High‐resolution episcopic microscopy enables three‐dimensional visualization of plant morphology and development

Abstract The study of plant anatomy, which can be traced back to the seventeenth century, advanced hand in hand with light microscopy technology and relies on traditional histologic techniques, which are based on serial two‐dimensional (2D) sections. However, these valuable techniques lack spatial arrangement of the tissue and hence provide only partial information. A new technique of whole‐mount three‐dimensional (3D) imaging termed high‐resolution episcopic microscopy (HREM) can overcome this obstacle and generate a 3D model of the specimen at a near‐histological resolution. Here, we describe the application of HREM technique in plants by analyzing two plant developmental processes in woody plants: oil secretory cavity development in citrus fruit and adventitious root formation in persimmon rootstock cuttings. HREM 3D models of citrus fruit peel showed that oil cavities were initiated schizogenously during the early stages of fruitlet development. Citrus secretory cavity formation, shape, volume, and distribution were analyzed, and new insights are presented. HREM 3D model comparison of persimmon rootstock clones, which differ in their rooting ability, revealed that difficult‐to‐root clones failed to develop adventitious roots due to their inability to initiate root primordia.

HREM is a microscope-microtome-based imaging system, which generates hundreds to thousands of perfectly aligned thin-section images of an embedded tissue. A fluorescence stereomicroscope and a digital camera facilitate a block-face image capture of fluorescent dyes mixture, which stains the embedded specimen in the block. Following every section (typically ~ 1-5 µm thick), the camera captures the image of the block cut surface. The images are stacked and processed using 3D visualization software to generate a 3D model of the specimen at a near-histological resolution (~2 µm 3 per voxel). The 3D model can then be virtually sectioned in any plane and enables metric analysis of the visualized structures (Geyer et al., 2017). Since its first introduction (Weninger & Mohun, 2007), the HREM method was used in various medical applications, such as morphological analysis of model organisms and human tissue visualization (Gershon et al., 2018;Geyer et al., 2014Geyer et al., , 2015Henkelman, Friedel, Lerch, Wilson, & Mohun, 2016;Mohun & Weninger, 2011;Pokhrel, Ben-Tal Cohen, Genin, Sela-Donenfeld, & Cinnamon, 2017;Weninger et al., 2014). The HREM method was shown to yield 3D images with higher resolution compared with high-resolution X-ray computed tomography (HRXCT), magnetic resonance imaging (MRI), and optical projection tomography (OPT) (Geyer et al., 2017;Geyer, Mohun, & Weninger, 2009). In computerized tomography-based imaging systems, such as OPT and HRXCT, the raw data are a set of projections which are then processed by an algorithm to create virtual sections that are later used for 3D reconstruction, leading to potential artifacts given this indirect multistep process. In contrast, the data set generated by HREM for 3D modeling are the actual images of thousands of perfectly aligned sections, thus obviating the need for virtualization and overriding artifacts generated by computerized tomography algorithms. Though HREM fails to reach confocal, light sheet, or electron microscopy resolutions, it can provide 3D imaging of much larger specimens (effectively up to 10 × 10 × 15 mm) (Geyer et al., 2017(Geyer et al., , 2009. Despite the wide and growing use of HREM in 3D imaging of animal model organisms and human tissues, this technique has not yet been applied in plants. Here, we describe the application of the HREM technique in studying two plant developmental processes: oil cavity development in citrus fruit peel and adventitious root formation in persimmon cuttings. Citrus essential oils, which consist mainly of volatile monoterpenoid and sesquiterpenoid compounds, accumulate in specialized secretory cavities occurring in the stem, leaves, flower organs (except the stamens), and fruit (Knight, Klieber, & Sedgley, 2001).
Biosynthesis of the essential oil compounds was suggested to take place in the epithelial cells lining the cavity with subsequent secretion into the cavity lumen (Voo, Grimes, & Lange, 2012). Anatomical studies of citrus secretory cavities using 2D sections were carried out to characterize cavity formation and development. Cell divisions in the epidermal and sub-epidermal layers initiate the formation of the secretory cavity (Liang, Wu, Lun, & Lu, 2006). At the next stage, a lumen is formed at the center of the secretory cavity, which gradually expands and accumulates essential oils. Studies of the mechanism leading to the formation of the central cavity began more than a hundred years ago, but remain controversial to this day. Several studies and most text books have described the formation of the secretory cavity as a lysigenous process, which involves cell wall degradation to initiate lumen formation (Esau, 1977;Fahn, 1988;Turner, 1999). Other studies have shown that the cavities are initiated schizogenously by cell wall separation (knight et al., 2001;Liang et al., 2006;Thomson, Platt-Aloia, & Endress, 1976). Cavity formation by overlapping schizogenous and lysigenous events was also described in Citrus sinesis and Citrus limon secretory cavities (Bennici & Tani, 2004). However, it was suggested that cell lysigeny observed during cavity formation was an artifact of the tissue fixation which resulted in poor tissue preservation (Turner, 1999). Citrus secretory cavities are initiated in the flavedo, the pigmented region of the pericarp, at the early stages of fruitlet development and reach their final structure in the immature green fruit (Bennici & Tani, 2004;Knight et al., 2001;Voo et al., 2012). During fruit maturation, secretory cavities continue to expand and gain a spherical or a pyriform shape (Bennucu & Tani, 2004;Knight et al., 2001;Liang et al., 2006;Voo et al., 2012). To estimate the essential oil amount in citrus fruit glands, secretory cavity volumes and density were calculated using 2D sections (Voo et al., 2012). Significant variability in the cavity volumes was found during all stages of fruit development. It was suggested that the differences in the secretory cavity volumes reflect continuous development and expansion of the cavities (Voo et al., 2012).
However, the variability in the cavity volumes may stem from the limited spatial information gained from the 2D sections, which may lead to inaccurate calculations. We utilized the HREM 3D imaging to study Citrus limon secretory cavity anatomy during different stages of fruit development. Citrus oil cavity formation, shape, volume, density, and distribution were analyzed, and new insights regarding their development are presented.
Development of adventitious roots from stem cutting is an important trait in herbaceous and woody plants, which provides a powerful tool for clonal propagation (Haissig & Riemenschneider, 1988;Hartman, Kester, Davies, & Geneve, 2014

