Comparison of 3D Imaging Methods in Electron Microscopy for Biomaterials

Introduction

It is important to know the 3D structure and localization of organelle and protein complex that are components of cells to understand the functions of cells and tissues. This information plays important roles not only in academic research but also in development of therapy for lethal diseases. Previously, TEM tomography has been used for 3D structure observation of these nano-structures. In TEM tomography, we take micrographs of serial tilting images, and reconstruct the 3D image by back-projection of obtained image series. TEM tomography provides high resolution 3D images, and also has a thickness limitation defined by transmission of the electron beam. Therefore, it is difficult to observe the 3D image of whole cell or tissues by TEM tomography. Recent histology or cell biology is strongly connected with molecular biology, and thus it is important to reveal the influence of the morphology and distribution of nano-structures on tissues and whole cell. For that reason, new microscopic methods have been developed, which enable us to observe a large area where the whole cell or tissues is recognized, and can maintain high-resolution while recognizing organelle and protein complexes. In this study, we focus on three kinds of new 3D observation methods, FIB-SEM, SBF-SEM, and Array tomography [1, 2, 3] (Fig. 1).

Fig.1 Schematic diagram of 3D reconstruction methods.

• SBF-SEM:
  SBF-SEM repeats the sample cutting by a diamond knife and the observation of backscattered electron image of the new sample surface in the specimen chamber.
• FIB-SEM:
  FIB-SEM repeats the sample sputtering by a focused gallium ion beam and the observation of backscattered electron image of the new sample surface in the specimen chamber.
• Array tomography:
  Serial sections are prepared with an ultramicrotome and mounted on a hard substrate such as a silicon wafer. These serial sections are taken for a backscattered electron image one by one and stacked in order.
• TEM tomography:
  TEM tomography takes serial-tilting projection images by TEM, and reconstructs the internal 3D structure by back projection of these images.

FIB-SEM is a SEM equipped with a Focused ion beam (FIB) column. A sample in the specimen chamber is sputtered by FIB, and the sputtered surface can be observed by SEM. FIB is not popular in biology but is popular in materials science. This is because specimens in materials science (metals, ceramics, etc.) are too hard to slice by a diamond knife, but can be made into a thin section or exposed new surface by sputtering with FIB. To observe the 3D structure with the FIB-SEM, we sputter the surface of the specimen, and observe the new exposed surface by backscattered electron imaging with SEM, and repeat the process of sputtering and observation. Finally, we can reconstruct the 3D structure by stacking the images of serial section series [1]. The advantages of FIB-SEM are high-precision determination of sputtering sites and preparation of thin sections from hard samples like bones and metals. On the other hand, the disadvantages are narrow observation area and slow sputtering speed.
In the SBF-SEM, we use a SEM that has a mechanism for cutting samples with a diamond knife in its specimen chamber. To reconstruct the 3D image, the surface of the sample is sliced by the diamond knife in the specimen chamber, and a new exposed surface is observed by the SEM. 3D images are reconstructed by stacking the images of serial section series like the FIB-SEM [2]. Advantages of SBF-SEM are high cutting speed and wide cutting area, since samples are cut by the diamond knife. On the other hand, there is a disadvantage that SBF-SEM needs special sample preparation (ex. NCMIR method) because samples are required to be prepared with high electron conductivity and strong heavy metal staining [3]. The preservative quality of nano-structures by this special sample preparation is poorer than the conventional preparation for TEM. Although the conductivity of the sample is increased by the special sample preparation, occasionally the sample may become charged.
In Array tomography, first we prepare the specimen of ultra-thin section ribbon by an ultramicrotome, and serial sections are mounted on the conductive wafer like a silicon wafer. These serial sections are observed piece by piece with a SEM, and 3D images are reconstructed to stack the images in order [4]. Advantages of this method are as follows; Suppression of charging of samples that are very thin and mounted on the conductive wafer, Low initial cost because this method needs only a conventional SEM and an ultramicrotome, and we can use good samples prepared by the conventional preparation techniques for TEM. A disadvantage of this method is that a lot of manual efforts are required, for example, preparation of serial section series.
There is little discussion to compare the features and specifications of these three methods, because the principle of these 3D reconstruction methods with SEM is almost the same. That is, the three methods enable us to make serial sections and observe new sections. In this study, we observed 3D structures of the same sample by these methods, and compared the results. Finally, we discuss and summarize the features of these methods and which sample and analysis are suitable. Furthermore, we attempted analysis of some sample using the suitable methods.

