Scanning transmission electron tomography to study virus assembly: Review for the retirement of Paul Walther

Abstract

In this short and popular review, we summarise some of our findings analysing the replication cycles of large DNA viruses using scanning transmission electron tomography (STEM tomography) that we applied in the laboratory of Paul Walther. It is also a tribute to a very kind and expert scientist, who recently retired. Transmission electron microscopy (TEM), in particular cryo‐EM, has benefited tremendously from recent developments in instrumentation. However, TEM imaging remains limited by the thickness of the specimen and classical thin‐section TEM typically generates 2D representations of 3D volumes. Although TEM tomography can partly overcome this limitation, the thickness of the sample, the volume that can be analysed in 3D, remains limiting. STEM tomography can partly overcome this problem, as it allows for the analysis of thicker samples, up to 1 µm in thickness. As such, it is an interesting imaging technique to analyse large DNA viruses, some of which measure 1 µm or more, and which is the focus of our research interest.

1. PERSONAL WORDS ABOUT PAUL WALTHER

Paul Walther is retiring and leaves behind both a deep personal and scientific impression. I (JKL) got to know Paul many years ago when he applied at the EMBL to head the electron microscopy facility. Soon after, he was heading the EM facility of the Ulm University as Professor Paul! Many people subsequently benefited from his expertise, including our team, in a longstanding, extremely pleasant and efficient collaboration. Until this date, we apply his workflow for high‐pressure freezing and freeze substitution we commonly call the ‘Paul‐protocol’. Adding a bit of water to the substitution mix, controversial among some EM experts, not quite understood how it works, but it results in a better contrasting of membranes, essential if one studies membrane biogenesis of viruses.

With the acquisition of the Titan 300 kV scanning transmission electron microscope (STEM) at the university of Ulm, we and others benefited by analysing thick sections by STEM tomography. The acquisition of tilt series in STEM‐mode using the newly acquired instrument was clearly facilitated by Paul's deep understanding of STEM imaging in 3D. ¹ For us the advantage of the technique was immediately obvious, as now the viruses, 300 to 1000 nm in diameter, fitted into the section! We spend many hours together acquiring tilt series in STEM‐mode resulting in four joint publications. ² , ³ , ⁴ , ⁵ Below we summarise some of our major findings using STEM tomography. It is a personal thank‐you to Paul, a wonderful scientist and friend, and we apologise for not citing other people's work applying STEM tomography to study important biological questions.

2. INTRODUCTION TO STEM TOMOGRAPHY

Transmission electron microscopy (TEM) is an important tool for the analysis of biological structures at high resolution. ⁶ , ⁷ While powerful, a main limitation is the thickness of the sample that can be imaged in routine TEM, which is typically around 70 nm. Classical room temperature TEM results in images that are two‐dimensional projections of a 3D volume. ⁸ This limitation can partially be overcome by tilting the sample to obtain a 3D impression by TEM tomography. ⁷ , ⁹ However, even in TEM tomography the section thickness is limiting, 500 nm max, ideally more around 200–300 nm. This limitation can be overcome by other 3D EM methods, one of which is STEM tomography. ¹⁰ , ¹¹ , ¹² In contrast to TEM, STEM does not suffer from chromatic aberration since there is no image‐forming lens between the sample and the detector. ¹³ , ¹⁴ Moreover, the STEM beam can be tuned to a relatively narrow semi‐convergence angle ⁷ , ¹⁴ and has the possibility of a dynamic focus. ¹³ , ¹⁴ , ¹⁵ This combined allows STEM to overcome limitations of focus gradients in titled samples and thus to resolve structures within sections with a thickness of up to 1 µm. ¹³ With the option of using such section thickness, STEM tomography becomes an extremely powerful imaging technique to study large DNA viruses, which are characterised not only by a large particle size but also by the formation of large cytoplasmic replication organelles.

3. HOW OUR WORK BENEFITED FROM STEM TOMOGRAPHY

We study the replicative cycle of large DNA viruses and our pet virus is the poxvirus vaccinia virus (VACV), the vaccine strain that led to the eradication of smallpox. ¹⁶ It is a brick‐shaped enveloped virus, measuring 250 × 200 × 300 nm. ¹⁷ , ¹⁸ Its membrane acquisition is complex; membrane biogenesis starts with the formation in the cellular cytoplasm of short membrane arcs (crescents) that grow to form 300 nm wide spheres (Figure 1). The membrane arcs and spheres are bended by a viral scaffold protein, the gene product of D13, which associates on the convex side of the newly formed viral membrane. ¹⁸ When the newly formed viral spheres, surrounding the viral core proteins take up the DNA, they undergo a major morphological change from spherical to brick‐shaped, the latter being infectious ¹⁷ (Figure 1). For a long time, we were unsure how the arcs are formed; we speculated they are membrane cisternae where the two lipid bilayers become tightly apposed. However, using cryo‐EM of vitreous sections we showed that the arcs are single, open, membranes likely derived from rupture of cellular ER‐derived vesicles. ¹⁹

FIGURE 1.