Significance Statement
We describe the application of a technique for wholemount three-dimensional imaging termed high-resolution episcopic microscopy (HREM) in plants. Analysis of secretory cavity development and adventitious root formation demonstrated the potential use of the HREM application for studying developmental processes in plants.
1995). The ability to form adventitious roots is affected by various factors, namely the ontogenic stage of the mother plant and the clone genotype (Haissig & Riemenschneider, 1988;Hartmannet al., 2014). Histological studies of adventitious root formation in cuttings demonstrated that root primordia are originated in many woody plants from the secondary phloem adjacent to the cambium layer (Bellini, Pacurar, & Perrone, 2014;Hartmann et al., 2014;Izhaki et al., 2018;Naija, Elloumi, Jbir, Ammar, & Kevers, 2008). Comparative anatomical studies of rooting and non-rooting cuttings were carried out to decipher the basis for poor rooting ability. Some studies attributed the inhibition of adventitious root development in difficult-to-root cuttings to the presence of an anatomical barrier manifested by a lignified sclerenchyma layer, which was suggested to act as a physical barrier preventing adventitious root emergence (Beakbane, 1961;Edwards & Thomas, 1980;Goodin, 1965), while other studies showed no correlation between lignified sclerenchyma and low rooting ability (Davies et al., 1982;Sachs, Loreti, & Bie, 1964). Several studies suggested that the inhibition of adventitious root formation was related to the capacity of the tissue to initiate root primordia rather than to the presence of an anatomical barrier (Amissah, Paolillo, & Bassuk, 2008;Davies & Hartmann, 1988;White & Lovell, 1984).  (Izhaki et al., 2018) were analyzed by HREM and detailed 3D histology images of the cuttings were generated. Comparison of the 3D models revealed that difficultto-root clones failed to develop adventitious roots due to their inability to initiate root primordia.

| Citrus growth conditions
Citrus limon fruits at different developmental stages were collected from twenty-year-old trees grown in an orchard at the Volcani Center, ARO, Rishon LeZion, Israel.

| Persimmon rootstock growth
Dyospyros virginiana genotypes were grown in the field under intensive irrigation and fertilization and used as mother plants for cuttings collection. Difficult-to-root genotypes were grafted on seedling scions. The mother plants were pruned in February at a height of about 40 cm above the ground level. Shoots of the current year growth were used as the source of cuttings for the rooting experiments.

