The mitochondrial cycle of Arabidopsis shoot apical meristem and leaf primordium meristematic cells is defined by a perinuclear tentaculate/cage-like mitochondrion

Plant cells exhibit a high rate of mitochondrial DNA recombination. This implies that before cytokinesis, the different mitochondrial compartments must fuse to allow for mtDNA intermixing. When and how the conditions for mtDNA intermixing are established is largely unknown. We have investigated the cell cycle-dependent changes in mitochondrial architecture in different Arabidopsis thaliana cell types using confocal microscopy, conventional, and three-dimensional electron microscopy techniques. Whereas mitochondria of cells from most plant organs are always small and dispersed, shoot apical and leaf primordial meristematic cells contain small, discrete mitochondria in the cell periphery, and one large, mitochondrial mass in the perinuclear region. Serial thin section reconstructions of high pressure frozen shoot apical meristem cells demonstrate that during G1 through S phase, the large, central mitochondrion has a tentaculate morphology and wraps around one nuclear pole. In G2, both types of mitochondria double their volume, and the large mitochondrion extends around the nucleus to establish a second sheet-like domain at the opposite nuclear pole. During mitosis, ~60% of the smaller mitochondria fuse with the large mitochondrion, whose volume increases to 80% of the total mitochondrial volume, and reorganizes into a cage-like structure encompassing first the mitotic spindle and then the entire cytokinetic apparatus. During cytokinesis, the cage-like mitochondrion divides into two independent tentacular mitochondria from which new, small mitochondria arise by fission. These cell cycle-dependent changes in mitochondrial architecture explain how these meristematic cells can achieve a high rate of mtDNA recombination and ensure the even partitioning of mitochondria between daughter cells. cytoskeletal arrays, and possesses sheet-like domains at the spindle poles. During interphase, about half of the cage-like mitochondrion fragments into smaller, physically discrete mitochondria, while the other half adopts a tentaculate architecture with its central, sheet-like domain forming a perinuclear cap. Based on the information contained in the 3D reconstructions, we have developed a model of how the architectural changes of the large mitochondrion relate to cell cycle-dependent functions of these meristematic cells.

. Furthermore, changes in their architecture and their ability to translocate rapidly throughout the cytoplasm appear to be of critical importance for executing their cellular functions. For example, it has long been known that mitochondria congregate around cellular areas with high energy requirements (Bakeeva et al., 1978;Bawa and Werner, 1988;Bereiter-Hahn, 1990;Bereiter-Hahn and Voth, 1994;Logan, 2006). In addition, in both mammalian and plant cells, they constantly undergo fission, fusion and branching changes while sliding to different cellular locations (Bereiter-Hahn, 1990;Bereiter-Hahn and Voth, 1994;Logan and Leaver, 2000;Arimura et al., 2004;Logan, 2006Logan, , 2006.

Serial thin section reconstruction of mitochondria of the unicellular algae
Chlorella (Atkinson et al., 1974) and Chlamydomonas reinhardii (Blank and Arnold, 1981), and analysis of the 3-D architecture of thin sectioned and of GFP-tagged mitochondria in yeast (Calvayrac et al., 1972;Osafune et al., 1975Osafune et al., , 1975Osafune et al., , 1975Yaffe, 1999Yaffe, , 2003 have shown that in each of these cell types, the mitochondria are joined into a single, reticulate structure. Extended tubules and networks have also been observed in the chondriome of different fungal and algal species (Floyd et al., 1972;Pickett-Heaps, 1974;Howard, 1981;Zadworny et al., 2007).
