TEM imaging of proteins in liquid-phase solution
For a proof-of-concept, the liquid-phase TEM specimen preparation method used for material science research can be effectively used for examining a biological sample. We first examined the protein GroEL, a standard sample that is commonly used for testing new approaches in the EM field. GroEL protein is a chaperone protein with a molecular mass of 800 kDa and a cylinder-shaped D7 symmetric structure with dimensions of ~13.5 nm x ~14.5 nm. A small amount (~0.2 µl at ~0.1 mg/ml) of label-free, stain-free, native GroEL protein in buffer was sandwiched between two formvar films and examined at room temperature (Fig. 1a). Survey micrographs showed donut-shaped particles with a diameter of ~14 nm. However, the shapes of the particles can be easily damaged by the cumulative effect the electron beam (Fig. 1b, and Extended Data Fig. 1). This damage includes sample deformation and bubbling caused by ionizing radiation that induces radiolysis of liquid molecules and leads to evaporation and the formation of gaseous byproducts 44. The bubbling phenomenon induced by electron beam is often observed in cryo-EM and liquid-phase TEM 52,53, but not in negative-staining (NS) or dried samples, suggesting that the cryo-electron microscopy (cryo-EM) sample is wet. Under the optimized illumination conditions, the high-resolution images of GroEL in liquid buffer reveal rich detailed features, such as the domain structures (Fig. 1c-e), which are consistent with those from the crystal structure (PDB entry: 1XCK,45 Fig. 1f) and cryo-EM images (Fig. 1g) but richer than those from NS (Fig. 1h). Notably, the biomolecular image contrast is same as that obtained by NS (Fig. 1h) and the silicon-nitride film encapsulated liquid-phase TEM sample, such as the simian rotavirus double-layered particles (DLPs) 24. However, the contrast is inverted to that obtained by cryo-EM (Fig. 1g). Several reasons may cause that the increased salt concentration due to surrounding water partially evaporation during specimen preparation. A small amount of water evaporation from a tiny amount of solution (~0.2 µl) will significantly increase the salt concentration, resulting the electron scattering capability by the surrounding solution to be higher than from proteins. The heavier elements or ions in the salts have a much higher scattering than the light elements in the proteins 46,47. A higher solvent concentration from native buffer is usually not fatal to the protein structure, which often occurred in protein crystallization and chromatography48,49. During protein crystallization, the surrounding layers of water maintaining the protein structure were intended to evaporate gradually50. The dark shadow surrounding the particles shown in NS images, but not shown in liquid TEM or cryo-EM, suggests that the high solvent content is evenly distributed in the background solution instead of accumulating in the protein particle surroundings. Our liquid TEM images showed a higher resolution structural details than that from NS or silicon-nitride film based liquid TEM, 24 suggesting that this simple method is suitable for examining proteins, virus and cells.
Liquid-phase TEM images of HeLa cells and HIV infected HeLa cells
HeLa cells alone and the HIV infected HeLa cells in growth medium were then encapsulated and examined by TEM respectively (Fig. 2a). The survey micrographs of cells exhibited soft-irregular shaped HeLa cells (Fig. 2b) consistent to the living HeLa cells imaged by super-resolution fluence light microscopy 51. Moreover, bubbling induced by electron beam can be often observed (Extended Data Fig. 2a and b) as that of other liquid-phase TEM 52,53. Plus, the radiation sensitivity of the cell membrane (Fig. 2c and Extended Data Fig. 2c) further confirmed that the specimen was wet.
The survey micrographs of HIV infected cells confirmed the soft-irregular shapes of HeLa cells. The irregular shape of cell membrane was often overlapped with each other, making the virus difficult to be clearly identified. However, on some cell surface, the roundish viruses were clearly visible on the cell surface (Fig. 2d and e). The virus images have a much stronger contrast than that from the cryo-EM (Fig. 2f-i), suggesting a unique advancement of this method in studying the virus and cell.
Surrounding the cells, there are many particles in diameters from ~20 nm to ~200 nm (Fig. 3a-c). The statistical analysis showed that smaller particles (<90 nm, Fig. 3b) have a peak population with a diameter of ~55 nm (Fig. 3d), while larger particles (Fig. 3c) have a peak population with a diameter of ~109 nm (Fig. 3d). Considering HIV usually has a spherical shape with varying diameters ranging from 64 to 170 nm 54,55, the large particles are similar in size to HIV11,12, and consistent with the observation of HIV by cryo-EM (Fig. 2f-i).
