The Arabidopsis thaliana pollen cell wall is a complex structure consisting of an outer sporopollenin framework and lipid-rich coat, as well as an inner cellulosic wall. Although mutant analysis has been a useful tool to study pollen cell walls, the ultrastructure of the arabidopsis anther has proved to be challenging to preserve for electron microscopy.
In this work, high-pressure freezing/freeze substitution and transmission electron microscopy were used to examine the sequence of developmental events in the anther that lead to sporopollenin deposition to form the exine and the dramatic differentiation and death of the tapetum, which produces the pollen coat.
Cryo-fixation revealed a new view of the interplay between sporophytic anther tissues and gametophytic microspores over the course of pollen development, especially with respect to the intact microspore/pollen wall and the continuous tapetum epithelium. These data reveal the ultrastructure of tapetosomes and elaioplasts, highly specialized tapetum organelles that accumulate pollen coat components. The tapetum and middle layer of the anther also remain intact into the tricellular pollen and late uninucleate microspore stages, respectively.
This high-quality structural information, interpreted in the context of recent functional studies, provides the groundwork for future mutant studies where tapetum and microspore ultrastructure is assessed.
The highly reduced gametophyte of flowering plant species forms within the sporophytic tissues of the anther. These microgametophytes, or pollen grains, play a central role in plant reproduction, carrying the sperm cells to the female reproductive structures for double fertilization. The critical role fulfilled by pollen grains in the life cycle of flowering plants has required that these structures be fortified against the elements of terrestrial life. The wall encasing the pollen is highly specialized, consisting of an inner (intine) cellulosic wall, and an outer (exine) wall containing a highly stable and recalcitrant biopolymer called sporopollenin (Southworth, 1974; Scott, 1994). In many species, including Arabidopsis thaliana, the exine additionally contains a pollen coat, which covers the patterned sporopollenin framework and aids in hydration and species-specific recognition by the stigma.
Pollen development relies heavily on the surrounding tissues of the anther wall, especially the tapetum, but also the middle layer, endothecium and epidermal cell layers. The epidermis prevents anther water loss and aids gas exchange and, together with endothecial cells, provides structural support to the anther wall (Goldberg et al., 1993). Secondary wall thickenings in the endothecial cells are required for pollen dispersal by anther dehiscence along specialized cells of the anther epidermis (Keijzer, 1987; Mitsuda et al., 2005). In contrast, cells of the middle layer and tapetum undergo cell death and are absent from mature anthers prior to dehiscence. Although the precise role of the middle layer is not known, the middle layer becomes tapetum-like if its degeneration is postponed, as in the arabidopsis fat tapetum mutant (Sanders et al., 1999). The middle layer also has secretory activity and in dioecious plants such as Actinidia deliciosa (kiwifruit) may regulate the production of male-sterile and male-fertile flowers by altering the timing of its own cell death (Falasca et al., 2013). The gametophytic reliance on sporophytic tissues is particularly clear in the case of tapetal cells. Tapetum-specific cell ablation studies demonstrated the specific and critical role served by tapetal cells in the development of pollen grains, as microspores aborted in their absence, while other anther cell types appeared unaffected (Mariani et al., 1990; Goldberg et al., 1995). The delay of tapetum degeneration that occurs in mutants, such as arabidopsis MALE STERILE1 (MS1) and rice TAPETUM DEGENERATION RETARDATION (TDR), results in collapsed microspores (Li et al., 2006; Yang et al., 2007). Signalling between microspore and tapetal cells and the complex transcriptional regulation of tapetum proliferation and cell death are both postulated to be necessary for the critical timing of development in the tapetum (reviewed by Parish and Li, 2010).
Among the many essential roles of tapetal cells, this sporophytic cell layer is a key player in forming the highly sculptured sporopollenin wall of the exine. The role of the tapetum in microspore wall formation is evident early in anther development, when microspore tetrads are freed from their callose encasement by tapetum-secreted callase (Dong et al., 2005). In the absence of normal callose dissolution in arabidopsis, as in the myb103 mutant, defective exine patterning is observed (Zhang et al., 2007). The tapetum-specific expression of genes required for sporopollenin synthesis indicates that an abundance of sporopollenin is exported from intact tapetal cells during the free uninucleate microspore stage of pollen development (see reviews by Ariizumi and Toriyama, 2011; Liu and Fan, 2013). Recently, immuno-localization and protein–protein interaction studies have supported a model in which the enzymes producing sporopollenin precursors are organized in a metabolon within the tapetum (Lallemand et al., 2013). However, as our knowledge of these functional components has progressed, our understanding of the dynamic cellular structure of the tapetum and the ultrastructure of the developing pollen wall over the course of anther development has lagged. The tapetum is a transitory tissue and, together with the many structures associated with early microspores, appears to be sensitive to the chemical fixation protocols employed for light and electron microscopy (Sanders et al., 1999; Ariizumi et al., 2005; de Azevedo Souza et al., 2009). A study in Brassica campestris used cryo-fixation to document early microspore wall development in the tetrad stage, but examination of tapetum ultrastructure was not within the scope of the work (Fitzgerald and Knox, 1995). In a study characterizing DEFECTIVE EXINE1 (DEX1) protein function in pollen wall formation, the formation of structured components within the primexine matrix, known as probaculae and protecta, and the importance of microspore undulations in the formation of these structures was demonstrated by comparing high-pressure frozen and freeze-substituted arabidopsis wild-type and dex1 mutant anthers in the tetrad stage (Paxson-Sowders et al., 2001). These studies validated cryo-fixation as a useful approach to preserve the delicate and transitory structures of early microspore walls in the tetrad stage, which serve as a scaffold for the subsequent polymerization of tapetum-derived sporopollenin into baculae and tecta.
