Visualization of yeast cells by electron microscopy

In the 1970s, hydrocarbon or methanol utilizable yeasts were considered as a material for foods and ethanol production. During the course of studies into the physiology of yeasts, we found that these systems provide a suitable model for the biogenesis and ultrastructure research of microbodies (peroxisomes). Microbodies of hydrocarbon utilizing Candida tropicalis multiply profusely from the preexisting microbody. β oxidation enzymes in the microbody were determined by means of immunoelectron microscopy. We examined the ultrastructure of Candida boidinii microbodies grown on methanol, and found a composite crystalloid of two enzymes, alcohol oxidase and catalase, by analyzing using the optical diffraction and filtering technique and computer simulation. We established methods for preparing the protoplasts of Schizosaccharomyces pombe and conditions for the complete regeneration of the cell wall. The dynamic process of cell wall formation was clarified through our study of the protoplasts, using an improved ultra high resolution (UHR) FESEM S900 and an S-900LV. It was found that β-1,3-glucan, β-1,6-glucan and α-1,3glucan, as well as α-galactomannan, are ingredients of the cell wall. The process of septum formation during cell division was examined after cryo-fixation by high pressure freezing (HPF). It was also found that α-1,3and β-1,3-glucans were located in the invaginating nascent septum, and later, highly branched β-1,6-glucan also appeared on the second septum. The micro-sampling method, using a focused ion beam (FIB), has been applied to our yeast cell wall research. A combination of FIB and scanning transmission electron microscopy is useful in constructing 3D images and analyzing the molecular architecture of cells, as well as for electron tomography of thick sections of biological specimens.


Introduction
Yeasts are unicellular eukaryotes, and have fundamentally the same subcellular structure as higher animal and plant cells, and are used widely as a model system in basic and applied fields of life science, medicine and biotechnology. The National Institutes of Health, USA, designates yeasts as one of model organisms for biomedical research.
The ultrastructure of yeast cells was first studied by transmission electron microscopy (TEM) using thin sections in 1957 [1], and the freeze-etching replica method was introduced in 1969; at that point the fine structure of yeast cells was obtained [2]. During the next 50 years, techniques for analyzing ultrastructure of yeasts advanced greatly. Here, I will review the development of fixation methods, rapid-freeze techniques and instruments for studying yeasts by use of electron microscopy (EM). Our discovery of fine structure and of enzymes of microbodies is shown, and the process and mechanism of septum formation as well as the ultrastructure, components and mechanism of formation of cell wall, will be described in this review. Finally, I will introduce a perspective on yeast ultrastructure research in the 21st century.

