Skip to main content
NCBI home page
Biophysical Journal logoLink to Biophysical Journal
. 2016 Feb 23;110(4):776–784. doi: 10.1016/j.bpj.2016.01.023

Benefits and Limitations of Low-kV Macromolecular Imaging of Frozen-Hydrated Biological Samples

Endre Majorovits 1, Isabel Angert 1, Ute Kaiser 2, Rasmus R Schröder 3,4,∗
PMCID: PMC4776044  PMID: 26910420

Abstract

Object contrast is one of the most important parameters of macromolecular imaging. Low-voltage transmission electron microscopy has shown an increased atom contrast for carbon materials, indicating that amplitude contrast contributions increase at a higher rate than phase contrast and inelastic scattering. Here, we studied image contrast using ice-embedded tobacco mosaic virus particles as test samples at 20–80 keV electron energy. The particles showed the expected increase in contrast for lower energies, but at the same time the 2.3-nm-resolution measure decayed more rapidly. We found a pronounced signal loss below 60 keV, and therefore we conclude that increased inelastic scattering counteracts increased amplitude contrast. This model also implies that as long as the amplitude contrast does not increase with resolution, beam damage and multiple scattering will always win over increased contrast at the lowest energies. Therefore, we cannot expect that low-energy imaging of conventionally prepared samples would provide better data than state-of-the-art 200–300 keV imaging.

Introduction

In the last 2 years, the introduction of direct electron detection has revolutionized macromolecular imaging of frozen-hydrated biological samples (1), and several recent studies have shown that quasi-atomic imaging of biological macromolecules is no longer a fantasy (2, 3). At the same time, one has to ask, what instrumental improvements could still have a major impact on the quality of structural imaging, and what novel experiments on single molecules and in tomography might be possible in future? A comprehensive overview of available but not yet commonly used novel methods has already pointed to the use of low electron energies as one interesting approach (4). Low-voltage imaging has largely been promoted by the carbon-materials community (5, 6) because it prohibits one of the beam-damaging mechanisms for typical carbon-material specimens, i.e., knock-on damage, which occurs when incident electrons hit an atom such that it can be freed from its chemical bond, e.g., in graphene (7) or at graphene edges (8). With this low-energy technology, it is even possible to record movies of single, freestanding organic molecules without obvious damage, allowing the analysis of dynamical behavior such as cis-trans isomerization (9). It should be noted that the accumulated electron dose for such dynamic experiments is many orders of magnitude higher than the dose that is known to damage ice-embedded samples in a single recorded image. Thus, one might expect that low voltages would also be useful for macromolecular imaging of ice-embedded biological samples. However, it is well known from the work of Egerton (10) that for the typical ice-embedded biological sample, ionization is the dominant damaging mechanism, and that this damage increases with larger electron sample interactions for lower electron energies.

The only effect that counteracts the increased beam damage for lower electron energy is a higher object contrast, which can be expected from the nonlinear contributions of second-order phase and amplitude contrast terms (11) and an effective amplitude contrast, which increases faster with decreasing electron energy than phase contrast and beam damage (aka inelastic scattering) (12). This increased atomic contrast has been verified in numerical simulations as well as experimentally for high-resolution atomic imaging of carbon materials (13, 14).

In this study, we investigated ice-embedded tobacco mosaic virus (TMV) particles as a typical test sample, using their dominant 2.3 nm helical repeat signal as a quantitative readout. Our study covered a range of incident electron energies from 80 keV down to 20 keV. When decreasing the electron energy, it is essential to use aberration-corrected transmission electron microscopes (TEMs), as imaging aberrations become the dominant factors for reduced resolution at low electron energies (15, 16). Recently, it was shown that even at an electron energy of 20 keV, the information limit reaches 0.5 nm (5) if Cs correction is applied. The sample contrast analyzed in our study was on a larger scale, whereas changes in atomic contrast localization, which certainly are affected by aberration correction (17), played only a minor role in our imaging conditions.

Materials and Methods

Sample preparation for low-dose cryo experiments

TMV (kindly provided by Dr. Anan Kadri, Biologisches Institut, Universität Stuttgart, Stuttgart, Germany) was used as the test object for all experiments. TMV particles were suspended and diluted in water, and the suspension was then pipetted onto hydrophilic (glow-discharged) TEM copper grids covered with holey carbon film (Quantifoil R2/2 grids; Quantifoil Micro Tools, Germany). After plunge-freezing in liquid ethane using a GATAN Cp3 plunge freezer (Gatan, Pleasanton, CA) or an FEI vitrobot Mark IV plunge freezer (FEI, Hillsboro, OR), the frozen grids were stored and transported under liquid nitrogen. For imaging in the TEM, they were mounted into either an Oxford CT3500 or a Gatan 626 cryo holder (both from Gatan).

