Dynamics of endoreduplication in developing barley seeds

Abstract

Seeds are complex biological systems comprising three genetically distinct tissues: embryo, endosperm, and maternal tissues (including seed coats and pericarp) nested inside one another. Cereal grains represent a special type of seeds, with the largest part formed by the endosperm, a specialized triploid tissue ensuring embryo protection and nourishment. We investigated dynamic changes in DNA content in three of the major seed tissues from the time of pollination up to the dry seed. We show that the cell cycle is under strict developmental control in different seed compartments. After an initial wave of active cell division, cells switch to endocycle and most endoreduplication events are observed in the endosperm and seed maternal tissues. Using different barley cultivars, we show that there is natural variation in the kinetics of this process. During the terminal stages of seed development, specific and selective loss of endoreduplicated nuclei occurs in the endosperm. This is accompanied by reduced stability of the nuclear genome, progressive loss of cell viability, and finally programmed cell death. In summary, our study shows that endopolyploidization and cell death are linked phenomena that frame barley grain development.

Cell cycle, embryo, endoreduplication, endosperm, Hordeum vulgare, cell death, seed development, super cycle value

Introduction

Seeds represent an encapsulated embryonic stage unique to Angiosperms and Gymnosperms. Seed functions include protection of the embryo, plant dispersal, survival in harsh conditions, and feeding of the embryo during germination (Bewley et al., 2006). High energy content and long storability make seeds one of the most valuable plant products. Cereal crops provide ~60% of the global food resources in the form of seeds (Food and Agriculture Organization of the United Nations, May 2020, http://www.fao.org/). Knowledge of molecular processes governing seed development and maturation is thus important to support the development of crops with higher yield and quality to secure enough food for the growing population. Here, we studied the dynamics of the cell cycle during seed formation in cultivated barley (Hordeum vulgare, 2n=2x=14, 5.1 Gbp/1C). The species ranks the fourth cereal in global production (Langridge, 2018), and its grains are used for feed (70%), production of alcoholic beverages (21%), and food (6%) (reviewed in Tricase et al., 2018). Furthermore, there is a growing interest in using barley for biofuel, cosmetics, and molecular farming (3%) (Hicks et al., 2014; Holásková et al., 2018; Langridge, 2018; Tricase et al., 2018). Finally, the reference genome, extensive genetic resources, and economic importance make barley an ideal temperate zone cereal model species (Mascher et al., 2017).

Cereal grain development includes three major stages (Sabelli and Larkins, 2009; Sreenivasulu et al., 2010; Dante et al., 2014). Stage I is initiated by double fertilization and characterized by cell proliferation and a slight weight gain; stage II involves differentiation of the main tissue types and a large weight increase accompanied by accumulation of storage compounds; stage III corresponds to seed maturation, weight reduction by desiccation, and finally physiological maturation and dormancy. The three phases partially overlap with water, milk, and dough caryopsis growth stages, respectively (Tottman et al., 1979).

The diploid embryo tissues of maternal and paternal origin proliferate and differentiate into the embryonic root, shoot apical meristem, cotyledon, and plumule (primary leaf) (Dante et al., 2014; Rodríguez et al., 2015). Triploid (3x) endosperm originates from fertilization of the diploid central cell by a haploid sperm nucleus. Initially, endosperm forms a syncytium, with nuclei being pushed to the cell periphery by a central vacuole. Later, microtubules form a radial network around nuclei and the anticlinal cell wall formation marks the onset of cellularization and the beginning of differentiation into the starchy endosperm and aleurone layer (Olsen, 2001). The mature barley endosperm comprises: the central starchy endosperm (CSE), aleurone layer (AL), the subaleuorne layer, the basal endosperm transfer layer (BETL), and the embryo-surrounding region (ESR) (Olsen, 2001; Sabelli and Larkins, 2009). The barley grain is covered by seed coats of maternal origin and pericarp [seed maternal tissues (SMTs)], which contain a high amount of starch, serve as feeding and protective structures, and participate in photosynthesis (Sreenivasulu et al., 2010; Sabelli et al., 2013; Radchuk and Borisjuk, 2014). The entire barley seed is protected by diploid hulls (the lemma and palea) of maternal origin, which remain tightly attached to the grain even after ripening (Rodríguez et al., 2015).

Rapid grain development requires strict regulation of gene expression, cell cycle dynamics, and accumulation of storage molecules (Sabelli and Larkins, 2009; Dante et al., 2014). Here, we estimated the dynamics of endoreduplication during grain development by measuring nuclear DNA content in the major seed tissues. Special attention was paid to endopolyploidization, a modified mitotic cycle during which G₂-phase nuclei undergo one or more additional rounds of DNA replication—endoreduplication—leading to chromosomes with four or more chromatids (reviewed in D’amato, 1964; Nagl, 1976). Endoreduplication is common in plants. Endopolyploid cells are larger, perform specialized functions, and/or are highly metabolically active (reviewed in Chevalier et al., 2011; Sabelli, 2012). In some species, endoreduplication is partially stress inducible, which may represent a bypass mechanism for tissue growth by cell expansion in cells with potentially unstable chromosomes (Adachi et al., 2011; Liu et al., 2015; Bhosale et al., 2018).

Endoreduplication is a developmentally controlled process. The occurrence of endoreduplication in fruits or seeds of many cultivated plants suggests that it might have been positively selected during plant domestication and breeding. However, the phenotypic and molecular consequences of endoreduplication in these tissues remain unclear and are very likely to be species dependent. In cultivated tomato, the amount of DNA in nuclei of some pericarp cells reaches up to 256C (reviewed in Chevalier et al., 2014) where 1C corresponds to the DNA content of an unreplicated haploid chromosome set. Suppression of two major positive regulators of endopolyploidization, WEE1 kinase and CELL CYCLE SWITCH PROTEIN 52 A (CCS52A), resulted in a reduction in fruit size (Gonzalez et al., 2007; Mathieu-Rivet et al., 2010), suggesting a direct role for endoreduplication in cell and organ size determination in tomato. In contrast, RNAi-induced down-regulation of S phase and the DNA replication repressor RETINOBLASTIOMA-RELATED GENE 1 (RBR1) resulted in more endoreduplicated but, surprisingly, smaller nuclei and cells (Sabelli et al., 2013). However, the total seed size was not affected, possibly due to an increased number of endosperm cells in RBR1 knockdown maize plants. This shows that in maize grains, endoreduplication can be decoupled from endosperm growth. This still leaves an open question on the functional significance of endoreduplication during seed development. However, to what extent endoreduplication is involved in barley seed development remains unknown.