General features of yeast cells and the development of their fixing techniques
The yeast is classified into the budding yeast and the fission yeast. The budding yeast ( Fig. 1a) has an ellipsoidal shape (6 × 5 μm), and divides by budding. On the other hand, fission yeast (Fig. 1c) has a cylindrical rod-shape (7-8 × 2.5 μm), and divides by medial fission. Although yeasts have these two types of cell proliferation, the cellular structures are basically the same as each other: nucleus with nuclear pores, mitochondria, microbodies, vacuoles, endoplasmic reticulum, and so on. Golgi apparatus is rarely observed as a 'fence' in budding yeast, but its stacks appear usually in fission yeast ( Fig. 1b and d).
At the late 1950s, I began to develop the technique of fixation of yeast cells for the study by use of EM, because yeasts have a thick cell wall composed of an inner network of glucans and an outer layer of mannoproteins, which is a barrier to penetration of osmium tetroxide (OsO 4 ) solution, which is widely used as a fixative for many organisms. Figure 2 shows the progress of fixation methods and corresponding TEM images of yeasts from 1957 to 1985. Yeast cells were fixed solely with potassium permanganate (KMnO 4 ) and EM photographs were taken by me in the early 1960s ( Fig. 2a and b) [3]. More fine EM images that were obtained by using double fixation with and KMnO 4 (Fig. 2c) made us able to study the ultrastructure in relation to cellular function [4]. To remove the yeast cell wall, we digested cells with Zymolyase (β-1,3-glucanase) [5], and obtained complete protoplasts. In the middle of the 1970s, our research team succeeded in obtaining a good EM image of the spheroplast or protoplast fixed with glutaraldehyde (GA) and OsO 4 ( Fig. 2d) [6].
However, the information on the cell wall was lost using this method. Then, the rapid freezing of yeast cells using a sandwich method was achieved [7], and finally we were able to realize the whole cell image of yeast cells with natural cell wall (Fig. 2e). Thus, we could now easily analyze images of the same specimen by use of both TEM and scanning electron microscopy (SEM), and hence we started studies on morphology of the yeast cell by high resolution (HR) and low voltage (LV) SEM [8].
To get a large quantity of frozen yeast cells for diversified analyses, we developed the HPF method [9][10][11][12][13][14]. Then, freeze-etching method and microsampling methods, using focused ion beam [FIB, see FIB-scanning transmission electron microscopy (STEM) is useful for cell biology Section] as well as thin sectioning, can be applied for both budding and fission yeast cells, for observation with TEM and SEM, the same as other organisms. Figure 3 shows the progress of model image of yeast cell ultrastructure, and changes in fixation methods during the last 30 years [9]. KMnO 4 fixation is still useful for visualizing cell walls, in order to examine mating and germination processes [15].
Fine structure and biogenesis of microbodies ( peroxisomes) in hydrocarbon utilizable yeast, Candida tropicalis, grown under different carbon sources Fine structure of microbody In the early stages of biotechnology, hydrocarbon or methanol utilizable yeasts were considered as a material for foods and ethanol production. Although the metabolism and fermentation of hydrocarbons producing useful substances have been extensively studied in this period, only few reports have been published about the relation between ultrastructure and metabolism [16,17]. During the course of studies of the physiology of hydrocarbon-utilizing yeasts, we found that microbodies that have a homogeneous matrix and are surrounded by single unit membranes appeared profusely in various strains of Candida  yeasts grown in n-alkanes [18]. Such microbodies in mammalian and plant cells have been identified as 'peroxisome' and /or 'glyoxysome', and catalase has been confirmed as the specific marker enzyme [19,20]. In the case of Candida yeasts growing in n-alkane as a sole carbon source, we found that the catalase activity is significantly higher than that growing in the usual carbon source, glucose, especially at the exponentially growing stage [21].
During the basic studies of the relationship between the appearance of microbodies and high catalase activity in Candida yeasts grown on n-alkanes, we established the isolation method of microbodies from the cells to investigate their ultrastructure [6].
The catalase activity in cells of various strains of Candida yeasts grown in n-alkane was markedly higher than that in the cells growing on glucose, ethanol or acetate. EM images revealed an abundant appearance of microbodies in the cells, and EM cytochemistry using 3,3-diaminobenzidine reagents confirmed the localization of catalase activity in the microbodies, as well as mitochondria [22][23][24].
Candida tropicalis cells, precultured on malt extract, contained microbodies scarcely, and showed very low levels of catalase activity. When the precultured cells were transferred to an n-alkane medium and incubated with shaking, the cell shape changed to the filamentous form over 16 h ( Fig. 4a and b), and concomitantly the number and volume of microbodies and the catalase activity increased [22].
In the 1970s we wondered why microbodies gathered or aggregated with each other even there was enough space in the cytoplasm ( Fig. 4a-d) [22]. In connection with this, a distinct structure was observed in the area where two microbodies touched each other, which we termed a 'zipper-like structure' (Fig. 4f-h). The presence of this structure was also confirmed by the freeze-fracture technique (see Supplementary data) [25]. This structure may cause aggregation of microbodies at the early stage of biogenesis, and may be concerned with the physiological stability of organelles or the efficiency of metabolism. Recently, the coiled-coil protein VIG1 has become to be essential for tethering vacuoles to mitochondria during vacuole inheritance of Cyanidioschyzon merolae [26]. Similar molecules to VIG1 may concern in association of microbodies. In recent years, in Aspergillus oryzae,

