Metabolic and transcriptional transitions in barley glumes reveal a role as transitory resource buffers during endosperm filling

Highlight The development and metabolism of barley glumes is tightly associated with grain filling and filial sink strength, which coordinate developmental phase changes in the glumes via metabolic, hormonal, and transcriptional control.


Introduction
During seed filling carbon (C) and nitrogen (N) compounds are remobilized from vegetative organs and transported to the seeds. Vegetative organs in barley and wheat are photosynthetically active, providing carbohydrates until late grain filling. By contrast, 60-90% of grain N originates from remobilization out of vegetative organs (Hirel et al., 2007). Among the photosynthetically active tissues, glumes hold an exceptional role, as the organs nearest the grains. In wheat, glumes have a unique cellular and chloroplast distribution associated with their particular metabolism and supporting grain maturation (Waters et al., 1980;Lopes et al., 2006). Up to 30% of photosynthates imported into grains derive from photosynthetic activity of glumes (Grundbacher, 1963).
Most N derived from other source tissues is not directly transported into developing grains, but instead transported through and eventually converted within glumes (Waters et al., 1980;Simpson et al., 1983).
Compared to flag leaves, glumes senesce late and thus could be important for N translocation during later grain filling. Thereby, the capacity of glumes to convert and translocate N during senescence is an important trait to assess in breeding for higher grain protein content.
To fulfil these tasks, glume metabolism must be coordinated with the different phases of grain development. In barley endosperm, cellularization begins around 4 days after pollination (DAP) and is completed 1-2 days later. The prestorage/cellularization phase, from anthesis to 6 DAP, and the storage phase, starting at 8-10 DAP, are separated by a transition stage characterized by transcriptional reprogramming and a switch into storage mode (Sreenivasulu et al., 2004). Between 8 and 10 DAP, the endosperm starts accumulating storage products, develops high sink strength for sucrose and N, and enters the linear phase of dry matter accumulation between 10 and 20 DAP (Weschke et al., 2000). Physiological maturity is reached around 24 DAP followed by desiccation.
Endosperm phase changes are accompanied by differences in sink strength, which affect metabolism, remobilization, and transport of resources into and from vegetative organs. As would be expected, central metabolism in glumes is coordinated with such changes during grain filling (Lopes et al., 2006). There is a lack of detailed knowledge about this cross-talk at the molecular level, and of metabolic and transcriptional adjustments according to the specific demands of grains. The remobilization of assimilates and reserves during seed filling is highly regulated (Watanabe et al., 2013). The WRKY (contains the WRKY amino acid signature at the N-terminus and zinc-finger structure at the C-terminus) and NAC (NAM, ATAF1,2, CUC) transcription factors are involved in regulating remobilization and senescence (Balazadeh et al., 2010;Breeze et al., 2011;Fischer, 2012). In barley, specific members of the NAC transcription factor gene family are co-regulated with senescence-associated genes in senescing flag leaves (Gregersen and Holm, 2007;Christiansen and Gregersen, 2014). In wheat the NAC transcription factor Gpc-B1 accelerates senescence and increases nutrient remobilization from leaves (Uauy et al., 2006).
Remobilized N from protein degradation has to be transported across membranes by specific transporters (Tegeder and Rentsch, 2010). Members of the amino acid transporter family (ATF) and nitrate/peptide transporter family (NPF) are key components in remobilization, and functionally characterized transporters are expressed in tissue and development specific-manners. AtAAP1 is involved in amino acid uptake into embryos (Hirner et al., 1998;Sanders et al., 2009), AtAAP8 is involved in amino acid uptake into endosperm (Schmidt et al., 2007), and AtPTR5 is preferentially expressed during early seed development (Komarova et al., 2008). Barley HvPTR1 transports peptides from endosperm to growing embryos during germination (West et al., 1998). Transporters involved in amino acid uptake into cells have mainly been characterized (Tegeder, 2012), but recently AtBAT1 (Dündar and Bush, 2009) and AtSIAR1 (Ladwig et al., 2012) have been shown to export amino acids out of cells with apparently opposing functions. While AtBAT1 shows preferential expression in sink tissues (Dündar, 2009), AtSIAR1 expression is associated with source tissues, and Arabidopsis mutants have lower contents and disturbed homeostasis of amino acids in siliques (Ladwig et al., 2012).
The aim of this study was to analyse temporal changes of transcript and metabolite abundances in glumes and endosperm during barley grain development. Such parallel profiling allows a correlation of shifts in glume metabolism and remobilization events with distinct phases of grain development. Furthermore, possible signals and transporters involved in coordinating metabolism and N translocation between glumes and endosperm are presented and discussed.