Materials and Methods

We used E.coli infected by T4 phage, λ phage, and φ174 phage as samples. Phages are a kind of virus infecting bacteria. The size is 200 nm or less, and it is impossible to observe their morphology without using an electron microscopy (Fig. 2 a-c).
Phages inject their genome such as DNA or RNA into E.coli when phages attach to E.coli. The infected phages take over the transcription system, the translation system, and the replication system of host cells, and proliferate itself in host cells. In this stage, we can observe proliferating phages in E.coli (Fig. 2 d-f). When phages proliferate sufficiently in the host, phages burst and go out from the host cell. New proliferated phages attach and infect the new host cell.
First, we piled up the soft LB agar containing E.coli on hard LB agar. Phages were infected with E.coli by dropping the phage containing solution onto this agar. We were able to recognize the area where the phages were infected with E.coli , as a plaque after incubation of this agar, and obtained the phage infected E.coli by taking out these plaques.
For TEM tomography and Array tomography, samples were prefixed by 2.5% glutaraldehyde in 0.1 M sodium cacodylate (pH 7.2), and postfixed 1% OsO₄ in 0.1 M sodium cacodylate (pH 7.2). Fixed samples were embedded in epoxy resin. The samples were sliced by an ultramicrotome into thin sections which were mounted on silicon wafers for Array tomography. These thin sections were stained by uranium acetate and lead citrate.
For FIB-SEM and SBF-SEM, we applied the NCMIR method for strong heavy metal stain [3]. FIB-SEM can also use samples that are prepared by the conventional method, and in this study we applied the NCMIR method to obtain the strong contrast image. After fixation and staining, the samples were embedded in epoxy resin. Observation conditions are shown in Table 1.
Serial section images obtained by FIB-SEM, SBF-SEM and Array tomography were aligned by Fiji [5], and 3D images were reconstructed by stacker (System In Frontier Inc. Japan). In TEM tomography, TEMography was used for acquisition of serial tilting images and 3D reconstructions. Obtained 3D images were segmented and analyzed by Colorist (System In Frontier Inc. Japan).

Fig.2 TEM images of phages.

Negative stained images of phage (a-c) and proliferating in E.coli (d-f, arrow heads).

Results and Discussions

Comparison of 3D reconstruction method using electron microscopy

We observed 2D images using FIB-SEM, SBF-SEM (backscattered electron images of the surface of bulk sample, accelerating voltages were 1.0 kV and 3.0 kV, respectively), Array tomography (backscattered electron images of thin section sample mounted on silicon wafer, accelerating voltage was 7.0 kV), and TEM tomography (projection image of thin section sample, no tilting, accelerating voltage was 120 kV). FIB-SEM and SBF-SEM for observation of the surface of bulk sample were inferior to Array tomography and TEM tomography for observation of thin section in the sharpness of images (Fig. 3). This is because, in this sample, most observation area was occupied only with low conductivity resin and we had to observe with low accelerating voltage to suppress the charging. The image quality of Array tomography is equal to that of TEM tomography. This result suggested that improvement of sample conductivity and observation at high accelerating voltage are important factors for observing biological samples in 3D structural observation with SEM (Fig. 3 c). In addition, the image quality of FIB-SEM and SBF-SEM is low compared with the other two methods, but they kept the resolution to recognize phage particles.
Next, we compare results of 3D reconstruction. In TEM tomography, the thickness of the slice (200 nm) is smaller than the diameter of E.coli (about 2 μm), so that the whole image of one E.coli cannot be reconstructed into 3D (Fig. 4 d). On the other hand, SBF-SEM and FIB-SEM succeeded in 3D observation of a large volume (Fig. 4 a, b). In Array tomography, the observation area of the XY plane is comparable to those of SBF-method and FIB-SEM. However, the thickness of observation volume was reduced because the prepared slices are only 20 in this study (Fig. 4 c). Although mature skills and long acquisition time were required, we consider that it is possible to increase the thickness of 3D reconstruction image by preparing mass serial sections and increasing the number of micrographs taken. In results of extracting one E.coli and segmentation of phages contained in the E.coli and the extracellular membrane of E.coli , we could recognize the morphology of the E.coli and 3D distribution of phages using any method, although there are differences in spatial resolution (Fig. 4 e-h). Focusing on the one phage particle, the shape of phage is extended in the Z direction in SBF-SEM and Array tomography, since the resolution along the Z is insufficient against the size of phages (Fig. 4 i, k). In FIB-SEM, although a phage was reconstructed as a ball shape, however, details of the 3D reconstruction image are lost compared to the result of TEM tomography (Fig. 4 j, l). Considering the result of the comparisons, it was found that the quantitative analysis of morphology of E.coli and the analysis of distribution of phages in E.coli can be performed by any 3D reconstruction method with SEM, meanwhile, TEM tomography provides high spatial resolution necessary for observing the nano-structure such as phages. Very similar results were obtained with SBF-SEM and FIB-SEM, which have the common basis of slicing and observing bulk samples in their specimen chambers. Almost the same volumes were obtained by SBF-SEM and FIB-SEM in this study, and the acquisition time of SBF-SEM (2.5 hours) is about 20 times as fast as that obtained by FIB-SEM (48 hours) (Table 1). However, FIB-SEM provided the reconstructed 3D structure image with higher Z resolution than SBF-SEM. The difference of this resolution is due to the difference of cutting thickness: 20 nm for FIB-SEM and 50 nm for SBF-SEM.
These results were summarized in Table 2. The image quality of SBF-SEM was inferior to that of other methods, however the wide range of observation area and the speed of observation were unrivaled with SBF-SEM. Furthermore, in samples with high conductivity, the image quality would be improved because the accelerating voltage can be raised more. For the above features, SBF-SEM is suitable for very large samples with high conductivity and for comprehensive analysis. For example, tissue samples are large, the resin area is small, and conductivity is high. Comprehensive analysis is required in connectome analysis.
FIB-SEM needs long acquisition time compared with SBF-SEM, however, it is advantageous for FIB-SEM that Z resolution is very high. This advantage is only valid for a narrow observation area. FIB-SEM is suitable for small samples, for example, small tissues or one whole cell. FIB-SEM has high affinity for quantitative analysis, since high quantitativeness needs the high resolution. In addition, it is possible to cut bones and metals, enabling 3D analysis of hard tissues and contact portion between metals and tissues.
In Array tomography, samples remain after the observation; meanwhile, matured techniques that make serial sections and long acquisition time were required to deepen the observable depth. Array tomography is suitable for detailed analysis of rare phenomena in a large observation area. We expected that the transparent substrate such as glass allows for various correlative methods, including correlative light and electron microscopy (CLEM). Array tomography is effective for observation of floating cells, bacteria and embryos which have low conductivity, surrounded by resin around samples, since the samples are thin sections on a conductive substrate, charging are suppressed and the samples can be observed with high accelerating voltage.