Data collection with Cs-corrected, energy-filtering, field emission gun TEMs

Images of vitrified TMV were acquired with two different Zeiss Libra 200 energy-filtering, field emission gun TEMs (EFTEMs, with highly aberration-corrected energy filters), each equipped with a second-order-corrected, in-column energy filter and imaging Cs corrector (CEOS, Heidelberg, Germany). We collected 20 keV, 40 keV, and 80 keV data using a dedicated Cs-corrected Libra 200 TEM operating at low voltages (5, 6) at Carl Zeiss Microscopy (Oberkochen, Germany). We collected 60 keV data using the Cs-corrected Libra 200 Kronos TEM at the University of Heidelberg. Some 60 keV images (see Fig. 2, A and B) were recorded on a 2k × 2k slow-scan CCD camera (UltraScan 1000; Gatan), and all other data were acquired on CMOS-based 4k × 4k cameras (TemCam-F416; TVIPS, Gauting, Germany).

Figure 2.

Figure 2

Open in a new tab

(A and B) Zero-loss filtered images of TMV particles in vitreous ice, recorded with a Cs-corrected EFTEM at 60 keV at an underfocus of 1.9 μm (A) and an overfocus of 2.1 μm (B). The contrast of the TMV particle with respect to the embedding ice layer is not inverted when going from underfocus to overfocus, as one would expect for conventional imaging at an electron energy above 100 keV: the TMV strand appears dark on a bright background for both under and overfocus. (C and D) The extracted intensity profiles across the TMV particles quantitatively support this observation.

All images were acquired with an inserted objective aperture. Unfiltered images were acquired without inserting an energy filter slit aperture, whereas zero-loss filtered images were acquired by inserting a filter slit aperture to remove the inelastically scattered part of the electron wave (electrons with energy loss higher than 10eV were removed). It should be noted that aberrations of the energy filter can affect high-resolution imaging if no zero-loss filtering is applied. Thus, our unfiltered images had a reduced resolution at the atomic level, but relative measurements in a dose series at 2.3 nm resolution were not affected. All image data were recorded at a primary magnification between 40,000 and 57,000.

Electron dose measurements

We determined the electron dose by measuring the current transmitted through a certain condenser aperture as well as the object area that was illuminated using the condenser aperture. A Faraday cup integrated into the viewing screen of the TEM allows one to measure the current transmitted through the aperture when the beam is focused to a spot within the cup. Thanks to the dedicated Köhler illumination in the Carl Zeiss Libra TEM, the condenser aperture defines the object area independently of the illumination setting used (i.e., independently of the convergence angle or illumination intensity). Therefore, the illuminated object area can be determined from an image of the condenser aperture. The ratio of current to illuminated area gives the electron dose rate for a certain illumination setting. Multiplication with the exposure time yields the applied electron dose per image area.

Image processing and data analysis

Images were processed and analyzed using the image-processing software DigitalMicrograph (DM, version 2.31.734.0; Gatan). Intensity profiles across TMV particles were generated using the profile tool of DM, which calculates integrated intensity profiles. We quantified the contrast between sample areas with embedded TMV particles and the surrounding ice area by comparing the interpolated gray value of TMV in the profile with the gray value of the pure amorphous ice. In the same way, we used the profile tool to determine the spatial-frequency-dependent signal (layer line intensities) contained in the power spectra (calculated via the FFT tool of DM) of individual TMV particles.

Results

Guided by the enhanced contrast at energies of ≤80 keV, we collected data to investigate the following questions: What level of spatial resolution can be directly visualized, and can it be used as a readout for contrast and beam-damage studies? How large is the contrast enhancement with decreasing electron energy? Does radiation damage limit the acquisition of useful low-kV data? If we combine contrast enhancement with increased radiation damage, how large is the yield of signal in data collected at low electron energies?

High resolution in low-kV, low-dose images

Fig. 1 A shows a typical zero-loss filtered image of TMV embedded in vitreous ice, recorded with a Cs-corrected EFTEM with an electron energy of 40 kV at an underfocus of ∼100 nm. The image was acquired with an electron dose of 1.2 e−/Å2. The 2.3 nm helical repeat of TMV is well visible in the image. Note the high virus particle contrast even at imaging conditions close to focus. At an accumulated electron dose of 2.3 e−/A2, the power spectrum of the image reveals the presence of the second layer line (Fig. 1 B), which corresponds to a spatial resolution of 11.5 Å.