An important factor playing a role during cereal grain development is programmed cell death. At stage I, several SMTs and the nucellar projection cells die (An and You, 2004; Domínguez and Cejudo, 2014; Tran et al., 2014; Radchuk et al., 2018). At later stages of seed development, ESR and CSE tissues experience cell death, but remain intact in the mature grain and their content will not be mobilized until germination. Finally, a mature grain consists mainly of dead material, where only the embryo, BETL, and AL remain alive (Young and Gallie, 2000; Sreenivasulu et al., 2010; Yifang et al., 2012; Kobayashi et al., 2013; Wu et al., 2016).

Even though barley is a cereal model species, the cell cycle and endoreduplication dynamics are poorly described during its seed development. The seed develops inside hulls of maternal origin (see Supplementary Fig. S1 at JXB online). Barley seed consists of the embryo with one maternal and one paternal genome; the endosperm with two maternal genomes and one paternal genome; and SMTs with two maternal genomes. Therefore, we investigated the dynamics of the mitotic cycle and endoreduplication in individual seed tissues during 7 weeks of barley grain development. We found a high degree of endoreduplication in endosperm and preferential elimination of endopolyploid nuclei in terminally developed tissues during the second half of the seed growth period. Collectively, this study provides comprehensive characteristics of cellular processes during the entire period of barley grain development.

Materials and methods

Plant materials and growth conditions

Six cultivars (cv.) of two-rowed spring barley (Hordeum vulgare subsp. vulgare) were used in this study: Betzes (PI 129430), Compana (PI 539111), Golden Promise (GP; PI 343079), Ingrid (PI 263574), Klages (CIho 15478), and Mars (PI 599629). The seeds were obtained from the National Small Grains Collection of the National Plant Germplasm System of the United States Department of Agriculture-Agricultural Research Service. Seeds were germinated for 3 d on wet filter paper at 25 °C in the dark. Germinating kernels were planted into 12×12 cm pots filled with a mixture of soil and sand (3/1; v/v) and grown in a climatic chamber under controlled long-day conditions (16 h day with 200 μmol m⁻² s⁻¹ light intensity and 20 °C; 8 h night with 16 °C) with 60% humidity. The day of pollination was defined using the morphology of the stigma and anthers according to the Waddington scale (W10) (Waddington et al., 1983) in the center of the spike (Weschke et al., 2000). The spiklets on the day of pollination were characterized by extended hulls, extended and widely branched stigma, and the presence of pollen grains on stigmatic hairs. Five seeds from each row corresponding to the center of the spikelet were collected at 2 d and 4 d intervals starting from anthesis until 48 days after pollination (DAP). Dry seeds were analyzed after at least 3 months of storage.

Estimation of nuclear DNA content and calculations of the super cycle value (SCV)

Nuclear DNA content was estimated using flow cytometry. Leaves collected from growing plants (from 10 d old till ~2 months old), coleoptile (a protective sheath covering the emerging shoot), and root tips dissected from seedlings at 4 days after germination (DAG) served as somatic tissue controls and were analyzed in ≥10 replicates. Each seed tissue was measured from ≥5 seeds, and the measurements were repeated three or more times on different days. Embryos were dissected using a needle and forceps under an SZX16 binocular microscope (Olympus). Five embryos were collected and used as one sample at 4 and 6 DAP. Cell nuclei were isolated from 4–8 DAP embryos by homogenizing them with a pestle in a 1.5 ml Eppendorf tube containing 300 µl of Otto I solution (0.1 M citric acid, 0.5% Tween-20). The crude suspension was filtered through a 50 µm nylon mesh (Sysmex-Partec) and stained after adding 600 µl of Otto II solution (0.4 M Na₂HPO₄·12H₂O) containing 2 µg ml⁻¹ DAPI. At 12 DAP, embryos and the other tissues were homogenized with a razor blade in a Petri dish containing 500 µl of Otto I solution and stained with 1 ml of Otto II solution containing DAPI. Nuclear samples were analyzed using a CyFlow Space flow cytometer (Sysmex-Partec) equipped with a UV-LED diode array. At least 5000 particles were acquired per sample, using a log₃ scale. Histograms were evaluated by the FloMax program (Sysmex-Partec) and interpreted as shown in Supplementary Fig. S4.

The percentages of nuclei in the embryo, endosperm, and SMTs were calculated based on the number of measured particles (counts) in (i) whole seeds and (ii) seeds containing only endosperm and SMTs (embryos were removed). Based on (ii), we estimated the proportion of endosperm and SMT nuclei, and next we also used (i) values to calculate the percentage of embryo nuclei according to the formula: embryo (%)=100–endosperm (%)–SMTs (%).

To estimate the amount of endoreduplication in individual samples, we introduce a new formula termed the super cycle value (SCV). Compared with other existing approaches, this formula estimates the frequency of endoreduplicated nuclei more conservatively (see the Results for more details). For the SCV, 8C in the diploid and 12C in the triploid tissues are considered as the first endoreduplicated levels. For diploid tissues, the SCV=[(n 2C×0)+(n 4C×0)+(n 8C×1)+(n 16C×2)]/(n 2C+n 4C+n 8C+n 16C) was calculated; n=number of counts per given C-value content. For triploid endosperm, 3C and 6C received a value of 0, 12C received a value of 1, and 24C received a value of 2.