Biogenesis of microbody
There are several hypotheses about the origin of microbodies. De Duve proposed that these organelles are derived from a dilated region of ER [28], although many of the published micrographs failed to demonstrate a direct continuity between the membrane of the nascent microbodies and the ER membrane. Our observation in the 1970s suggested it seems most likely that yeast microbodies would originate from preexisting microbodies. This assumption may be supported by the fact that microbodies exist as aggregations in cells incubated for 6-8 h on the n-alkane medium, and in 46 h-old cells the cytoplasm is filled with the organelles [22]. In contrast, ER does not arise in cells during the early incubation period with the n-alkane medium.
Recently, dumbbell-shaped microbodies were observed which had electron dense particles inside a constricted area (10 nm φ, arrows in Fig. 4e) in 2.5 h-cultured cells. It is possible that these particles are involved in the control of the division of microbodies [29,30]. Furthermore it has been shown that two daughter microbodies arise by binary fission of the preexisting microbody in alga Cyanidioschyzon, which contains one microbody per cell [31].
On the other hand, in yeast Saccharomyces cerevisiae, peroxisome fuses distinct vesicles derived from the ER [32], indicating that research of peroxisome biogenesis now reopens the discussion of the participation of ER [33,34]. In near future the process of multiplication, segregation and breakdown of peroxisome will be elucidated by means of the latest visualization techniques: UHR EM and super resolution fluorescence microscopy.

Enzymes in microbodies of Candida tropicalis
We determined by immunoelectron microscopy (IEM) that the microbodies of C. tropicalis cells grown in n-alkane included thiolase, acetyl CoA (T-I) and 3-ketoacetyl CoA thiolase (T-III), which are enzymes for β-oxidation of fatty acid. The T-III locates specifically in the microbodies, whereas T-I was detected both in the cytoplasm and microbodies ( Fig. 4c and d). These results were consistent with the biochemical data based on subcellular fractionation, indicating that the yeast β-oxidation system operates efficiently only in the microbody. Other peroxisomal enzymes, acetyl-CoA oxidase, catalase, cartinin acetyltransferase, isocitrate lyase (ICL) and malate synthase, were also detected in microbodies [35,36].
The ICL gene of n-alkane utilizable yeast C. tropicalis could be expressed in S. cerevisiae MT8-1 cells grown on an oleic acid medium, and the enzyme proteins transported into the microbody. To identify the signal sequence responsible for the proper subcellular location of C. tropicalis ICL (CT-ICL) in S. cerevisiae, the localization of truncated mutants of CT-ICL were examined by IEM. Three amino acids residues at the C-terminal end were not an essential sequence for the location in the peroxisomes. In cells expressing the internal region of the CT-ICL gene, the expressed polypeptides did not transport to the peroxisomes; rather, the polypeptides were mis-targeted to mitochondria or formed a protein aggregate body (PAB) in the cytoplasm [37]. PAB is now generally termed 'aggresome'.

Composite crystalloid in the microbody of methanol utilizable yeast Candida boidinii
In the course of the studies of the physiology of methanol utilizing yeasts, Kloeckera sp. no 2201 (also known as C. boidinii), microbodies of small size, were found in cells grown in a glucose medium, transferred into methanol medium and then incubated for 2 h. After 4 h of cultivation, 5-6 microbodies were observed on the section of the cell (Fig. 5a). Although the number did not change during prolonged cultivation, their size increased with the passage of cultivation time. The microbodies were successfully isolated by means of sucrose gradient centrifugations from the methanol-grown cells [38], and microbody-specific enzyme activities were detected in the fraction by biochemical techniques. The activity of catalase and alcohol oxidase, the key enzymes of methanol oxidation, was also found in the fraction throughout the cultivation period [39].
The ultrastructure of microbodies in methanolgrown cells differed from that of hydrocarbon utilizing cells, and they have a lattice structure (Fig. 5b). The use of a cryo-sectioning technique, together with electron diffraction, optical diffraction and filtering and computer simulation, determined the intrinsic structure of the crystalloid. The activity of alcohol oxidase and catalase was found uniformly in the crystalloid, using electron probe microanalysis [40] of cryo-sections, as well as by EM cytochemistry [41-47].
Profiles of the microbody crystalloid in the cryosections with negative staining, which was a popular technique for detecting the finest structure at that time, were classified into three types: tetragonal, hexagonal and rectangular lattice (Fig. 5d, i1-i3). When a figure of Fig. 5c-i was superimposed 81 times (which was the old analytical method), large and small particles were found (yellow arrow and arrowhead, respectively in Fig. 5c-iv). The large particle was composed of 4 subunits. However, the number of subunits in small particles was impossible to determine by this method. Then, the particle was analyzed by an optical filtering technique [48], which was the most advanced technique at that time, and two small subunits were found in the small particles ( Fig. 5c-v, red).
Their optical diffraction (Fig. 5d, ii1-ii3) and also electron diffractions (Fig. 5d, iii1-iii3) revealed that the crystalloid has a cubic structure. By tilting experiments, one of the three profiles of the lattice was converted to another ( Fig. 5d, i1-i3). Optical diffraction images (Fig. 5d, ii1-ii3) clearly showed that the lattice was composed of two types of particles, with different diameters of 10 and 7 nm, which were arranged alternately. Computer simulations of tilting images of the crystalloid conclusively showed that the crystalloid comprises composite crystals of two particles, and that large and small particles were composed of eight (blue) and four (red) subunits, respectively ( Fig. 5e-iii). Large particles seemed to consist of alcohol oxidase molecules, while the small particles were presumably made up of catalase according to their molecular weight. We proposed the 3D model of the microbody crystalloid ( Fig. 5f) [45]. The distance between particles was 11 nm with a unit cell length of 22 nm, and the shortest distance between two large (or small) particles measured 15.6 nm. They located alternately to make up the composite crystals, the same as the structure of rock salt. These composite crystals can participate most effectively in methanol oxidation in the cell.