In contrast to sporopollenin, which appears to be actively exported from the intact tapetum early in microspore development, components of the lipidic pollen coat are produced and stored in the tapetum and deposited onto pollen upon tapetum programmed cell death (Heslop-Harrison, 1968a). The pollen coat components are liberated upon the breakdown of two specialized organelles in the tapetum, elaioplasts and tapetosomes, which accumulate at late stages of pollen development (Piffanelli et al., 1998; Hsieh et al., 2003). Elaioplasts are specialized plastids that are rich in steryl esters, free polar lipids and plastid lipid-associated proteins and are filled with globuli (Piffanelli and Murphy, 1998; Ting et al., 1998; Kim et al., 2001). Tapetosomes are unique, densely packed organelles that are intimately associated with the endoplasmic reticulum (ER), and consist of a fibrous meshwork of vesicles, fibrils and oil bodies containing oleosin proteins, alkanes and flavonoids (Hsieh and Huang, 2004, 2005, 2007). The biochemical data on the components of tapetosomes and elaioplasts have been correlated with their ultrastructure in chemically fixed specimens; however, these structures are rich in membranes and lipids, so structural features are sometimes ambiguous.
The objective of this study was to define the ultrastructure of the developing anther of arabidopsis using cryo-fixation techniques in order to preserve the delicate tapetum tissue and developing pollen grains during production of the complex pollen cell wall. The goal is to provide a detailed analysis of wild-type cell structure in arabidopsis anthers, which will serve as a baseline for studies examining cellular and cell wall architecture in mutants affected in microspore and tapetum development.
MATERIALS AND METHODS
Wild-type arabidopsis (Arabidopsis thaliana, Columbia-0) seeds were sterilized in ethanol and germinated on agar plates with Murashige and Skoog medium, pH 5·7. Seeds grown under 24 h light at 28 °C for 7 days were then transferred to soil (Sunshine Mix 4, Sungrow Horticulture), and grown to maturity at 20 °C under long-day conditions (20 °C, 16 h light/8 h dark cycles).
Cryo-scanning electron microscopy analysis
Stamens were dissected from multiple immature buds (0·7–1·0 mm long) and submerged in sterile water within copper sample holders (type B hats, Ted Pella Inc.). The stamen-filled chamber was sealed with a second hat and secured to a cryo-electron microscopy specimen block with Tissue-Tek optimal cutting temperature compound (Ted Pella, Inc.). The block was submerged in liquid nitrogen, transferred under vacuum to a fracturing chamber and fractured by removing the upper hat from the sample. The sample block was moved under vacuum (using the VCT100 system, Leica Microsystems) to a pre-cooled scanning cryo-electron microscope (cryo-SEM; −135 °C, S-4700 Field Emission SEM, Hitachi), sublimed to remove surface ice (−95 °C for 20 min) and imaged in an uncoated state at −118 °C with an accelerating voltage of 5·0 kV.
High-pressure freezing and freeze substitution
For high-pressure freezing, buds representing a range of anther developmental stages were dissected from flowering wild-type plants (4–6 weeks old). The perianth was removed from each bud and remaining stamens and gynoecium were submerged in extracellular cryoprotectant [0·2 m sucrose between two sample holders that formed a chamber (type B hats, Ted Pella, Inc.)]. High-pressure freezing was performed immediately (EM HPM 100, Leica Microsystems), and samples were transferred under liquid nitrogen to cryovials of freeze-substitution medium containing 2 % w/v osmium tetroxide in anhydrous acetone with 8 % v/v dimethoxypropane. Sample vials underwent cryosubstitution within a slurry of dry ice and acetone at –80 °C for 3 days; then, the vials were warmed to −20 °C by the end of 1 week. Sample vials were warmed to room temperature, and the substitution medium was replaced with acetone. Samples were embedded in Spurr's epoxy resin over 4 days, reaching 10 % v/v resin in acetone on day 1, 60 % on day 2 and 100 % on day 4. Resin-embedded samples were transferred to capsules (BEEM Embedding Capsules #70020, Electron Microscopy Sciences) and polymerized at 60 °C for 2 days. Four sets of wild-type stamen samples were independently prepared by the methods described and used for light microscopy and transmission electron microscopy (TEM) analyses.
