Barley (Hordeum vulgare cv. Navigator) plants were grown under standard greenhouse conditions using day/night temperatures of 23°C/15°C. Navigator is an Australian malting barley cultivar. The date of anthesis for each barley head was assessed visually by the presence of free pollen on anthers. Grains were collected every 4 days from 6 to 42 days post anthesis (DPA) from the middle of the head. Each individual sample contained tissues collected from three grains. At 6 DPA and 42 DPA, whole grain tissue was analyzed. At 10 DPA, internal grain contents (embryo plus endosperm) were collected separately from external maternal tissues. From 14 to 38 DPA, the embryo was manually dissected from the grain and the remaining endosperm/maternal tissues were collected together; as much husk as possible was removed from the samples.

For developing grain, RNA was prepared and sequenced from three replicate samples of each tissue at each time point, using the Spectrum^TM Plant Total RNA kit (Sigma-Aldrich, St Louis, MO, United States), as described by Betts et al. (2017).

RNA integrity was assessed using an Agilent 2200 TapeStation system (Agilent Technologies, Inc., Waldbronn, Germany). Libraries were prepared from total RNA using the TruSeq Stranded Total RNA with Ribo−Zero Plant kit according to the manufacturer’s instructions (Illumina¹) for three biological replicates of each tissue type. Transcript abundances and count estimates (transcripts per million TPM) were determined using a k-mer index build from the representative transcript models (Mascher et al., 2017) using a k-mer length of 31 and the kallisto program (version 0.46.0) with 100 bootstraps (Bray et al., 2016). Principal component analysis (PCA) was performed using TPM values for all genes and the prcomp function in R. Hierarchical clustering was performed with the Partek Genomics software suite version 6.16 (Partek Incorporated, St. Louis, MO, United States). Functional gene categories were analyzed using the PageMan tool (Usadel et al., 2006) following gene annotation using Mercator (Lohse et al., 2014) of reference plant genome sequences, with manual curation.

Gene ontology (GO) enrichment analyses were undertaken using the “BiNGO” plugin program for Cytoscape (PMID 15972284); enrichment was considered statistically significant if a p < 0.05 was observed after Bonferroni correction. In silico gene discovery and annotation was performed as described in Betts et al. (2020).

Mature and germinated grains were cut approximately in half along their longitudinal axes and fixed in 0.25% glutaraldehyde, 4% paraformaldehyde, and 4% sucrose in phosphate-buffered saline (PBS), pH 7.2. The grains were subsequently dehydrated and embedded in LR White resin, as described in Burton et al. (2011). Sections (1 μm) were cut on an ultramicrotome using a diamond knife and dried onto polylysine-coated microscope slides (Thermo Fisher Scientific). Sections of mature grain and 96 h germinated grain were treated with or without α-amylase (Megazyme Bacillus licheniformis; E-BLAAM) for 1 h at room temperature. Lugol’s solution (2% w/w KI, 1% w/w I₂) was applied to the slides for 30 s and rinsed off with water. Sections were imaged using a Carl Zeiss microscope (Axio Imager M2, Zeiss, Germany) equipped with an AxioCam Mrm camera and processed using ZEN 2012 software (Carl Zeiss, North Ryde, Australia).

For scanning electron microscopy (SEM), samples were analyzed at 10 kV at Adelaide Microscopy (University of Adelaide) using a Quanta 450 Field Emission Gun Environmental SEM (FEI, Hillsboro, United States) for 24 h-germinated grain or using an XL-30 FEGSEM (Philips, Amsterdam) for 96 h-germinated grain. For environmental SEM at 24 h, grains were cut in half longitudinally and examined directly. For SEM at 96 h, grains were fixed for 16 h as above, washed in PBS/sucrose solution and dehydrated in an ascending ethanol series from 70 to 100% (v/v). Samples were dried using hexamethyldisilazane, mounted on aluminum stubs, and sputter coated with 5 nm platinum prior to examination.