Plant growth and harvest
Barley (Hordeum vulgare L. cv. Barke) was grown in greenhouses with 16 h light/8 h dark. Stages of grain development were determined as described previously (Weschke et al., 2000). Glumes and endosperm tissue were collected between 10am and 12pm in 2-or 4-day intervals starting at anthesis (glumes) and 4 DAP (endosperm) until 24 DAP. Endosperm was manually separated from pericarp between 4 and 14 DAP, and whole caryopses were sampled between 16 and 24 DAP.

RNA isolation, labelling, and array hybridization
Glume and endosperm material for three biological replicates was harvested from 0 (only glumes), 4, 8, 10, 14, 18, and 24 (glumes and endosperm) DAP; and total RNA was extracted with a Spectrum™ Plant Total RNA Kit (Sigma Aldrich, Steinheim, Germany). RNA integrity was confirmed using the Bioanalyser system (Agilent Technologies). 100 ng RNA was used for cRNA synthesis and Cy3-labelling with a Low Input Quick Amp Labelling Kit (Agilent Technologies). Labelling efficiency, and amount and quality of cRNA, were assured using an ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, USA) and Bioanalyser system. 600 ng labelled cRNA was used for fragmentation and array loading (Gene Expression Hybridization Kit, Agilent Technologies). Hybridization was done for 17 h at 65°C. After washing (Gene Expression Wash Buffer Kit, Agilent Technologies) and drying, arrays were scanned at 5 µm resolution using an Agilent Technologies Scanner G2505C. Resulting images were evaluated (determination of spot intensities, background correction) with Feature Extraction V11.5 (Agilent Technologies).

Data evaluation
Data evaluation was done with Genespring V12.5 (Agilent Technologies). Values were log 2 transformed and quantile normalized, before relative expression values were calculated by subtracting the median expression of each probe from the other values of this specific probe (baseline transformation). After removing outliers and transcripts without significant expression at any time point, ANOVA (P ≤ 0.005, FC ≥ 3) and FDR correction (Benjamini-Hochberg) was performed. These stringent parameters were chosen in order to identify important transcripts without (unnecessarily) expanding the data set.

Preparation of vascular tissue
Seeds were harvested (eight biological replicates) at 4, 8, 10, 14, 18, and 24 DAP according to Thiel et al. (2009). Vascular tissues were micro-dissected from the middle of the grain with 30 µm thickness per section (Supplementary Figure S1).

UPLC measurements
Extraction and measurement are described in Thiel et al. (2009), with the following changes: 10 mg of dried material was used; before extraction, 5 µl Norvalin (5mM) were added; 100 µl aliquots of extract were concentrated under vacuum and dissolved in 200 µl water.

C/N and starch measurements
Total C and N were determined from dried and ground material using a Vario EL Elementar analyser (Elementar Analysensysteme GmbH, Hanau, Germany). Starch was measured as described (Weschke et al., 2000).

Growth parameters, starch and N content
Growth parameters were analysed for glumes between 0 and 24 DAP and for endosperm from 4 to 24 DAP. Glume dry weight increased steadily by 40% from 0 to 8 DAP followed by a transient decrease of 20% between 8 and 10 DAP. Thereafter, dry weight rose slightly, remained constant until 20 DAP and declined thereafter (Fig. 1A). Total N in glumes increased until 8 DAP, followed by a sharp decline at 8 DAP without further changes until 20 DAP and a slight decrease thereafter (Fig. 1B). Starch content in glumes was generally low compared to endosperm (Supplementary Table S1) and decreased by 65% between 0 and 10 DAP, before increasing by 15% from 18 to 24 DAP (Fig. 1C). Endosperm dry weight and starch content rose at ~8 DAP followed by linear accumulation during the main storage phase (10 to 20 DAP), levelling off afterwards (Fig. 1A, C). Total N increased linearly from 8 DAP until around 20 DAP (Fig. 1B).
The results show that from anthesis glumes accumulated dry weight and N until approximately 8 DAP, followed by a considerable decrease, coinciding with the start of starch and dry weight accumulation in the endosperm.