Light microscopy and TEM
Resin-embedded samples were sectioned to a thickness of 0·3–0·5 μm for light microscopy and 50–70 nm for TEM, with a microtome (Ultracut E, Leica-Reichert) and diamond knife (Ultra 45°, Diatome). For viewing and staging anther development by light microscopy (with a Leica DMR microscope), sections were stained on glass slides with toluidine blue in 1 % sodium borate, and sealed beneath a Permount-attached coverslip. TEM sections were collected on copper 100-hex grids with a 0·5 % Formvar coating with silver–grey interference colour, and stained first with uranyl acetate (2 % in 70 % v/v methanol, 15 min), then with Reynolds lead citrate (6–8 minutes). Grid-mounted samples were viewed on a Hitachi H7600 TEM and images were captured with an AMT Advantage (1 megapixel) CCD camera (Advanced Microscope Technologies).
The classic study by Sanders et al. (1999) defined the stages of pollen and anther development in arabidopsis using light microscopy. Here, we re-examined these stages using high-pressure freezing and freeze substitution combined with TEM. We first undertook an overview of pollen development using light microscopy to follow the development of all cell layers of the anther through the meiotic events of microsporogenesis and mitotic events of microgametogenesis (Fig. 1). In addition, corresponding pollen morphology and nuclei associated with these development stages are shown (Supplementary Data Fig. S1). Following cryopreservation, freeze substitution and resin embedding, the sectioned anthers captured near the first microspore meiotic division were cytoplasmically dense and the cell layers were not separated (Fig. 1A, B), as seen in earlier studies employing chemical fixation. Tapetal cells at the tetrad stage (Fig. 1C) contained enlarged vacuoles and were easily distinguished from surrounding sporophytic layers by their dense cytoplasmic contents and binucleate nature. The tapetum had a tendency to separate from the surrounding middle layer cells, but our fixation method, unlike chemical fixation, showed that the tapetum and locule were tightly associated. At more advanced stages of microsporogenesis, the tapetum became less vacuolated and began to accumulate dense organelles (Fig. 1D, E). Throughout these stages, the tapetal cells formed a continuous ring of tightly packed cells (Fig. 1A–E), unlike some chemically fixed samples, in which tapetal cells were separated by gaps (Sanders et al., 1999). The preservation method employed here additionally revealed an abundance of locule fluid around microspores through to the early tricellular stage of pollen development (Fig. 1D–G), in contrast to previous studies, in which the medium between microspores and tapetal cells appeared as an empty space (Sanders et al., 1999).
After two meiotic divisions and cytokinesis had transformed callose-encased MMCs to tetrads of haploid microspores (Fig. 1A–C), microsporogenesis concluded with the release of free haploid microspores from their callose encasement (Fig. 1D). At this stage of development, a dramatic change in the free microspores was observed, as a structured exine wall of sporopollenin became visible (Fig. 1D, E). At early stages, free uninucleate microspores appeared cytoplasmically dense and contained small but numerous vacuoles around a centrally located nucleus (Fig. 1D, Supplementary Data Fig. S1). In more mature uninucleate microspores, the cytoplasmic contents were in a peripheral location as the vacuoles coalesced into one enlarged vacuole (Fig. 1E, Supplementary Data Fig. S1). In our samples, free microspores appeared round, and the cytoplasmic contents filled the area formed by the microspore wall, in contrast to previous studies in which microspores frequently appeared shrunken or detached from their walls. With their densely staining exine wall, polarized uninucleate microspores, often called ring-vacuolate microspores (Blackmore et al., 2007), represent the final stage of microsporogenesis before the divisions of microgametogenesis that will form mature pollen grains (Fitzgerald and Knox, 1995).