Samples of scutellum or embryo were frozen in liquid nitrogen, ground to a powder and lyophilized. To approximately 20 mg of each powdered sample was added 900 μL 1% (w/v) SDS containing 5 mM (w/v) DTT. Samples were mixed and centrifuged (1 min, 10,000 g), the supernatant was removed and the SDS/DTT extraction process was repeated twice. The insoluble material was resuspended in 1 mL 70% ethanol, mixed, centrifuged and the supernatant was removed. This treatment was repeated twice. The powder was subsequently resuspended in 500 μL 20 mM Na₂HPO₄/NaH₂PO₄ buffer (pH 6.5) and 12.5 U lichenase (Megazyme, Ireland) was added for 90 min with mixing at 50°C to hydrolyze (1,3;1,4)-β-glucan. Samples were centrifuged, supernatants were removed and the insoluble material was washed in water three times and lyophilized.

The dried samples were resuspended in 100 μL 0.5 M NaOH, to which were added 800 μL water, 50 μL 1 M HCL, 50 μL of 200 mM NaOAc-HOAc buffer (pH 4.0), 1.4 U isoamylase (Megazyme, E-ISAMY, 1,000 U/mL) and 1.4 U pullulanase M2 (Megazyme, E-PULBL, 900 U/mL from Bacillus licheniformis). The mixture was incubated for 90 min at 40°C, neutralized with 60 μL 1 M Tris-HCl buffer (pH 8.5), heated to 100°C and centrifuged at 5,000 g for 1 min. The supernatants were collected, diluted fivefold with water and their chain length distributions were measured by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) as described in Botticella et al. (2018).

Western blot detection of SSI and SBEI was performed essentially as described and using the same antibodies as Cuesta-Seijo et al. (2016), with an HRP-coupled goat anti-rabbit secondary antibody (Invitrogen) and Pierce ECL Plus Western Blotting Substrate (Thermo Fisher Scientific). Extracts from equal amounts of flour (approximately 100 μg) were loaded per lane.

For the developing grain, high quality RNA with an average integrity number of more than 7.0 was isolated from each tissue at each time point. The complete dataset of 81,280 gene sequences for developing grain, expressed as TPM, is shown in Supplementary Table S2. RNA-seq transcript data for germinated grain was from the same dataset described by Betts et al. (2020), in which transcripts from 33,421 genes were quantitated from the scutellum, the residual embryo and three aleurone sections (the proximal al1 section, the central al2 section and the distal al3 section) for 0–4 days after the initiation of germination.

PCA of all genes from developing grain (Figure 2A) and germinated grain (Figure 2B) show that in each case samples are separated by tissue (principal component 1 PC1, x-axis) and time (PC2, y-axis). In developing grain, transcripts of the embryo/endosperm tissues are clearly separated from the endosperm/maternal tissues, and both also separate with time. In germinated grain, biological replicates of scutellum- and embryo-transcribed genes clustered tightly, while gene transcripts in the three aleurone tissues separated more widely with time. It should be noted that Betts et al. (2017) performed multi-dimensional scaling analyses on a similar dataset from germinated grain and also showed that samples were separated predominantly by tissue and time, respectively.

The HORVU gene identifiers (Mascher et al., 2017) for the most abundantly transcribed genes involved in starch synthesis and degradation are listed in Table 1. Starch genes were annotated using the Morex 2017 genome assembly (Mascher et al., 2017), based on comparisons with known sequences from the 2012 scaffold (International Barley Genome Sequencing Consortium [IBGS] et al., 2012), the NCBI database, and the literature (Frandsen et al., 2000; Radchuk et al., 2009; Mangelsen et al., 2011; Ma et al., 2014, 2016; Aubert et al., 2018). Where several, nearby gene models contained identical or near-identical sequences, all gene model (HORVU) identifiers are provided for completeness. A summary of updated annotations for genes encoding enzymes involved in starch synthesis and degradation is presented in Supplementary Table S1.

The heat map in Figure 3A shows the most highly transcribed genes during grain development and allows the identification of important individual genes and their expression patterns over the developmental period.