In contrast, in mammalian cells, where large numbers of mitochondria can be resolved by light microscopy techniques, their architecture can vary from small (1-2 µm in diameter) spheres/ovals to 10µm long, sausage-like and sometimes branched organelles whose longitudinal axis tends to parallel the orientation of the radial microtubules (Bereiter-Hahn, 1990). However, in specialized cells types such as muscle fibers (Bakeeva et al., 1978) and COS-7 cells (Yaffe, 1999), some of the mitochondria have been shown to posses a more network/reticulate type of organization. All of these changes in mitochondrial shape and the accompanying fusion and fission events can occur within minutes (Bereiter-Hahn and Voth, 1994). Higher plant mitochondria are generally portrayed as being oval or sausage-like, with only few studies reporting the existence of branched mitochondria that undergo fusion, fission and ameboid-like changes over short periods of time (Logan and Leaver, 2000;Logan et al., 2003;Arimura et al., 2004;Foissner, 2004;Logan, 2006). To account for the high rate of recombination of the plant mtDNA (Lonsdale et al., 1988;Gillham, 1994), the "discontinuous whole" hypothesis postulates that the individual, small mitochondria must transiently fuse to transfer their mtDNA into a common physical space (Logan, 2006(Logan, , 2006. The relationship between mitochondrial propagation and the cell cycle has also been investigated for many years. New mitochondria do not arise de novo; they derive from preexisting ones (Bereiter-Hahn, 1990;Yaffe, 1999;Logan, 2006). It is generally assumed that the doubling of the mitochondrial volume and numbers occurs during the S and G2 stages of the cell cycle, as evidenced by the abundance of dumbbell-shaped mitochondrial profiles in electron micrographs during these stages of the cell cycle (Bereiter-Hahn, 1990). For many years, segregation of the mitochondria into the two daughter cells has been assumed to be a stochastic process. However, in yeast cells, the transfer of some of the reticulate mitochondrial domains to the growing buds has been shown to require intact cytoskeletal systems and GTPase-mediated activities Shepard and Yaffe, 1999;Fekkes et al., 2000;Boldogh and Pon, 2006). Although the information about mitochondrial propagation in plants is more limited, there are several reports on the coupling of the mitochondrial cycle to the cell cycle (Arimura et al., 2004;Sheahan et al., 2005;Zottini et al., 2006).
In this study, we have analyzed the changes in shape, number and distribution of the mitochondrial population in different cell types of Arabidopsis thaliana by means of conventional transmission electron microscopy (TEM), confocal laser scanning microscopy (CLSM) imaging, and three-dimensional (3D) TEM modeling. Whereas in most of the cell types the shape of the mitochondria corresponds to the accepted pattern of oval or sausage-like organelles, the chondriome of shoot apical meristem (SAM) and leaf primordium (LP) meristematic cells is defined by the existence of a large mitochondrial mass that coexists with smaller mitochondria and undergoes changes in shape and size during the cell cycle. To obtain a more precise view of these mitochondrial masses, we have reconstructed entire cryofixed SAM meristematic cells at representative stages of the cell cycle by serial thin section electron microscopy (EM) imaging and 3D modeling. These reconstructions confirm that Arabidopsis SAM and LP meristematic cells contain a large, network-forming mitochondrion that surrounds the nucleus and interacts with a set of smaller, cortical mitochondria in quantifiable ways. During mitosis and cytokinesis most of the mitochondria fuse into a large, coherent, cage-like mitochondrion that encompasses the entire spindle/phragmoplast

Results
During the course of a series of studies designed to elucidate the structural basis of plant cell division (Seguí-Simarro et al., 2004;Seguí-Simarro and Staehelin, 2006;Seguí-Simarro et al., 2007), we have analyzed thousands of electron micrographs of Arabidopsis SAM, LP and root meristem cells preserved by high pressure freezing/freeze-substitution techniques. One of the unexpected findings was a striking dichotomy in mitochondrial architecture. As illustrated in Figure 1, some of the crosssectioned mitochondria exhibited a classical round, oval, dumbbell or rod-like configuration. In the shoot apex ( Figure S1A), this pattern was consistently observed in differentiated cells of the cotyledons and the apical region of the LP (Figures S1B, C).
