Introduction of glucan synthase into the cytosol in wheat endosperm causes massive maltose accumulation and represses starch synthesis

SUMMARY We expressed a bacterial glucan synthase ( Agrobacterium GlgA ) in the cytosol of developing endosperm cells in wheat grains, to discover whether it could generate a glucan from cytosolic ADP-glucose. Transgenic lines had high glucan synthase activity during grain ﬁlling, but did not accumulate glucan. Instead, grains accumulated very high concentrations of maltose. They had large volumes during development due to high water content, and very shrivelled grains at maturity. Starch synthesis was severely reduced. We propose that cytosolic glucan synthesized by the glucan synthase was immediately hydrolysed to maltose by cytosolic b -amylase(s). Maltose accumulation resulted in a high osmotic potential in developing grain, drawing in excess water that stretched the seed coat and pericarp. Loss of water during grain maturation then led to shrinkage when the grains matured. Maltose accumulation is likely to account for the reduced starch synthesis in transgenic grains, through signalling and toxic effects. Using bioinformatics, we identify an isoform of b -amylase likely to be responsible for maltose accumulation. Removal of this isoform through identiﬁcation of TILLING mutants or genome editing, combined with co-expression of heterologous glucan synthase and a glucan branching enzyme, may in future enable elevated yields of carbohydrate through simultaneous accumulation of starch and cytosolic glucan.


INTRODUCTION
There is a pressing need to increase yields of the world's major cereal crops to meet the needs of the growing population and to combat the negative effects of climate change. Although manipulation of rates of photosynthesis can contribute to this goal, it is generally agreed that sink limitation (i.e., the capacity of grains to store assimilates transported from the leaves) is an important factor in determining yield. Numerous attempts have been made to improve sink capacity through manipulation of the rate of synthesis of starch, the major component of cereal grains. It is argued that a higher capacity to sequester incoming assimilate as starch will result in more/larger grains and hence improved yield, although there is little evidence for a strong correlation between grain weight and grain starch content (Cakir et al., 2015;Fahy et al., 2018).
Efforts to increase starch synthesis have focused largely on increasing the activity of ADP-glucose pyrophosphorylase (AGPase), the first committed enzyme of starch synthesis and the source of the ADP-glucose (ADPGlc) substrate for starch synthases (Tuncel and Okita, 2013). Unlike the situation in other plant organs in which AGPase is exclusively plastidial, AGPase activity in the endosperm of the grain is largely cytosolic. Its ADPGlc product is transported from the cytosol into the plastid for starch synthesis. Unfortunately, attempts to increase yield by increasing cytosolic AGPase activity (through expression of more AGPase protein or AGPase with altered regulatory properties) have met with limited success. Higher grain weights in controlled conditions have not been shown to result in significant yield improvement in the field except under some stress conditions, including high temperatures immediately after pollination (Hannah et al., 2017;Kang et al., 2013;Smidansky et al., 2002;Tuncel and Okita, 2013). Here we describe the first steps towards employing an alternative strategy to increase grain sink strength, by introducing enzymes that can convert the cytosolic AGPGlc into a cytosolic glucan. This approach might potentially provide a second major sink for ADPGlc, allowing synthesis of both starch and glucan and thus a higher total capacity for carbohydrate storage. Support for such an approach comes from a previous study of leaves expressing cytosolic enzymes of glucan synthesis. Transient expression of yeast or red-algal glycogen-synthesizing enzymes in the cytosol of Nicotiana benthamiana leaf cells resulted in accumulation of levels of glucan many times higher than the levels of starch in these cells (Eicke et al., 2017). Eicke et al. expressed UDPGlc-dependent glycogen synthases that presumably used the endogenous cytosol UDPGlc pool as the substrate for glucan synthesis. It seems possible that stable rather than transient expression of these enzymes will have a negative impact on endogenous pathways that use UDPGlc, including cell wall synthesis. The fact that cereal endosperm has an endogenous ADPGlc pool entirely committed to storage carbohydrate synthesis meant that we could introduce an ADPGlc-dependent glycogen synthase with a low risk of interference with other metabolic pathways. Accordingly, we expressed ADPGlc-dependent GlgA-encoded glycogen synthase of Agrobacterium in cytosol of developing grain, an enzyme shown previously to be highly active when expressed in Arabidopsis (Crumpton-Taylor et al., 2013;Lu et al., 2018) and reported to initiate glucan synthesis without a requirement for a primer (Ugalde et al., 2003).