Comparative gene expression analysis in glumes and endosperm
Comparative transcript analysis was performed in glumes and endosperm to analyse changes in central metabolic pathways, remobilization, and transport processes, as well as putative regulatory elements. Labelled cRNA from glume (0-24 DAP) and endosperm (4-24 DAP) fractions were hybridized to Agilent microarrays. In endosperm and glumes, 8998 and 3999 transcripts were identified as differentially expressed (significant differences between at least two stages; Supplementary Tables S2,  S3). General profiles for both tissues are similar, revealing three distinct phases: (i) high differential expression between 0 and 8-10 DAP in glumes and endosperm; (ii) low differential expression between 8-10 and 14 DAP for glumes and endosperm; (iii) high differential expression after 14 DAP ( Fig. 2A, B).

Central carbohydrate and N metabolism
In glumes, gene expression related to glycolysis (e.g. Glc-6-P epimerase, enolase, cytosolic/plastidic pyruvate kinase, and phosphoglycerate mutase) decreased steeply from 0 to 8 DAP and slightly thereafter. This was similar to the expression of the main starch metabolism genes, such as sucrose synthase, various starch synthases, and ADP-Glc pyrophosphorylase. By contrast, expression of seven genes related to glycolysis and seven associated with starch biosynthesis strongly increased in endosperm from 4 to 8 DAP, remained constantly high up to 14 DAP, and declined thereafter (Fig. 3).
In glumes, nine genes associated with amino acid biosynthesis were most highly expressed at 0 and 4 DAP, followed by decreasing expression. Four are involved in the aspartate pathway towards lysine, methionine, and threonine biosynthesis; two others are involved in arginine biosynthesis. By contrast, endosperm expression of nine genes related to amino acid biosynthesis increased at 8 DAP and decreased after 18 DAP (Fig. 3). Endosperm expression of storage protein genes increased strongly from 4 to 8 DAP, and remained at a high level until 24 DAP (Fig. 3).
Transcript analysis revealed opposing trends for certain metabolic pathways such as glycolysis, and starch and amino acid synthesis, these being downregulated in glumes but upregulated in the endosperm during grain filling. TCA cycle and mETC-activities were strongly upregulated in glumes at 8 DAP, the beginning of grain filling.

Carbohydrate and N transporters
Remobilization of reserves from glumes and accumulation in the endosperm depends on efficient transport from sink to source. In glumes, expression of several carbohydrate transporter genes increased at 8 DAP and then further until 24 DAP. This involved hexose/sugar transporters and members of the SWEET family, which potentially export sugars from Arabidopsis leaves (Chen et al., 2012). In the endosperm, solute transporters related to storage product synthesis (e.g. HvSUT1, plastidic translocators for ADP-Glc and phosphoenol pyruvate) were upregulated at 8 DAP with decreasing expression levels towards 24 DAP (Fig. 4).
HvAAP2 was upregulated between 0 and 8 DAP and is homologous to AtAAP2, which is involved in xylem to phloem transfer and is important for sink N supply (Zhang et al., 2010). HvAAP7 and HvAAP11 were upregulated from 14 and 18 DAP and are possibly involved in senescencerelated remobilization. The second group contained members from other ATF subgroups like HvLHT1, specifically expressed in glumes and probably important for amino acid re-translocation (Kohl et al., 2012), HvCAT1, HvANT4, and HvGAT1. Barley homologues of AtBAT1 (Dündar and Bush, 2009) and AtSIAR1 (Ladwig et al., 2012) showed opposing expression profiles. HvBAT3 was upregulated between 0 and 14 DAP followed by downregulation, while HvSIAR1-like was downregulated between 0 and 10 DAP followed by strong upregulation.
Bias towards upregulation is less pronounced for putative nitrate/peptide transporters, where 13 from 30 candidates showed decreasing expression (Supplementary Table  S4). Within upregulated candidates, three major profiles present in the ATF transporters could be observed (Fig. 4). Endosperm-expressed N transporters probably facilitate N import or distribution. 80% and >73% of ATF and NPF transporters, respectively, were upregulated during development (Supplementary Table S5). Major patterns showed upregulation until 14 DAP, transient upregulation at 8 or 10 DAP, and constant upregulation after 8 DAP (Fig. 4).
Among transiently upregulated transcripts, HvAAP3 is closely related to AtAAP8 and AtAAP1, importing amino acids into seeds (Schmidt et al., 2007;Sanders et al., 2009). Expression of a BAT-like transcript was also steadily increasing during development, and is possibly involved in phloem unloading (Dündar, 2009), while increasing expression of OsPUT1-like after 8 DAP indicates polyamine import into grains (Mulangi et al., 2012).