Isolation of nuclei and TUNEL assays

Around 100 embryos were manually dissected from 8 DAP seeds using an SZX16 binocular microscope (Olympus). Precisely 80 roots of seedlings at 4 DAG were cut ~1 cm from the apex. Both types of tissues were rinsed immediately in 10 mM Tris buffer pH 7.0, fixed in 2% (v/v) formaldehyde/Tris buffer for 30 min at 4 °C, and then washed 3× 5 min in Tris buffer at 4 °C. Embryos were homogenized with a pellet pestle in 1.5 ml Eppendorf tubes in LB01 buffer (Doležel et al., 1989). Root apices were excised (~1 mm from the tip) and homogenized in LB01 buffer for 13 s at 15 000 rpm using a Polytron PT1300D homogenizer (Kinematica AG). The crude homogenates were filtered through a 50 µm pore size mesh. From at least 30 dissected embryos older than 8 DAP, peeled seeds without the embryo were rinsed in Tris buffer, pre-fixed by vacuum infiltration in 4% formaldehyde/Tris buffer for 20–30 min at 4 °C, followed by fixation and washing as above. Tissues were chopped with a razor blade in LB01 buffer on a Petri dish and filtered through a 50 µm nylon mesh. Suspensions of nuclei were stained with 2 µg ml⁻¹ DAPI.

Approximately 500 nuclei for each determination of DNA content were sorted into a drop of LB01 buffer on microscopic slides using a FACSAria II SORP flow cytometer and sorter (BD Biosciences, Santa Clara, CA, USA), air-dried, and stored at –20 °C until use. Double-strand breaks (DSBs) were detected by TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) assay using the in situ cell death detection kit with a fluorescein-dUTP label following the manufacturer’s instructions (11684795910, Roche). Nuclei which had been flow-sorted from root tips were used as controls. The TUNEL negative (–) background noise control was prepared by omitting terminal deoxynucleotidyl transferase treatment, while the high signal control nuclei [TUNEL positive (+)] were digested with 100 U ml⁻¹ DNase I for 10 min at 21 °C. Nuclei were counterstained with DAPI (1 µg ml⁻¹) in Vectashield (H-1000, Vector Laboratories). Analysis of fluorescence signals was performed with an AxioImager Z2 (Zeiss) epifluorescence microscope equipped with a DSD2 spinning disk confocal imaging module and monochromatic Zyla 4.2 camera (both Andor). Z-stacks were captured separately for each fluorochrome using the appropriate excitation and emission filters with the IQ3 (Andor) system. At least 50 nuclei were evaluated per individual sample, each in three biological replicates. TUNEL (–) nuclei were characterized by pale-green fluorescence signals, while TUNEL (+) nuclei had bright green signals (Ghasemzadeh et al., 2015; Palermo et al., 2017).

Fluorescein diacetate and Evans blue staining assays

Hulls were removed from fresh seeds with forceps. Peeled seeds were cut to half in phosphate-buffered saline (PBS) along the longitudinal and transverse axes with a razor blade. At least 20 seeds were tested per experimental point. One half of the sample was stained with 2 mg l⁻¹ fluorescein diacetate (FDA, F7378, Sigma-Aldrich) in PBS for 15 min in the dark, and then washed twice for 10 min with PBS (Kobayashi et al., 2013). The other half of the samples were stained in 0.1% (w/v) Evanse blue (314-13-6, Sigma-Aldrich) for 2 min. Stained sections were washed twice for 10 min with distilled water (Wu et al., 2016). Transverse and sagittal sections of samples were observed with an SZX16 binocular microscope (Olympus) equipped with a GFP (green fluorescent protein) filter. Images were captured with a Regita 1300 QImaging camera and QCapture ×64 software (Olympus) using the same exposure times. The sagittal section area of whole peeled seed, endosperm, and embryo was measured in ImageJ (Schneider et al., 2012).

Light microscopy and determination of seed growth parameters

For phenotypic analysis, seeds from 0 (ovary) to 48 DAP and dry seeds were peeled off, weighed with an analytical scale (Sartorius), and then photographed using an SZX16 binocular microscope (Olympus) equipped with a Regita 1300 QImaging camera and QCapture ×64 software (Olympus). A minimum of 50 seeds were used per individual replicate. To analyze transverse and sagittal plans, the seeds were cut with a razor blade along the longitudinal and transverse axis, respectively. Seed length, width, and sagittal section area were measured using ImageJ. The growth rate (GR) was calculated according to the formula: GR (%)=(x₂–x₁)/x₂×100, where x₁ was the previous time point measurement (e.g. at 0 DAP) and x₂ was the next time point measurement of seed growth (e.g. at 2 DAP).

Confocal microscopy, morphology of the nuclei, and calculation of the mitotic index

Whole seeds were fixed in acetic acid/alcohol (1:3) for 2–4 h and stored in 70% ethanol at –20 °C until use. The transverse slides were cut with a razor blade and stained with 2 µg ml⁻¹ DAPI in Vectashield. The images were captured with an AxioImager Z2 epifluorescence microscope (Zeiss) equipped with a DSD2 confocal module, monochromatic camera Zyla 4.2, and IQ3 program (all Andor). The photos were processed with Imaris 9.2 software (Bitplane, Oxford Instruments). Three to five slides, each corresponding to one endosperm, were evaluated per experimental point. In total, ~600 cells per slide were scored for quantification of mitotic divisions. The mitotic index (MI) was calculated according to the formula: MI (%)=cells in mitosis/total number of cells×100. For calculation of the proportion of deformed nuclei, flow-sorted nuclei were used. Three to four slides were evaluated per time point, each with at least 250 nuclei per individual C-value.

Statistical analysis

Statistical significance was examined by one-way ANOVA and post-hoc comparison by Tukey’s multiple range test (P≤0.05) using Minitab v. 18 (Minitab, LLC) software. Principal component analysis (PCA) and correlation analysis were performed with Statistica v. 13 (StatSoft) and RStudio programs, respectively.

Results

Endoreduplication occurs in barley endosperm and SMTs

To assess the dynamics of barley seed growth, we monitored grain weight, length, and width of peeled (after manual hull removal) growing seeds of the cv. Compana from 0 (ovary) to 48 DAP and at 3 months after harvest (Fig. 1; Supplementary Fig. S2). From 0 to 4 DAP, the seeds increased their size and weight mainly due to the growth of SMTs. From 6 DAP onwards, sagittal sections revealed endosperm expansion and acceleration of seed growth (Supplementary Fig. S2C). At 12 DAP, seeds reached the maximum length of ~9 mm, and from 12 to 28 DAP, the weight of the caryopsis increased mostly because of seed widening, presumably due to CSE expansion. During this period, the embryo also grew rapidly (Supplementary Fig. S2B, C). At 32 DAP, the seed reached maximum weight, length, and width of 0.076 g, 9.37 mm, and 4.07 mm per grain, respectively (Fig. 1B). Subsequently, the seeds started desiccating, leading to their reduction in size and weight.