Figure 1.

Figure 1

Open in a new tab

(A) Zero-loss filtered image of TMV particles embedded in vitreous ice, recorded with a Cs-corrected EFTEM with a primary electron energy of 40 kV at an underfocus of ∼100 nm. The image was acquired with an electron dose of 1.2 e−/Å2. The same TMV particles were imaged repeatedly to analyze beam damage with increasing accumulated electron dose. (B–F) Power spectra (FFT) of the images after accumulating the electron dose displayed in the upper-right corners. At an accumulated electron dose of 2.3 e−/Å2 (B), the layer lines corresponding to the 1.15 nm and 2.3 nm helical repeats are visible, whereas with an increasing accumulated electron dose, the 2.3 nm layer line is fading away (C–F).

This result implies that it may be possible to achieve a sufficiently high resolution to perform a state-of-the-art structural analysis under low-kV conditions, but the high-resolution signal will decay very rapidly. The expected enhanced object contrast is verified in close-to-focus images, which points to the expected increased amplitude contrast at lower imaging energies.

Increased amplitude contrast for weak-phase/weak-amplitude objects at low kV

Biological specimens are weak-phase objects. Conventional low-dose images of this type of specimen at higher imaging energies therefore suffer from low contrast and poor visibility due to the low contribution of amplitude contrast (absorption) and Z contrast (objective aperture).

The traditional way to obtain sufficient contrast is to employ defocusing, thereby generating phase contrast. Images are usually recorded in underfocus because in this case the phase contrast and the amplitude contrast add up at low spatial frequencies, whereas in overfocus the phase contrast and amplitude contrast are reversed and can even cancel out.

At conventionally used electron energies of ≥100 keV, one observes a contrast reversal when the image passes through the focus. The reason for this contrast reversal is the dominating defocusing effect for the contrast transfer function (CTF) and the almost negligible amplitude contrast at higher imaging energies.

This behavior changes at low electron energies, as shown in images of vitrified TMV recorded at 60 kV at an underfocus of 1.9 μm (Fig. 2 A) and an overfocus of 2.1 μm (Fig. 2 B). This observation can be explained by the assumption that the amplitude contrast increases significantly with decreasing electron energy. In this case, the phase contrast does not dominate the CTF at low spatial frequencies. This increase in amplitude contrast is predicted by image formation theory (11) and was experimentally verified for different carbon materials (12, 14). Our experiments show an analogous behavior for ice-embedded biological specimens. Of interest, according to carbon studies, this amplitude contribution at low electron energies contains atomic information, indicating the possible potential for macromolecular imaging at the highest resolution.

Energy dependence of the increased amplitude contrast at low kV and the effect of increased inelastic scattering

Scattering theory predicts an increasing interaction between electrons and matter as the electron energy decreases. This holds not only for the elastic amplitude and phase contribution but also for the inelastic (sample-damaging) scattering. It is important to test these different contributions for different energies. In a simple Gedanken experiment, it becomes obvious that at very low electron energies, there will no longer be an elastically scattered electron wave transmitted through the specimen, and therefore no usable image of the sample can be recorded. At the same time, all electrons will contribute to damaging the specimen. Therefore, one could expect that the balance between the increasing amplitude contrast and decreasing transmitted elastic wave would lead to an optimal specimen contrast at a certain electron energy.

Fig. 3, A–D, show zero-loss filtered images of TMV embedded in vitreous ice recorded in focus with Cs-corrected EFTEMs. The images were acquired at electron energies of 80 keV, 60 keV, 40 keV, and 20 keV, respectively. The applied electron dose and the ice thickness around the dashed areas are similar in all four cases. The increase of image contrast with decreasing electron energy is obvious.

Figure 3.

Figure 3

Open in a new tab

(A–D) Zero-loss filtered images of TMV particles embedded in vitreous ice, recorded with Cs-corrected EFTEMs in focus at electron energies of 80 keV, 60 keV, 40 keV, and 20 keV, respectively. (E–L) Intensity profiles extracted from the regions of interest depicted as dashed boxes in the images in (A)–(D). (E)–(H) correspond to the zero-loss filtered data, and (I)–(L) correspond to the unfiltered data. The contrast of the TMV particle against the amorphous ice layer is indicated in each case. Table 1 displays the average contrast in zero-loss filtered and unfiltered images of TMV depending on the electron energy. The contrast ratio of zero-loss filtered/unfiltered images is given in the bottom row. Fig 3. (M) Visualization of the results shown in Table 1.

