Correlative microscopy – illuminating the endomembrane system of plant seeds Free

Plant cell biology research has taken great advantage of complementary visual techniques that employ a diverse array of cellular exploration methods at the single-cell level and at the highest resolution. Although fluorescence microscopy can be combined with a large variety of fluorescent dyes and markers to observe cellular structures over a large sample area and to obtain insights into their distribution and dynamics, the resolution of conventional live-cell imaging is constrained by diffraction (Huang et al., 2010; Schubert, 2017; Shaw et al., 2019). The development of super-resolution fluorescence imaging techniques has significantly overcome the limitations of light microscopy to reach nanoscale resolution ranges. These methods can reach a lateral resolution of ∼50–70 nm [in the case of stimulated emission depletion (STED) microscopy] or even ∼20–30 nm [in the case of photoactivated localization (PALM) microscopy] and have successfully been used in plants (Colin et al., 2022). Other methods like super-resolution confocal live-imaging microscopy (SCLIM) provide ultrafast high-resolution imaging and have been used, for example, to identify highly specialized subdomains within the trans-Golgi network (TGN) in plants (Shimizu et al., 2021). In contrast, electron microscopy provides access to ultrastructural cellular organization at nanometre-scale resolution and captures the entire structural information, not only fluorescent structures (either intrinsically fluorescent or labelled by genetically encoded fluorescent tags), in static images through contrast added during sample preparation (Kuchenbrod et al., 2021; Martell et al., 2017; Hell, 2009; Weiner et al., 2022). The superior resolving power at high magnification [using transmission electron microscopy (TEM)] and the possibility to readily image large sample sections with high resolution using backscattered electrons [using scanning electron microscopy (SEM)] can be complemented by different volume electron microscopy approaches including electron tomography (ET) and serial block-face SEM (SBF-SEM) (Peddie et al., 2022).

Combining the abovementioned techniques allows the acquisition of complementary information by maximizing their advantages and mitigating their limitations. However, using both fluorescence microscopy and electron microscopy on the very same sample or tissue section and region of interest has long posed a methodological obstacle in the field of cell biology. Correlative light and electron microscopy (CLEM) methods provide the ability to observe the same sample section at both levels (Marion et al., 2017). Moreover, CLEM integrates the information that can be obtained from fluorescence microscopy with the high spatial resolution of TEM or SEM, enabling a comprehensive visualization of cellular structures at different scales. In the past few years, CLEM has seen a huge development in various fields of cell biology, including plant cell research (Bell et al., 2013; Chambaud et al., 2023; Wang and Kang, 2020; Lübben et al., 2024; Chung et al., 2024).