Fig. 1.

Phenotype of developing barley seeds of cv. Compana. (A) Grain developmental series from 0 (ovary) to 48 DAP and in dry seeds. For analysis of sagittal and transverse planes, the seeds were cut in half. Scale bar=5 mm. The yellow insets show early-stage embryos with scale bar=500 µm. (B) Quantitative data for 100-seed weight, seed length, and seed width calculated as absolute values (black line, left y-axis) and relative to the ovary (columns, right y-axis). Data represent the means (±SD) from three biological replicates, each with at least 50 seeds. ANOVA was carried out separately for each tested parameter. Values marked with the same letter do not differ according to Tukey’s multiple range test (P≤0.05).

To test the relationship between seed parameters and the degree of endopolyploidy, we estimated nuclear C-values using flow cytometry. Somatic tissue controls represented by the root apical meristem (RAM), coleoptile, and young leaves contained 96–97% 2C and 4C nuclei, respresenting G₁- and G₂-phase nuclei, respectively, and only ≤3.5% 8C endopolyploid nuclei (Supplementary Fig. S3). The RAM tissues had an almost equal ratio of 2C (46.7%) and 4C (48.8%) nuclei, indicating a high cell division activity. In contrast, the leaves contained 80% of 2C nuclei, suggesting a lower cell division rate. However, the proportion of endoreduplicated nuclei remained low in somatic tissues, as indicated by a very low SCV (0.04–0.06).

We measured C-values of nuclei isolated from (i) dissected embryos and (ii) a mixture of SMTs and endosperm (Fig. 2A; Supplementary Figs S4, S5). This allowed us to estimate the relative proportion of embryo, SMTs, and endosperm nuclei during grain development (Fig. 2B). Directly after pollination, SMTs represented the majority of nuclei, but the proportion of endosperm nuclei quickly increased and reached the majority (58%) of all seed nuclei at 8 DAP. Afterwards, the proportion of endosperm nuclei started decreasing, and the percentage of embryo nuclei continuously increased up to 34% at 44 DAP. Finally, the dry barley grain contained 31% of nuclei from the embryo, 41% from SMTs, and 28% from the endosperm. Interestingly, the low proportion of endosperm nuclei contrasts with the fact that this tissue makes most of the seed mass (Fig. 1A).

Fig. 2.

Nuclear C-values in developing barley seed tissues. (A) Representative histograms of nuclear DNA contents obtained from the dissected embryo (EMB; left graph) and seed maternal tissues and endosperm (SMTs and END, respectively; right graph) at 12 DAP. Representative histograms of nuclear DNA contents obtained at 0–24 DAP and dry seeds are shown in Supplementary Fig. S4. (B) Cumulative percentage of nuclei from the major seed tissues. Stacks marked with the same letter do not differ within the respective group according to Tukey’s test (P≤0.05). (C) Percentage of nuclei with different C-values in major seed tissues at different DAP. Data are the means (±SD) from three biological replicates, each with at least five individual measurements. ANOVA was carried out separately for each C-value. Values marked with the same letter do not differ according to Tukey’s test (P≤0.05). n.a. indicates that samples were not analyzed. (D) Super cell cycle values during seed development based on the data from (C). ANOVA was carried out as described for (C).

Next, we estimated C-values of nuclei in individual tissues (Fig. 2C; Supplementary Fig. S5). Embryos contained ~75% of 2C nuclei during the first 12 DAP and subsequently their proportion decreased to ~60%. Reverse dynamics were observed for 4C nuclei. The endoreduplicated 8C nuclei reached 9% at 16 DAP and remained at this level. SMTs showed large dynamics of 2C and 4C nuclei starting at 80% (2C) and then oscillating between 25% and 50%. The amount of endoreduplicated 8C and 16C SMT nuclei reached 20% at 44 DAP, but then strongly decreased during seed desiccation. Endosperm contained the highest percentage of endoreduplicated nuclei (up to 40%), and 2-day-old syncytium already had 8% of endoreduplicated 12C and 24C nuclei (Fig. 2C; Supplementary Fig. S5). The frequency of endoreduplicated (12C and 24C) nuclei continuously increased up to ~50% from 20 to 28 DAP. Subsequently, the frequency of endoreduplicated nuclei rapidly decreased, and endosperm tissues of dry seeds contained only 7% of such nuclei. This indicated a programmed and preferential loss of endopolyploid nuclei during the terminal stages of barley endosperm development. Finally, we estimated C-values of nuclei from hulls, but this tissue showed little variation and its profile strongly resembled that of leaves Supplementary Figs S3, S6).

There are multiple methods for quantifying nuclear C-values. This includes the mean C-value, which is an indicator describing the average C-value of all measured nuclei (Rewers et al., 2009). However, the mean C-value does not allow a direct comparison of the tissues with different ploidy such as the diploid embryo and triploid endosperm. Here, the cell cycle value (i.e. endoreduplication index), calculated and averaged from the number of endoreduplication cycles for all nuclei, is a better solution (Barow and Meister, 2003; Parker et al., 2018). Nevertheless, the existing cell cycle value formulae consider 4C nuclei as already endoreduplicated once. Although some 4C nuclei might already be programmed for endoreduplication, others will not be. The latter may be particularly true for embryonic and RAM tissues used in this study, where many cells with 4C nuclei will be regularly cycling. Therefore, we introduced a new formula called the SCV (see the Materials and methods for details), which conservatively compares the degree of endoreduplication in individual tissues irrespective of their basic ploidy (Fig. 2D; Supplementary Figs S3C, S6C). For diploid tissues, SCV=0 means that all nuclei are 2C or 4C while SCV=1 means that the nuclei are on average 8C. The SCV of 4 DAP embryos was 0.03 and remained stable until 12 DAP, then gradually increased to ~0.15 and reached a plateau. The SCV of SMTs started from 0.02, constantly increased to 0.47 at 12 DAP, and then gradually decreased to 0.27 in dry seeds. The endosperm SCV curve had a profile similar to that of SMTs, but started from 0.13 at 2 DAP, peaked at 0.58 at 20 DAP, and then decreased to 0.08 in dry seeds.