Transcriptional transitions in glume metabolism during grain filling
Photosynthesis-associated transcripts in glumes were highly expressed at 0-4 DAP, with decreasing levels after 8 DAP (Fig. 5A). Chlorophyll is degraded during leaf senescence by a pathway involving pheophorbide a oxygenase (PAO) (Hörtensteiner, 2006). In glumes, HvPAO was upregulated from 14 DAP onwards (Fig. 5A), indicating chlorophyll degradation only at late seed filling. Protein degradation is a prerequisite for N remobilization, and several proteases were transcriptionally upregulated in glumes at two distinct phases, between 0 and 8 DAP and from 14 DAP onwards. This involved several serine-, aspartyl-, and cysteinelike proteinases including a homologue to Arabidopsis senescence-associated gene-12 (SAG12), which is specifically activated by developmentally controlled senescence but is not stress-or hormone-controlled (Noh and Amasino, 1999).
Whereas several amino acid biosynthesis genes were downregulated (Fig. 3), others involved in different parts of amino acid metabolism (e.g. homogentisate 1,2-dioxygenase and pyrroline-5-carboxylate reductase) were upregulated in glumes and may participate in aromatic amino acid and proline degradation. Alanine aminotransferase and glutamine synthetase-1 (GS1) could be involved in glutamine biosynthesis for export (Thiel et al., 2009). Thus, expression patterns indicate interconversion and/or degradation of certain amino acid species in glumes during later grain filling. Transcripts of glutamine-dependent asparagine synthase (Gln-ASN) increased by 150-fold between 14 and 24 DAP, indicating an important role for remobilization (Fig. 5).
Potential metal transporters, like members of natural resistance-associated macrophage proteins (NRAMP), zinc transporters, yellow stripe-like (YSL), and oligopeptide transporters (OPT) were transcriptionally upregulated at 8 DAP and/or at 24 DAP. Specific genes of lipid biosynthesis/ degradation were differently expressed during glume development. Two members degrading phospholipids, phosphatidylcholine phospholipase D, and glycero-P-diesterase were strongly upregulated between 0 and 8/10 DAP. Candidates involved in degradation of mono-and triacylglycerides, like enoyl-CoA hydratase, palmitoyl protein thioesterase, and mono-and triacylglycerol lipase were upregulated after 14 DAP. Enzymes synthesizing long-chain fatty acids, acylactivating enzyme, very-long-chain fatty acid-condensing enzyme, and 3-ketoacyl-acyl carrier protein were most highly expressed at 18 and 24 DAP.
Genes involved in the phenylpropanoid/lignin pathway were upregulated in glumes from 8 to 24 DAP, including polyphenol oxidase, flavonoid 7-O-methyltransferase, and hydroxycinnamoyl-CoA shikimate/quinate hydroxycinnamoyl transferase. Two genes encoding laccases were constantly upregulated during development. Laccases catalyse lignin polymerization from the precursors coniferyl-and sinapyl-alcohol, and respective knockouts in Arabidopsis drastically reduce the lignin content (Zhao et al., 2013), indicating an important role for lignification and secondary cell wall thickening in mechanical support and water transport (Zhao et al., 2013).
Whereas mitochondrial activity was transcriptionally activated at 8 DAP (Fig. 3), specific genes encoding TCAcycle and mETC-enzymes, such as pyruvate and NADH dehydrogenases, electrontransfer flavoprotein, and alternative oxidase, were upregulated only after 14 DAP.
Results from transcript profiling suggest metabolic transitions in glumes from sink to source in accordance to grain filling, namely downregulated photosynthesis and upregulated proteolysis, lipid and phenylpropanoid metabolism, and mitochondrial activities and metal transport.
Nineteen out of 22 WRKY transcription factors showed profiles according to (i) and (ii). The latter includes HvWRKY12, which is probably involved in age-dependent senescence rather than N remobilization (Hollmann et al., 2014).
JA biosynthesis depends on the subsequent action of lipoxygenase (LOX), allene oxide synthase and cyclase, 12-oxophytodienoate reductase (OPR), followed by beta oxidation (Lyons et al., 2013). In glumes, continuously upregulated genes encode specific LOX isoforms and enzymes involved in late steps of JA biosynthesis (OPRs) and a homologue to AtOPLC1, involved in beta oxidation (Koo et al., 2006).