The first enzyme in committing glucose to starch synthesis is AGP, which consists of two large (L) and two small (S) subunits (Stark et al., 1992). Of the four AGP subunit genes, the AGP-L1 gene is most highly transcribed in the endosperm, where a TPM value of 4608 is observed at 10 DPA (Table 1). This value decreases steadily to 166 TPM at 38 DPA (Figure 3A and Supplementary Table S2). In the embryo, TPM values for this gene stay approximately constant in the range 195–489. Transcript levels for the AGP-S2 gene peak at 896 TPM at 10 DPA and subsequently remain approximately constant in the range of 136–309 TPM in both the embryo and the endosperm for the duration of grain development. Much lower levels of transcripts for the AGP-L2 and AGP-S1 subunit genes are detected, mainly at the early time points (Supplementary Table S2). We could detect no clear correlations between the abundances of transcripts that encode the two large and two small subunits of the enzyme.

Quite distinct transcript profiles are observed for the starch synthase genes. Transcripts for the SS1 gene, after peaking at 828 TPM at 10 DPA, are found in approximately equal abundance (range of 89–244 TPM) in both the endosperm and embryo tissues. In contrast, SS2a (TPM peak of 654 at 10 DPA decreasing to low levels at 38 DPA), SS3a (234 TPM at 10 DPA down to very low levels at 38 DPA), and GBSS1a (816 TPM at 10 DPA, decreasing to very low levels at 38 DPA) transcripts are found almost exclusively in the endosperm/maternal tissues. Finally, the SS2c (5–17 TPM), SS3b (23–70 TPM), and GBSS1b (126–244) transcripts are found mostly in the embryo (Figure 3A and Supplementary Table S2).

Transcripts of the starch branching enzyme (SBE1) are detected mainly in the endosperm (peak of 2747 TPM at 14 DPA, down to 462 TPM at 38 DPA). The SBE2a (peaking at 1382 TPM at 10 DPA in the range of about 100–400 TPM throughout) transcripts are found in both the embryo and endosperm, while transcripts of SBE2b (peaking at 1599 TPM at 10 DPA and decreasing to very low levels at 38 DPA) genes are transcribed mostly in the endosperm (Supplementary Table S2).

Transcripts for genes encoding the debranching enzymes ISA1 and ISA3 are detected in the developing grain (Table 1) at peak levels of 300 TPM (10 DPA) and 127 TPM (18 DPA), respectively. The ISA1 transcripts are found in both the embryo and endosperm, while ISA3 transcripts are found mostly in the embryo (Figure 3A and Supplementary Table S2). Transcripts for the single LD gene peaked at 78 TPM at 14 DPA in the endosperm and thereafter decreased (Supplementary Table S2). Although phosphorylation levels in cereal starches are generally low, transcripts of the α-glucan water dikinase gene (GWD1) were detected in the embryo of developing grain (Table 1).

In contrast to the high levels of transcripts for many starch synthetic enzymes, transcript levels for the major starch degrading enzymes are relatively low in the developing grain, as expected. Of the Amy genes, only transcripts for Amy4_2 are detectable and these are predominantly in the embryo tissue extracts (Supplementary Table S2). However, very high levels of transcripts for the Bmy1 are detected in the endosperm of developing grain, with a maximum of 8394 TPM at 18 DPA (Table 1). Transcripts for this gene are maintained at high levels from 14 DPA (5,667 TPM) until 34 DPA (5,336 TPM), and decrease slightly by 38 DPA (3,693 TPM). The high levels of transcription of this gene are also evident in the heat map shown in Figure 3B. With the exception of the AGL3 gene, which is transcribed mainly in the embryo and maternal tissues at levels of 382 TPM at 18 DPA decreasing to 59 TPM at 38 DPA, genes encoding other enzymes involved in starch degradation are expressed at very low levels in both tissue preparations from the developing grain (Figure 3B and Supplementary Table S2).