Within plant cells, cereal endosperm cells present several unique and intriguing features that distinguish them from other cell types. To begin with, cereal endosperm is a short-lived triploid tissue that does not survive germination, as only the aleurone cells in its periphery are viable in the mature seed. The endosperm undergoes programmed cell death (PCD) from early developmental stages (Sabelli et al., 2013; Young and Gallie, 2000) and, most puzzlingly, the progression of PCD in endosperm tissue runs parallel with the deposition of the storage reserves (Young et al., 1997). Furthermore, 50% of the total protein content in mature barley seeds consists of seed storage proteins (SSPs); these SSPs are mainly proteins of the prolamin group, which in barley is represented by the hordeins (Gubatz and Shewry, 2010). Hordeins comprise four closely related protein families: the sulphur-rich B and γ hordeins, the sulphur-poor C hordein and the high molecular weight (HMW) D hordein (which is homologous to the HMW glutenins of wheat; Kreis and Shewry, 1989). Hordein synthesis in the barley endosperm starts as soon as 10 days post anthesis and continues until around 30 days post anthesis (Brandt, 1976; Roustan et al., 2018; Sørensen et al., 1989). The dynamics of the endomembrane system in endosperm cells respond to these elevated rates of protein synthesis, which are sustained throughout development (Arcalis et al., 2014; Roustan et al., 2018). The endoplasmic reticulum (ER) in younger endosperm cells expands to support protein synthesis and folding, favouring cisternae over tubules, as has been observed for maize (Arcalis et al., 2020). At the same time, the first organelles for the deposition of the SSPs appear. Indeed, hordeins (monomeric and polymeric) form insoluble aggregates within the ER that are transport incompatible. These aggregates bud off the ER forming the so-called protein bodies (PBs) that will eventually be deposited into large central vacuoles within endosperm cells (Ibl et al., 2014; Rechinger et al., 1993). Although the exact pathways through which barley hordeins are trafficked are poorly described, Levanony et al. (1992) have proposed a transport route for wheat prolamins that does not involve the Golgi complex. According to their model, a significant portion of the prolamins exits directly from the ER and is transported to the vacuole, bypassing the Golgi by following an autophagy-like pathway. Different to conventional autophagy, the delivery of prolamins to storage vacuoles in cereal endosperm seems to be selective for SSPs (Michaeli et al., 2014) and does not involve the formation of phagophores. Additionally, the participation of the core ATG8 protein machinery (which is involved in conventional autophagic processes) in this ER-to-vacuole pathway is still unclear (Reyes et al., 2011). Such a transport route has also been proposed for barley, where the presence of internal membranes in the protein storage vacuoles (PSVs) suggests that the transport of hordeins to the PSV at least partially involves a Golgi-independent autophagy-like pathway (Ibl et al., 2014). The plasticity and rapid structural changes of the endomembrane system in cereal endosperm cells concurrent with the synthesis of SSPs make it a most suitable subject for exploring the limits and capacities of endomembrane systems in general, and in plant cells in particular. In this study, we build upon previous studies mostly based on in vivo imaging (Ibl et al., 2014) by utilizing an electron microscopy approach to obtain structural information at high resolution and investigate the development of the endomembrane compartments over time. Furthermore, we introduce a CLEM protocol that combines the use of fluorescence-tagged marker proteins with ultrastructural data to deepen our understanding of the progressive changes in the endomembrane system within endosperm tissue. Our protocol is based on the in-resin fluorescence CLEM method developed by Kukulski et al. (2012) and later adapted for plant samples by Chambaud et al. (2023). This protocol preserves fluorescence and ultrastructure and is suitable for immunostaining, and we have adapted it for use in barley (Hordeum vulgare). Here, transgenic barley lines expressing recombinant marker proteins to label endomembrane compartments were used. By optimizing the sample preparation workflow by adding a high-pressure freezing step and adapting other parameters like section thickness, sample mounting for light microscopy and light microscopy image acquisition (Chambaud et al., 2023), we were able to observe fluorescence signals and precisely select the endomembrane compartments under the light microscope before revealing their ultrastructure using TEM or SEM.

Barley seed endosperm arises through a double fertilization event, begins development as a multinucleate syncytium and completes cellularization at around 6 days after pollination (dap) (Wilson et al., 2006). We generated a marker line constitutively expressing GFP attached to a signal peptide for entry into the ER (SP-GFP) as a luminal marker to identify secretory compartments in vivo during endosperm development. We followed the first stages of seed development by combining live-cell imaging data on the one hand and ultrastructural data obtained from electron microscopy observations on the other. The combination of these two datasets allowed us to define three stages in early seed development, based on the morphology of the endomembrane system and storage organelles rather than the days after pollination: stage 1 (6–8 dap), stage 2 (∼8–12 dap) and stage 3 (∼12–14 dap) (Fig. 1).

At 6–8 dap (stage 1), the differentiation of the aleurone layer has already occurred (Fig. S2A). Sub-aleurone cells harbour abundant electron-transparent compartments that show different sizes and shapes (Fig. 1A). The first PBs (up to 500 nm in diameter) appear in the lumen of such electron-transparent compartments (Fig. 1A,B) but can also be found within ER cisternae, where they can accumulate, dilating the ER lumen. It is interesting to remark that contrary to previous observations of maize zein bodies (Arcalis et al., 2020), barley hordeins that are found within the ER are not in contact with the ER membrane (Fig. 1C). The ER in young sub-aleurone cells consists of few cisternae (Fig. 1A–D), which sometimes can be found forming large dilations of up to 4 µm (Fig. 1D).