In stages of bicellular pollen grain development, a small generative cell was visible within a large vegetative cell (indicated by arrows in Fig. 1F, G; for enlargement see Supplementary Data Fig. S1), consistent with the asymmetry of the first mitotic division. Each pollen grain contained multiple tiny vacuoles, unlike the enlarged vacuole of the previous stage, and the large vegetative cells appeared rich in subcellular organelles (Supplementary Data Fig. S1). After the second mitotic division, during which the generative cell had produced two male gametes (or sperm cells), tricellular pollen grains were larger in volume and filled with numerous lipid bodies and starch granules (Fig. 1H, Supplementary Data Fig. S1). Numerous darkly staining organelles were prominent in tapetal cells surrounding pollen in the bicellular and early tricellular stages of development (Fig. 1F–H). These organelles, although not clearly discernable in the light microscope, were consistent in size and shape with elaioplasts and tapetosomes (Fig. 1F–H). The locule fluid, which stained with toluidine blue in early bicellular pollen-containing anthers (Fig. 1F), was weakly stained in the late bicellular pollen stage (Fig. 1G) and appeared absent in locules containing tricellular pollen (Fig. 1H). Simultaneously, tapetal cells gradually appeared thinner and became almost entirely occupied by densely staining organelles (Fig. 1H). The surrounding endothecial cells became highly vacuolated, and the cell corners stained light blue with toluidine blue, which is typical of lignified cell walls (see cells marked by asterisks in Fig. 1H). Also in the tricellular pollen stage, the intine, or inner pollen wall, was clearly visible by light microscopy as a lightly staining band between the darkly staining exine and pollen cytoplasm (arrows in Fig. 1I). In the late tricellular pollen stage, the presence of tapetum cytoplasmic constituents in the locule indicated cell disintegration after programmed cell death had occurred (Fig. 1I). Finally, mature, slightly oblong tricellular pollen grains, ready for release by anther dehiscence, filled the locule space formed by the endothecium and epidermis (Fig. 1J).
From this analysis of anther development, most anther cell types over the course of microsporogenesis and microgametogenesis were clearly visible and preserved in high quality by the cryo-fixation and embedding techniques employed. However, this analysis provided limited information on the middle layer, as these cells appear to be easily damaged in the embedding process and were poorly resolved by light microscopy. To view the middle layer of the anther wall and microspores with associated locule fluid in a near-native and hydrated state, immature anthers were cryo-fractured longitudinally and visualized by cryo- SEM (Fig. 2). In two of the four microspore-filled locules, all anther wall cell layers were visible (Fig. 2A). Each locule revealed uninucleate microspores in the ring-vacuolate stage of development (Fig. 2B, equivalent to the stage depicted in Fig. 1E). Surrounding microspores, the fluid of the locule appeared semi-ordered with a lattice-like substructure. While this may reflect some ice crystal formation after plunge freezing, with separation of solutes into the eutectic phase between ice crystals, the regularity of the pattern is intriguing and is worthy of further investigation. Furthermore, this analysis showed that the tapetum bulges into the locule in close association with nearby microspores, rather than forming the typical rigid shape of cell wall-encased plant cells. Interestingly, in contrast to the current models that describe loss of the middle layer early in development, the middle layer cells, directly external to the tapetum, clearly contained cytosolic contents, in agreement with findings by Owen and Makaroff (1995) (Fig. 2B). A wide band of endothecial cells was visible, external to the thin band of middle layer cells and below a surface layer of epidermal cells. A waxy cuticle was observed on the outermost face of epidermal cells. The cryo-SEM method employed here revealed two poorly characterized features of immature anthers: the lattice-like substructure of the locular fluid and persistence of the middle layer into the latest stage of uninucleate microspore development.
The light microscopy analysis suggested that the cryo-fixed anthers retained structural elements that were not preserved in earlier studies using chemical fixation, but the sub-micron details were beyond the limit of light microscopy resolution. In order to trace the development of the multilayered pollen wall, we used cryo-fixation combined with TEM to examine its origins, beginning with the primary cell wall surrounding meiotic microspore mother cells (MMCs), through primexine formation around microspores in tetrads, and sporopollenin-based exine wall formation around free uninucleate microspores. MMCs demonstrated a high degree of symplastic continuity, with abundant cytomictic channels interconnecting the cytoplasm of MMCs within each anther locule (Fig. 3A; arrowheads indicate cytomictic channels). The plasma membrane and primary cell wall of meiotic MMCs (Fig. 3B, C) were separated by the previously characterized callose wall (Dong et al., 2005), and no primexine wall was visible. In the tetrads, the primexine wall first became visible on the surface of individual microspores (Fig. 3D–F). With the exception of aperture sites, the microspore plasma membrane appeared invaginated at regular intervals and acquired ‘spacers’ in each of the pockets formed by its undulations (Fig. 3E), which appeared electron-dense, consistent with previous reports (Fitzgerald and Knox, 1995; Paxson-Sowders et al., 1997). Separated by spacers, the peaks along the microspore plasma membrane correlated with sites of probacula assembly (arrowheads in Fig. 3E, F). In a late tetrad stage, probaculae and protecta were distinguished as electron-dense cones topped by caps, respectively, and extended from the microspore plasma membrane at regular intervals within the primexine matrix (arrowheads in Fig. 3F). Although little is known about the composition of probaculae and protecta, these structures appear to contain an early form of sporopollenin (Heslop-Harrison, 1968b).