Mechanism of cell wall formation
Since the protoplasts of Schizosaccharomyces pombe can completely regenerate cell walls to revert to normal cells in liquid media, the regenerating process is a good model system to examine a de novo process for cell wall synthesis: (i) initiation of the cell wall, (ii) formation of microfibrils, (iii) deposition of an amorphous wall substances and (iv) involvement of the actin cytoskeleton.

Summary of the cytoskeletal changes during cytokinesis of Schizosaccharomyces pombe
Fission yeasts are useful model organisms for research into the cell division cycle [4,49,50]. In the 1980 and 1990 s, the behavior of the cytoskeleton during the yeast cell cycle was intensively studied in many laboratories, including the author's [51-59].
We proposed that the actin cytoskeleton plays a leading role in the formation of the cell wall through studies, starting in 1989, of cell wall formation and the behavior of filasomes [60,61] using a fluorescence microscopy. We also found that prior to septum formation, actin localized as if to lead invagination of the cell membrane, and formed an actin ring with the progress of septum expansion [14,62,63]. The actin cytoskeleton may play the role of conductor in the cytokinesis [64][65][66][67]. Figure 6 summarizes the cytoskeletal changes through the cell division cycle in S. pombe [51-59,68,69].

Development of UHR-LVSEM
To visualize details of a regenerating cell wall, we developed an improved UHR FESEM S-900LV, collaborating with Hitachi Co., which gives us better resolution at low beam energy ( 5 kV or below) with small aberrations of the objective lens and a shorter working distance [70,71]. This modified in-lens FESEM S-900, referred to here as S-900LV (commercial model SEM S-900H), is a type of UHR LVSEM, which enables us to observe directly the surface topographs of uncoated biological samples with maximum fidelity (Fig. 7) [72].