The dynamics of endoreduplication in developing barley seeds prompted us to analyze what the degree of correlation is between endoreduplication (i.e. SCV) and seed growth parameters (i.e. seed weight, length, and width). We performed Pearson correlation coefficientanalysis and visualized it using a heat map (see Supplementary Fig. S7). When we looked at the endocycle during embryo development, we observed a strong positive correlation of SCV with seed weight and width (r=0.78 and 0.83, respectively). For SMTs, the strongest correlation existed between SCV and seed length (r=0.83). Surprisingly, in developing endosperm, only a moderate correlation appeared between compared parameters (r=0.29–0.46). These correlations have an indicative value (e.g. the phase when seeds greatly gain weight and width is the time when the embryo undergoes most of its endoreduplication), but do not describe causality.

To summarize, barley embryo shows a moderate degree, while SMTs and endosperm show high degrees of endoreduplication. Surprisingly, the amount of SMTs and endosperm endopolyploid nuclei becomes strongly reduced during seed maturation and drying.

Endoreduplicated nuclei accumulate DNA damage

The loss of endoreduplicated nuclei at later stages of seed development raised the question of their fate. We hypothesized that they may have reduced genome stability and are removed. Therefore, we performed TUNEL assays using flow-sorted nuclei from the embryo, SMTs, and endosperm at 4, 8, 16, and 24 DAP (Fig. 3). In the TUNEL assay, 3' termini of DNA are labeled by tagged nucleotides using terminal deoxynucleotidyl transferase, and then detected. The signals indicate the presence of DNA DSBs. The 2C and 4C nuclei from the RAM at 4 DAG were used as somatic tissue controls. The background negative signal controls were represented by 2C RAM nuclei without the terminal deoxynucleotidyl transferase treatment. In contrast, damaging the DNA of the same nuclei using 100 U ml⁻¹ DNase I for 10 min resulted in 80% of TUNEL (+) nuclei. For embryos, in 2C and 4C nuclei we observed ~15% TUNEL (+) cells. However, in 8C nuclei isolated from 24 DAP embryos, the accumulation of DSBs increased with increasing amounts of DNA [49% TUNEL (+) nuclei]. The nuclei of SMTs showed ~40% TUNEL (+) signal in 2C and 4C up to 16 DAP. However, for all C-values at 24 DAP, amounts of DSBs increased and ~60% of TUNEL (+) nuclei appeared. For endosperm, 3C, 6C, and endoreduplicated 12C nuclei accumulated DSBs in increasing C-value- and age-dependent manners.

Fig. 3.

Analysis of genome integrity using TUNEL assay. (A) Representative TUNEL negative (–) and positive (+) nuclei isolated from 24 DAP endosperm tissues. Nuclei were counterstained with DAPI (blue), and DNA double-strand breaks (DSBs; green) were detected using TUNEL assay. All nuclei were photographed with the same settings. Scale bar=10 µm. (B) Quantification of DNA damage in nuclei of different C-values and from different tissues. The TUNEL assay control was represented by RAM nuclei. The negative control (N) was prepared without terminal deoxynucleotidyl transferase and its fluorescence was considered as the background signal. Positive control (P) nuclei were treated with DNase I. Four DAP embryos were not analyzed (n.a.) for technical reasons. Values are the means (±SD) from three biological replicates, each with at least 50 individual nuclei. ANOVA was carried out for all C-values, time points, and tissues within the plot. Values marked with the same letter do not differ according to Tukey’s test (P≤0.05).

This demonstrates that endoreduplication is associated with reduced genome stability in barley seed tissues and the greatest damage was observed in the late SMTs and endosperm nuclei.

Seed maternal and endosperm tissues undergo cell death

Studies in maize (Young and Gallie, 2000), rice (Kobayashi et al., 2013; Wu et al., 2016), bread wheat (Yifang et al., 2012), and triticale (Li et al., 2010) identified that cell death is an essential process during cereal seed development. To detect viable and dead cells in developing barley grains, we performed FDA and Evans blue staining (Fig. 4A, B; Supplementary Figs S8, S9). FDA is hydrolyzed in living cells into the green fluorescent fluorescein, which indicates viable cells (Schnürer and Rosswall, 1982). In turn, Evans blue points to a loss of membrane integrity by dyeing the intracellular space of non-viable cells.

Fig. 4.

Time-course study of cell viability in developing seeds. (A) Representative sagittal and transverse sections of barley seeds showing fluorescein signals (green fluorescence) after staining with 2 mg l⁻¹ fluorescein diacetate (FDA). The inset shows FDA-stained pollen attached to the stigma, with scale bar=50 µm. Dashed white lines show the whole sagittal section area of the seed. A negative control without FDA staining is shown in Supplementary Fig. S8. (B) Representative sagittal and transverse sections of developing barley seeds stained with 0.1% Evans blue. Scale bar for (A) and (B)=5 mm. The seeds shown are representative of at least 20 individual stained seed. (C) DAPI staining of nuclei and cell membranes in barley endosperm tissues. Dark spots indicate nuclei and dark squares indicate the area of cells, as DAPI is a membrane-permeant dye. Green arrows mark dividing nuclei. Red arrows mark degenerating nuclei and nuclear residues. CSE=central starchy endosperm, AL=aleurone layer. Scale bar=20 µm. More time points are presented in Supplementary Fig. S10. (D) The frequency of dividing nuclei in developing endosperm. Every time point represents the mean of at least three independent measurements (slides), each with at least 600 counted nuclei. Values marked with the same letter do not differ according to Tukey’s test (P≤0.05). (E) Representative undeformed and deformed nuclei isolated from 24 DAP endosperm tissues. Nuclei were counterstained with DAPI (gray). Scale bar=20 µm. (F) Quantification of deformed nuclei of different C-values and from different tissues. Values are the means (±SD) from three biological replicates, each with at least 250 nuclei per individual C-value. ANOVA was carried out separately for each tissue. There were no significant differences for the RAM (P=0.486).Values marked with the same letter do not differ according to Tukey’s test (P≤0.05).