Fig.3 Comparison of observation area and image quality of each method.

In SBF-SEM, the specimen is easily charged, deteriorating the image quality, because block resin that is little conductive occupies a wide observation area (a). FIB-SEM also enables to observe the wide block resin area, but gallium ions that sputtered the sample surface suppress the charging (b). In Array tomography, we were able to observe high signal-to-noise (SN) ratio images, because the charging was suppressed for the sliced thin sections mounted on a silicon wafer that is highly conductive (c). In TEM tomography, we observed only one E.coli (d).

Fig.4 Comparison of 3D reconstruction results.

We reconstructed the 3D structure of E.coli infected with T4 phages. Volume rendering images of the whole observation area (a-d). Segmentation images of one E.coli cell that is infected with T4 phages (green), T4 phages that is grown in E.coli (orange), and infected (pink) (e-h). Segmentation images of one T4 phages in E.coli (i-l).

Analysis of 3D reconstruction image

① The relationship of morphology of E.coli that is infected by T4 phages and internal phage particle

SBF-SEM was enough to analyze the morphology of E.coli and distribution of T4 phage. We analyzed the relationship of the morphology of E.coli and internal phages by using the result of SBF-SEM. First we examined the relationship between the volume of E.coli and the number of phages (Fig. 5 a). The results show that the maximum number of internal phage in an E.coli is about 280 particles, and no E.coli containing more phage particles could be observed. We considered that as the number of phage particles increases further, this increase causes E.coli to burst. The number of contained phage particles tends to increase, as the volume of E.coli increases. However, this relationship is not linear; there was a difference in the number of phage particles even in the same volume E.coli (B and C, D, E in Fig. 5 a). Therefore, we selected characteristic E.coli (A-E in Fig. 5 a, c). Selected E.coli cells are as follows. The ones were small volume and they have few phage particles (A, B in Fig. 5 a, c). Next ones were also small volume, but they had many phage particles (C-E in Fig. 5 a, c). Another ones were large volume and they had a lot of phage particles (F in Fig. 5 a, c). E.coli cells were divided into three parts in the long axis direction, center part, middle part, and side part. We analyzed which part of the T4 phage particles was localized (Fig. 5 b). Most of phage particles localized at the center part, when phage particles were few. The phage particles localized in the center part were dispersed from the middle part to side part, as the internal phage particles increase. Furthermore, as the volume of E.coli increased, phage particles in the center part decreased more, and phage particles localized in the side part increased. Next, we focused on the shape of E.coli . The shape of E.coli that had few T4 phage particles was rod like shape (Fig. 5 c B). When internal T4 phage particles were increased, E.coli expanded to the short axis direction, and the shape became barrel like shape (Fig. 5 c C-E).