| Rooting of cuttings
Rooting of cuttings was performed as described by Izhaki et al., (2018). Briefly, Shoots were collected from the mother plants early in the morning, wrapped in a moist filter paper, and placed in plastic bags in a cooler until cutting preparation. Sub-apical semi-hardwood cuttings having 3-4 nodes and 2-3 mm in diameter were used. The leaves close to the base of the cuttings were removed, and the remaining upper leaves were cut in half. Before rooting, the cuttings were treated by dipping for 2 min in a solution containing 0.3% merpan (Adama), 0.25% sportak (Gadot-Agro), and 0.05% Triton X-100 (Adama) to prevent inoculation by pathogenic fungi. The base of the cuttings was dipped in a solution containing 6,000 mg/L indole-3-

| Sample preparation and embedding
Sample preparation and embedding were performed as described by Geyer et al. (2017). Briefly, tissues were fixed in 4% paraformaldehyde (w/v) under vacuum for 1-2 hr at room temperature and then incubated overnight at 4°C. Fixed tissues were rinsed twice with phosphate buffered saline and dehydrated through a graded series of ethanol (30%, 50%, 70%, 80%, and 95%). The dehydrated tissues were placed in an infiltration solution containing catalyzed monomer A of the JB-4 (2-hydroxyethyl methacrylate) embedding kit (Polysciences), 0.275% w/v Eosin B (Sigma-Aldrich), and 0.056% w/v Acridine orange (Sigma-Aldrich). Samples were vacuumed for 2-3 hr, infiltration solution was replaced with a fresh solution, and the samples were incubated at 4°C on rotating shaker for 10 days (full solution infiltration is achieved within 5-10 days, according to the sample size, which ranges from 3-10 mm in diameter). Tissues were embedded in embedding solution containing the JB-4 infiltration solution and JB4-solution B, according to manufacturer guidelines. Samples were placed in a molding tray filled with embedding solution, aligned using forceps, and covered with a plastic stub block holder (Indigo Scientific). Following the setting of the JB-4, the samples were removed from the embedding mold and baked for 48 hr at 70°C to harden the blocks prior to sectioning. Blocks were stored in tightly closed and dry containers in a dry and cool place, to avoid humidity and warm temperatures which soften the blocks and prolonged exposure to ambient light which causes block surface bleaching.

| HREM sectioning and data generation
Data generation was performed as described by Geyer et al. (2017).
Briefly, blocks were mounted on the robotic microtome HREM unit (Indigo Scientific) equipped with MVX10 Olympus Microscope (Olympus) and ProgRes MFscan digital camera (Jenoptic). 2.5-µmthick sections were cut, and images of the block cut surface were captured after each section with a GFP fluorescent filter (excitation filter 470/40; emission filter 525/50) at a resolution of 2720 × 2048 pixels at 16 bit. For accurate magnification calculations, an image of 1000µm graticule bar (EMS) was taken following each sample.

| Data processing and 3D modeling
Data processing and 3D modeling were performed as described by Weninger et al., 2018. Briefly, individual images were automatically processed for leveling the pixels histogram, black and white inverted and converted to 8-bit grayscale mode using ImageJ (Rueden et al., 2017) or Fiji (Schindelin et al., 2012) software. If section thickness did not correspond with the pixel size to form a cubic voxel, scaling of the single images to match the section thickness was done using Photoshop (Adobe Systems). Alternatively, the parameters for cubic voxel size were directly loaded to the 3D visualization software.
Image stacks were loaded for 3D visualization and segmentation analysis using ImageJ Volume viewer or Amira (ThermoFisher Scientific).