RESULTS
Wheat plants cv Cadenza were transformed with a construct containing the gene encoding the glycogen synthase of Agrobacterium tumefaciens, GlgA, with expression driven by the promoter of a high molecular weight glutenin subunit, 1Dx5 (Lamacchia et al., 2001). We recovered 30 independent transformants that were shown by polymerase chain reaction (PCR) to contain the GlgA gene. We also selected as a control a line that was positive for the gene encoding the selectable marker, bar, but not for the GlgA gene. In some experiments, the untransformed parental line Cadenza was included as an additional control. From the transgenic plants containing the introduced GlgA gene, we chose three lines in which there was a segregation of 1:2:1 in the T1 generation for plants homozygous, heterozygous and null for GlgA. Two of these lines had single insertions of the transgene (lines 1 and 3); the third had multiple insertions at a single locus (five insertions, line 2). The three lines are referred to below as GlgA lines 1, 2 and 3.
Using quantitative PCR, we found that the transcript for GlgA was detectable in endosperm of developing grain on T1 plants at the late soft dough stage [30 days after anthesis (DAA) in controlled environment room conditions] for all three lines. No GlgA transcript was detectable in the bar-only control or in Cadenza (Figure 1a). To discover whether the GlgA lines contained active glycogen synthase, we measured the activity of starch/glucan synthase in the same pools of developing grain used for quantitative PCR analysis (Figure 1b). Activity in control lines was 28 nmol min À1 g À1 fresh weight, in line with values in the literature for soluble starch synthase activity at similar stages of grain development [e.g. mean values for fieldgrown genotypes in the range 15-40 nmol min À1 g À1 fresh weight (Fahy et al., 2018); values for plants in controlled environment rooms approximately 25 nmol min À1 g À1 fresh weight in two separate studies (Hawker and Jenner, 1993;Borrill et al., 2015)]. Activity in homozygous transgenic lines was between 8 and 40 times higher than in control lines. As expected, activity in pools of grains from heterozygous plants was lower than that from homozygous plants, but higher than that from control lines.
We used the same pool of grains for a preliminary analysis of the effects of elevated glycogen synthase activity on starch synthesis. Starch content was up to 93% lower in homozygous lines expressing glycogen synthase than in control lines. The activity of AGPase, the first committed enzyme of starch synthesis, was strongly reduced in homozygous transgenic grains, by up to 80%, relative to grains of the bar control (Figure 1c,d). Starch content and AGPase activity were also reduced in heterozygous relative to control grains, but the reductions were less severe than for grains from homozygous plants.
At maturity, grains from homozygous GS lines were markedly different from those of the control line and Cadenza. They were severely shrunken, and had low weights and very low starch contents (Figure 1e-g). Whereas starch accounted for about 67% of the weight of control grains, it made up less than 10% of the weight of grains on homozygous T1 plants. The effect of the transgenes on starch content at maturity was equally strong in grains that developed on T3 homozygous plants (see text below). Grain weight and starch content at maturity were also reduced in heterozygous plants, but the reductions were less severe than for homozygous plants (Figure 1eg). The grains of homozygous GlgA lines germinated more slowly than those of the control line and Cadenza did, but overall germination success was not affected (Figure 1h).
Glycogen synthase activity perturbs dry matter, starch and water content during grain development To uncover the reasons for the very large reductions in starch accumulation and hence in grain weight in lines with cytosolic glycogen synthase, we examined changes in glycogen synthase activity through grain development and compared these with changes in dry matter, water content and carbohydrates. These analyses were made on grain developing on homozygous T2 plants grown in a glasshouse in summer (between July and September, day lengths 16.5-14 h).