However, in SAM and LP meristematic (undifferentiated) cells, we found in many instances lobed mitochondria with narrow, elongated domains resembling the matrixules described by Logan, (2006) and Scott et al., (2007) together with mitochondria with even more complex cross-sectional profiles (Figures 2A, B). To gain a better understanding of the actual 3D morphology of these complex mitochondria, we analyzed randomly chosen sets of serial EM sections of SAM cells ( Figure 2C). To our surprise, these random sets of serial thin sections contained many more large and branched mitochondria than anticipated. They also showed that these mitochondria always possessed at least one sheet-like domain that was partly wrapped around the nucleus (Figures 2A, S1D, E), and other domains that partly encompassed organelles and other cytoplasmic structures. These persistent spatial relationship patterns suggested that the large, branched mitochondria might serve specialized functional needs of these proliferating cells. To address this question, we examined the morphology of mitochondria in SAM and LP cells, as well as in other cell types, not derived from the SAM, by means of confocal, conventional (2D) and 3D TEM techniques.

Mitochondria of living shoot apical meristem and leaf primordia meristematic cells display a network-like configuration
To determine if the morphology of the large, complex mitochondria of SAM and LP meristematic cells changed in a cell cycle-dependent manner, we analyzed their architecture and distribution in living cells (Figure 3). Excised SAM and LP tissues were incubated first in Mitotracker, a mitochondria-specific stain ( Figure 3A, green signal) and then in DAPI, a DNA-specific fluorescent stain ( Figure 3B, blue signal) prior to viewing in a CLSM. Images of the confocal sections were recorded in both green and blue channels, and merged ( Figure 3C). As seen in Figures 3D, E, some small, rounded mitochondria could be resolved in the periphery of the cells, but most of the staining was associated with the mitochondrial mass around the nucleus. In the majority of the smaller interphasic cells ( Figure 3D), the network-forming mass stained more intensely at one pole of the nucleus than at the other (arrowheads), whereas in larger cells the distribution around the nucleus appeared more balanced (arrows in Figure 3D). Since the size of meristematic cells increases as they progress through the cell cycle, the different distributions of mitochondria in differently sized cells suggested a cell cycle-dependent control of the mitochondrial networks. Thus, in the larger, dividing cells ( Figure 3E) the number of small, round mitochondria in the cell periphery appeared to be reduced and the size of the large, network-forming mitochondrial masses increased compared to the interphase cells.
To obtain more detailed information on the organization of the mitochondria in living cells, we produced 3D models of the fluorescing mitochondria of both SAM and LP meristematic cells, from serial CLSM sections of both interphasic (red arrows in  Figures 4H and 4I). Our CLSM data were consistent with the idea that the chondriome of SAM and LP meristematic cells was composed of two types of mitochondria, small, discrete mitochondria in the cell periphery, and a larger mitochondrial mass in the cell center whose size increased during mitosis and cytokinesis. To determine whether the mitochondrial architecture described above for SAM and LP meristematic cells is typical of all Arabidopsis cell types, we re-examined our extensive collection of EM pictures from different Arabidopsis cell types, including root tip, stem, mature leaf, meiocyte, microspore, pollen, endosperm, endothelial and embryo cells (Kiss et al., 1990;Staehelin et al., 1990;Otegui and Staehelin, 2000;Otegui et al., 2002;Otegui and Staehelin, 2004;Seguí-Simarro et al., 2004;Austin et al., 2005;Otegui et al., 2005;Austin et al., 2006;Otegui et al., 2006;Seguí-Simarro and Staehelin, 2006;Borsics et al., 2007;Christopher et al., 2007;Seguí-Simarro et al., 2007 and unpublished results). After an exhaustive viewing of thousands of micrographs, we could not find any evidence in any of these cell types for the presence of mitochondria with a complex 3D architecture resembling the structures seen in TEM micrographs of SAM and LP meristematic cells. Instead, all of the micrographs showed mitochondrial profiles reflecting the conventional and generally accepted round, oval or sausage-like morphology. Nevertheless, to confirm this notion we also examined by CLSM cells in two of these tissue types, stem ( Figure 6) and root tip cells ( Figure S2) stained with Mitotracker and DAPI. As expected, none of the cells in these tissues displayed mitochondrial structures that resembled the large, nucleus-associated mitochondrial masses of SAM cells. Instead, all of the mitochondria of these cells produced fluorescence signals that were both small and discrete (Figures 6A-G and S2A-G). When selected cells of these tissues were modeled, an oval or sausage-like morphology was evident for mitochondria of both stem ( Figure 6H) and root cells ( Figure S2H). Thus, modeling of the stem and root cell image stacks demonstrated that the mitochondria of these cell types differ both in their morphology, their distribution, and their numbers from those observed in SAM and LP meristematic cells. Most notably, these cells lacked a nucleus-associated mitochondrial mass.