A representative cross section of a barley seed in developmental stage 2, comprising the highly electron-dense aleurone cells, sub-aleurone cells and first starchy endosperm cells, is shown in Fig. S2B. At stage 2, the sub-aleurone cells have undergone substantial changes and are now filled with starch grains and vacuole-like compartments containing larger hordein bodies (Fig. 1E,F). In vivo imaging revealed the presence of SP-GFP within such compartments (Fig. 1G), confirming their vacuolar nature, since the default storage site for SP-GFP in cereal endosperm are the PSVs, as expected for secretory proteins in cereal endosperm (Arcalis et al., 2004). Interestingly, hordeins exclude SP-GFP and appear as non-labelled inclusions within the PSVs (Fig. 1G). In the classical literature (Cameron-Mills and von Wettstein, 1980), hordein fractions, which form multiphasic PBs, have been named and characterized based on their electron density as seen under the electron microscope. The spherical, mostly homogenous aggregates of different sizes and medium electron density are denoted as component A, component B corresponds to a reticulate matrix that surrounds component A and is relatively scarce, and the smaller, electron-dense globules correspond to component C. At developmental stage 2, multiphasic PBs can be found in the sub-aleurone cells. PBs with medium electron density (component A) and a diameter of less than 1.5 µm form clusters within the PSVs together with the first deposits of the highly electron-dense component C (Fig. 1E,F). SP-GFP can also be seen within the ER, whose presence, in contrast to earlier stages, is remarkable. Indeed, large stacks formed by several tightly packed parallel cisternae separated by 50 nm extend through the cytoplasm of the sub-aleurone cells (Fig. 1E,F).

As development continues (stage 3), cells in the sub-aleurone layer continue to accumulate starch and SSPs (Fig. S2C). Within the PBs, component A is now more abundant, although the single inclusions are now smaller in size (less than 1 µm diameter). Component C is now more apparent and can be observed together with the component A inclusions squashed within the PSVs (Fig. 1H–J). As a consequence of the accumulation of PBs, which fill the lumen of the PSVs (Fig. 1H–J), SP-GFP can only be observed trapped between the hordein inclusions that exclude SP-GFP, with few exceptions (Fig. 1K). It is noteworthy that although the stacks of ER cisternae are still present, they are now formed by shorter cisternae (Fig. 1H,I).

Despite the high degree of detail and clarity achieved under the electron microscope, unequivocally identifying the distinct compartments of the endomembrane system in developing endosperm is challenging. The endomembranes within the sub-aleurone cells are notably dense and intricate, comprising numerous compartments with various degrees of electron transparency (Figs 1E and 2A). Fig. 2A illustrates the difficulty of the task, as it shows multiple vacuole-like structures of diverse shapes and turgidity, each exhibiting distinct contents, intricately intermingled within the cytoplasm. The use of GFP-based marker lines would help to identify such endomembrane compartments, but imaging at a high-resolution level is hampered by the loss of fluorescence due to chemical fixation. To address this issue, we established a CLEM protocol for imaging the developing cereal endosperm. To explore the potential of our CLEM method, we chose a transgenic barley marker line that expresses high levels of SP-GFP and provides strong fluorescence signals. Following high-pressure freezing and freeze substitution (HPF-FS), a representative overview of the processed sample (stage 2) demonstrates well-preserved tissue structure with no significant osmotic artifacts and excellent fluorescence preservation (Fig. 2B; Figs S3 and S4). Consequently, numerous PSV-like structures containing both SP-GFP and non-fluorescent hordein bodies can be readily identified and correlated with electron microscopy images like those shown in Figs 1 and 2A. A higher magnification allowed us to fully assess the degree of ultrastructural preservation. The cell compartmentalization is preserved, particularly the already mentioned PSV-like compartments (Fig. 2C). The trafficking of SP-GFP through the secretory pathway is readily understood, as strong SP-GFP fluorescence is evident within the ER and Golgi. These organelles can be identified in the electron micrographs owing to the good membrane preservation (Fig. 2C,D). As observed in live-cell imaging, hordeins within the PSVs exclude SP-GFP (Fig. 1G,K), with the exception of few very bright puncta (Fig. 1K). However, the use of CLEM gives more detailed information, as it becomes apparent that whereas component A does exclude SP-GFP, green fluorescence is evident within the highly electron-dense component C, perhaps indicating different trafficking pathways for the two hordein components stored in the PSVs (Fig. 2E).