Glycogen synthase activity was assessed on soluble extracts of grains subjected to electrophoresis on native gels containing glycogen, which were then incubated with ADPGlc followed by staining with iodine to reveal bands of glycogen synthase activity. Activity in control and GlgA lines was low at an early stage of grain development, and then it started to rise in GlgA lines. Activity was much higher in GlgA than control lines from 7 to 20 DAA, then declined to a low level as grains approached maturity at 30 DAA ( Figure 2a).
Up to about 15 DAA, the dry weights and water contents (percentage of fresh weight) of GlgA grains were not markedly different from those of controls. However, beyond this point, dry matter accumulation was much more rapid in the control than in GlgA grains so that by 30 DAA the dry weight of GlgA grains was only two-thirds that of the

Developing GlgA grains have very low starch and exceptionally high sugar contents
The low dry weights and starch contents of mature GlgA grains were attributable to an almost complete cessation of starch accumulation after the early stages of grain filling ( Figure 3a). Electron and light microscopy of developing grains confirmed that starch granules were much smaller and less abundant in grains from GlgA than control lines ( Figure 4). Granules in grains of GlgA and control lines were of similar dimensions at 6 DAA, but by 15 DAA, granule size was greater in the control than in GlgA grains. At maturity, scanning electron microscopy revealed large numbers of A and B granules in cells of control grains but very small numbers of granules in largely 'empty' cells in grains of GlgA lines.
The high water content of grains of GlgA lines at 15 and 20 DAA were associated with marked swelling of the grains relative to controls in this time interval (Figure 5a). At 20 DAA, grains from GlgA lines had a pocket of liquid beneath the integuments around the crease region, which pushed the vascular strand outwards. In cross-section, the region occupied by the liquid pocket appeared as a large empty space above the nucellar projection ( Figure 5b).
These data suggested that GlgA grains might accumulate large amounts of water-soluble metabolites in place of starch, increasing grain osmotic potential and thus drawing in excess water. We measured metabolites on a fresh weight basis. On a dry weight basis, the differences in metabolite levels between transgenic and control lines would be greater at 15 and 20 DAA because of the higher water contents of transgenic lines at these points (Figure 2b). The features of the data described below were seen on both a fresh weight and a dry weight basis.
Sucrose and hexose levels in endosperm tissue were higher in GlgA than control grains at 15 DAA, and substantially higher at 20 and 30 DAA (Figure 3b-d). The higher total sugar levels were largely contributed by elevated contents of sucrose. Although there was a general elevation of sucrose and hexoses in GlgA grains, both the extent and timing of this elevation indicated that it could not account for the very substantial swelling and liquid accumulation in GlgA grains. In particular, the swelling of GlgA grains was obvious by 15 DAA, but substantial elevation of sucrose and hexoses was not seen until 20 DAA. We reasoned that GlgA grains might be accumulating a different soluble metabolite as a direct consequence of cytosolic glucan synthase activity. We first tested whether the grains accumulated soluble glucan. Treatment of extracts containing soluble metabolites with methanol-KCl solution (a standard method for glucan precipitation) revealed that levels of precipitable, soluble glucan were low in grains from both control and GlgA lines (Figure 3f). The higher values in control lines at 30 and 40 DAA are probably due to precipitation of small amounts of starch that remained in the soluble fraction in these high-starch lines. We reasoned that failure to detect a glucan product of cytosolic glycogen synthase might be because the product was subject to hydrolysis by cytosolic b-amylases, resulting in accumulation of maltose in endosperm cells. This hypothesis was confirmed by measurements of maltose using a specific assay that employs heterologously expressed disproportionating enzyme (DPE2) from Arabidopsis (Smirnova et al., 2017) (Figure 3e). Control grains contained very low or undetectable maltose levels throughout development. This was also true in GlgA lines at 7 DAA, but levels of maltose then rose rapidly so that by 15 DAA levels were comparable with or higher than those of sucrose. Levels were even higher, and higher than levels of sucrosein GlgA lines at 20 DAA. By 30 DAA maltose levels had fallen steeply, and were three-to four-fold lower than levels of sucrose in the GlgA lines at this late developmental stage. Finally, we investigated the composition of the liquid in the pocket that appeared above the nucellar projection in GlgA lines at 15 and 20 DAA. A fine syringe was used to extract this liquid from five to seven grains at 20 DAA. The collections were pooled for each line, and sampled for sugar measurements. The liquid contained high concentrations of sucrose (85-90 mM) and maltose (27-50 mM), and much lower concentrations of glucose and fructose (7 and 1 mM respectively) (Figure 3).