Metabolite contents in developing glumes and endosperm
Metabolite levels were measured in glumes and endosperm (Fig. 6, Supplementary Table S6), while free amino acids were analysed by UPLC (Supplementary Table S7). Levels of sucrose, maltose, raffinose, xylulose, rhamnose, and fucose increased whereas xylose, arabinose, and trehalose decreased in glumes with progressing development. Several sugars are involved in cell wall biosynthesis, indicating alterations in cell wall dynamics in glumes. In endosperm, hexoses were highest during early development whereas sucrose peaked at 8 DAP (Weschke et al., 2000).
In glumes, almost all metabolites within the glycolytic, pentose phosphate, and shikimate pathways, like hexoses and their phosphates, 6-phospho-gluconate and 3-phospho-glycerate, decreased whereas endosperm levels were highest at 8 and 10 DAP. Inositol and inositol-1-phosphate and sugar alcohols, except glycerol, decreased during glume development.

Free amino acid concentrations in glumes, endosperm, and vascular tissue
In glumes between pollination and 8 DAP, the summarized concentrations of all free amino acids increased from 11 to 32 µmol g -1 fresh weight, followed by a drastic decrease between 8 and 10 DAP (~50%; Fig. 7A; Supplementary  Table S7). Thereafter, levels fluctuated at around 20 µmol g -1 . Most amino acids behaved similarly, showing distinct declines between 8 and 10 DAP, except Asp and Glu. Asp was generally low without larger changes, whereas Glu increased throughout. Ser was always higher by 10-to 15-fold compared to Gly. Asn was 2 to 3-fold higher than Gln between 2 and 10 DAP but not different later on.
In the endosperm, free amino acid levels decreased from 80 µmol g -1 fresh weight at 4 DAP to 25 µmol g -1 at 14 DAP, showing the largest decline between 8 and 10 DAP (Fig. 7B). Between 16 and 22 DAP levels remained constant, before decreasing at 24 DAP. While Ala, Gln, Glu, and Val followed this pattern, Arg, Asn, Pro, Ser, and Thr accumulated between 4 and 6 DAP, before their levels declined.
Concentrations of Gln and Pro most strongly declined (by ~80%) between 4 and 12 DAP. In contrast to Ser, Gly increased from 8 to 12 DAP, marking a switch from high to low Ser:Gly ratios during grain filling. Asp largely did not change. The results indicate a sudden decrease of most free amino acids (especially of Pro and Asn) in glumes at 8 DAP. This coincides with extensive use for storage protein synthesis in the endosperm.
As expected, high endosperm demand for amino acids due to storage activity is transmitted to the glumes, initiating remobilization. To analyse possible metabolic signals, GC-MS-based metabolic profiling was performed on microdissected regions comprising the main vascular bundles of grains between 4 and 24 DAP (Supplementary Figure  S1). Thirty-two unambiguous metabolites were detected (Supplementary Table S8). Temporal profiles of 12 amino acids were compared between glumes, endosperm, and vascular regions (Fig. 8). Amino acid profiles were highly correlated (Pearson correlation) between vasculature and endosperm with C vc, en between 0.98 and 0.81 for Gln, Ser, Glu, Thr, Pro, Ala, Val, and Asn. Less positive (C vc, en <0.76) or even negative correlation (Ala, Glu) occurred between vasculature and glumes. This indicates that endosperm amino acid demand is propagated via the vasculature. In beech, increased Gln and Asp levels in the phloem lead to reduced root NO 3 uptake (Geßler et al., 1998) and amino acid feeding negatively regulates expression of high-affinity NO 3 --uptake transporters in barley roots (Vidmar et al., 2000). In a reciprocal manner, depletion of certain amino acids such as Gln (C vc, en = 0.98) within the vasculature could signal and communicate endosperm demand to vegetative organs.

Discussion
Glumes are the vegetative organs closest to grains and are important for converting and translocating assimilates to them (Waters et al., 1980;Simpson et al., 1983;Lopes et al., 2006). Phase changes during endosperm development are accompanied by large variations of sink strength that greatly affect metabolism in and assimilate fluxes from glumes. Parallel transcript and metabolite profiling in glumes and endosperm during grain filling showed that glume metabolism was adjusted to the changing demands of the grains, reflected by specific signatures of metabolite and transcript abundances. Obviously, grain filling and filial sink strength coordinate phase changes in glumes via metabolic, hormonal, and transcriptional control.

Glumes are photosynthetically active sinks during the pre-storage phase
Transcript profiling in glumes at 0 and 4 DAP revealed high activity of photosynthesis, glycolysis, and starch and amino acid biosynthesis, but low mitochondrial activity and transport of sugars, peptides and metals. Accordingly, levels of hexoses, their phosphates, glycolytic intermediates, and starch were highest. By contrast, endosperm gene expression was low for central pathways glycolysis, and starch and amino acid biosynthesis, but high for mitochondrial activity before 8 DAP (Figs 3-5). Thus, during the pre-storage phase, glumes are photosynthetically active organs with a high level of biosynthesis, while the endosperm is a sink with high respiratory and mitochondrial activities.
During the pre-storage phase, glumes accumulated dry matter, total N, and free amino acids, with concentrations increasing nearly 3-fold between 0 and 8 DAP (Fig. 7). Likewise, α-amino N content in wheat glumes increases significantly until 5 DAP (Waters et al., 1980). This demonstrates that during the pre-storage phase glumes generate early and intermediary sinks before high endosperm sink strength is established.