Previous studies using cryo-fixation in arabidopsis and Brassica sp. have focused specifically on wall formation within the tetrad stage of pollen development. Although a number of studies have examined the microspore wall after tetrad release, the chemical fixation methods employed typically result in shrunken microspores and separation of the microspore plasma membrane and exine wall. Therefore, we extended our investigation of microspore wall ultrastructure in cryo-fixed anthers through the uninucleate stages of microspore development. At this stage, in contrast to clusters of microspores in the tetrad stage, individual microspores were surrounded by a homogeneous locule of medium electron density (Fig. 3G, J). Microspores at this early uninucleate stage typically exhibited an exine of baculae and tecta that was homogeneously electron-dense (Fig. 3I). Upon close examination, the exine wall of a subset of free microspores contained lamellae, visible as electron-translucent lines within an electron-dense wall, and situated roughly parallel to the microspore plasma membrane (arrows in Fig. 3H). Remnants of the primexine matrix were observed surrounding the sculptured exine wall, appearing as a loose mesh between the locule and the sporopollenin (Fig. 3I).
In the absence of cytoplasmic shrinkage and separation from the cell wall, the intine wall, a thin band of wall material between the exine and the microspore plasma membrane, was identified in early free microspores (Fig. 3I) and became clearly distinguished by the mid-free microspore stage (Fig. 3K). Surrounding the developing intine, the exine wall around mid-uninucleate microspores consisted of highly electron-dense baculae and tecta (Fig. 3K, asterisks and arrowheads, respectively). In sections cut tangential to the mid-uninucleate microspore through the exine, baculae appeared round and discrete, connected by tecta, where captured (Fig. 3L). In summary, the first evidence of the patterned outer wall was identified on the plasma membrane of microspores in tetrads. Deposition of the interior layer of the pollen wall, the intine, was observed by the end of the free uninucleate microspore stage. No gaps or locule fluid were observed between layers of the pollen wall and the callose wall or microspore plasma membrane, in contrast to many previous reports.
To obtain high-quality information on tapetum ultrastructure over the course of microspore development, we began by examining anthers in developmental stages near the time of microspore meiosis. Tapetal cells in arabidopsis are binucleate from the MMC stage (Supplementary Data Fig. S2A) to the late uninucleate microspore stage of anther development (Supplementary Data Fig. S2B). Although the binucleate state of tapetal cells presumably persists until the tapetal cell death programme is initiated, this state is difficult to capture as elaioplasts and tapetosomes rapidly consume the majority of the cellular space. As early as the MMC stage of pollen development, the tapetal cells were distinguishable from the middle layer, endothecial and epidermal cell layers of the anther wall by their dense cytoplasm and binucleate nature (Fig. 4A). These tapetal cells were cytoplasmically dense, with ER, ribosomes, mitochondria and vacuoles (Fig. 4B). In subsequent stages, the tapetum appeared metabolically active, based on the high abundance of ER, Golgi apparatus, mitochondria and proplastids in the tetrad and free uninucleate stages of microspore development (Fig. 4C–H). Where microspores were not closely appressed to the tapetum, the locule-facing edge of tapetal cells often appeared wavy with occasional invaginations (asterisks in Fig. 4D, F, H). Thus, during the period of peak sporopollenin synthesis, from the tetrad stage to the uninucleate microspore stage, the tapetum is rich in ER and the plasma membrane, lacking a cell wall, is often in contact with microspores.
In addition to sporopollenin biosynthesis, tapetal cells are known to synthesize numerous lipid-rich pollen coat constituents during the pollen mitotic divisions. We investigated tapetum ultrastructure in detail at these stages in cryo-fixed anthers (Fig. 5). Proplastids and abundant ER observed at earlier stages were replaced by elaioplasts, tapetosomes and morphologically distinct rough and smooth ER (Fig. 5A). At the tricellular pollen stage, elaioplasts and tapetosomes filled the majority of each tapetal cell, pushing lamellar smooth ER (asterisk in Fig. 5B) and rough ER (arrows in Fig. 5B) to the cortical cytoplasm. With the preservation methods we employed, tapetal cells remained intact through the bicellular and early tricellular pollen stages (Fig. 5A, B). Consistent with tapetum rupture due to programmed cell death, the contents of tapetal cells were scattered in the locule and appeared closely associated with the crevices of tricellular pollen exine (Fig. 5C), or were absent from locules, in which the tricellular pollen grains were engorged and exhibited pollen coats (Fig. 5D).