To allow a quantitative comparison of the dependence of amplitude contrast on electron energy, we extracted intensity profiles across TMV strands in several data sets (three TMV strands at 20 keV, eight at 40 keV, five at 60 keV, and five at 80 keV). All images were acquired in focus with Cs-corrected EFTEMs at a similar electron dose. The Cs correction ensures that the effect of spherical aberration on the CTF (even in the case of nonlinear image formation) is eliminated. When in addition images are recorded in focus, the phase contrast contribution is also largely eliminated. Thus, we can directly test the increased amplitude contrast contribution, which is known to increase at a higher rate than the phase contrast contribution (12). For our analysis, we chose areas with a similar ice thickness of ∼40 ± 10 nm.

Examples of the extracted intensity profiles are shown in Fig. 3, E–L. The intensity profiles were extracted from the regions of interest depicted as dashed boxes in Fig. 3, A–D. We defined contrast as the image intensity difference between areas of pure ice Iice and the intensity of areas with TMV ITMV divided by the intensity in areas of pure ice, C = (Iice – ITMV)/ Iice.

The contrast for the zero-loss filtered images in Fig. 3, A–D, was determined to be 4.3% at 80 keV, 6.0% at 60 keV, 8.5% at 40 keV, and 14% at 20 keV (i.e., an increase by almost a factor of 2 when going from 80 keV down to 40 keV).

To quantify the contrast enhancement generated by zero-loss filtering dependent on electron energy, we acquired unfiltered images of TMV embedded in amorphous ice in addition to the zero-loss filtered images and performed the same analysis as described in the previous section for the zero-loss filtered images (Fig. 3, I–L). Since we are studying contrast enhancement at low resolution, and since the influence of chromatic aberration of the filter unit influences only the high-resolution part of the images, we can neglect chromatic aberrations in further discussion (18). The contrast of the TMV particles for the unfiltered images was determined to be 2.1% at 80 keV, 3.4% at 60 keV, 3.8% at 40 keV, and 6.8% at 20 keV (Fig. 3, I–L; corresponding image data not shown). These data show that for decreasing electron energy, the same amount of contrast enhancement is found in unfiltered images as in zero-loss filtered data.

The results for a larger data set of unfiltered and zero-loss filtered images are summarized in Table 1 within Fig. 3, which shows the averaged results for electron energies of 20, 40, 60, and 80 keV. The bottom row shows the ratio of amplitude contrast obtained by zero-loss filtering versus unfiltered imaging. It can be concluded that zero-loss filtering enhances contrast by a factor of 2 as compared with unfiltered imaging, in agreement with previous results obtained for higher electron energies (18) and completely in accordance with basic scattering physics. Furthermore, a slightly higher enhancement of contrast is observed for lower electron energies. This further increase can be explained by the growing effect of multiple scattering events in a specimen of the same thickness at lower electron energies, as discussed below.

The results (summarized in Fig. 3, Table 1) are visualized in Fig. 3 M. The decrease of amplitude contrast with increasing electron energy can be fitted in a first-order approximation with a function proportional to the scattering cross section for electrons in matter, which is proportional to 1/U (where U is the electron energy) (12).

Decreased tolerable electron dose at low kV

In the same way as contrast increases with decreasing electron energy due to the rising scattering cross section, there is also an increase in radiation damage. Visible indications of beam damage in ice-embedded biological samples include loss of structural resolution, ice crystallization, and bubbling effects of the sample. To elucidate the advantages and disadvantages of low-kV, low-dose imaging, we studied the combined effect of contrast and beam damage. For this purpose, we analyzed the dependence of the observable beam damage on the electron energy.

Dependence of ice crystallization on electron dose and electron energy

Beyond a certain electron dose, amorphous ice begins to crystallize. The crystallization is initiated in crystal nuclei and spreads over the amorphous ice film as the total electron dose increases. When this happens, the specimen structure will be destroyed by the disruptive force of the ice crystals. In general, in cryo-electron microscopy (cryo-EM), ice crystallization is seen as a sign of an excessive electron dose and a completely destroyed sample. Therefore, we verified that ice crystallization started at electron doses that were beyond the doses that are critical for ionization damage of a biological object for electron energies down to 20 keV.