Glumes and endosperm display opposed metabolic shifts at the beginning of grain filling
At 8-10 DAP metabolic shifts occurred in glumes, indicated by decreasing dry matter, starch, total N, and most amino acids. At the transcript level this was reflected by downregulated photosynthesis, starch and amino acid biosynthesis, and glycolysis, but upregulated TCA cycle, mETC activities, and transport (Figs 3-5). Endosperm dry weight and starch increased linearly after 8 DAP together with storage-associated/sink-strength-related gene expression such as sucrose and amino acid transporters HvSUT1, HvAAP3, sucrose synthase, ADP-Glc pyrophosphorylase, and hordeins (Figs 3 and  4). Accordingly, levels of sucrose, glycolytic intermediates, Fig. 6. Changes in metabolite levels in barley glumes and endosperm during development. Samples were taken in 2-day steps between 0 and 24 DAP (glumes) and at 4,8,10,14,18,and 24 DAP (endosperm). Levels of proteogenic amino acids and GABA were measured using UPLC with two (Arg, Gly) and three biological replicates, respectively. All other metabolites were measured by GC-MS with six biological replicates per time point. Data were corrected by internal standard and fresh weight, subsequently maximum normalized and colour coded for each metabolite (maximum amount = 100%, values >90% coloured in dark blue, values <10% in dark red), as shown by the insert depicting a constant increase from 0 to 100% between 0 and 24 DAP. Metabolites without significant changes in one of the tissues are marked (1) for glumes and (2) for endosperm. and amino acids were highest at 8-10 DAP in endosperm but decreased in glumes (Fig. 6). To conclude, metabolic shifts implicate opposing trends for central pathways glycolysis, and starch and amino acid biosynthesis, namely downregulation in glumes but upregulation in endosperm at early grain filling.
Glumes undergo transition into remobilizing and exporting organs coinciding with the beginning of storage activity in the endosperm. Phase changes in glumes may be initiated by emerging endosperm sink strength. This is supported by the fact that removal of sink organs generally prevents remobilization and delays senescence. Amino acid concentrations (especially Gln) decrease upon grain filling in wheat flag leaves, but increase in response to ear excision (Peeters and Van Laere, 1994), indicating that grain sink strength induces this drain. Senescence is also delayed by diverse crop manipulations such as inhibiting kernel set in maize and depodding (Miceli et al., 1995;Borrás et al., 2003). Instead, senescence in spinach plants is induced by exhaustive reallocation of nutrients from leaves to flowers (Sklensky and Davies, 2011).