Serial thin section analysis of whole cells demonstrates that the large mitochondrion of SAM cells undergoes characteristic architectural changes during the cell cycle.
To overcome the spatial resolution limitations of fluorescence microscopy techniques, and in order to obtain higher resolution data of mitochondrial architecture during the cell cycle, we produced electron micrographs of serial thin sections of six entire cryofixed SAM cells (~130 sections/cell) at representative stages of the cell cycle, traced the mitochondria, nuclei and plasma membranes, and finally generated 3D models of the cells using the IMOD software program. Each cell was chosen according to well defined morphological criteria (see Table S1 and Experimental Procedures for details) to represent a specific stage in the cell cycle (G1, G1-S, G2, prometaphase, early telophase and late telophase). Arabidopsis SAM and LP meristematic cells are characterized, as other meristematic cells, by a small size (a major diameter of ~8 μ m in G1 cells and up to ~12 μ m in G2/dividing cells), a polyhedral shape, a nearly spherical large nucleus (from ~4 μ m in diameter in G1 nuclei to ~5 μ m in diameter for G2 nuclei), high organellar density, limited vacuolation, and a dense cytoplasm ( Figure 1).
All of these features, typical of highly active, proliferating cells make them perfect candidates to apply the above mentioned criteria for cell cycle staging (Table S1). Typically, this latter type of mitochondrion possesses tentacle-like tubular extensions that originate from the margins of the sheet-like central domain. Few individual mitochondria are seen in the vicinity of the tentaculate mitochondrion. As documented in Figures S1D, E, the nuclear-capping tentaculate mitochondrion comes fairly close to the nuclear envelope, but the two membranes never touch as evidenced by the presence of ribosomes between them. During G2, the first major 3-D architectural change of the tentaculate mitochondrion is observed ( Figure 7C). Besides increasing in size, this mitochondrion acquires a clamp-like morphology by forming a second sheet-like nuclear cap domain opposite to the first one. This duplication of the nuclear cap domain sets the stage for conversion of the large mitochondrion into a cage-like organelle that encompasses the spindle during mitosis, and the reforming daughter nuclei and phragmoplast during cytokinesis (Figures 7D, E).
During mitosis ( Figure 7D), the two sheet-like mitochondrial domains that bracket the spindle are connected to each other through large, tubular mitochondrial elements. Careful analysis of these tubular bridging elements shows that while some form continuous links between the two sheet-like pole domains, others exhibit distinct breaks and/or highly constricted domains ( Figure 7D, arrows). These structural features are consistent with the idea that the tubular bridging elements are unstable mitochondrial domains that undergo cycles of fission and fusion, while the overall cagelike conformation of the large mitochondrion around the spindle is maintained. After formation of the phragmoplast and initiation of cell plate growth, a number of individual mitochondria as well as tubular domains of the cage-like mitochondrion assemble in the plane of division around the margins of the forming cell plate in a beltlike conformation ( Figure 7E). As the cell plate expands, small, individual mitochondria begin to appear again in the cortical region of the cell, while the large mitochondrion is converted back to the tentaculate morphology of the G1-stage interphasic cells by severing of the tubular-bridging elements of the cage-like mitochondrion ( Figure 7F).

Prior to mitosis, most of the individual SAM mitochondria fuse to the cagelike mitochondrion, which comprises ~80% of the mitochondrial volume during cell division.