We collected several electron-dose image series of TMV in vitreous ice at electron energies of 20 keV (five experiments), 40 keV (seven experiments), 60 keV (four experiments), and 80 keV (five experiments) to monitor the accumulated electron dose within the observed specimen area. Fig. 4 shows one example of TMV strands in vitrified ice recorded at an electron energy of 60 keV with electron doses of 9 e−/Å2 up to 825 e−/Å2. The onset of ice crystallization becomes obvious in Fig. 4 B, where several crystallization nuclei can be detected at an accumulated electron dose of 245 e−/Å2. This electron dose is well beyond the electron dose applied in low-dose imaging. The appearance of the TMV strands at this electron dose shows that the TMV structure is already badly damaged. In addition, bubbling can be observed in Fig. 4, C and D, which show images acquired with an accumulated electron dose higher than 300 e−/Å2.

Figure 4.

Figure 4

Open in a new tab

(A–D) Images of TMV in vitreous ice, acquired with a Cs-corrected EFTEM at 60 keV, show beam damage after accumulation of a certain electron dose. (B) Ice crystallization starts at an accumulated electron dose of ∼245 e−/Å2. (C) The first signs of bubbling in the TMV can be observed at an even higher electron dose. (E) Measurements of the minimum accumulated electron dose at which crystallization of amorphous ice becomes detectable in images of vitrified TMV for the different electron energies. For every observed electron energy, ice crystallization starts at an electron dose that is well above the electron dose that is critical for structural damage.

For each of the collected image series, we determined the electron dose at which the first signs of ice crystallization could be detected. The results of our analysis are summarized in Fig. 4 E. As expected, the minimal electron dose for ice crystallization increased with increasing electron energy. At each observed electron energy, this electron dose was well above the critical electron dose for beam damage of our 2.3 nm structural signal. In the following text we will focus on this measure only, although we are well aware that higher resolution may already be affected at a lower electron dose.

Analysis of structural damage dependent on electron dose and electron energy

Radiation damage in organic specimens is mainly caused by ionization of their constituting molecules and ultimately leads to destruction of the object structure. With an increasing accumulated electron dose, the object structure is increasingly affected, until the object is totally deformed. The scattering cross section for inelastic scattering increases with decreasing electron energy in proportion to 1/U (where U is the electron energy). This should be reflected in the amount of electron dose at which a certain degree of structural damage is observed.

We acquired a significant number of electron dose image series of TMV embedded in vitreous ice at various electron energies (five series at 80 keV, two series at 60 keV, nine series at 40 keV, and four series at 20 keV) to analyze the structural damage depending on the accumulated electron dose at various electron energies. An example of our analysis is shown in Fig. 1, B–F. Each image was acquired with an electron dose of 2.3 e−/Å2, and the accumulated electron dose is indicated as an overlay in each image. A closer look at the data reveals that the layer line of the 2.3 nm helical repeat is visible in the first power spectrum (Fig. 1 B) but fades away after the third image (Fig. 1 D) and becomes invisible in the last image (Fig. 1 F), which corresponds to an accumulated electron dose of 21 e−/Å2.

The analysis of the entire data set is summarized in Fig. 5 A. The data points represent the accumulated electron dose at which the first layer line of TMV disappears. Obviously, the tolerable accumulated electron dose decreases with decreasing electron energy.

Figure 5.

Figure 5

Open in a new tab

(A) Decay of the helical periodicity of TMV in vitreous ice as a function of electron energy. The data points represent the accumulated electron dose at which the 2.3 nm layer line of TMV disappears in the power spectrum of the image, as shown for example in Fig. 1, B–F, where the layer line has disappeared at 21 e−/Å2. Each data point corresponds to one electron dose series. As expected, radiation damage appears at a lower accumulated electron dose with decreasing electron energy. (B) Decay of the SBR of the 2.3 nm helical layer line of TMV in vitreous ice as a function of accumulated electron dose and electron energy. The data points represent the SBR of the first layer line of TMV as determined from the power spectrum for a given accumulated electron dose and electron energy. The data imply an optimal yield of object signal at an energy of ∼60 keV for imaging ice-embedded TMV under the chosen experimental conditions. See text for details.

Signal/background ratio at low electron energies: specimen thickness becomes a limiting factor

To avoid beam damage, the applied electron dose must be decreased with decreasing electron energy. At the same time, contrast is increased significantly. The crucial issue, therefore, is the amount of signal that can be obtained from the object under investigation at a certain electron energy.