Close inspection of tapetal cells at the late stages of pollen development revealed organelle substructure not previously reported. Although tapetal cells were primarily filled by elaioplasts and tapetosomes in late bicellular and tricellular stages of pollen development, distinct forms of ER were also observed. Dilations of rough ER with fibrillar material in the lumen were observed throughout the tapetal cytoplasm (arrows in Figs 5A and 6A), alongside developing tapetosomes, elaioplasts, Golgi bodies and lamellar ER (asterisk in Figs 5B and 6B). Large sheets of lamellar smooth ER were also observed specifically in very late stages of tapetal cell development, just prior to their rupture (asterisk in Figs 5B and 6B). In the final stages of tapetum development, the tapetum exhibited fragmented cytoplasm and lacked free ribosomes (Fig. 6B–D).
Close inspection of elaioplasts (Fig. 6C) and tapetosomes (Fig. 6D) revealed clear substructure within these organelles. Elaioplasts contained numerous round globuli within their stroma. Within each elaioplast globule, two main components of varying texture and electron density were visible (specified by boxed regions in Fig. 6C, magnified in Fig. 6E, F). The electron-translucent component, presumably containing steryl esters, formed small and often circular inclusions within an electron-dense meshwork. The preservation methods used here revealed a large proportion of electron-dense constituents associated with elaioplast globuli. In addition to the globuli of elaioplasts, occasional membrane-like protrusions were observed between the globuli (Fig. 5A).
In contrast to elaioplasts, tapetosomes transitioned from a near circular appearance in bicellular pollen locules to amorphous multi-lobed shapes that occupied much of the cellular space and lacked an encasing membrane (Fig. 6A, C, D). Within tapetosomes, distinct components were differentiated, including clusters of rod-like, tightly packed membrane tubes (Fig. 6G) that occurred within a network of branching tubules in a dense matrix (Fig. 6H). The sample preservation methods employed here revealed a high level of substructure within tapetosomes that was difficult to correlate with previously reported tapetosome structures. The dense matrix with associated tubules may be ER-derived membranes housing flavonoids, as have been identified in Brassica sp. tapetosomes (Hsieh and Huang, 2005, 2007). The nature of the densely packed rod-like structures of tapetosomes (Fig. 6G) has not been described previously, and their chemical composition is unknown.
Finally, we employed TEM to view the subcellular details of the cryo-fixed mature pollen cell wall at high resolution. The ultrastructure of the mature pollen wall was striking: the detailed structures of the intine and exine, tightly appressed to the plasma membrane of the microspore, were revealed [Fig. 7, labelled blue in (B)]. The intine appeared as a thick electron-lucent band with microfibrils running parallel to the vegetative cell plasma membrane [Fig. 7, pink band in (B)]. The exine exhibited two main constituents, a homogeneous and sculpted component consistent with the sporopollenin-based wall, and a heterogeneous component consistent with the pollen coat [Fig. 7, false-coloured green and orange, respectively, in (B)]. The sporopollenin component of the exine wall appeared uniformly electron-dense, with baculae that extended radially from the intine and tecta that formed at the external end of baculae, roughly parallel to the pollen plasma membrane. A thin band of comparable electron density was also observed at the interface between the intine and pollen coat in the crevices between baculae. In contrast to the homogeneous appearance of the sculptured sporopollenin, the pollen coat component of the exine appeared highly heterogeneous. Within the pollen coat, two electron-translucent components were observed, one appearing as long and rectangular, the other appearing roughly star-shaped (arrowheads in Fig. 7B). As the images were taken in cross-section, it is possible that these two electron-translucent constituents represent the same component of the coat, captured in different view planes. While the remainder of the pollen coat was electron-dense, it exhibited previously undescribed regions of regularly patterned lamella (asterisk in Fig. 7B).
In this study, high quality cryo-preservation methods were employed to clarify the major events and morphological markers in arabidopsis anther development, while additionally defining novel features within tapetum organelles and microspore walls. High-pressure freezing and cryo-fixation methods enabled cellular ultrastructure to be rapidly preserved, preventing many of the artefacts and distortions observed when conventional methods of chemical fixation were used (Gilkey and Staehelin, 1986). In particular, cryo-fixation prevented the separation between cells of the tapetum, microspore shrinkage and wall detachment, the loss of locule fluid and the deformation of subcellular membranes and organelles. The application of anther plunge-freezing and cryo-SEM additionally enabled the observation of all layers of the anther wall without the artefacts caused by tissue embedding. This approach revealed novel information on the poorly characterized middle layer, locule fluid and the morphology of tapetal cells. This study highlights the numerous cellular events required to assemble the multi-layered pollen wall and the importance of optimized sample preparation when analysing pollen and tapetum development, which will be particularly important in mutants with altered tapetum and/or microspore wall development.