Metabolic transition of glumes occurs in two phases
Gene expression in glumes suggested metabolic transitions at two phases. The first, at around 8 DAP, is consistent with the onset of endosperm storage activity and is probably a consequence of increasing endosperm sink activity. Amino acid profiles in the grain vasculature were highly correlated to those of the endosperm, but differed from glumes (Fig. 8). Although the total amount of free amino acids in the endosperm increased until 18 DAP (data not shown), concentrations of most members decreased during early grain filling. Such depletion might be transmitted via the vasculature and specific amino acids could function as metabolic signals communicating endosperm demand to vegetative organs. Chlorophyll breakdown, regarded as a senescence marker (Hörtensteiner, 2006), did not occur at 8 Fig. 7. Concentrations of free amino acids in (A) barley glumes and (B) endosperm during development. Samples were measured via UPLC in 2-day steps from 0 DAP (glumes) and 4 DAP (endosperm), respectively, before data were corrected by internal standard and fresh weight. Each data point represents two (Arg, Gly) or three biological replicates ±SD.
DAP since pheophorbide a oxygenase (PAO) was transcriptionally activated after 14 DAP. Senescence-associated genes, like cysteine proteinase, homologous to Arabidopsis SAG12 (Guo and Gan, 2005) and an AtSWEET15 homologue (AtSAG29), a potential sugar exporter and senescence-related protein that accelerates senescence when overexpressed in Arabidopsis (Seo et al., 2011), were also not upregulated before 14 DAP. To conclude, developmental senescence in glumes was not initiated before 18 DAP (Fig. 5).
The second phase comprised developmental ageing and senescence denoting later grain filling. At transcript levels it was characterized by a further decrease of photosynthesis, glycolysis, and starch biosynthesis, whereas chlorophyll, lipid, and amino acid degradation increased together with proteolysis, mETC activity, and transport processes. Several genes involved in the final steps of amino acid biosynthesis were steadily downregulated, while others were upregulated, especially at later stages, such as homogentisate 1,2-dioxygenase and pyrroline-5-carboxylate reductase, which are probably engaged in Tyr, Phe, and Pro degradation.
To conclude, upon the first metabolic transition (8 DAP), glumes are converted into remobilizing and transporting organs for assimilates, providing for grain filling. The second transition (18 DAP) assigns developmental ageing and senescence. The period between is the main storage phase. It is obviously important that glumes remain fully functional at this stage. Sequential arrangements reflect the cascades of sink-induced remobilization at 8 DAP and developmental ageing after 14 DAP.
Biosynthesis and signalling of ABA are upregulated during senescence (van der Graaff et al., 2006). HvNCED3 was continuously upregulated and HvAAO4 expression increased from 0 to 8 DAP and after 14 DAP, indicating that late steps of ABA biosynthesis are upregulated in glumes. Accordingly, ABA signalling, HvPKABA1, and HvABF1 were bi-phasically upregulated, supporting ABA functions in glume phase transition. JA generally accelerates leaf senescence (Ueda et al., 1981), although mechanisms are still unclear. MeJA induces senescence-associated transcripts AtSAG12 (Xiao et al., 2004), AtCLH1/CORI1, and AtERD1/SAG15 (Jung et al., 2007). Enzymes involved in JA biosynthesis, OPRs, lipoxygenases, and the AtOPLC1-homologue, involved in JA-related β-oxidation (Koo et al., 2006), were continuously upregulated in glumes. In contrast to ABA, JA-related transcripts were not bi-phasically upregulated, indicating effects only on age-dependent senescence.  Es). Measurements were taken between 4 and 24 DAP using UPLC for glumes and endosperm, with three biological replicates, and using GC-MS for grain vasculature, with eight biological replicates. Concentrations were normalized and colour coded from dark red (low) to dark blue (high) values (see Fig. 6), before Pearson correlation (C) was determined for Vc and Gl and Vc and Es, respectively. Correlations are shown in orange (negative) and green (positive); statistically significant correlations (P < 0.05) are marked with an asterisk.

Glume transition from sink to source is accompanied by changed expression of N transporters
Switching from import to remobilization/export requires transport/re-translocation of N. HvAAP2, upregulated at 8/10 DAP, is homologous to AtAAP2, and involved in Arabidopsis xylemto-phloem transfer and sink N supply (Zhang et al., 2010). NPF-members hv_38962 (array contig) and hv_38267 were also upregulated at 8/10 DAP and are similar to tonoplast-localized AtPTR2/AtNPF8.3, functioning in flowering and seed development (Song et al., 1997), and to plasma membrane-localized AtPTR5/AtNPF8.2, important for peptide transport in seeds (Komarova et al., 2008). These transporters could potentially establish sink strength and intermediate storage in glumes (Fig. 4).
Putative amino acid exporters HvBAT3 and HvSIAR1like were opposingly expressed (Fig. 4). HvBAT3, downregulated after 10 DAP, is homologous to AtBAT1, putatively involved in sink phloem unloading (Dündar and Bush, 2009). HvSIAR1-like, upregulated after 14 DAP, is homologous to AtSIAR1, and involved in amino acid remobilization and homoeostasis of Arabidopsis leaves (Ladwig et al., 2012). Switching activities of HvBAT3 and HvSIAR1-like could reflect the transition of glumes from sink to source.
HvOPT6, with unique expression among oligopeptide transporters, was downregulated at 4 DAP and upregulated after 14 DAP, and is homologous to glutathione transporter OsGT1 (Zhang et al., 2004), suggesting combined N and sulphur remobilization.