The transcription patterns of the genes that mediate starch degradation in cv. Navigator are shown diagrammatically in the heat map of Figure 4A, where the different patterns of the various transcripts can be compared, together with the differences between the aleurone and scutellum tissues, and the wave of transcription from the al1 to the al3 tissues for specific Amy, LD, and AGL genes. The large and small starch granules of starchy endosperm cells show no obvious degradation at 24 h (Figure 4B), but at 96 h the small granules have disappeared and the large granules show extensive surface pitting and hollowing out of their cores (Figure 4C).

The most abundant transcripts in the germinated barley grain are those of the Amy1_2 gene (Betts et al., 2020; Figure 4A). Transcript levels in the aleurone al1 cells increase from 0 to 19,960 TPM between 0 and 72 h after the initiation of germination and decrease at 96 h (Table 1 and Supplementary Table S2). The transcripts also increase to high levels in the aleurone al2 tissue, but with an approximate 24 h lag behind the al1 values. There is a further lag before the transcripts increase in the al3 tissue (Figure 4A). The Amy1_2 gene is also transcribed in the scutellum, where transcripts peak at 3,099 TPM at 72 h (Figure 4A and Table 1).

The Amy2_3 gene is transcribed at high levels in the aleurone, with a peak 3,526 TPM in al1 at 72 h; relatively low levels of transcripts for this gene are found in the scutellum (Figure 4A and Table 1). Other α-amylase gene transcripts include Amy1_1a, which peak in al1 at 72 h, and are also found in the scutellum at relatively high levels. Lower levels of transcripts for the Amy2_1 gene (peak at 96 h in al1), the Amy3 gene (peak at 48 h in al1), and the Amy1_1d gene (peak of 515 TPM at 72 h in al1) are present, predominantly in the aleurone (Figure 4A and Supplementary Table S2).

Transcripts for Bmy1 are found right along the aleurone layer, with levels of 610 TPM at 0 h in the al3 tissue (Figure 4A and Table 1); these levels decrease with time (Figure 3A). Transcripts for the Bmy3 gene are present in the scutellum at a level of 460 TPM at 0 h and also decrease with time (Table 1 and Figure 4A). Of the other genes that encode starch-degrading enzymes, the LD gene is the most highly transcribed, with levels in al1 increasing to a peak of 6,408 TPM at 72 h before dropping back at 96 h (Table 1 and Figure 4A). Transcripts for this gene are also detected at 72 h in the scutellum, but at much lower levels. The AGL1 gene is transcribed at relatively high levels in both the aleurone and scutellum, with values peaking in al1 tissue at 2,201 TPM at 48 h and at 1,260 TPM in the scutellum at 24 h (Table 1). Following a lag of about 24 h, similar expression patterns of this gene are observed in the al2 and al3 tissue (Figure 4A).

Transcript abundances of important genes for starch synthesis in the germinated grain are compared in the heat map shown in Figure 5, where it is evident that these transcripts are found predominantly in the scutellum; their TPM values are much lower than in the developing grain (Table 1; Betts et al., 2020). Transcripts of several genes involved in starch synthesis are detected in the germinated grain, particularly in the scutellum; note that transcripts in the remaining embryo (without the scutellum; Betts et al., 2020) of the germinated grain are not considered here. SS1 gene transcripts increase in the scutellum to a peak of 316 TPM at 96 h (Figure 5 and Table 1). Transcripts for this SS1 gene are also detected in the aleurone during the 96 h period, but levels are less than about 100 TPM throughout. Other genes encoding starch biosynthetic enzymes are found mainly in the scutellum, but also at lower levels in the aleurone. These include transcripts from the AGP-L1 gene (471 TPM in the scutellum at 96 h), the AGP-S2 gene (218 TPM at 0 h in the scutellum), the SBE2a gene (306 TPM at 96 h in the scutellum), the ISA3 gene (85 TPM at 0 h rising to 420 TPM at 96 h), the GBSS1b gene (119 TPM at 96 h in the scutellum), and the SS4 gene (119 TPM at 96 h in the scutellum) (Figure 5 and Supplementary Table S2).