We have used genetically encoded fluorescent tags, as well as different fluorescent stains, in combination with confocal laser scanning microscopy (CLSM) in previous studies to investigate the endomembrane system of maize and barley endosperm cells (Ibl et al., 2014; Arcalis et al., 2022). However, conventional CLSM can only detect fluorescence signals, rendering unlabelled molecules or organelles undetectable (Weiner et al., 2022). This situation is illustrated in Fig. 3A, which shows several PSVs stained with Neutral Red. Neutral Red can be used to stain hordeins within storage compartments (Ibl et al., 2024), and indeed it clearly highlights the hordein bodies, which exclude SP-GFP and are unlabelled in the green channel. However, by staining the hordeins, it becomes clear that some of the unstained areas of the PSVS in the green channel are also not stained by Neutral Red and are therefore not easily identifiable. A similar structure was found in HPF-FS tissue (Fig. 3B). By using immunolabelling, we could identify several hordein bodies in the HPF-FS tissue section, but again, a structure similar in size and shape to the hordein bodies remained unlabelled by SP-GFP and hordein immunofluorescence. The electron microscopy image, however, offers additional information by revealing that the non-labelled feature corresponds to a membrane-bound structure within the vacuole (Fig. 3B,C). In a recent study, we documented the presence of similar membrane structures within the PSVs in developing maize endosperm. These structures were identified as autophagic bodies, indicating that starvation-induced autophagic bodies contribute to the content of PSVs in maize endosperm (Arcalis et al., 2022). Similarly, here we observed autophagic body-like structures containing mitochondria within a vacuole in developing barley endosperm (Fig. 3D,E). The use of monodansylcadaverine, (MDC), which is a selective stain for autophagosomes and autophagic bodies (Contento et al., 2005), confirmed the presence of autophagic bodies also within the PSVs of barley endosperm (Fig. 3F,G).

In addition to the SP-GFP marker line, we used a TIP3–GFP marker line to specifically label PSVs by visualizing their tonoplast. Thus, abundant PSVs showing homogenous size and shape could be identified within the aleurone layer, as expected (Ibl et al., 2014). In contrast, the numerous PSVs observed in the sub-aleurone cells present different shapes within the cells in the sub-aleurone layer. Consistent with previous findings (Ibl et al., 2014), these PSVs contained abundant membrane TIP3–GFP fragments and internal vesicles that could be a consequence of microautophagy events in which the tonoplast is involved in the sequestration of cytoplasmic regions, giving rise to an autophagic body within the PSV surrounded by a tonoplast membrane (Fig. 4A,B). This observation was further supported by a marker line co-expressing SP-GFP and the ER membrane marker SEC61–mCherry, which revealed the rare event of an ER-bound compartment engulfed by an SP-GFP-filled PSV, also suggesting the occurrence of microautophagic events (Fig. 4C). Although fluorescence microscopy allowed the identification of the different membranes involved in the process, possible cargo entrapped in these membranous compartments remained undetected. CLEM allowed us to observe tonoplast invaginations engulfing hordein bodies, thus confirming that at least some hordeins are incorporated into the PSV by microautophagy (Fig. 4D). After fusion of the single hordein bodies, tonoplast fragments could remain trapped within the hordein mass, resulting in the diffuse signal observed (Fig. 4D).