To examine this issue, we analyzed our data set of zero-loss filtered images of vitrified TMV by determining the signal/background ratio (SBR) of the first layer line in the power spectrum of the TMV strands depending on the accumulated electron dose and electron energy. Our data are summarized in Fig. 5 B. The data are corrected for length variations in the evaluated TMV strands. This means that the measured SBR is divided by the square root of the length of the TMV strand, since the observed SBR is proportional to the square root of the number of contributing helical repeats, N: SBR = SBRTMV/SQR(length TMV).

We observed a relatively broad scattering of the data points that might have come from 1) variation in the defocus value of the images, which can have a strong influence on the intensity of the layer lines; 2) imperfection of the helical structure along the averaged TMV rod; and 3) some inaccuracy in dose determination.

The analyzed data imply that the best SBR is obtained at an electron energy of 60 keV under the chosen experimental conditions. The 80 keV data points show an almost equally high SBR, but it is obvious that the SBR is significantly lower at 40 keV and weakest at 20 keV. This breakdown of SBR at 40 keV is supported by the fact that the electron detector we used (TVIPS TemCam-F416) has an optimal conversion factor for single electron events at 40 keV electron energy (Hans R. Tietz, TVIPS, personal communication). If the conversion factor were equal for all electron energies, the SBR at 40 keV would be even lower compared with other electron energies. Similarly, the favorable electron conversion factor for 60 kV electrons could explain why the SBR is slightly better than it is at an electron energy of 80 kV.

The decreased SBR in the low-kV data results from the increased scattering cross sections at low electron energies: around 40 keV the mean free path for inelastic scattering in amorphous water drops below 40 nm and the mean free path for inelastic scattering in carbon drops below 20 nm. In our analysis, we used data collected from TMV strands ∼15 nm thick embedded in a vitrified ice layer with a thickness of ∼40 nm. This means that multiple scattering becomes significant. The amount of elastic scattering that generates both phase and amplitude contrast is significantly reduced, leading to a decrease of SBR in the images.

Discussion

Investigators have long sought to develop technologies to improve image contrast for biological samples embedded in ice. High-brightness electron guns, zero-loss energy filtering (18, 19, 20), and more recently physical phase plates (21, 22, 23) have been tested, but so far none of these techniques has proven to be a game changer for high-resolution macromolecular imaging. In contrast, direct electron detectors have proven to be indispensable for any state-of-the-art high-resolution imaging, and now the use of low electron energies, which has tremendously improved transmission electron microscopy for typical carbon materials, could lead to another paradigm change.

Our experiments have shown that the situation for ice-embedded biological macromolecules is not as simple as that for freestanding carbon materials. From our data, it is obvious that lowering the energy of the imaging electrons does indeed improve contrast, but at the same time more beam damage is accumulated. This is to be expected for samples in which knock-on damage is not the predominant damaging mechanism.

Previous work (10, 24) and the data presented here can now be combined in a very intuitive picture. Any kind of scattering or excitation interaction of the electrons will increase when the energies of the imaging electrons are decreased. The elastic interaction will yield more phase and, according to image formation theory and experiment, even higher amplitude signal, but at the same time inelastic interactions will increase the level of damage and reduce the fraction of single elastic scattering events.

The expected larger increase of amplitude contrast (11, 12) is verified here. It leads to an optimal electron energy, such that amplitude contrast is increased more than inelastic scattering, and thus the net effect results in an optimal energy for object contrast versus beam damage. At very low electron energies, however, this effect cannot sufficiently counteract the increase in multiple scattering and the increased beam damage will win. We see this behavior even for a medium-resolution readout at 2.3 nm, which strongly suggests that higher resolution would also not gain from lower electron energies. In addition, multiple elastic/inelastic scattering increases with the sample thickness or lower electron energy, and in the end all electrons are scattered inelastically and damage the sample. In other words, the effective thickness of our conventionally prepared ice-embedded samples increases beyond that of any usable object. Therefore, we cannot expect that low-energy imaging of conventionally prepared samples would provide better data than state-of-the-art 200–300 keV imaging.

Two possible new approaches, however, could change this picture in the future. First, novel sample-preparation protocols, such as preparing biomolecules within a largely reduced embedding structural water layer, could lead to thinner samples. Second, inelastically scattered electrons could be used for high-resolution imaging. In that case, the embedding ice, which effectively increases the specimen thickness and leads to the majority of inelastically scattered electrons, could still be accepted. During the imaging process, it would act as a diffuse scatterer, providing an incoherent illumination of the biological sample.