Dynamics of the ultrastructure during cell wall formation in Schizosaccharomyces pombe observed by SEM and TEM
Using a UHR LVSEM S-900 LV and S-900, we can study the ultrastructure of regenerating cell walls of S. pombe protoplasts with uncoated or slightly coated [73] surfaces, respectively. Just after digesting the cell wall of the cylindrical S. pombe cells with Novozyme 234, no cell wall material was seen on the surface of the round shaped protoplast ( Fig. 8a). The absence of staining with calcofluor confirmed that the surface of freshly prepared protoplasts was free from cell wall materials, glucans and α-garactomannnan [59]. TEM images of thin sections also confirmed the absence of the remnants of old wall materials on the surface [74]. Many invaginations of the cell membrane, which is unique to S. pombe, were found on the surface.
After 10 min incubation, a portion of the surface accumulated invaginations (Fig. 8b, !), and at 1.5 h  fibrous materials appeared at the area (Fig. 8c, !). After 5 h, fibrous network covered the whole surface of the protoplast, and the shape of the protoplast changed to the elongated form (Fig. 8e). The microfibrils twisted around each other and developed into thick fibrils forming flat bundles. Intrafibrillar spaces were gradually filled with particles ( Fig. 8d, arrowhead). After 10 h, the granular materials filled up intrafibrillar spaces on the surface (Fig. 8f). Finally, after 12 h the complete cell wall was regenerated ( Fig. 8g and h) [75].
In the cell-free system using isolated cell membrane of C. albicans and the negative staining method, secreted particles (2 nm φ) subsequently converted to microfibrils 4 nm thick [76]. Uncoated UHR LVSEM images of the regenerating protoplast showed scattered particles, 3 nm ϕ (arrowhead) and 2 nm thick ( ) microfibrils on the cell surface (Fig. 8i). These microfibrils may be joined end to end, twisted around each other, attached side-by-side to 8 nm thick ( ) fibrils, and then developed to a ribbon shaped network 16 nm thick 16 nm thick ( ). The ribbon shaped structures further contacted side by side and finally formed diversified structures with various width and length, up to 200 nm wide and 1 μm long (arrow) [72]. SEM images of microfibrils and ribbon shaped structures shown here were similar to TEM images of a negatively stained specimen of the reverting protoplast for 3 h (Fig. 10a) [9, 72,75].