Developing grains contain transcripts for a cytosolic bamylase
Given that maltose is the major product of degradation of linear glucans by the exoamylase b-amylase, we examined whether developing wheat grains contain a b-amylase that might be responsible for maltose production in our transgenic lines. The rice b-amylase OsBAM1 (LOC_Os07g35940 or Os07g0543300) and the Arabidopsis b-amylase AtBAM5 are reported as cytosolic (Hirano et al., 2011;Laby et al., 2001). Both proteins belong to clade VI of b-amylases, almost all members of which are predicted to be cytosolic (Thalmann et al., 2019). Interestingly, OsBAM1 is expressed almost exclusively in seeds (Hirano et al., 2011;Saika et al., 2005). Using information from Ensembl Plants (https://plants.ensembl.org/Triticum_aestivum/) on shared duplications in the rice and wheat genomes (Salse et al., 2008) and from the Wheat Expression Browser (expVIP: www.wheat-expression.com; Borrill et al., 2016), we identified nine wheat genes putatively encoding clade VI OsBAM1/AtBAM5 orthologues ( Figure 6). Phylogenetic analysis suggests that these form three sets of homoeologues, namely TaBAM5.1, TaBAM5.2 and TaBAM5.3. Expression data from expVIP indicated that expression of TaBAM5.3 homoeologues is essentially confined to grains, occurring during development in the period 9-30 DAA with a possible peak at 12-14 DAA (Table 1). No expression of TaBAM5.2 homoeologues was detected in grains. While TaBAM5.1 homoeologues are weakly expressed in developing grains, data from expVIP indicated that expression is primarily in the seed coat, whereas TaBAM5.3 is highly expressed in developing endosperm. We confirmed the cytosolic localization of TaBAM5.3 by transiently expressing the protein in N. benthamiana leaves (Figure 7). Taken as a whole, these data are consistent with the idea that a single, cytosolic isoform of b-amylase, TaBAM5.3, is expressed in developing wheat endosperm, at a point at which there is massive accumulation of maltose in transgenic grains expressing cytosolic glucan synthase.  3) are highlighted in red, green and yellow respectively. The TaBAM5.3-5A homoeologue (TraesCS5A02G554200) indicated in bold is most highly expressed in wheat endosperm during grain development (Table 1)  The three TaBAM5.3 homoeologues TraesCS5A02G55420, TraesCSU02G03270 and TraesCS4B02G393500 are expressed most highly in the grain, particularly in the starchy endosperm 12-14 days after anthesis (DAA). Note that there is a well-known translocation between chr 4 and 5 that explains why the A homoeologue of BAM5.3 is on chr 5 and the B homoeologue is on chr 4. Expression data are shown as transcripts per million (tpm). Leaf and root expression data at 14 days old are shown for reference. Expression data were obtained from the public database www.wheat-expression.com. ©

DISCUSSION
Despite the high activity of glycogen synthase in the cytosol in developing grains of the transgenic lines, we did not detect the presence of any novel polymeric glucans. Introduction of the Agrobacterium ADPGlc-dependent glucan synthase into a cellular compartment with high endogenous levels of ADPGlc might be expected to result in accumulation of linear glucans, likely to associate to form insoluble material. The absence of any such material could be attributable either to failure of the heterologous glycogen synthase to generate glucans, or their immediate degradation by endogenous cytosolic glucan hydrolases. The former explanation seems unlikely. First, we have shown previously that the Agrobacterium GlgA protein is active and capable of synthesizing glucans when expressed in plants [in Arabidopsis chloroplasts, directed by a transit peptide from the small subunit of Rubisco (Crumpton-Taylor et al., 2013;Lu et al., 2018)]. Second, it is reasonable to assume that the large amounts of maltose in transgenic lines are derived from a product