The dependence of developing pollen grains on tapetal cells has been well established (Ariizumi and Toriyama, 2011; Liu and Fan, 2013). However, a detailed analysis of arabidopsis tapetum ultrastructure through pollen development is lacking in the literature, likely due to the challenges associated with preservation. The TEM analysis in this study revealed several important features of tapetal cells. First, the tapetum lacks a cell wall as early as the tetrad stage (Figs 1 and 2), as hypothesized by Owen and Makaroff (1995). Second, the tapetum ultrastructure during exine deposition showed no signs of an active endomembrane secretion system, e.g. post-Golgi vesicles, consistent with a plasma membrane transporter-mediated export of sporopollenin components. These observations are consistent with the model that ATP-binding cassette (ABC) transporters, such as arabidopsis ABCG26 and rice ABCG15 (Quilichini et al., 2010; Qin et al., 2013), export sporopollenin components from the tapetum. Finally, these data provide evidence that tapetal cells were distinguishable into the tricellular pollen stage (anther stage 12; Fig. 1H), in contrast to previous reports indicating that tapetal cells are absent from locules with tricellular pollen (Sanders et al., 1999).
The most striking features of the tapetum revealed in this TEM analysis were tapetosomes and elaioplasts, two highly specialized organelles that distinguish late-stage tapetal cells from all other plant cells, and are the major contributors to pollen coat formation. Tapetosomes and elaioplasts are typically described as spherical and approximately 3 μm in diameter (Wu et al., 1997). In contrast, we found that the size, shape and number of tapetosomes varied dramatically based on the developmental stage sampled. They appeared numerous and round in bicellular pollen stages, and amorphous and branching through a large portion of the tapetum cytoplasm in tricellular pollen stages. The biogenesis of tapetosomes has been proposed to resemble oil body biogenesis (Hsieh and Huang, 2004), but arabidopsis tapetosomes are morphologically distinct, consisting of tightly bunched membrane tubules in a dense matrix rather than discrete spherical oil bodies, as found in seeds (Fig. 6). While tapetosomes were contiguous with ER, their relationship to the dilated rough ER cisternae and the abundant smooth ER is presently unknown. In Brassica napus (brassica), cell fractionation, biochemical analysis, and confocal microscopy have revealed that flavonoids co-localize with ER and tapetosomes, while alkanes are associated with tapetosome oil bodies (Hsieh and Huang, 2007). Tapetosome oil droplets additionally contain triacylglycerols and are believed to be coated by a layer of phospholipids and oleosin proteins (Hsieh and Huang, 2004). The distribution of flavonoids and triacylglycerols within the branching tubules and rod-like structures observed in arabidopsis tapetosomes are not known. The two distinct regions of the tapetosomes observed in our study, i.e. branched tubules and rod-like clusters, are consistent with earlier work in which Platt et al. (1998) observed that chemically fixed brassica tapetosomes (then called lipid bodies) had two distinct regions, as differentiated by their relative electron densities, although in that study the substructure was not clear.
As in tapetosomes, substructural features within arabidopsis tapetum elaioplasts complement the previous biochemical and ultrastructural data. In B. napus, elaioplasts contain abundant steryl esters, which constitute the predominant neutral lipids in these specialized plastids and are believed to form the contents of globuli (Platt et al., 1998; Wu et al., 1999). Based on their presence and abundance in elaioplasts, phospholipids, glycolipids, monogalactosyldiacylglycerols and plastid-associated proteins (PAPs) are additionally predicted to associate with and/or maintain the surface of globuli (Wu et al., 1999; Kim et al., 2001). We identified two components of varying texture and electron density within elaioplast globuli not previously described. Based on the chemical data available, we predict that the electron-dense meshwork we observed in globuli represents a lipid and PAP-based protein scaffold that encases globuli of the electron-translucent (or extracted) steryl esters. Instead of rectangular crystals within globuli of elaioplasts (Platt et al., 1998), the electron-translucent areas of globuli exhibited a variety of shapes in our samples (Fig. 6).
Although the cryo-fixation and embedding methods we employed preserved novel features of tapetum cellular content and membranes, the fragility of middle layer cells still presented challenges. Although the middle layer is often cited as a cell layer that ruptures in the tetrad stage of development (Sanders et al., 1999), there is evidence that the middle layer cells persist into the tricellular stage of pollen development in arabidopsis (Owen and Makaroff, 1995). With cryo-SEM, it was clear that middle layer cells persist after the release of free microspores and into the late stages of uninucleate microspore development, where they appear alive based on their subcellular contents. Although the function of middle layer cells in arabidopsis pollen development is unknown, our data suggest that their cell walls, as in tapetal cells, may be reduced or altered. In our cryo-preserved samples, and in chemically preserved samples (Owen and Makaroff, 1995), there was spatial separation of tapetal cells from the anther wall. Given that cryo-fractured samples examined by cryo-SEM did not show this separation, the previously observed separation may be an artefact of the embedding process rather than a native state that aids secretion, as proposed previously (Owen and Makaroff, 1995).