To further characterize the two types of mitochondria of these meristematic cells, we have determined their volume, surface area, number, and contribution to the total cell volume during the cell cycle. To obtain accurate quantitative data, these calculations were made exclusively over the serial sectioned, reconstructed and modeled cells. The volume occupied by mitochondria was found to fluctuate slightly from 8-12% of total cell volume ( Figure 8A), indicating that the mitochondrial volume expands in parallel with the cell volume during the cell cycle. Doubling of the total mitochondrial volume and of the total surface area of the outer envelope membrane (data not shown) occurs during G2. Thereafter, they remain relatively constant during mitosis and cytokinesis ( Figure 8A). The observed doubling of the total mitochondrial volume and of the surface area during G2 occurs in parallel in the individual mitochondria and in the perinuclear tentaculate mitochondrion ( Figure 8B) and is consistent with the observed morphological changes (Figure 7). This suggests that the increase in volume is not only due to a simple change in mitochondrial shape such as swelling, but reflects mitochondrial growth related to the duplication of these organelles.
Quantitative comparisons of the volumes of the two types of mitochondria during the cell cycle have also yielded functionally relevant insights. Thus, whereas the tentaculate mitochondrion contributes from 42 to 44% of the total mitochondrial volume in interphase (G1, G1-S and G2) cells, it contributes ~80% of the total mitochondrial volume during mitosis and cytokinesis, i.e., when it assumes a cage-like organization.
Conversely, the volume percent of the small, dispersed mitochondria reaches a maximum during interphase (56-58%), and drops to a low of ~20% during mitosis.
These volumetric changes suggest that the dramatic enlargement of the tentaculate/cagelike mitochondrion shortly before the onset of mitosis involves the fusion of a significant number of small, individual mitochondria with the large mitochondrion. The putative fusion of small mitochondria with the large, cage-like mitochondrion prior to mitosis is also reflected in a ~60% reduction in the number of small, individual mitochondria as the cells progress from interphase to mitosis and cytokinesis ( Figure   8B).
Taken together, these data indicate that during G2, both the tentaculate mitochondrion and the individual mitochondria double their volume, but without a net increase in the number of individual mitochondria. Instead, the enlarged, individual G2

Discussion
In Arabidopsis, only SAM and LP meristematic cells contain two types of mitochondria that undergo changes during the cell cycle.
The main finding of this study is the discovery that in SAM and LP meristematic cells of Arabidopsis the chondriome consists of two structurally distinct types of mitochondria that undergo cell cycle-dependent changes in shape, size and distribution, whereas in all of the other cell types (~10) studied by us, only one mitochondrial morphotype could be identified in TEM micrographs. This finding is based (1)  The two types of mitochondria of SAM and LP meristematic cells are termed small, round/tubular and large, tentaculate/cage-like mitochondria. The population of conventional, small, round or tubular mitochondria is located in the cell periphery. In contrast, the single, very large, tentaculate/cage-like mitochondrion is seen to be closely apposed to the nucleus and to occupy the cell center.The most notable cell cycledependent changes in mitochondrial organization are all associated with this latter mitochondrial type (Figures 7, 8), which nearly doubles its relative volume as the cells progress from interphase to mitosis. By integrating structural and quantitative data, we have been able to identify the growth, fission and fusion, and morphological remodeling activities that bring about the architectural changes during the cell cycle. A schematic summary of these events is presented in the diagram of Figure 9. mitochondrion, and the dramatic changes in architecture of the large mitochondrion prior to mitosis and after cytokinesis. During G2, the mass of the mitochondria doubles ( Figure 9, arrow 2) at a rate that appears to correspond to the rate of doubling of the cytoplasmic volume, as evidenced by the constancy of the percentage of cell volume (~10%) occupied by mitochondria during the entire cell cycle.
Mitochondrial enlargement during G2 appears to be a common feature of cells of many multicellular organisms (Bereiter-Hahn, 1990;Kennady et al., 2004), and thus our data on mitochondrial growth fit into this general pattern. As the nuclear envelope breaks down and the cells enter mitosis, both the number and the volume percent of the small, individual mitochondria population decreases to ~20%, and the volume percent of the cage-like large mitochondrion increases to ~80%, which most likely reflects a change in the balance of fission and fusion events between the small mitochondria and the large mitochondrion ( Figure 9, arrow 3). During late cytokinesis, this shift in net volume from the small mitochondria to the large mitochondrion is reversed. Thus, after the large mitochondrion has divided into two, the balance of fission and fusion events reverts to an increase in fission events (Figure 9, arrow 4), leading to both a significantly increased (~3x) number of small, individual mitochondria per cell and an increase (~3x) in their percent volume. This new balance between the two mitochondria types is maintained throughout the G1 to S stages of the cell cycle ( Figure 9, arrow 1).

