. (B) Overlay of cryo-LM and cryo-SEM images of a neuronal cell body (white box) containing a poly-GA inclusion. (C) Side-view of the cell marked in (B) from an angle using FIB-induced secondary electrons. Yellow rectangles indicate the areas that will be removed by FIB milling during lamella preparation. (D) Overlay of cryo-LM and cryo-SEM images on the final lamella. (E) Overlay of cryo-LM and cryo-TEM images on the final lamella. (F) Tomographic slice of a tomogram recorded in the area boxed in (E). Red arrowheads indicate poly-GA ribbons. In the inset, green arrowheads mark GFP-associated densities decorating poly-GA ribbons. Scale bars: (A) 300 µm, (B) 40 µm, (C) 10 µm, (D, E) 4 µm, (F) 300 nm (inset 20 nm). From 6 with permission.
Surprisingly, poly-GA aggregates exhibited a completely different morphology compared to polyQ inclusions. Poly-GA aggregates did not consist of filaments but of planar twisted ribbons (Fig. 4A), which did not interact with cellular membranes. Cryo-ET, in combination with subtomogram averaging and classification, revealed that poly-GA aggregates interfere with the ubiquitin-proteasome system, which is of critical importance for cellular proteostasis (a cellular network that ensures proteome integrity). Proteasomes are sequestered and functionally stalled by poly-GA inclusions, leaving them unable to resume their role in protein degradation (Fig. 4B, C).
Thus, this study suggested a completely different mechanism for cytotoxicity compared to the findings described in the study on polyQ inclusions.

Figure 4.
(A) 3D rendering showing a snapshot from a cellular tomogram of a poly-GA aggregate. Poly-GA forms planar twisted ribbons (red) and sequesters proteasomes (green). (B) Two distinct conformations of the 26S proteasome obtained by subtomogram averaging and classification. Upper row exhibiting two slightly different conformations of proteasomes in the ground state (green) and lower row showing two substrate processing configurations (blue). Note that in ground state proteasomes, the Rpn5 and Rpn6 subunits are parallel to each other, but diverge in substrate processing states. (C) Detailed view of 26S proteasomes stalled while engaging poly-GA ribbons. Scale bar: (A) 200 nm.  Modified from6 with permission.
Without the aforementioned workflow (Fig. 1) — faithfully preserving molecular structures by sample vitrification, targeting by cryo-correlative microscopy, preparing thin samples with cryo-FIB and 3D imaging by electron tomography — it would not have been possible to obtain significant new insights into the structural alterations that take place in the cell upon protein aggregation.10, 11 One key to success is the combination of cryo-FIB milling with cryo-ET (Fig. 5) that has game-changing potential for cell biology, as it allows overcoming the thickness limitations of EM and at the same time guarantees high resolution with the best possible structural preservation. The cryo-ET workflow empowers biologists to literally open up windows into the interior of vitrified cells and enable in situ imaging of functional cellular environments with molecular resolution.12
The unique combination of specimen preparation and imaging technologies (see glossary box) employed in the studies described above demonstrates the potential of such a workflow for unveiling disease-related ultrastructural changes with an unprecedented level of detail. Results from these studies suggested two distinct mechanisms for cytotoxicity of protein-aggregates inside cells. The ability to see these entities and their interactions with other cellular components holds great promise for increasing our understanding of physiological and disease processes. Cryo-ET is on its way to becoming a key technique for exploring the native cellular environment, while controlling sample thickness by cryo-FIB milling will likely start another development cycle analogous to the advances allowing the use of this technology in materials science and semiconductor physics decades ago.

Figure 5.
Illustrating the combination of sample thinning and electron tomography. (Left) A thin in situ cryo-lamella is prepared from frozen-hydrated cellular specimens by removing material above and below the target section with the ion beam (FIB) while observing the operation with the scanning electron beam (SEM). (Right) The electron-transparent lamella remains on the EM grid, which is transferred to a cryo-TEM, where a tomographic image series is acquired as the lamella is tilted incrementally. Finally, the images are combined computationally to reconstruct a 3D tomogram. Example shown: inclusion (red).
  1. Beck M. & Baumeister W., “Cryo-Electron Tomography: can it reveal the molecular sociology of cells in atomic detail?”, 2016, Trends in Cell Biology 26(11): 825-837.

  2. Oikonomou C.M. & Jensen G.J., “Cellular electron cryotomography: towards structural biology in situ”, 2017, Annual Review of Biochemistry 86: 873-896.

  3. Cheng Y. “Single-particle cryo-EM—How did it get here and where will it go”, 2018, Science  Vol. 361, Issue 6405, pp. 876-880.

  4. Rigort A. & Plitzko J.M., “Cryo-focused-ion-beam applications in structural biology”, 2015, Archives of Biochemistry and Biophysics 581: 122-130.

  5. Bäuerlein F. et al., “In Situ Architecture and Cellular Interactions of PolyQ Inclusions”, 2017, Cell 171, 179–187.

  6. Guo Q. et al., “In Situ Structure of Neuronal C9orf72 Poly-GA Aggregates Reveals Proteasome Recruitment”, 2018, Cell 172, 1–10.

  7. Briggs, J.A., “Structural biology in situ – the potential of subtomogram averaging”, 2013, Curr. Opin. Struct. Biol. 23:261–267.

  8. Danev R. et al., “Volta potential phase plate for in-focus phase contrast transmission electron microscopy”, 2014, Proc Natl Acad Sci USA., 111(44):15635-40.

  9. Ross, C.A. & Poirier, M.A. “Protein aggregation and neurodegenerative disease”, 2004, Nature Medicine Vol. 10, S10–S17.

  10. Guedes-Dias, P. & Holzbaur, E., “Huntingtin Fibrils Poke Membranes”, 2018, Cell , Vol. 171 (1) , 32-33.

  11. Pontano-Vaites L. & Harper W., “Protein aggregates caught stalling”, 2018, Nature 555, 449-451.

  12. Mahamid J. et al., “Visualizing the molecular sociology at the HeLa cell nuclear periphery” 2016, Science Vol. 351 (6276), 969-972.

Cryo-FIB/SEM - Allows preparation of thin cryo-lamellas from vitrified cells that can be transferred to a cryo-TEM.
Cryo-Correlative Microscopy - Provides a means to localize cellular features prior to cryo-FIB targeting by imaging them in a cryo-fluorescence light microscope.
Phase Plates - Enhance contrast by modifying the contrast transfer function of the microscope such that enhanced contrast is obtained at lower spatial frequencies.
Direct Electron Detectors - Increase the efficiency of the detection process by avoiding the electron to photon conversion step used in CCD cameras. 
Sub-Tomogram Averaging - Method in which repeating structures within a tomogram are computationally extracted and averaged to increase contrast and resolution. 
Cryo-ET Workflow - Multimodal workflow including vitrification, cryo-fluorescence, cryo-FIB, cryo-ET and data analysis.
Website developed by S8080 Digital Media