Our CLEM approach further confirmed the presence of autophagic bodies within the PSVs. Non-fluorescent areas within the PSVs, which are easily observed in live-cell imaging, are also observed in HPF-FS samples. The electron microscopy imaging revealed that at least some of these areas correlate with membrane-bound structures. Moreover, the possibility to observe the whole tissue, and not only the fluorescence-labelled structures, revealed the engulfment of hordeins by the tonoplast in our TIP3–GFP marker line, which is easily identified in live-cell imaging, in a clear example of microautophagy. Exploiting the advantages of the different imaging techniques available provides a new perspective on PSVs: rather than a static protein storage site, they are versatile organelles with multiple functions.

We present here high-quality electron microscopy images to track the development of the endomembrane system in barley endosperm cells. These images allow for the visualization of large tissue sections with exceptional detail, including close-up views at the cellular and subcellular levels. Our observations reveal that in the early stages of development, abundant electron-translucent structures are present within the sub-aleurone cells, occupying most of the cell volume. At least some of them correspond to ER dilations, similar to those observed in other cereal endosperm cells in early developmental stages. In rice, ER dilations have been linked to the formation of disulphide bonds during synthesis of SSPs (Onda et al., 2009). Given that cereal prolamins are rich in Cys residues, and since barley prolamins account for 50% of the seed proteome (Gubatz and Shewry, 2010), it is most likely that the ER dilations in young barley endosperm cells are triggered by hordein synthesis.

Later in development, and only within 1–2 days, the ER develops to form noticeable stacks of densely packed cisternae, while PSVs emerge as the primary storage site of the PBs. Although not as prominent as in the initial stages of endosperm development, smaller ER dilations are still observed in later stages. A wide variety of membrane compartments of different electron density coexist within the cytoplasm, and they are often arranged so closely together that they resemble interlocking puzzle pieces, making their identification very difficult. As development progresses, the changes in the endomembrane system are restricted to the progressive accumulation of PBs within the PSVs, and the fragmentation of the ER stacks, resulting in short cisternae stacks similar to those previously reported in developing maize endosperm cells (Arcalis et al., 2020).

In addition to the electron microscopy studies, we used different fluorescently tagged recombinant proteins (including a previously undescribed transgenic barley line expressing GFP with a signal peptide) and fluorescent dyes to identify compartments and follow the endomembrane system in vivo over the course of endosperm development. Barley sub-aleurone cells expressing SP-GFP showed vacuolar localization of the GFP signal, which could also be found within the ER lumen, as expected (Arcalis et al., 2004). In addition, our TIP3–GFP marker line identified not only the abundant PSVs in the aleurone and sub-aleurone layers, but also the numerous vesicles and membrane fragments in the PSV lumen, consistent with previous reports (Ibl et al., 2014).

Although fascinating, monitoring the numerous changes occurring during development at a subcellular level poses significant challenges, emphasizing the need for a multiscale approach. The images and techniques discussed so far do not allow for direct correlation, which is essential to obtain maximum information. To address this issue, we aimed to establish a CLEM protocol for cereal seeds. Given the high complexity of the endomembrane system and the dense arrangement of subcellular structures in barley endosperm, as observed using TEM, a tool like CLEM could provide valuable insights into the endomembrane system in cereal seeds.

We based our protocol on previous studies by Otegui (2021) and Chambaud et al. (2023), both of which focused on Arabidopsis thaliana. Cereal seeds are large organs that require sectioning prior to high-pressure freezing. Consequently, these seed sections have a high ratio of injured to intact tissue. Combined with the challenges posed, among others, by cell walls and the presence of numerous starch grains, achieving optimal ultrastructural preservation of cereal endosperm – whether through chemical fixatives or high-pressure freezing – is not straightforward. Despite these challenges, we successfully obtained high-quality results by preserving both the fluorescence and the ultrastructure of the tissue after HPF-FS.