Reticular mitochondria have not been observed in unperturbed wild type higher plant meristematic cells to date.
For over 30 years, reticular type mitochondria have been reported as a characteristic feature of unicellular organisms such as trypanosomes, yeast, fungi, Chlamydomonas, Chlorella and Euglena, (Floyd et al., 1972;Osafune et al., 1972;Hoffmann and Avers, 1973;Atkinson et al., 1974;Pickett-Heaps, 1974;Howard, 1981;Bereiter-Hahn, 1990;Yaffe, 1999Yaffe, , 1999Jakobs, 2006;Hoog et al., 2007). In contrast, the mitochondria of animal cells are typically discrete round or sausage-like organelles (reviewed in Bereiter-Hahn, 1990;Bereiter-Hahn and Voth, 1994). Only in certain types of differentiated mammalian cells have reticulate mitochondria been observed (Bakeeva et al., 1978;Bawa and Werner, 1988;Bereiter-Hahn, 1990;Smirnova et al., 1998;Yaffe, 1999 have been typically portrayed as round to sausage-like organelles (Olyslaegers and Verbelen, 1998;Logan and Leaver, 2000;Nebenführ et al., 2000;Arimura and Tsutsumi, 2002;Van Gestel and Verbelen, 2002;Logan et al., 2003;Arimura et al., 2004;Logan, 2006). Although some studies report on the presence of reticulate mitochondria in unperturbed eggs (Kuroiwa et al., 1996;Kuroiwa et al., 2002) and vascular cell types (Gamalei and Pakhomova, 1981), most of the documented examples of reticular or disc-shaped mitochondria come from mitochondrial mutants (Arimura and Tsutsumi, 2002;Logan et al., 2003), or from cells subjected to experimental perturbations such as low oxygen pressure (Bereiter-Hahn, 1990;Van Gestel and Verbelen, 2002;Logan, 2006), prolonged cell culturing (Rohr, 1978) or protoplasting (Sheahan et al., 2004;Sheahan et al., 2005). It may be argued that the exposure of the root, stem and shoot tissues to sucrose (a cryoprotectant) in the growth and freezing medium prior to high pressure freezing might have produced the "reticulate" (tentaculate/cage-like) mitochondrial architecture observed in this study. This seems unlikely, since plants grown for the CLSM studies, which were never exposed to 150 mM sucrose, displayed the same mitochondrial configuration as seen in the high pressure frozen samples (compare Figures 2, 4, 5 and 7). Furthermore, in all of our different studies, tentaculate, cage-like mitochondria were seen exclusively in SAM and LP meristematic cells.
The presence of a large tentaculate/cage-like mitochondrion in SAM and LP meristematic cells makes these cell types unique. This begs the question as to why these cells contain mitochondria with this type of morphology? For reasons detailed below, we postulate that the proliferative function of SAM cells and the fact that they are the precursors of, among others, the germ cell lines, can explain the presence of a reticulate mitochondrion in these cells.

The tentaculate/cage-like mitochondrial architecture provides a means for efficient delivery of ATP to cell proliferation-related activities.