Fig.5 Relationship of morphology of E.coli and involved phages.

The left graph shows the relationship of the volume of E.coli infected with T4 phage and the number of involved T4 phages (a). We selected six characteristic E.coli cells (pink dot in the left graph, A-F). The morphology of these 6 E.coli cells that involved T4 phages were visualized by segmentation (c), the right graph shows the relationship of the volume of E.coli and the distribution of T4 phage (b).

② The relationship between type of phage and distribution of phage in E.coli

T4 phage could be recognized by any 3D structure analysis with SEM since it has strong contrast in E.coli . However,λ phage andφ174 phage were difficult to recognize in E.coli because they had low contrast in E.coli , andφ174 phage was too small (Fig. 2 e, c, f). Therefore, the comparison of the E.coli infected by T4 phage,λphage andφ174 phage was done using 3D reconstruction images by Array tomography. First we focused on the morphology of the E.coli . The shape of the E.coli infected byλphage changed from a rod-like structure to a ball-like structure. We separated E.coli infected phages into three layers; inner layer, middle layer, and outer layer, and investigated which phages are localized in which layer (Fig. 6 a). The result show that T4 phage andφ174 phage dispersed in E.coli without bias, however, almostλ phage particle localized at the outer layer. T4 phage andφ174 phage are lytic phage. As soon as they are infected with E.coli , they begin to proliferate in host cells and host cells burst (lyse). On the other hand,λphage is a temperate phage. When they are infected with E.coli and inject their genome into host cells, they do not lyse immediately. After infection, they insert their genome into the genome of host and behave as part of the host (lysogenization). However, the repressor which suppresses the proliferation of the phage is inactivated by the change of environment, phage start to proliferate and the host lyse. The difference in observed morphological change and localization of phages may originate from proliferation area, lytic phages proliferate in infected area immediately and the temperate phages proliferate on the host genome.φ174 phages did not show the biased distribution in the three layers, but they localized in specific part (Fig. 6 d). In this experiment, we could not identify the first infected place, but the localized place may be related to the first infected place.

Fig.6 3D distribution of phages in E.coli

We compared distributions of phages from the results of 3D structure reconstruction of T4 phage, λphage, and φ174 phage form Array tomography. We separated E.coli into three parts; the inner area, middle area and outer area, and counted the number of phages involved in each part. T4 phages and φ174 phages have little bias of distribution, however almostλphages localized at the outer area (a). E.coli that infected withλphage changed their shapes from rod to globular (c).

Conclusion

Each 3D reconstruction method with SEM has its own strength and weakness depending on the sample preparation and acquisition conditions such as accelerating voltage. In SBF-SEM and FIB-SEM, data acquisition is performed automatically, on the other hand, after segmentation analysis is often performed manually. Segmentation that extracts tissues and regions of interest is often performed based on image contrast. However, biological samples have small differences in contrast, and sometimes the contrast alone may not determine the segmentation. It is possible to segment with the morphology as a landmark by keeping high resolution of the image, and thus it helps improve segmentation efficiency. SBF-SEM and FIB-SEM need to introduce the dedicated instruments, while Array tomography can be started with conventional SEM and ultramicrotome, but some skills are required. The choice of which method to use is an important strategy in promoting research efficiently. We hope this paper is helpful for researchers who are planning to start the 3D analysis with electron microscopy.

Acknowledgments

The author would like to thank JEOL people who support this study; H. Nishioka for supporting the management in this study, M. Suga for the support of data acquisition of Array tomography, H. Matsushima for the support of data acquisition with FIB-SEM, Y. Yamaguchi for useful discussion and technical guidance of SBF-SEM, K. Hasumi for the support of the segmentation. The author is also grateful to National Institute of Technology and Evaluation, National Institute of Advanced Industrial Science and Technology for providing the E.coli and phages.

References

• Knott G, Marchman H, Wall D, Lich B. Serial section scanning electron microscopy of adult brain tissue using focused ion beam milling. J Neurosci. (2008).
• Micheva KD, Smith SJ. Array tomography: a new tool for imaging the molecular architecture and ultrastructure of neural circuits. Neuron. (2007).
• Deerinck, T. J. & Bushong, Eric & Thor, A & Ellisman, Mark. NCMIR methods for 3D EM: A new protocol for preparation of biological specimens for serial block face scanning electron microscopy. Nat Center Microsc Imag Res. 6-8. (2010).
• Denk W, Horstmann H. Serial block-face scanning electron microscopy to reconstruct three-dimensional tissue nanostructure. PLoS Biol. (2004).
• Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez JY, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A. Fiji: an open-source platform for biological-image analysis. Nat Methods. (2012).