CLEM enabled us to precisely detect endomembrane compartments and facilitate the acquisition of electron microscopy data guided by fluorescence information. Specifically, we could gain some information about the PSVs and the hordein bodies contained within. Firstly, we observed that the electron-dense hordein fraction in component C colocalized with soluble SP-GFP. It is well accepted that the bulk of the prolamins form PBs within the ER and are subsequently taken up into PSVs in an autophagy-like manner bypassing the Golgi (Levanony et al., 1992). Notably, we observed that most hordeins, packed in the medium-electron-dense component A within the PSVs excluded SP-GFP, which could only be found within the vacuolar lumen or together with the high electron-dense component C in PSVs. The presence of SP-GFP within component C suggests a common trafficking pathway at least for a part of the hordein fraction. Our SP-GFP was detected in the Golgi, as expected. Trafficking of prolamins through the Golgi has been shown in wheat and barley (Parker and Hawes, 1982; Bechtel et al., 1982; Møgelsvang and Simpson, 1998), whereby this pathway seems to be restricted to the early developmental stages, as Golgi organelles and transcripts encoding Golgi-associated proteins decrease with seed maturation (Shy et al., 2001). Although it might be tempting to assume Golgi trafficking for the hordeins forming component C, based on their colocalization with SP-GFP within the PSVs, the fact that this component becomes more abundant in later stages of sub-aleurone development suggests that a putative common trafficking pathway is primarily active during the earlier stages covered in our study.

Secondly, the use of CLEM allowed us to acquire preliminary data of autophagy in barley PSVs. We were able to ascribe lytic vacuole properties to the PSV compartments in barley endosperm, as autophagic bodies could be identified, in some cases associated with the uptake and degradation of cytosolic material such as mitochondria at an intermediate developmental stage. The use of MDC revealed bright punctate signals within barley endosperm PSVs, thus reinforcing the involvement of PSVs in autophagic processes by identifying typical starvation-induced autophagic bodies in developing barley endosperm. The onset of SSP synthesis in cereal endosperm is concomitant with a rise in levels of ER-resident BiP chaperones, indicating that the protein folding capacity of the ER is also affected (Muench et al., 1997; Wakasa et al., 2011). One way to alleviate ER stress and return to a balanced state (Kim et al., 2017; Vitale and Pedrazzini, 2022) is the transport of cellular components to lytic compartments via autophagy and autophagy-like processes (van Anken et al., 2021). In this context, as we have proposed previously for maize (Arcalis et al., 2022), the mitophagy events observed within the PSVs could be a consequence of imbalanced cellular stress (Vitale and Pedrazzini, 2022).

In addition to starvation-like autophagy, we also observed some indications of microautophagy in barley endosperm cells. Similarly, the transport of zeins to PSVs in the maize aleurone has been characterized as microautophagy, occurring independently of the ATG8 and ATG12 conjugation machinery and autophagosome formation (Ding et al., 2022). Microautophagy events would also explain the abundant vesicles and membrane remnants that we observed in vivo, in agreement with previous observations (Ibl et al., 2014). These membranes would be part of the intravacuolar membrane system that extends throughout the entire PSV matrix observed in electron microscopy images (Ibl et al., 2014; Rechinger et al., 1993).

The results presented in this paper (summarized in Fig. 5) reflect the complexity of the endomembrane system of cereal seeds. We have shown that a multiscale approach, including fluorescence and electron microscopy, is highly suitable to study such a dense and intricate endomembrane system. Together with other methods like immunogold staining (Pan et al., 2021), super-resolution microscopy (Schubert, 2017; Shaw et al., 2019; Bayle et al., 2021), expansion microscopy (Kao and Nodine, 2021) and three-dimensional (3D) electron microscopy techniques (Feeney et al., 2018; Ding et al., 2022), which have been successfully used to study the endomembrane system in plant seeds and plant tissues, we anticipate an exciting multiscale future for plant cell imaging. Here, we opted for a CLEM approach, combining the advantage of fluorescence labelling for structure identification and electron microscopy resolution for simultaneous in-depth analysis. We have shown that luminal and membrane markers work equally well in cereal endosperm, and these promising results open up further perspectives for the study of cereal endosperm with the prospect of using and combining different marker lines. The results also pave the way for future CLEM applications in cereal seeds involving 3D structure determination (3D CLEM).