Mitochondrial shape, number and distribution has been shown to be affected by the developmental state and the physiological status of a cell (Bereiter-Hahn, 1990;Stickens and Verbelen, 1996). From this perspective, the cell cycle-dependent changes in the spatial organization of the tentaculate/cage-like mitochondrion of SAM and LP cells can be viewed as a mechanism for becoming more energy-efficient in their primary function, cell proliferation. As shown in Figure 8, over 40% of the total mitochondrial volume is associated with the tentaculate mitochondrion that wraps itself around one pole of the nucleus ( Figures S1D, E and 7). This spatial relationship between the nucleus and mitochondria would appear to be optimal for funneling large quantities of ATP into the nucleus to support the reassembly of the interphasic nucleus in G1, the replication of the nuclear genome in S (Figure 9, G1-S) and the high rate of transcription in the G2 phase (Figure 9, G2). Similarly, the transformation of the The mitochondrial DNA (mtDNA) of flowering plants exhibits a higher frequency of intraorganelle recombination than the mtDNA of mammalian cells (Lonsdale et al., 1988;Gillham, 1994). Indeed, the angiosperm mitochondrial genome has evolved to become recombinationally active, a condition that promotes extensive genomic rearrangements (Gray et al., 1999). A prerequisite for intraorganelle recombination is mitochondrial fusion, which allows for the genomes (nucleoids) from different mitochondria to intermix. To reconcile these facts with the finding that plant mitochondria tend to be small, structurally independent units it has been proposed that plant mitochondria are "discontinuously interconnected" by means of transient fusion and fission events as defined by the "discontinuous whole" hypothesis (Logan, 2006).
However, not all plants use the same strategy to produce this discontinuous interconnected state. For example, in non dividing, highly vacuolated cells such as onion epidermal cells, mtDNA intermixing between the small, grain-shaped mitochondria appears to be achieved by means of a high rate of fusion and fission events (Arimura et al., 2004). Nevertheless, this mechanism does not guarantee an equal distribution of the mtDNA after fission. In tobacco mesophyll protoplasts induced to fuse, the fusion process triggers the formation of a large reticular mitochondrion from smaller mitochondria to homogenize the mtDNA of the two cells (Sheahan et al., 2005).
In cultured Medicago truncatula cells, the mitochondrial cycle is defined by the presence of punctate mitochondria at the onset of the log phase, the formation of more reticulate mitochondria during cell growth, and the accumulation of mitochondria around the cytokinetic apparatus during the division phase (Zottini et al., 2006).
The mitochondrial cycle reported here for SAM cells constitutes another variation on this theme. As discussed above, the organization of the SAM cell chondriome is optimized for the delivery of ATP energy to the structures involved in cell proliferation. Equally important, however, is the fact that the tentaculate/cage-like architecture also provides an efficient means for the intermixing of mtDNA and for the equal partitioning of the intermixed DNA to the two daughter cells. Thus, as the tentaculate mitochondrion is converted to a cage-like mitochondrion during the G2 stage of the cell cycle by growth and the fusion of smaller mitochondria (Figure 9), its volume increases from ~40% to ~80% of the total mitochondrial volume. Thus, during mitosis and early cytokinesis ~80% of the mtDNA is in the same mitochondrial  micrographs is often insufficient to allow for an unambiguous interpretation of the cellular structures that produce the fluorescence signals. We have overcome this limitation by producing 3D mitochondrial models based on serial thin sections of high pressure frozen tissues. This approach yields 3D models with an ~100 fold increase in resolution in the x/y-axes, and an ~4 fold increase in resolution in the z-axis (Staehelin and Kang, 2008). Prior to the advent of computer-assisted reconstruction and modeling with programs such as those included in the IMOD package (Kremer et al., 1996), the generation of 3D physical models of organelles from serial thin sections typically involved many months of labor, which limited the interest in and the use of this type of experimental approach. With current desktop computers, alignment of the thin section images, correction for section distortions and the tracing of individual organelles can be achieved in weeks, and quantitative data from the 3D models can be done in minutes. In the last years we have shown that the production of 3D models from serial thin section micrographs is a viable option for visualizing the 3D organization of cellular structures than cannot be adequately resolved by means of confocal microscopy (Seguí-Simarro and Staehelin, 2006 and this study). We believe that the use of this approach has the potential to provide novel insights on the 3D architecture of mitochondria in other cell types where perinuclear congregations of mitochondria have been reported by using CLSM.
www.plantphysiol.org on August 23, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved.