Investigators have obtained very thin samples of carbon material, e.g., by embedding the material within a sandwich of graphene layers (25). Future work will have to reveal whether such samples can also be prepared from biological macromolecules without changing the macromolecular structure. In addition, it remains to be tested how changing the electron dose rate can affect beam damage. Such experiments are now possible because state-of-the-art direct electron detectors allow for movie-data collection at high frame rates. An analysis of the obtainable resolution and accumulated beam damage, comparable to that reported by Fromm et al. (2), should yield conclusive data regarding high and low electron flux at different electron energies ranging from 80 to 20 keV. Less obvious is the idea of using inelastically scattered electrons for high-resolution imaging. It has been shown that with the use of a TEM corrected for chromatic aberration, very thick samples can be imaged in three dimensions (26), but a careful analysis of the resolution obtained from such samples shows that resolution might be limited to the nanometer range (27).

Author Contributions

E.M., I.A., and R.R.S. conceived, designed, and performed experiments; analyzed data; prepared digital images; and drafted the manuscript. U.K. conceived the original low-voltage SALVE Project at Ulm University. All authors were involved in discussing the data and manuscript.

Acknowledgments

The authors thank Gerd Benner and Alexander Orchowski (Carl Zeiss Microscopy GmbH, Oberkochen, Germany) for discussions and support in converting conventional TEMs into low-voltage microscopes suitable for cryo-EM. We thank Max Haider (CEOS GmbH, Heidelberg, Germany) and Harald Rose (University Ulm, Ulm, Germany) for stimulating discussions about microscope usage and image formation at low electron energies. The TMV test sample was kindly provided by Dr. Anan Kadri (Biologisches Institut, Universität Stuttgart, Stuttgart, Germany). We also thank Dr. Götz Hofhaus (Cryo Electron Microscopy CellNetworks, BioQuant, Universitätsklinikum Heidelberg, Germany) for expert help with sample preparation.