Cell death followed a specific pattern in developing barley seeds (Fig. 4A, B; Supplementary Fig. S8). During early seed development (0–8 DAP), we observed fluorescein signals in the middle part of SMTs (pericarp, and seed coats), but not in the embryo sac or in the developing endosperm. From 12 DAP, the fluorescence appeared in the endosperm and its intensity increased during CSE formation until 32 DAP, suggesting high metabolic activity of this tissue. From 36 DAP, the fluorescence intensity started decreasing from the periphery towards the central zone of the CSE, indicating reduced viability of cells in this outer region. During maturation, the area of fluorescein-labeled cells further shrunk. This pointed to a progressive loss of cell viability during seed desiccation (Fig. 4A; Supplementary Fig. S8). Evans blue staining revealed a complementary pattern. During early seed development, we detected increasing regions of blue staining in the top (in the region surrounding the brush) and bottom parts of maternal tissues, but not in the longitudinal elongation zone. Staining in the endosperm was first detected in a few dispersed cells at 8 DAP, and the number of stained cells increased over time. The AL was free of stain until 48 DAP, but some signal could be detected in AL of dry seeds (Supplementary Fig. S9). No staining was observed within the embryo at any stage of seed development (Fig. 4B; Supplementary Fig. S9).

To verify a potential link between cell death and the loss of endoreduplicated nuclei, we investigated the morphology of CSE nuclei after DAPI staining of sliced seeds (Fig. 4C; Supplementary Fig. S10). The nuclei frequently divided at 6 and 8 DAP (MI ~20%) but the frequency of mitoses decreased over time and no cell divisions were observed at 20 DAP or after this time point (Fig. 4D). From 16 DAP, the density of nuclei in the central part of the endosperm progressively decreased (indicating their elimination by cell death) and many larger nuclei were progressively degenerating, as indicated by their deformed shape (Fig. 4E, F), and finally only remnants of nuclei (pieces of chromatin) were observed. Conversely, the nuclei in the three layers of AL cells started to be clearly visible at ~12 DAP and persisted until maturation.

This showed that some SMTs and endosperm cells undergo cell death from as early as 2 and 8 DAP, respectively; endoreduplicated CSE nuclei are the first to be degraded during seed maturation, while AL nuclei remain alive.

Dynamics of endoreduplication in six two-row barley cultivars

To challenge the validity of our findings in a broader context, we compared the data from the US cv. Compana with five European two-row spring barley cultivars: Mars (Czech Republic), Betzes, Klages (both Germany), Ingrid (Sweden), and GP (UK) (Fig. 5A). The cultivars differed as to their dry seed weight and morphology (Fig. 5B; Supplementary Fig. S11).

Fig. 5.

Estimation of nuclear DNA contents in seed tissues of six two-row barley cultivars. (A) Geographical origin of the used cultivars. GP=Golden Promise. (B) 100-seed weight of hulled and peeled dry kernels. Data are the means (±SD) from three biological replicates, each with at least 100 seeds. ANOVA was carried out separately for hulled and peeled seeds. Values marked with the same letter do not differ according to Tukey’s test (P≤0.05). (C) A heat map of the P-values of AVOVA between cultivars for individual C-values in diploid and triploid tissues at a given DAP. Source data are shown in Supplementary Figs S12 and S13. (D) Super cell cycle values of the embryo and seed maternal tissues (EMB+SMTs) versus endosperm (END) at a given DAP. Values are the means (±SD) from three to five biological replicates, each with at least three individual measurements. ANOVA was carried out separately for each time point between cultivars, and gray shading shows the periods of significant differences between cultivars. P-values are shown in Supplementary Fig. S14. Mean=mean percentages of super cycle value calculated between all cultivars. Data are the averages (±SD) between cultivars. Points marked with the same letter do not differ according to Tukey’s test (P≤0.05).

First, we estimated whether there was a statistically significant difference between individual C-values at different DAPs for diploid tissues (embryo and SMTs) and triploid endosperm for all cultivars from 4 DAP until 24 DAP and in dry seeds, and plotted the P-values to reveal the major trends (Fig. 5C; Supplementary Figs S12 and S13 show the source data for the statistics). There were few differences until 8 DAP for the diploid tissues. From 12 DAP, the intercultivar variation increased for 4C, 8C, and 16C nuclei and continued mainly with the 16C nuclei at 16–24 DAP. From 24 DAP, the intercultivar differences shifted towards the nuclei with lower C-values (2C and 4C) and also persisted in dry seeds. However, it should be noted that the differences were often caused by one or two cultivars (Supplementary Fig. S12); nevertheless, these data point to an existing variation in endoreduplication during barley embryo and/or SMT development.

During endosperm growth, the frequencies of almost all C-values varied in the cultivars from 12 to 24 DAP (Fig. 5C; Supplementary Fig. S13). The frequency of 3C nuclei differed from 24% to 48% at 12 DAP (Compana and GP versus Ingrid, respectively) and from 28% to 39% at 24 DAP (Mars versus Klages, respectively). The variation in 6C nuclei was mainly due to the cv. Mars, which had more 6C nuclei than other genotypes (e.g. 49% versus ~34% in other cultivars at 16 DAP). For the endoreduplicated endosperm nuclei (12C and 24C), the major differences occurred from 12 to 20 DAP, corresponding to the time of rapid endosperm expansion, the endoreduplication maximum, and then its decrease. Interestingly, there were no significant differences in C-values in the endosperm of dry seeds between the cultivars, suggesting that it is rather the dynamics of the whole process that were affected. To provide an average picture, we calculated the mean percentage of nuclei of individual C-values for all genotypes, which summarizes the trends in DNA content changes during barley seed development (Supplementary Figs S12C, S13C).