| HREM 3D analysis of citrus secretory cavity development
To characterize Citrus limon secretory cavity initiation and development, we investigated cavity anatomy during different stages of fruit development using HREM 3D imaging. Peels from 5-, 8-, and 15-mmdiameter fruitlets as well as peels from mature fruits were fixed and sectioned, and 3D histology models were generated. In 5-and 8-mm fruitlets, initiation of secretory cavities was observed, reflected by sub-epidermal cell divisions, which gave rise to elliptical cell clusters ( Figure 1b,c,e,f, Movies S1-S3). At the next stage of cavity development, a lumen began to form schizogenously at the center of the cell cluster. Parallel initiation of lumens was observed in each cell cluster, which joined to generate a single lumen (Figure 1c,f, Movies S1-S3).
The lumen continued to expand as the cavity developed (Figure 1c,f, F I G U R E 1 Secretory cavities in young Citrus limon fruitlets. 3D histology model of a 5-mm Citrus limon fruitlet (a). 5-mm fruitlet showing secretory cavities at the cell cluster stage and cavities with lumens (b). Early stages of secretory cavity initiation in 5-mm fruitlet peel: cell clusters, cell clusters with lumen initiation, and secretory cavities with lumens (c). 3D histology model of an 8-mm fruitlet (d). 8-mm fruitlet showing secretory cavities at different developmental stages: cell clusters, and cavities with lumens (e). Different stages of secretory cavity initiation in 8-mm fruitlet peel: cell clusters, cell clusters with lumen initiation, and secretory cavities with lumens containing oil droplets (f). Black arrowhead-cell cluster; black asterisk-cavity with lumen; black arrow-lumen initiation; white arrowhead-oil droplet. Bars (a-b, d-e) = 500 µm, (c, f) = 100 µm. For 3D models, see Movies S1-S3 Movie S3). No evidence of lysigenous events during lumen formation was observed (Figure 1c,f, Movie S3). Similarly, previous studies in citrus showed that secretory cavity initiation is restricted to the early stages of fruitlet development and is formed schizogenously (Knight et al., 2001;Liang et al., 2006). However, parallel events of cell separation leading to lumen initiation have not been described before, demonstrating the unique ability of tissue 3D visualization to capture subtle events occurring repeatedly during development.
Secretory cavities at three different developmental stages were observed in 5-and 8-mm fruitlets: cavities at the cell cluster stage, cell clusters with initial cell separation, and cavities with defined lumen (Figure 1b,c,e,f, Movies S1-S3). In 15-mm fruitlets, all cavities contained a defined lumen and no cavity initiation was observed (Movie S4). Using the HREM models, essential oil droplets could be observed in the cavity lumens. Oil droplets were already detected in immature cavities in 5-mm fruitlets and continued to accumulate as the lumen expanded (Figure 1b,c,e,f, Movies S1-S3). The different 3D HREM models demonstrate the spatial distribution of the secretory cavities in the fruit peel, featuring a single sub-epidermal layer of secretory cavities which encompass the fruitlet circumferences ( Figure 1, Movie S1, S2, S4). The 3D models generated from the different stages of fruit development enabled accurate calculation of spatial oil cavity density. Fruit development was character- F I G U R E 4 Adventitious root development in easy-and difficult-to-root persimmon (Dyospyros virginiana) cuttings. 3D histology model of a 10-day-old easy-to-root persimmon cutting (a). Young root primordium in a 10-day-old easyto-root cutting (b). 3D histology model of a 20-day-old easy-to-root cutting with an adventitious root emerging from the stem cortex (c). 3D histology model close up of a root primordium emerging from the stem cortex in a 20-day-old easy-to-root cutting (d). Section through a 20-day-old difficult-to-root stem cutting (e). Close up of a 20-day-old difficultto-root stem cutting (f). E, epidermis; C, cortex, P, phloem; VC, vascular cambium; X, xylem, arrowhead-root primordium.
Bars (a-f) = 500 µm. For 3D models, see Movies S7-S10 The data presented show that the 3D models generated by the HREM technique enabled a unique spatial analysis of secretory cavity development at the cellular level, as well as accurate evaluation of oil cavity distribution, density, shape, and metric analyses.

| Histological characterization of rooting and non-rooting persimmon rootstock cuttings
To analyze the developmental stage at which adventitious root formation is inhibited in non-rooting persimmon rootstocks and to explore the relationship between stem anatomy and root de- In this study, the HREM method contributed an innovative spatial characterization of the entire cutting base. The 3D models generated enabled acquisition of homologous microscopic sections of rooting and non-rooting cuttings to allow accurate comparative anatomical analyses.

| Prospective HREM uses
To our knowledge, the application of the HREM technique in plants is novel, providing precise spatial visualization of plant tissue anatomy and architecture. HREM systems can also be equipped with automatically revolving fluorescent filters, thus allowing for multichannel imaging. While excitation at the green channel is normally used for gross morphology, the red channel can be utilized for gene expression pattern analysis using opaque dyes such as NBT/BCIP which is commonly used for RNA in situ hybridization or X-gal staining to detect β-galactosidase activity. The UV channel can be applied for nuclear staining using DAPI and the far-red wavelength for whole-mount antibody staining. Thus, the application of HREM demonstrated here for plant research has good potential to be further developed and to provide a strong set of tools to support studies in plant development.

ACK N OWLED G M ENTS
We thank Yoram Eyal (Volcani Center, ARO) for insightful discussions and valuable comments on this manuscript and Chris Hunter (Indigo Scientific, UK) for introducing the commercialized High-Resolution Episcopic Microscopy system.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest associated with the work described in this manuscript.