By zooming in a rectangular shaped HeLa cell with smooth cell membrane (in dimension of ~17 µm × ~6 µm, Fig. 3e), we found two chain-shaped densities present near the center (Fig. 3f). ET tilt series of images acquired under a low magnification (~1,000 ×) showed two distal ends having concave surface shape between two films in distance of ~ 1 um (Fig. 3g, Extended Data Fig. 3g and Supplementary Video S1). The IPET 3D reconstruction of the cell confirmed two chain-shaped densities with similar size and shape near the cell center (Fig. 3g, and Supplementary Video S1). The cell shape and two “copies” of center densities suggest this cell is in the middle of cell division, i.e., extending or lengthening itself to split its chromosomes into two halves.
The intermediates of viral entry into cell
The dividing cell provided an ideal thin cell membrane with a smooth edge at its distal ends, which was easy to distinguish the viruses from the cell surface (Fig. 2d and e, and Supplementary Video 2). Taking advantage of these properties, we observed the several intermediate stages of viral entry (Fig. 4a-c) as following: i) The landing stage (right before viral penetration): ~104-nm spherical viral particle landed on the cell surface and formed a concave surface by wrapping ~¼ of the virus surface (Fig. 4a). Between the virus and cell surface boundaries, there appeared with a concave-shaped hair-pin structure with a thickness of ~6 nm with two legs, ~67 and ~40 nm in length (indicated by two cyan arrows in Fig. 4a). The second object shows a quatrefoil-shaped structure ~21 nm in size (indicated by a yellow arrow in Fig. 4a). These two objects may represent two types of cell receptors that accumulate between viral and cell surfaces in response to viral binding through protein-protein interactions. Moreover, some spike-like particles (diameter of ~7 nm, black arrows indicated in Fig. 4a) were observed departing from the viral surface and merging into nanoparticles nearby (diameter of ~50 nm). We believed the ~7 nm spike-like particles represent to the viral glycoproteins such as glycoprotein VSV-G. Considering the interaction between HIV-1 spike protein and cell receptors triggered the exposure of a high-affinity binding site for a coreceptor,57 followed by the initiation of the membrane fusion process, the observed concave-shaped hair-pin structure and quatrefoil-shaped structure represent to the cell receptors and co-receptors.
To confirm above observations, the 3D map of this virus was reconstructed by IPET. IPET 3D projection and its superimposed 3D density map (a combination of positive and negative isosurface maps) confirmed that the spherical virus-like particle had spike-like proteins on its surface (Fig. 4a, Extended Data Fig. 4, and Supplementary Video 3). The accumulated densities observed within the concave surface of the cell may be due to cell receptor binding (hairpin-shaped density indicated by cyan arrows in Fig. 4a) with viral glycoprotein (white dots on virus surface) and the triggered downstream binding of proteins, such as the quatrefoil-shaped cell receptor (indicated by orange arrow in Fig. 4a). Moreover, next to the virus, the cell membrane displayed an abnormally flat, smooth and thick surface (indicated by pink arrows in Fig. 4a), which may indicate the formation of lipid rafts involved in viral entry, as reported.58
ii) In the penetrating stage, a globular viral particle with a diameter of ~130-nm half embeds into the cell membrane (Fig. 4b, Extended Data Fig. 5 and Supplementary Video S3). The half of the virus outside the cell membrane showed a smooth surface absent with coated spike-like particles. In contrast, the viral surface exhibited three attached nanoparticles with diameters of ~47 nm, ~64 nm and ~65 nm (black arrow in Fig. 4b), while these nanoparticles exhibited spike-like particles (similar size and shape), suggesting that the spikes were transferred from the virus to the nanoparticles. For the half of the virus embedded in the cell membrane, both the viral and cell membranes disappeared, which made it difficult to identify the boundary between virus and cell. Considering that these disappeared membranes are relatively large in size (the area is equal to three times the cross-sectional area of the virus), these membranes should have merged with the cell membrane and increased the local surface area of the cell based on the conventional mechanism of the nonendocytic route of viral fusion. In this mechanism, the viral membrane with its extra unbound spike proteins merged into the cell membrane and became a part of the cell membrane.1 However, considering that lipid rafts are also involved in virus entry34,35 via localization of cell receptor for HIV-1 entry,59 the accumulation of lipid rafts surrounding the virus should reduce the local mobility of the cell membrane, which should dramatically change the local landscape of the cell membrane after the great increase in its area by merging with the viral membrane. Changes in the landscape of the local cell membrane should be easily observed. However, while we did not observe significant changes in the cell membrane, such as wrinkles or distortions, we did observe three attached ~40-60-nm nanoparticles. Moreover, these extra unbound spikes from the merged viral membrane should preserve their biological functions in binding to nearby cell receptors on the targeted cell surface or the nearby cell surface. As a result, the cell surface morphology should be further modified by increasing the connectivity among the soft-irregular-shaped cell membrane boundaries or increasing the interactions and connectivity among the cells. However, the absence of observation of the accumulation of lipids or any change in cell surface morphology challenges the conventional mechanism of viral-cell membrane fusion.