To check whether full length enzymes encoded by the starch biosynthesis genes mentioned above were actually present in the scutellum, extracts of the scutellum, the embryo and the rest of the germinated grain were subjected to western hybridization analyses (Supplementary Figure S1). Antibodies against the SS1 and SBE1 were available (Cuesta-Seijo et al., 2016) for these analyses and showed that proteins of the expected molecular mass of 68 and 102 kDa, respectively, were present in the embryo and scutellum tissue extracts. Extensive degradation of the enzymes, which might be expected if they originated from the developing grain, was not evident. The apparent increase of both enzymes between 24 and 96 h is consistent with de novo synthesis in the germinated grain (Supplementary Figure S1).

The scutellum cells were also examined for the presence of starch granules, using Lugol’s iodine stain (Figure 6). Some starch granules are visible in mature grain at 0 h, particularly in the parenchyma cells of the scutellum (Figures 6A,C), but become much more abundant at 96 h, when they are again more abundant in the parenchyma cells of the scutellum, compared with the scutellum epithelial layer (Figures 6B,D). The starch granules in the scutellum are relatively small, generally about 2–3 μm in diameter.

Given the differences in transcript levels of genes involved in starch synthesis in the developing and germinated grain (such as SBE, GBSS, and ISA; Figure 3A cf. Figure 5) and the observation that different combinations of enzymes can lead to different amylopectin structures (Smith, 2001; Zhong et al., 2020), we investigated the internal chain lengths of this component of the starch granule. Starch was extracted and purified from the scutellum, the embryo and the starchy endosperm of germinated grain and the chain length distributions of the starches were determined following pullulanase/isoamylase hydrolysis and HPAEC-PAD (Botticella et al., 2018). The differences were generally small, although the endosperm starch appeared to have slightly more chains of DP 10–20 than the embryo/scutellum starches, while the scutellum starch at 24 h after the initiation of germination appeared to have more chains in the DP 7–10 range (Supplementary Figure S2). In view of the small differences observed, the fine structures of these amylopectins were not studied further.

To illustrate more clearly the major findings shown as heat maps in Figures 4, 5, we have extracted TPM data for highly transcribed genes that mediate starch biosynthesis and degradation in germinated grain from Betts et al. (2020) and plotted their tissue and temporal transcription patterns in quantitative terms. Thus, Figure 7 shows the spatiotemporal development of gene transcripts encoding starch degrading enzymes in germinated grain, where Amy1_2, LD and AGL1 transcripts peak in abundance at 72 h after the initiation of germination. The progression of transcription from al1 (proximal region of the aleurone) to al3 (distal region of the aleurone) is apparent, although there were also relatively high levels of the AGL1 transcripts in the scutellum from 0 to 48 h after the initiation of germination. Transcripts of the Bmy1 gene decreased with time (Figure 7) and are likely, in large part, to reflect residual transcripts from the developing grain, where Bmy1 transcripts are exceptionally high (Table 1).

The micrographs presented in Figure 6 indicate that starch is synthesized in the scutellum of the germinated grain. Consistent with this observation, highly transcribed genes involved in starch synthesis (SBE2a, ISA3, GBSS1b, and SS1) are found predominantly in the scutellum of germinated grain (Figure 8). In each case, transcript abundances decrease in the first 24 h after imbibition, but thereafter increase. The transcripts observed at 0 h presumably represent residual mRNA fragments from the developing grain. It should be noted that the ISA3 enzyme could be involved in either or both starch synthesis and degradation in the germinated grain (Figure 1).

To detect any temporal differences between the transcription of genes that encode hydrolytic enzymes involved in cell wall, starch and storage protein degradation in the germinated grain, transcript levels of major genes that mediate these processes were investigated (Figure 9). The gene encoding (1,3;1,4)-β-glucanase isoenzyme E1 (HORVU1Hr1G057680) was chosen to represent cell wall degradation, cysteine endoprotease B (HORVU3Hr1G091800) to represent storage protein degradation and Amy1_2 to represent starch degradation. High transcript levels for these genes were observed and in each case they reached a maximum at 4 days after imbibition (Figure 9).