Barley (H. vulgare L.) wild-type variety Golden Promise (GP) and the transgenic lines tonoplast intrinsic protein 3 (TIP3)–GFP (Ibl et al., 2014), SEC61–mCherry (Ibl et al., 2014) and SP-GFP were cultivated as previously described (Ibl et al., 2014). In detail, the plants were grown in soil in incubation rooms with a 16 h photoperiod, 16°C and 70% relative humidity for 2 months. After tillering, cultivation was continued at 22°C. Seeds were collected at different developmental stages prior to desiccation from the middle of the ears of the plants (Ibl et al., 2014; Shabrangy et al., 2018). T3-generation seeds were screened for fluorescence, and those that were fluorescence positive were selected for imaging.

For the generation of hybrid transgenic lines carrying the genes encoding SP-GFP and SEC61–mCherry, flowers of a spike were detasselled 2–3 days before pollination. At the time of anthesis, fertilization was carried out by shedding pollen of a selected transgenic line into the emasculated flower. Progeny carrying the respective transgenes were selected by GFP and mCherry fluorescence.

The fluorescent marker construct was generated by inserting the SP-GFP coding sequence under the control of the maize ubiquitin promoter (Kapusi et al., 2013). Sequence encoding the zein signal peptide (5′-ATGAGGGTGTTGCTCGTTGCCCTCGCTCTCCTGGCTCTCGCTGCGAGCGCCACCTCC-3′) was fused to the sequence encoding ATG-less GFP (Chiu et al., 1996) via a linker sequence (5′-GCCGGCGCCGGCGCCGGC-3′) (Fig. S1). Transgenic plants were generated by particle bombardment according to a previously reported protocol (Kapusi and Stoger, 2022). Hygromycin phosphotransferase (hpt) was used as plant selectable marker.

For in vivo imaging, developing barley seeds at different developmental stages were harvested, and cross sections were obtained with a razor blade. Neutral Red (Thermo Fisher Scientific 229810250; Waltham, MA, USA) was used for the staining of PSVs at a concentration of 4 µM (Ibl et al., 2024; Stefano et al., 2018). Monodansylcadaverine (MDC; Sigma-Aldrich, 30432; Steinheim, Germany) was used to stain autophagosomes and autophagic bodies, as described previously (Contento et al., 2005). Briefly, seed sections were incubated in 50 µM MDC in phosphate-buffered saline (PBS) for 15 min and observed under the CLSM after three washes in PBS, following incubation with 1 µM concanamycin A (Sigma-Aldrich 27689; Steinheim, Germany) for 6 h (Xiong et al., 2007). Sections were mounted in tap water (for GFP and Neutral Red ) or in PBS (MDC) using petroleum jelly as a spacer between slide and coverslip (thickness 0.13–0.16 mm) and immediately analysed with a Leica SP5 CLSM (Leica Microsystems, Wetzlar, Germany) with filter settings for MDC (excitation 405 nm, emission 413–479 nm), GFP (excitation 488 nm, emission 496–536 nm) and Neutral Red (excitation 561 nm, emission 570–600 nm). Representative images from at least three biological replicates were analysed using LASX (Leica Microsystems, UK), ImageJ (National Institutes of Health, Bethesda, MD, USA), Adobe Photoshop CS6 or Affinity Photo software.

Developing wild-type seeds were fixed and processed as described previously (Arcalis et al., 2020). In brief, tissue sections were fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.15 M cacodylate buffer overnight at 4°C. Samples were then subjected to double osmium impregnation by incubation in 2% osmium tetroxide (Electron Microscopy Sciences 19150; Hatfield, PA, USA) and 0.2% Ruthenium Red (Sigma-Aldrich 00541-1G; Steinheim, Germany) in 0.15 M cacodylate buffer (pH 7.4) for 1 h, followed by 1% (w/v) thiocarbohydrazide (Sigma-Aldrich 88535-25G; Steinheim, Germany) in ultrapure water for 45 min and 2% aqueous osmium tetroxide (Electron Microscopy Sciences 19150; Hatfield, PA, USA) for an additional hour. After overnight incubation in uranyl acetate replacement stain (UAR-EMS, Science services), tissue sections were then further incubated in Walton's lead aspartate (Walton, 1979) prior to dehydration in an ethanol series. Following final dehydration in acetone, samples were gradually infiltrated with LV Resin (Agar Low Viscosity Resin Kit, Agar Scientific) and polymerized at 60°C for 48 h.