One hypothesis to explain the above phenomena is that the extra membrane and unbound spike proteins formed attached ~50-nm nanoparticles surrounding the viral entry area. Based on the viral surface area calculation, the extra membrane from an embedded virus with a diameter of 109 nm (the peak population of the virus, shown in Fig. 3d) can produce ~2.9 particles with a diameter of ~55 nm (the secondary peak population of the surrounding particles, shown in Fig. 3d). The match between the calculation and our observation of the above three nearby nanoparticles (with diameters of ~45-65 nm) supports our hypothesis.
Additionally, in the EM images, the disappearance of the lipid membrane made it difficult to identify the boundary between the virus and cell. However, because of the uniform density of the virus core and its embedded half surrounded by tiny particles (~3-4 nm in diameter, white arrow indicated in Fig. 4b), the hemispherical shape of the virus embedded in the cell membrane was outlined, suggesting that the virus remained spherical after penetrating the cell.
iii) In the embedded stage, a virus particle with a diameter of ~96 nm was embedded underneath the cell membrane (Fig. 4c, Extended Data Fig. 6 and Video S3). The globular-shaped virus was surrounded by a layer of tiny particles with diameters of ~5 nm (arrows indicated in Fig. 4c). Considering i) the ~50 nm nanoparticles observed surrounding the cell (black arrows indicated in Fig. 4ab), ii) the lack of observation of the significant wrinkles or distortion of the local membrane (Fig. 4a), and iii) the lack of observation of the unbound spike-like proteins on the cell surface (Fig. 4a), these phenomena are consistent with our hypothesis, i.e., the viral membrane was unwrapped from the virus surface with unbound spike-like proteins, formed into ~50-nm lipid-protein nanoparticles, and released into the surrounding solution. Underneath HIV surface membrane envelope, a layer of matrix protein p17 encapsulates a cone-shaped nucleocapsid that is formed by a layer of nucleocapsid protein p24 60,61. Inside the capsid, the viral RNA is condensed with ribonucleoproteins and enzymes62,63. Remarkably, we found that this embedded virus remained in the form of a spherical shell of matrix proteins (Fig. 4b and c) instead of a cone-shaped viral capsid 64. This result adds another detail to the conventional model, in which the matrix protein shell underneath the viral membrane continues to protect the cone-shaped viral capsid from exposure to the cytoplasmic solution during the process of membrane fusion.
To further confirm the shape of the virus in the endosome, we examined two endocytosed particles (Fig. 4d and e, Extended Data Fig. 7 and 8 and Supplementary Video 3) respectively. One endocytosed particle has a globular virus-like (in diameter of ~62 x ~74 nm, indicated by the white arrows in Fig. 4d) adhered to a polygon-shaped density (in size of ~30 × ~45 nm, indicated by the orange arrows in Fig. 4d). The other endocytosed particle has a polygon-shaped density (in size of ~45 × ~50 nm) (Fig. 4b). Considering that these HIV capsids are in cone or polygon shapes instead of spheres,65 the observed polygon-shaped densities should be the capsid. The location of the capsid was observed inside the cell instead of near the boundary of the cell membrane, which suggests that the viral capsid is released inside the cell rather than near the cell membrane boundary during membrane fusion, consistent with our hypothesis.
The hypothesis of the viral entry mechanism
Considering that the three lipid-spike-formed nanoparticles attached to the viral surface have diameters of ~47- ~65 nm (black arrow indicated in Fig. 4b), we asked whether the similarly sized nanoparticles that formed a majority in the extracellular solution (with a major peak of size population at ~55 nm, shown in Fig. 3b and c) also contained spike-like proteins. We zoomed into the structures of two extracellular particles (Fig. 4f and g, Extended Data Fig. 9 and 10, and Supplementary Video 3). One extracellular particle contains two adhered nanoparticles with diameters of ~50 nm and ~60 nm attached to the cell surface (Fig. 4f), while the other particle also contains a pair of adhered nanoparticles in diameter of ~50 nm and ~60 nm but not attached to the cell surface (Fig. 4g). Both of their 3D reconstructions showed that all these nanoparticles contained ~7 nm spike-like protein particles (indicated by the cyan color in Fig. 4f and g) based on their similar image contrasts and particle sizes to those inside the 3D maps of the virus-attached nanoparticles (black arrows indicated in Fig. 4b). These tests suggested that the ~50 nm extracellular nanoparticles are similar to the virus-attached nanoparticles, and both contain spike-like proteins, supporting our hypothesis.