While most cell layers of the sporophytic anther wall have defined roles in pollen development and pollen cell wall formation, the medium that fills the locule and is in direct contact with both the tapetal cells and developing pollen grains is poorly understood. Our images clearly show that locule fluid was present from the tetrad stage to the early bicellular pollen stage, after which it appeared to become less dense, and was minimal or absent from locules filled with tricellular pollen grains. Interestingly, this is also the time frame in which sporopollenin is likely to traffic from tapetal cells to microspores, and, during this time frame, lipid transfer proteins are abundant in the locule fluid (Huang et al., 2013). Although some studies have suggested that traffic of pollen wall components occurs through the locule in vesicles, we found no evidence to support this hypothesis. While the locule fluid appears homogeneous when embedded anthers are sectioned for light microscopy (Fig. 1) and TEM (Figs 3–6), it exhibited a lattice-like appearance in cryo-fractured anther locules at comparable stages (Fig. 2). This is likely due to the formation of ice crystals upon freezing; however, the ordered nature of these ice crystals may point to structural organization within the locule, such as the viscin thread-like structures between microspores and tapetal cells described in Betula pendula anther locules (Rowley and Morbelli, 2009). If this interpretation is correct, the locule in arabidopsis, or other flowering species with secretory tapeta, and its protein contents, could aid traffic to the developing microspores.
The uninucleate microspore stage represents a critical transition, when the sporopollenin framework assembles. Intermediate stages of this process were observed, suggesting that visible transitions occur in the exine as the framework matures. Judging by the maturity of the patterned wall, the first step in sporopollenin polymerization is characterized by lamellae within the exine, followed by the transition of the exine into the homogeneous baculae and tecta in all subsequent stages (Fig. 3). The lamellar appearance in exine formation in some species has been hypothesized to indicate early sporopollenin polymerization, much like the models for suberin and cutin assembly, but early lamellar exine has not been described in arabidopsis or other dicot microspore walls previously (Scott, 1994). In Poa annua and Lilium sp., the initial lamellar appearance of the exine is replaced by a homogeneous exine and is hypothesized to indicate the progression from partially polymerized to polymerized acetolysis-resistant sporopollenin (Rowley, 1962; Dickinson and Heslop-Harrison, 1968). This hypothesis is supported by our data, as the lamellar appearance associated with exine was only observed in early free microspores (Fig. 3H), while the homogenous sporopollenin wall was observed in all mid-late uninucleate microspore walls (Fig. 3 K). While the exact time frame over which sporopollenin export and assembly occurs is not known, the sporopollenin framework built in uninucleate microspore stages did not appear altered in subsequent stages, suggesting that this wall is complete by late-uninucleate pollen development stages. Recent analyses of the timing of gene expression and mutant phenotypes in the polyketide synthase/sporopollenin biosynthetic pathway are consistent with this model (de Azevedo Souza et al., 2009; Grienenberger et al., 2010; Kim et al., 2010).
After tapetum programmed cell death, the deposition of the pollen coat marks the completion of wall formation on arabidopsis tricellular pollen grains. Under our fixation methods, mature pollen grains exhibited extensive heterogeneity within the pollen coat. The structures of the coat observed here are likely to include neutral lipids, alkanes and very long chain wax esters (Preuss et al., 1993; Fiebig et al., 2000; Ariizumi et al., 2003), as well as flavonoids (Hsieh and Huang, 2007), and hydroxycinnamoyl spermidine metabolites (Grienenberger et al., 2009), together with proteins such as lipid transfer proteins, oleosins and self-incompatibility proteins (Piffanelli et al., 1998; Huang et al., 2013).
In conclusion, we applied cryo-fixation and diverse imaging techniques to the analysis of anther, pollen wall and tapetum development in wild-type arabidopsis anthers. These data provide a baseline of the morphological features in developing arabidopsis anthers, especially in the tapetum. Future application of these methods to functional studies of mutant plants with altered pollen wall formation will be particularly revealing.
Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Fig. S1: light microscope images of pollen morphology through microsporogenesis and microgametogenesis. Fig. S2: the binucleate tapetum in arabidopsis.
Funding for this study was provided by the Canadian Natural Sciences and Engineering Research Council, Collaborative Research and Training Experience Grant, and Discovery Grants, to A.L.S. and C.J.D; and a Post Graduate Scholarship-Doctoral to T.D.Q. The assistance provided by the BioImaging Facility staff (University of British Columbia) is gratefully acknowledged. We thank Rebecca Smith, Heather McFarlane, Anika Benske and Miranda Meents (University of British Columbia) for thoughtful comments on the manuscript.