Confocal Laser Scanning Microscopy and reconstruction
Seeds of Arabidopsis thaliana (Landsberg erecta) were grown on 0.8% (w/v) agar plates with MS medium for 5 days at 24ºC (16h photoperiod). Shoot apical meristems, leaf primordia, stems and root tips were excised, incubated in 2 µM MitoTracker were provided by a UV laser and an Ar laser, respectively. The X and Y resolution of the CCD chip was 4096 pixels with a dynamic range of 12 bits per channel, while the resolution of the z-axis control (galvanometric stage) was 40 nm. Zseries images were obtained through the collection of serial, confocal sections at 0.5 µm intervals. For partial cell confocal section reconstructions, up to 11 confocal sections were converted into stacks of serial images, aligned and modeled with the IMOD software package (Kremer et al., 1996). To adjust the reconstructions to the actual cell volume, a correction factor was applied only to the Z-axis, calculated by dividing the Zinterval length of each virtual section (0.5 μ m) by the pixel size in the confocal micrographs.

Processing of EM samples by high-pressure freezing and freeze substitution
Seeds of Arabidopsis thaliana (Landsberg erecta) were grown and acclimated to increasing sucrose concentrations as previously described (Seguí-Simarro et al., 2004).
Shoot apices were excised, transferred to aluminum sample holders, cryoprotected with 150 mM sucrose and frozen in a Baltec HPM 010 high pressure freezer (Technotrade, Manchester, NH), and transferred to liquid nitrogen. The samples were freeze substituted in 4% OsO 4 in anhydrous acetone at -80ºC for 5 days, followed by slow warming to room temperature over 2 days. After rinsing in several acetone washes, they were removed from the holders, incubated in propylene oxide for 30 min, rinsed again in acetone and infiltrated with increasing concentrations of Epon resin (Ted Pella, Inc., Redding, CA) in acetone, according to the following schedule: 4 h in 5% resin, 4 h in 10% resin, 12 hr in 25% resin and 24 h in 50, 75 and 100% resin, respectively.
Polymerization was carried out at 60ºC for 2 days under vacuum.

Electron microscopy and serial section reconstruction
Ribbons of Epon (~80-100 nm thick) serials sections were collected and mounted on formvar-coated copper slot grids and stained with uranyl acetate and lead citrate (Seguí-Simarro et al., 2004). Meristematic cells in G1, S, G2, mitosis and cytokinesis were selected from the sections according to previously described morphological criteria (Seguí-Simarro and Staehelin, 2006; Table S1): nuclear/cytoplasmic surface area ratio, which progressively decreases from G1 to G2 (Seguí-Simarro and Staehelin, 2006); nuclear size, which increases as the nuclear genome replicates at the S phase (Jovtchev et al., 2006); architecture of the nucleolus and its different subdomains (dense fibrillar component, fibrillar centers, granular component and nucleolar vacuoles), which change during the cell cycle (Risueño and Medina, 1986;Risueño et al., 1988); and cell wall thickness, which progressively increases after the new cell wall is formed (Risueño et al., 1968;Seguí-Simarro and Staehelin, 2006). A prometaphase cell with a discontinuous nuclear envelope and condensed chromosmes was also selected. The cytokinetic stages of early telophase and late telophase were selected according to the differential architecture of the developing cell plate (Seguí-Simarro et al., 2004;Seguí-Simarro et al., 2007). For whole cell serial section reconstructions, single digital micrographs were converted into stacks of serial images and aligned with the IMOD software package (Kremer et al., 1996). To adjust the reconstructions to the actual cell volume, a correction factor was applied only to the Z-axis, calculated by dividing the section physical thickness (100 nm) by the pixel size in the digital micrographs. Serial EM section reconstructions were displayed and modeled with the 3DMOD program of the IMOD software package, as described previously (Seguí-Simarro et al., 2004). Since two morphologically different mitochondrial subpopulations were observed, they were modeled as different objects. This enabled us to independently acquire numerical data from each subpopulation. Once a model was completed, it was meshed and numerically analyzed using the imodmesh and imodinfo programs from the IMOD software package. Total volumes and surface areas of organelles were calculated with the imodinfo program, and the numerical data processed in a spreadsheet.       that is partly wrapped around the nucleus (n) but does not make contact with the nuclear envelope (ne) (see also Figure S1C). This central sheet-like domain connects to more tubular, peripheral domains as shown in (B)