Next, we calculated the SCV for the cultivars (Fig. 5D; Supplementary Fig. S14A). For embryos/SMTs, it varied significantly from 12 DAP to 24 DAP, and in dry seeds. For instance, at 16 DAP, the SCV ranged from 0.19 (Klages) to 0.32 (GP). In endosperm tissues, the SCV differed from 12 to 20 DAP. To gain insight into the relationships among the cultivars, we performed PCA using SCV data (Supplementary Fig. S14B). The first component (PC1) separated samples based on DAP and revealed which time points of diploid tissues and endosperm development were more similar. The second component (PC2) showed the relationships between the cultivars. In the mixture of diploid tissues, the PCA revealed the relationship between cultivars; Mars, Klages, and Compana formed one group, and Betzes, GP, and Ingrid were far from them. Subsequently, the SCV analysis in endosperm revealed two groups; the first group included the interval from 4 to 8 DAP (the mean SCV ranged from 0.10 to 0.29), and the second group contained the period from 16 to 24 DAP (the mean SCV ranged from 0.26 to 0.18), excluding 12 DAP (mean SCV=0.33) and dry seed (mean SCV=0.17) (Fig. 5D; Supplementary Fig. S14B).

Collectively, these results demonstrate genetic variation in the endosperm endocycle kinetics during barley seed development.

Discussion

Barley caryopsis development has been explored extensively from the biochemical and metabolic perspective (Radchuk et al., 2009; Sreenivasulu et al., 2010; Peukert et al., 2014). In contrast, cell cycle dynamics and endoreduplication during barley seed formation have received much less attention. This motivated us to investigate the spatiotemporal dynamics of endoreduplication and loss of nuclei in the embryo, endosperm, and SMTs from pollination until dry seeds (Fig. 6).

Fig. 6.

The model of endoreduplication and cell death dynamics during barley seed development. (A) Phases of barley seed development. Thick gray horizontal lines mark three previously described general cereal developmental stages (Sabelli and Larkins, 2009; Sreenivasulu et al., 2010; Dante et al., 2014). The black curved line shows the dynamics of fresh weight changes. (B) Overview of developmental events in barley seed development. Data regarding nuclei numbers have been published (Bennett et al., 1975; Olsen, 2001, and references therein; Radchuk et al., 2011). (C) Schematic dynamics of endoreduplication in maternal tissues (SMTs; green), endosperm (END; blue), and embryo (EMB; red). (D) The dynamics of programmed cell death in barley seed tissues.

Endoreduplication is receiving increasing attention as one of the mechanisms that leads to cell differentiation and specialization (Bhosale et al., 2018). To quantify endoreduplication frequency in diploid and triploid tissues, we introduced the new concept of SCV. Compared with previous studies, the formula for the SCV is more conservative and considers as endoreduplicated only from 8C nuclei in diploid and from 12C nuclei in triploid tissues. We found that barley endosperm tissues underwent one to two rounds of endoreduplication resulting in 12C and 24C nuclei, respectively. This is similar to wheat (Chojecki et al., 1986) and rice (Kobayashi, 2019). More endoreduplication cycles were found in the endosperm of sorghum with four (Kladnik et al., 2006) and in maize with up to seven rounds of endocycles (Engelen-Eigles et al., 2000; Sabelli and Larkins, 2009). In barley endosperm, the major wave of endoreduplication started from ~6 DAP and increased linearly to 20 DAP, which corresponds to seed developmental stage II characterized by production of storage components (Dante et al., 2014). This is similar to sorghum with the onset of endosperm endoreduplication at 5 DAP (Kladnik et al., 2006), but earlier than that at 10 DAP in maize (Kowles et al., 1990; Leiva-Neto et al., 2004; Sabelli and Larkins, 2009) or 12 DAP in wheat (Chojecki et al., 1986). Some differences also occur in the peak of endosperm endoreduplication between cereals. Endoreduplication peaked at 20 DAP under our experimental conditions, while it peaked at 15–18 DAP in maize and at 24 DAP in wheat (Chojecki et al., 1986; Sabelli and Larkins, 2009).

How much the dynamics are affected by environment or basal ploidy remains unknown. A common observation for cereals is that endoreduplication is initiated in the endosperm central region from where it spreads toward the periphery, excluding the AL (Kladnik et al., 2006; Sabelli, 2012; Kobayashi, 2019). Here, barley is an exception as it contains endoreduplicated cells also in the AL (Keown et al., 1977). We detected one to two rounds of endocycles (leading to 8C and 16C nuclei) also in SMTs and embryos, and their endoreduplication peaked at ~12 and 20 DAP, respectively. Hence, our data support the notion that endoreduplication is associated specifically with the reproductive organs in grasses (Sabelli, 2012). Barley somatic tissues generally have a low degree of endopolyploidy and most cells are found at the G₁ phase of the cell cycle (Trafford et al., 2013).

Despite the fact that endoreduplication appears to be ubiquitous among cereals, its role in seed development remains poorly understood (Sabelli, 2012). We detected a moderate to high positive correlation between the SCV of the embryo and SMTs and the biomass parameters of the growing grain. Suprisingly, only a weak correlation appeared between an increasing endoreduplication in endosperm and seed biomass parameters. This is in agreement with the observations in maize, where modification of the cell cycle did not affect the final seed size (Sabelli et al., 2013). The pericarp is the major site of starch deposition during the first days after pollination in barley (Radchuk and Borisjuk, 2014). At later time points, the main tissues accumulating starch, proteins, and lipids are the CSE and AL (Radchuk et al., 2009; Sreenivasulu et al., 2010; Peukert et al., 2014). By combining these metabolomics studies with our morphometric analysis and SCV, we conclude that endopolyploidization in the pericarp and endosperm will be associated with the accumulation of storage components. This is further supported by the observation of a correlation between high transcriptional activity of storage protein and starch genes and increasing amounts of nuclear DNA during barley endosperm development (Giese, 1992; Sreenivasulu et al., 2004). In turn, endopolypidization in barley embryo coincides in time with the differentiation of embryo tissues, which is evidence for its a role in sustaining cell fate (Bramsiepe et al., 2010).