Editor: Andreas Engel

References

  • 1.Scanlan D. Ecology. Bacterial vesicles in the ocean. Science. 2014;343:143–144. doi: 10.1126/science.1248566. [DOI] [PubMed] [Google Scholar]
  • 2.Fromm S.A., Bharat T.A.M., Sachse C. Seeing tobacco mosaic virus through direct electron detectors. J. Struct. Biol. 2015;189:87–97. doi: 10.1016/j.jsb.2014.12.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Bartesaghi A., Merk A., Subramaniam S. 2.2 Å resolution cryo-EM structure of β-galactosidase in complex with a cell-permeant inhibitor. Science. 2015;348:1147–1151. doi: 10.1126/science.aab1576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Schröder R.R. Advances in electron microscopy: A qualitative view of instrumentation development for macromolecular imaging and tomography. Arch. Biochem. Biophys. 2015;581:25–38. doi: 10.1016/j.abb.2015.05.010. [DOI] [PubMed] [Google Scholar]
  • 5.Kaiser U., Biskupek J., Benner G. Transmission electron microscopy at 20 kV for imaging and spectroscopy. Ultramicroscopy. 2011;111:1239–1246. doi: 10.1016/j.ultramic.2011.03.012. [DOI] [PubMed] [Google Scholar]
  • 6.SALVE III Project. http://www.salve-project.de/, accessed January 29, 2016.
  • 7.Kotakoski J., Krasheninnikov A.V., Meyer J.C. From point defects in graphene to two-dimensional amorphous carbon. Phys. Rev. Lett. 2011;106:105505. doi: 10.1103/PhysRevLett.106.105505. [DOI] [PubMed] [Google Scholar]
  • 8.Girit C.Ö., Meyer J.C., Zettl A. Graphene at the edge: stability and dynamics. Science. 2009;323:1705–1708. doi: 10.1126/science.1166999. [DOI] [PubMed] [Google Scholar]
  • 9.Nakamura E. Movies of molecular motions and reactions: the single-molecule, real-time transmission electron microscope imaging technique. Angew. Chem. Int. Ed. 2013;52:236–252. doi: 10.1002/anie.201205693. [DOI] [PubMed] [Google Scholar]
  • 10.Egerton R.F. Control of radiation damage in the TEM. Ultramicroscopy. 2013;127:100–108. doi: 10.1016/j.ultramic.2012.07.006. [DOI] [PubMed] [Google Scholar]
  • 11.Rose H. Theoretical aspects of image formation in the aberration-corrected electron microscope. Ultramicroscopy. 2010;110:488–499. doi: 10.1016/j.ultramic.2009.10.003. [DOI] [PubMed] [Google Scholar]
  • 12.Hayashida M., Kawasaki T., Takai Y. Estimation of suitable condition for observing copper-phthalocyanine crystalline film by transmission electron microscopy. Nucl. Instrum. Methods Phys. Res. B. 2006;248:273–278. [Google Scholar]
  • 13.Lee Z., Meyer J.C., Kaiser U. Optimum HRTEM image contrast at 20 kV and 80 kV—exemplified by graphene. Ultramicroscopy. 2012;112:39–46. doi: 10.1016/j.ultramic.2011.10.009. [DOI] [PubMed] [Google Scholar]
  • 14.Meyer J.C., Kisielowski C., Zettl A. Graphene at the edge: stability and dynamics. Science. 2009;323:1705–1708. doi: 10.1126/science.1166999. [DOI] [PubMed] [Google Scholar]
  • 15.Rose H. Direct imaging of lattice atoms and topological defects in graphene membranes. Nano Lett. 2008;8:3582–3586. doi: 10.1021/nl801386m. [DOI] [PubMed] [Google Scholar]
  • 16.Bell D.C., Russo C.J., Kolmykov D.V. 40 keV atomic resolution TEM. Ultramicroscopy. 2012;114:31–37. doi: 10.1016/j.ultramic.2011.12.001. [DOI] [PubMed] [Google Scholar]
  • 17.Haider M., Uhlemann S., Urban K. Electron microscopy image enhanced. Nature. 1998;392:768–769. [Google Scholar]
  • 18.Langmore J.P., Smith M.F. Quantitative energy-filtered electron microscopy of biological molecules in ice. Ultramicroscopy. 1992;46:349–373. doi: 10.1016/0304-3991(92)90024-e. [DOI] [PubMed] [Google Scholar]
  • 19.Angert I., Majorovits E., Schröder R.R. Zero-loss image formation and modified contrast transfer theory in EFTEM. Ultramicroscopy. 2000;81:203–222. doi: 10.1016/s0304-3991(99)00190-4. [DOI] [PubMed] [Google Scholar]
  • 20.Yonekura K., Braunfeld M.B., Agard D.A. Electron energy filtering significantly improves amplitude contrast of frozen-hydrated protein at 300kV. J. Struct. Biol. 2006;156:524–536. doi: 10.1016/j.jsb.2006.07.016. [DOI] [PubMed] [Google Scholar]
  • 21.Murata K., Liu X., Chiu W. Zernike phase contrast cryo-electron microscopy and tomography for structure determination at nanometer and subnanometer resolutions. Structure. 2010;18:903–912. doi: 10.1016/j.str.2010.06.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Frindt N., Oster M., Schröder R.R. In-focus electrostatic Zach phase plate imaging for transmission electron microscopy with tunable phase contrast of frozen hydrated biological samples. Microsc. Microanal. 2014;20:175–183. doi: 10.1017/S1431927613013901. [DOI] [PubMed] [Google Scholar]
  • 23.Danev R., Buijsse B., Baumeister W. Volta potential phase plate for in-focus phase contrast transmission electron microscopy. Proc. Natl. Acad. Sci. USA. 2014;111:15635–15640. doi: 10.1073/pnas.1418377111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Egerton R.F., Rauf I. Dose-rate dependence of electron-induced mass loss from organic specimens. Ultramicroscopy. 1999;80:247–254. [Google Scholar]
  • 25.Algara-Siller G., Kurasch S., Kaiser U. The pristine atomic structure of MoS2 monolayer protected from electron radiation damage by graphene. Appl. Phys. Lett. 2013;103:203107. [Google Scholar]
  • 26.Wacker I.U., Röder I.V., Schröder R. Studying membrane topology of the neuromuscular junction by electron tomography of thick resin slices in a Cs/Cc-corrected TEM. Microsc. Microanal. 2010;16(Suppl 2):972–973. [Google Scholar]
  • 27.Baudoin J.-P., Jinschek J.R., de Jonge N. Chromatic aberration-corrected tilt series transmission electron microscopy of nanoparticles in a whole mount macrophage cell. Microsc. Microanal. 2013;19:814–820. doi: 10.1017/S1431927613001475. [DOI] [PubMed] [Google Scholar]

Articles from Biophysical Journal are provided here courtesy of The Biophysical Society

RESOURCES