Schematics of membrane assembly of VACV (the organelles are not to scale). Virus assembly is initiated within viral factory (VF), where viral DNA replication takes place. The first evidence of virion assembly is the formation of short membrane arcs, bended by the scaffold protein D13 (red dotted lines). The membrane‐arcs grow to form spheres, surrounding the viral core proteins (yellow), thereby creating the immature virion (IV). After taking up the viral genome, the IV with nucleoids (IVN) undergoes a major structural change to form the brick‐shaped mature virus (MV). The MV contains an oval core and two lateral bodies (violet). nu: cell nucleus; ER: endoplasmic reticulum; mc: mitochondria. Created with BioRender.com.

Membrane acquisition by open membrane precursors seems to be a hallmark of at least some other members of the family of large DNA viruses. Most of those viruses exceed VACV in size and in the number of proteins they encode for, some encoding more than 1000 proteins and measuring up to 1 µm in diameter. ²⁰ , ²¹ The first member to be identified among these giant viruses is Mimivirus, an icosahedral shaped virus roughly 500 nm in size, infecting amoeba. ²² Interestingly, its membrane acquisition shows striking similarities to VACV. ²³ Its precursor membranes are small open membrane units which grow to form an icosahedral membrane shaped by a viral scaffold protein, also called major capsid protein (MCP), on its convex side. The MCP shows structural similarity to the VACV D13 protein. ²⁴ , ²⁵ , ²⁶ Interestingly, the precursor membranes were not resolved when samples were heavily contrasted such as during classical epoxy resin embedding or after freeze substitution. ²³ The membranes only stood out in thawed cryo‐sections according to Tokuyasu where membranes are negatively stained and appear as white lines by EM (Figure 2). Apparently, the contrasting used in classical embedding (osmium and uranyl acetate) did not allow us to discriminate the highly contrasted content of the viral factory (VF) from the contrasted membranes located at their periphery. ²³ In thawed cryo‐sections, the negatively contrasted membranes generally appear as a continuous white line. In classical epoxy resin embedded samples, however, membranes sometimes appear discontinuous, likely due to embedding artefacts. As explained below, this inspired us to try STEM tomography on thick thawed cryo‐sections.

FIGURE 2.

Thawed cryo‐sections of Acantamoeba polyphaga infected for 10 h with Mimi virus, labelled in A with anti‐DNA. (A) Various maturation stages of Mimivirus; close to the viral factory (VF) immature viruses (IV) accumulate, whereas further away from the VF the icosahedral viruses become more and more complex displaying several layers. Those viruses labelled with anti‐DNA have taken up the genome and display an additional internal structure likely representing the viral core. The arrowheads point to lateral body‐like structures, the black arrow to the stargate and the white arrow to the DNA‐portal of the mature virus (MV). (B) A higher magnification view of the small membrane structures (white arrows) that accumulate close to the VF. We showed using TEM tomography that these small membranes contribute to the formation of the viral membrane and are open. The icosahedral‐shaped newly formed viral membranes at the periphery of the VF consist of two layers; a membrane (mb) and a second layer likely the viral major capsid protein (pcp) shaping the membrane. Scale bars: 200 nm (original figure: A: figure 4 and B: figure 3 from Ref. [23]).

The origin and biogenesis of the membranes of African swine fever virus (ASFV), another member of the large DNA virus family, was also highly debated. In fact, data obtained with VACV were used for ASFV as argument for one or the other model of membrane acquisition and vice versa. ²⁷ We applied several complementary EM methods, including STEM tomography to study its membrane biogenesis and in this study, we tried for the first time STEM tomography on thawed cryo‐sections. In a major effort, two very skilful technicians produced 1 µm thick cryo‐sections, placing the flat, thick, ribbon in the middle of a slot grid for STEM tomography. Sections were contrasted as usual, using methylcellulose and uranyl acetate and prior to STEM tomography image acquisition, the grids were coated with a thin layer of carbon. And, surprisingly, it worked! Collectively, our results indicated that the ASFV‐membrane also originates from the ER and open membrane precursor are the source of this membrane, shaped on its convex side by the capsid protein (Figure 3). ⁴ This was likely one of the first and rare examples where STEM tomography was performed on thick thawed cryo‐sections.

FIGURE 3.