Glume-specific remobilization of assimilates and resources
Gene expression and metabolite profiles indicate glume-specific mechanisms of assimilate conversion and translocation towards grains. Induced L-alanine:2-oxoglutarate aminotransferase (Ala:OG-AT) and GS1 together may convert Ala by Ala:OG-AT to Glu and further to Gln by GS1 using amino groups from protein/amino acid degradation. Gln could then either be exported or converted to Asn by Gln-ASN, one of the most upregulated genes at 24 DAP (150-fold). This would mobilize N as Asn at the expense of Ala and amino N. Gln-ASN is important for remobilizing N during senescence of Medicago truncatula leaves (De Michele et al., 2009). In rice and tobacco, Gln-ASN is located in vascular tissues (Gaufichon et al., 2010). Assuming such a location in glumes suggests that Asn is preferentially synthesized in vascular tissue for export (Fig. 9A). Accordingly, levels of Ala and Glu remain high in glumes during grain filling, whereas Asn decreases 5-fold with similar profiles in glumes, vasculature, and endosperm.
Cystathionine-γ-synthase (CGS), involved in Met biosynthesis (Hacham et al., 2013), was upregulated at 24 DAP in glumes (100-fold) together with serine acetyl transferase (SAT) and cysteine synthase (OAS1), involved in Cys biosynthesis from Ser. Upregulated Cys and Met biosynthesis contributes to possible conversion of amino N from Ser, Asp, and sulphur to phloem-mobile S-methyl-methionine (SMM), (Bourgis et al., 1999). The pathway (Fig. 9B), involves additional enzymes, like aspartate kinase and methionine-S-methyltransferase, which were also upregulated in glumes. Similar mechanisms are suggested for synthesis and transport of SMM within barley nucellar projections to translocate reduced sulphur from senescing tissue into endosperm (Thiel et al., 2009).
Differential transcription indicated degradation of phospholipids by phosphatidylcholine phospholipase D and glycero-P-diesterase, in glumes already at 8 DAP, together with biosynthesis of long-chain fatty acids by acyl-activating enzyme and very-long-chain fatty acid-condensing enzyme. However, degradation of mono-and triacylglycerides (TCGs) did not occur before 18 DAP, indicated by upregulated monoand triacylglycerol lipases. These processes are probably involved in mobilizing C from membrane lipids into phloemmobile sucrose (Kaup et al., 2002;Troncoso-Ponce et al., 2013). To conclude, in glumes, phospholipids from chloroplasts are degraded early, accompanied by de novo biosynthesis of TCGs. Accordingly, levels of stearate and palmitate increased in glumes (Fig. 6). During developmental senescence TCGs are degraded and C is converted to phloem-mobile sucrose or respired by mETC. Alternatively, acetyl-CoA could be converted into amino acids and to SMM for export (Fig. 9B).
In glumes, mitochondrial metabolism was upregulated at early grain filling involving TCA cycle activity, mETC, and ATP synthesis and transport, consistent with increased levels of citrate, isocitrate, and succinate. As expected, most ATP generated in mitochondria is needed to energize transport of sugars, amino acids, and peptides. Accordingly, respective transporters were co-induced (Fig. 4). Similarly, in senescing Arabidopsis leaves, mitochondrial respiration has to supply ATP and C skeletons to redistribute N (Hörtensteiner and Feller, 2002). Several mitochondrial genes were only upregulated later on, involving respiration and mETC (cytochrome c oxidase, electron transfer flavoprotein, NADH dehydrogenase) and energy dissipation (alternative oxidase). These genes are probably involved in developmental senescence and amino acid degradation, also upregulated at the transcript level at 24 DAP. Mitochondrial alternative oxidase could balance senescence-related stress responses from excess degradation of sugars and/or amino acids as shown in legume embryos with perturbed metabolism (Gregersen and Holm, 2007;Weigelt et al., 2008;Weigelt et al., 2009;Araújo et al., 2014).
Our results show that development of barley glumes after anthesis is separated into three phases associated with grain development, and these mark the transition from sink to source tissue. Until 8 DAP, glumes are growing and photosynthetically active tissues accumulating dry weight, total N, and free amino acids. Furthermore, decreasing levels of starch and glycolytic metabolites as well as the corresponding transcripts are observed. Between 8 and 10 DAP, coinciding with the beginning of storage protein synthesis in grains, total N and free amino acids decrease significantly, which probably represents relocation of nutrients to meet the demands of developing grains. Concentrations of free amino acids in endosperm and grain vasculature decrease at this stage, which could signal increasing N demand to glumes and trigger remobilization. Accordingly, expression of specific transporters in glumes is upregulated together with enzymes from the TCA cycle and mETC providing energy for transport. After 18 DAP, glumes undergo developmental ageing and senescence, involving chlorophyll degradation by PAO, specific proteases, and N transporters. Transition between these phases is probably governed by transcription factors from the NAC and WRKY families and influenced by ABA.

Supplementary material
Supplementary data can be found at JXB online. Supplementary