Identification of cell wall components
Bundles formed on the surface of reverting protoplasts were cleaved into small pieces when the protoplasts were treated with β-1,3-glucanase for 2 h ( Fig. 10a and b) [9,72] and completely digested with longer treatment. Furthermore, fibrous networks are not formed on the protoplasts in the presence of aculeacin A, an antifungal antibiotic and a potent inhibitor of β-1, 3-glucan synthesis [76][77][78]. Immunogold signals against for β-1,3-glucan were detected on IEM ( Fig. 10d and e arrows). Thus, fibrous structures are composed of microfibrils of β-1,3-glucan as in a budding yeast cell wall [76,78]. By using gold-labeled lectin, α-galactomannan was detected on particles locating in the intrafibrillar space of the reverting protoplast for 3 h (arrowhead in Fig. 8d, f and j) [79]. The intrafibrillar spaces were filled with the particles after 12 h (Fig. 8g) indicating α-galactomannan was also an ingredient of the cell wall.
In addition to β-1,3-glucan, it was found by NMR spectroscopy [80] and IEM using an antibody against α-1,3-glucan that the S. pombe cell wall included α-1,3-glucan. Most of all, the α-1,3-glucan distributed just outside the cell membrane, supporting previous reports that the α-glucan is synthesized on the cell membrane [81]. Recently we also detected Mok 1p, the protein of α-1,3-glucan synthetase, locating on the cell membrane. Mok 1p moves from the cell tip to the medial region during the cell cycle [14]. These results show that α-1,3-glucan is required for the primary step of glucan bundle formation in new cell wall synthesis during vegetative growth.
Based on the experiment using anti-β-1, 6 glucan antibodies, we proposed that β-1, 6 glucan in a highly branched form is also an ingredient of the cell wall. However, such β-1, 6 glucan was not recognized on the Golgi body [66,82]; rather, they existed in the cytoplasm and beneath of the cell membrane. From these results and the fact that the low molecular α-1,3-glucan was transported through the secretory pathway to the cell membrane where the synthase located, and then synthesized into higher molecules, it was speculated that linear β-1, 6 glucan was synthesized in the endoplasmic reticulum-Golgi system, and that highly branched structures were formed on the cell membrane ( Fig. 9) [66,67,82]. Thus β-1,6-glucan, as well as α-1,3-glucan and β-1,3-glucan, is an ingredient of the cell wall of S. pombe. Those glucans and α-galactomannan are characteristic components of this yeast [83]. Throughout these experiments, we used antibodies against α-1,3-glucan, and highly branched β-1,6-glucan that we had prepared [80].
Just beneath the cell membrane filasomes are easily identified (Fig. 10i) by heavy staining with RuO 4 [61]. They are 100-300 nm in diameter and consist of one microvesicle (35-70 nm in diameter) and fine filaments surrounding the vesicle [65]. The location of filasomes in reverting protoplasts coincides with the site where F-actin patches exist [65]. It seems that filasomes may contribute to transporting the cell wall materials from the cytoplasm to underneath the cell membrane.
Mechanism of cell wall formation and 3D images of intracellular structure during regeneration of the protoplast The cell wall and septum are abnormal in actin point mutant cps8, and the distribution pattern of the actin cytoskeleton differs from wild-type [87,88]. In the mutant cells, the development of crosslinks between β-1,3-glucan fibrils was defective, and the intrafibrillar space on the reverting protoplast surface was not filled with amorphous particles composed of α-galactomannan [89]. In the wild type cells, formation of the fibril network was completely inhibited in the presence of 40 μg ml -1 cytochalasin D, and this inhibition was reversible [59]. These results suggest that the actin cytoskeleton may control secretion of cell wall materials such as α-galactomannan. In addition, the actin cytoskeleton requires localization of α-1,3-glucan synthase to the site of growth and plays an essential role in cell morphogenesis [90].
Protoplast treated with periodic acid-thiocarbohydrazide-silver protein staining for polysaccharides (PATAg) shows a higher electron density of Golgi bodies [74]. The intracellular secretion machinery, especially the Golgi apparatus and secretory vesicles, was analyzed by 3D reconstruction from 40 to 80 serial sections of protoplasts stained with PATAg at five distinct reverting stages (Fig. 11). The total volume of reverting protoplast increased by 3.8 and 4.3 times at 3 and 5 h, respectively, and the volume of Golgi apparatus at the corresponding stages increased 2.3-and 2.5-fold, respectively. The number of secretory vesicles also increased markedly by 3.4 and 5.8 times, respectively. Filasomes may correlate with this secretory system. Combined data, obtained by using UHR LVSEM and TEM techniques, suggest the mechanism of cell wall formation in yeast as shown in Fig. 12 [9]. In addition, our data together with other studies [83,[91][92][93][94][95][96][97] proposes the composition and structure of the cell wall of S. pombe (Fig. 13) [9,98].   Significance of the HPF technique in structural biologywith reference to a study of septum formation in fission yeast HPF was first used as an alternative method of cryo-fixation in the 1960s, and has advanced since then. The commercially available unit is reliable with respect to its physical performance in yeast cells as well as plant tissues [99]. Cells of S. pombe were cryo-fixed by HPF using HPM 010 (BAL-TEC AG, Liechtenstein) under 210 MPa. They were observed at −140°C by Philips XL30 FEG and Hitachi S-4700 SEM, using the cryo-system Alto 2500 (Oxford), after cryo-cutting at −175 to −185°C and magnetron spatter coating with platinum (2 mm). The high quality image by ultra-low temperature (ULT)-LVSEM showed an ultrastructure corresponding to the thin section of the same specimen, and 3D vision could observed from the outside and inside of the cell, allowing the stereo view of initial septum formation [68].
Figure14a and b show a fracture image of an outer side and the inside of a cell, respectively. A fracture image of a central plane along the longitudinal axis of a cell (Fig. 14c), closely resembles the thin-section image of the same HPF specimens (Fig. 14d). On the PF face of the cell membrane, invaginations found on the protoplast surface ( Fig. 8a) are also observed (Fig. 14a). Fracture images of S. pombe cells forming septum during cytokinesis are shown on Fig. 15. Newly formed primary septum keeps expanding and is lined with secondary septum until the cytoplasm separates (see Fig. 6).
Throughout the septum formation, the immunogold labeled α-1,3-glucan molecules were observed on the expanding incomplete septum [14, 68,69]. The α-1,3-glucan synthase Mok 1p is an essential enzyme for the synthesis of α-1,3-glucan. To analyze the behavior of Mok 1p during septum formation, frozen cells by HPF were prepared by freezefracture replica after SDS treatment and then labeled with antibodies (SDS-FRL) [100]. The immunogold-labeled Mok 1p molecules (arrowhead) accumulated on the cell membrane around the region where the septum was formed (Fig. 16a,  arrow). Mok 1p were also localized on the membrane beneath complete septum (Fig. 16b,  arrowhead). After daughter cells separated, Mok 1p was observed on their swollen tip (Fig. 16c,  arrowhead).
Cryo-fixed specimens by HPF were also available for conventional TEM and IEM. The thin sections of cells showed a superior image of the fine structure of all subcellular organelles. The IEM images displayed in situ localization of glucans in S. pombe cells [66][67][68][69]. During septum formation β-1,3and α-1,3-glucans were located in the invaginating primary septum, providing the first morphological evidence at the initial stage of septum formation. At this time, highly branched β-1,6-glucan was not detected, whereas other glucans were observed. Later, highly branched β-1,6-glucan also appeared on the secondary septum. Throughout the septum formation, actin was located just beneath the invaginated cell membrane and also accumulated at the medial region of the cell [63,64,67]. From these facts it is concluded that the actin cytoskeleton governs septum formation.
Very recently we found the essential function of α-1,3-glucan in the septation and cell separation during cell division [101].