COS cells infected for 14 h with African swine fever virus (ASFV), fixed with 1% glutaraldehyde and prepared for cryo‐sectioning; 450 thick thawed cryo‐sections were imaged by HAADF‐STEM tomography, where the contrast is reversed and the membranes appear black. The images 1 to 4 are different virtual slices 3.4 nm in thickness of tomograms illustrating the fine structure of the immature virus of ASFV. The inner layer appears as a continuous black line (white arrow) and represents the viral membrane (white arrows) and is shaped by a layer on its convex side that displays tiny spikes (arrowheads in 3 and 4). Scale bars: 100 nm (original figure: figure 6 from Ref. [4]).

Until this date, the molecular requirements of VACV membrane biogenesis remain incompletely understood. Roles have been implicated for some of the major viral membrane proteins as well as for the viral scaffold protein D13. ¹⁷ Less clear is the role of so‐called viral membrane associated proteins (VMAPs); these are minor proteins essential for the production of infectious progeny and with some role for membrane assembly. To address this knowledge gap, we studied recombinant viruses in which the protein of interest, one of the five known VMAPS, can be conditionally synthesised.

VACV infection in the absence of A11, a 40 kDa protein and one of the VMAPs, completely blocks the formation of the viral membrane arcs, the crescents, and hence no infectious virus is produced. Instead, the scaffold protein accumulates in larger areas that are surrounded by cisterna derived from the endoplasmic reticulum. ⁵ Without anti‐D13 labelling on thawed cryo‐sections these areas are hard to distinguish from the rest of the cytoplasm. After our skilful technicians acquired enough routine and confidence placing thick cryo‐sections on slot grids, they proceeded to label these thick sections with antibodies to our structures of interest, with anti‐D13 and an antibody to an ER‐protein. They managed to do all of these steps without destroying the film of the grid or the sectioned ribbon.

STEM tomography of immunolabelled cryo‐sections revealed that the D13 structures are surrounded by membranes that label positive for the ER marker and that these membranes were closed. These findings suggested to us that A11 is essential for membrane rupture and that membrane‐rupture of ER membranes is necessary to initiate VACV membrane biogenesis, the formation of the crescent membranes. ⁵ Given the size of the D13 accumulations in the cellular cytoplasm (1 to 2 µm in diameter) STEM tomography was a very convenient complementary technique to make us confident that the membranes surrounding the D13 accumulations were closed.

Infection of HeLa cells with VACV in the absence of the H7 protein, another VMAP of 17 kDa, results in the formation of D13‐positive structures that look quite different from those formed in the absence of A11. In epoxy resin‐embedded samples with high contrast, networks structures are seen to accumulate, that contain small membrane tubes connected by electron‐dense spots. ³ Parallel EM immunolabelling show that these networks are D13 positive. STEM tomography appeared to be essential to understand the organisation and the biogenesis of these networks. These are connected to, and surrounded by, ER‐cisterna; thus, the tubes originate from the ER (Figure 4). Tomography of thawed, D13‐labelled and refrozen cryo‐sections strongly suggested that the electron‐dense spots corresponded to D13‐trimers, the structure of which is known from X‐ray crystallography. ²⁵ However, image analyses of these sections showed that the trimers failed to form hexamers. The latter are thus likely essential for the formation of the arcs, the crescents and thus for VACV membrane biogenesis. Thus, we proposed that the VMAP H7 is needed for the proper organisation of D13 and thus for membrane biogenesis. ³

FIGURE 4.

HeLa cells infected for 12 h with a vaccinia virus recombinant conditionally expressing the VMAP H7. In the absence of H7 expression D13 accumulates in structures surrounded by the endoplasmic reticulum (black arrows). These appear by STEM tomography on 750 nm thick sections as electron‐dense spots (white arrows) connected by short membrane tubes (white arrowheads). The images are two 3.6 nm thick virtual slices of the reconstructed tomogram (Z:41 and Z:51, 36 nm apart) of the 750 nm section, imaged by STEM tomography (original figure: figure 3 from Ref. [3]).

4. SUMMARY AND CONCLUSION

In this short tribute to Paul, we want to illustrate the power of STEM tomography for the study of membranes upon viral infection, exemplified by some of our own data. In our opinion, the power of STEM tomography is underrated and applied by too little EM facilities. Reasons for this might be many‐fold, the lack of specific expertise to image in STEM‐mode or of a specific biological question, such as studying virus‐host interactions. Together with Paul, we have demonstrated that it can also be applied to thick thawed cryo‐sections. On a personal side, we are already missing our joint long hours on the scope, discussing about EM methods and about life. Paul's contributions to research, in particular to the development of STEM tomography, will be missed tremendously in the scientific EM community.

Wieczorek, S. , & Krijnse Locker, J. (2025). Scanning transmission electron tomography to study virus assembly: Review for the retirement of Paul Walther. Journal of Microscopy, 299, 206–211. 10.1111/jmi.13374