Ultrathin sections were collected on copper grids (Electron Microscopy Sciences HO75-Cu; Hatfield, PA, USA) and imaged in a FEI Tecnai G2 transmission electron microscope operating at 200 kV. For SEM, 200 nm sections were mounted on silica glass and an SEM stub and imaged under an Apreo SEM (Thermo Fisher Scientific), operating in Optiplan mode (2 kV, 0.1 nA). Representative images from three biological replicates are shown.

Transgenic barley seeds (SP-GFP or TIP3–GFP) at different developmental stages were harvested and fixed and processed as described previously (Otegui, 2021), with few adaptations for our samples. 200 µm sections were made with a VT1200 Vibratome (Leica Microsystems) and collected in 0.15 M cacodylate buffer. Consecutive sections were alternatively chemically fixed as described above or high-pressure frozen. For the latter, seed sections were rapidly placed in the well of aluminium Type B carriers (3 mm×0.52 mm×0.3 mm; Science Services), using hexadecane (Roth) as filler and cryoprotectant. Rapidly, seed sections were frozen using an EM HPM100 high-pressure freezer (Leica). Samples were then cryosubstituted as described previously (Otegui, 2021) in anhydrous 0.2% (w/v) glutaraldehyde and 0.2% (w/v) uranyl acetate (Merck K24460773; Darmstadt, Germany) in anhydrous acetone at −90°C for 72 h and subsequently infiltrated and embedded in HM20 resin (Polysciences 23994-225; Warrington, PA, USA) using gelatine capsules (Electron Microscopy Sciences 70102; Hatfield, PA, USA). After UV polymerization, the obtained blocks were stored protected from light at 4°C.

For the localization of hordeins in resin-embedded samples, sections were incubated with 5% (w/v) bovine serum albumin in 0.1 M phosphate buffer (pH 7.4). Then, a 1:100 dilution of polyclonal rabbit anti-wheat gliadin (Sigma-Aldrich G9144; Carlsbad, CA, USA) was added, as described previously (Hensel et al., 2015). As a secondary antibody, Alexa Fluor 546-conjugated donkey anti-rabbit IgG (Life Technologies A10040; Carlsbad, CA, USA) diluted 1:50 in 0.1 M phosphate buffer (pH 7.4) was used. Representative images from at least three technical replicates are presented.

Thin sections (90 nm for SP-GFP samples, 150 nm for TIP3–GFP) were collected onto formvar coated finder grids (Science Services G200F2-Cu; München, Germany) to ease region of interest search. Grids were mounted for confocal imaging following a previously reported method (Chambaud et al., 2023), carefully recovered after confocal imaging and allowed to air dry prior to TEM observations. Additional to the protocol from Chambaud et al. (2023), we also prepared sections for SEM imaging, which allows imaging of larger fields of view. 200–300 nm-thick sections were collected onto silicon wafers (Science Services Sc4Czp-525-10; München, Germany). These were placed on a slide, covered with a drop of distilled water and carefully covered with a coverslip. After confocal imaging (filter settings as abovementioned), silicon wafers were carefully recovered, air dried and mounted onto a stub for SEM imaging as described above. Correlation of images obtained from the CLSM and the TEM or SEM was based on natural landmarks such as cell shapes, plastids and the nucleus. Correlated images were obtained manually, with the ec-Clem plugin (Paul-Gilloteaux et al., 2017) of the Icy software (de Chaumont et al., 2012) and the landmark correlation tool of ImageJ. Representative images showing colocalization of GFP fluorescence and PSVs, ER or Golgi from at least three technical replicates (three different resin blocks, corresponding to three different seed samples) are shown.

The authors acknowledge the HRSM project NANOBILD for infrastructure support and the Core Facility Multiscale Imaging (Boku University, Vienna, Austria) for technical advice and assistance.