We found that many SMTs and endosperm nuclei were lost during the later stages of seed development. The results of TUNEL assays suggested that the degeneration initiated with highly endopolyploid nuclei. However, nuclear DNA content analysis of dry seeds revealed the presence of 3C and 6C (major ploidy level) nuclei and only ~6% of endoreduplicated 12C nuclei. Microscopic observations confirmed that these populations of nuclei originated from the AL (Keown et al., 1977; Sreenivasulu et al., 2010). Before degeneration, we observed accumulation of DNA damage in endoreduplicated nuclei. Data from Arabidopsis showed that endopolyploid nuclei in somatic cells and endosperm possess less condensed heterochromatin, which could be (i) a consequence of or (ii) the reason for lower genome stability (Schubert et al., 2006; Baroux et al., 2007). Hence, another potential function or a consequence of endoreduplication could be disabling specific nuclear functions and finally marking the cells for cell death (Sabelli, 2012). This is supported by the overlap of the spatial patterns of endoreduplication and cell death in the CSE in maize and rice (Young and Gallie, 2000; Kobayashi et al., 2013; Kobayashi, 2019).

The experiments performed in this study, including the analysis of cell viability and microscopic analysis of seed tissues, showed that both barley SMTs and endosperm underwent cell death. Although the pericarp tissues were highly metabolically active during the first 8 DAP, Evans blue staining revealed a loss of viability already at 2 DAP in some regions. The previous study focusing on cell death in barley SMTs showed that cell divisions already decreased at 2 DAP in the pericarp, and the expansion of the tissue occurred by cell elongation in longitudinal directions (Radchuk et al., 2011). We observed that cell death first appeared in the top and the bottom parts of the pericarp, and later (as the pericarp became very thin) expanded to the elongation zone. This pattern is consistent with the results described in bread wheat (Young and Gallie, 1999) and triticale (Li et al., 2010). To summarize, SMTs first synthesize, temporarily store, and finally transport the nutrients to the developing embryo and endosperm (Li et al., 2010; Radchuk et al., 2011), and this is accompanied by the elimination of the tissues by cell death.

In barley, wheat, and triticale, cell death proceeded stochastically throughout the CSE (Young and Gallie, 1999; Li et al., 2010). Conversely, in maize and rice, dying cells first appear in the upper CSE and expand outwards (Young and Gallie, 2000). These results suggest that the cell death patterns vary greatly depending on the species. However, there seems to be a link between endoreduplication, cell death, and storage compound deposition (Fig. 6). At 24 DAP, when the whole CSE has undergone cell death, there was still detectable mRNA and the starch accumulation continued (Radchuk et al., 2009). This indicates that specific precursors, regulatory molecules, and/or enzymes are synthesized in excess before degradation of the nuclei.

By analyzing six elite barley cultivars, we found variation in the kinetics of endosperm endoploidization. On the one hand, some varieties reached the peak earlier (the variation was between 16 and 20 DAP) and, on the other hand, there were varieties with a relatively low SCV throughout the entire endosperm development. Similar variability was observed in rice (Kobayashi, 2019). This indicates the presence of genetic variation in the endopolyploidy dynamics during barley endosperm development. In our data set, we did not detect a link between the level of endoreduplication and the geographical origin of the cultivars. However, it has to be noted that the cultivars used in our study represent only a limited genetic diversity and that the variation is probably much greater in genetically distant landraces (Milner et al., 2019). Furthermore, our study suggests that cell cycle dynamics during barley development are a genetically strictly controlled process. Although endoreduplication plays an important role in determining cell size in cereal endosperm (Kladnik et al., 2006; Kobayashi, 2019), there was no significant correlation between the SCV in developing endosperm and weight of the mature caryopsis. Because organ size is determined not only by cell size, but also by cell number (Robinson et al., 2018), we speculate that the variation in cell size associated with endoreduplication is probably cancelled out by the diversity in cell number in the examined barley cultivars.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. Morphology of barley seeds 16 days after pollination (DAP).

Fig. S2. Growth ratio of developing barley seed.

Fig. S3. Flow cytometric estimation of nuclear C-values in barley somatic tissues.

Fig. S4. An example showing the interpretation of histograms of nuclear DNA content.

Fig. S5. Representative gated flow cytometric histogram of barley seed tissues of cv. Compana.

Fig. S6. Estimation of nuclear C-values in barley hulls.

Fig. S7. Analysis of correlation between seed endopolyploidy and biomass parameters.

Fig. S8. Negative control of fluorescein diacetate assay.

Fig. S9. Analysis the aleurone layer viability.

Fig. S10. DAPI staining of nuclei and cell membranes in barley endosperm tissues.

Fig. S11. Phenotypic analysis of dry seeds from six two-row barley cultivars.

Fig. S12. Comparison of nuclear C-values in diploid seed tissues from two-row barley cultivars.

Fig. S13. Comparison of endosperm individual C-values from two-row barley cultivars.

Fig. S14. Super cell cycle value analysis of embryo and seed maternal tissues versus endosperm.

Abbreviations

Abbreviations
 
• AL

  aleurone layer

 
• BETL

  basal endosperm transfer layer

 
• CSE

  central starchy endosperm

 
• DSB

  double-strand breaks

 
• DAG

  days after germination

 
• DAP

  days after pollination

 
• ESR

  embryo-surrounding region

 
• FDA

  fluorescein diacetate

 
• MI

  mitotic index

 
• PCA

  principal component analysis

 
• RAM

  root apical meristem

 
• SCV

  super cycle value

 
• SMT

  seed maternal, tissue

 
• TUNEL

  terminal deoxynucleotidyl transferase dUTP nick end labeling

Acknowledgments

We thank Eva Jahnová and for technical assistance and Zdenka Bursová for plant care. AN, MK, and AP were supported primarily by The Czech Science Foundation grant 18-12197S. AP was further supported by a Purkyně Fellowship from the Czech Academy of Sciences. Multiple co-authors were funded from the European Regional Development Fund project ‘Plants as a tool for sustainable global development’. (no. CZ.02.1.01/0.0/0.0/16_019/0000827).

Author contributions

AP and AN designed the research. AN, MK, BT, YZ, and DW performed experiments. JV flow-sorted the nuclei. AN analyzed the data, and designed and prepared the figures. JD discussed the data. AN and AP wrote the manuscript with contributions from all authors. All authors approved the final version of this article.

Data availability

All data supporting the findings of this study are available within the paper and within its supplementary data published online.