FIB-scanning transmission electron microscopy (STEM) is useful for cell biology
To analyze the molecular architecture of cells, micro-sampling methods using an FIB, which are the main techniques for semiconductor device analysis, have been applied to yeast cells. This is the first case of applying FIB to the biological specimen [102].
The 0.3 μm thick specimen was obtained by milling with FIB, and the dark field image was observed by scanning transmission electron microscopy (STEM) (Fig. 17). Then, 3D images were constructed from 50 views taken by tilting 5°on a STEM. Six images are picked up and shown in Fig. 17a-c, and a clear image at high magnification in Fig. 17d [103-106]. On the outer surface of the cell wall, brush-like structures were recognized, clearly consisting of the presence of mannnoproteins (See Fig. 13), and fibrous structures were observed near the cell membrane side of the cell wall, consisting of glucan networks.
We tried to examine in more detail the cell wall architecture. The cell wall of S. cerevisiae was treated with acid after solubilizing the cytoplasm and prepared by FIB micro-sampling, and then the stereo images by STEM were taken at an original magnification of ×250 000. Fine fibrils were recognized inside the cell wall (circles in Fig. 18e and f ) [98,105,106]. In addition, STEM is also useful for electron tomography of thick sections of biological specimens [107][108][109]. We succeeded determining the polarity of actin cytoskeleton modified with heavy meromyosin in cells of S. pombe that had been quite difficult to detect by usual IEM (Fig. 19) [98].

Future directions
I have reached the conclusion, from my long experience, that it is important to use the most suitable model system for each study. For example, I used hydrocarbon-or methanol-utilizable yeasts to study the biogenesis of microbodies, and the regeneration system of protoplast was used to resolve the mechanism of cell wall formation. I found the best condition for conversion of the intact cells to protoplast at 100% yield even in a liquid medium, and could establish the system during the work. I also established some methods of analysis such as: (i) development of fixation, (ii) adoption of image reconstruction and analysis, even at the early period of EM, (iii) conception of rapid freezing by the sandwich method for yeasts, (iv) development of UHR LVSEM, by which uncoated biological specimens can be observed under LV and (v) establishment of the ULT/LVSEM method.
The studies by means of EM will continue as important and powerful techniques for obtaining extensive information on biological structure in the 21st century, since the study of 3D will become subjective. The identification of the gene expression and proteins will become possible by the established cryo-technique, especially freeze replica combined with SDS-FRL and STEM of the same specimen. The juncture of CEMOVIS [110] with IEM will be the key to the resolution of research in cell biology. Furthermore, the integration of live cell imaging and super resolution fluorescence microscopy, which has been developed by a new theory, will be aniticipated as a new method for resolving the movement of functional molecules in cells the near future.
In the past, we constructed 3D images by successive serial sectioning of the biological specimen [65,74]. In contrast, the methods for reconstructing 3D images by FIB SEM [111] or SBF SEM are powerful and quick techniques. Recently, we have been able automatically to obtain 600 serial images from a whole S. pombe cells for only 20 h by FIB SEM [112]. This makes us able to analyze the 3D architecture of the cell.
Thus the techniques for observing and analyzing ultrastructures of molecules as well as intracellular organelles have developed significantly, and now it is possible to obtain image information with high quality. However, to understand the meaning of observations, it is necessary to accumulate basic knowledge of morphology rather than to rely upon the technique.

Funding
The work was supported by grant-in-aid for scientific research from the Ministry of Education, Culture, Sport, Science and Technology of Japan