of the heterologous glucan synthase. Glucan synthase activity in GlgA grain rose from 5 DAA, reaching a peak at 15 DAA and remaining high at 20 DAA, coinciding with profile reported for the high molecular weight glutenin promoter (Furtado et al., 2009;Lamacchia et al., 2001). Maltose levels rose very steeply from 7 to 15 DAA, and remained high at 20 DAA. Essentially no maltose was detected in control grain at any point. We propose that a linear glucan product of glucan synthase in the cytosol is immediately hydrolysed to maltose by an endogenous b-amylase. Only one cytosolic b-amylase encoded in the wheat genome is expressed in endosperm at the time of accumulation of maltose in our transgenic plants, hence we propose that this enzyme, BAM5.3, is responsible for the conversion to maltose of the glucan synthase product.
A parallel situation occurred when UDPGlc-dependent glucan synthases were transiently expressed in N.
benthamiana leaves (Eicke et al., 2017). Maltose and maltotriose, which are present in only low amounts in untransformed leaves, were accumulated to high levels, presumably as products of metabolism of linear polymers produced by the glucan synthases.
In addition to the massive accumulation of maltose, transgenic grains differed from control grains in undergoing an almost complete cessation of starch accumulation at an early stage of development. This could be due to a depletion of the ADPGlc pool by the introduced glucan synthase, or to deleterious effects of maltose on the pathway of starch synthesis. Our results strongly favour the latter explanation. Accumulation of maltose was accompanied by low AGPase activity and the accumulation of high concentrations of sucrose in the endosperm and in the apoplast above the phloem-unloading zone. The liquidfilled pocket above the nucellar projection presumably arises because of an influx of water into the grain brought about by the apoplastic accumulation of sucrose and the leakage of maltose from endosperm cells. These features imply that maltose accumulation reduces the overall capacity for sucrose metabolism in the developing grain, limiting starch synthesis, the synthesis of other dry matter components, and ultimately its own synthesis.
Consistent with our findings, maltose accumulation has been shown previously to have strongly deleterious effects on metabolism in leaves. Arabidopsis mutants that accumulate maltose in the cytosol and/or in the chloroplast the dpe2, mex1 and be2 be3 mutants, are slow-growing and somewhat chlorotic (Dumez et al., 2006;Lu et al., 2006;Niittyl€ a et al. (2004);Stettler et al., 2009). A study of the mex1 mutant, lacking the chloroplast envelope protein MEX1 that allows export of maltose from the chloroplast to the cytosol during starch degradation, found reduced starch turnover, high levels of chloroplast degradation and transcriptional changes consistent with accelerated protein turnover. These widespread metabolic effects were specifically caused by maltose accumulation: reducing the Full coding sequence of TaBAM5.3-5A was generated by gene synthesis and inserted into the pB7YWG2 vector (with a 35S promoter and C-terminal eYFP tag; Karimi et al., 2002). eYFP, enhanced yellow fluorescent protein. The construct was introduced into N. benthamiana leaves via Agrobacterium infiltration, and imaged 3 days after infiltration with confocal microscopy. Chlorophyll autofluorescence (red), eYFP (yellow) and bright field (BF). An uninfiltrated N. benthamiana leaf was used as control. Scale bar: 20 µm. accumulation of maltose by knocking out a chloroplast bamylase in mex1 (the mex1 bam3 double mutant) reduced the severity of the mex1 phenotype (Stettler et al., 2009). The accumulation of maltose and maltotriose in N. benthamiana leaves transiently expressing a cytosolic glucan synthase also impaired normal starch turnover (Eicke et al., 2017).
Although maltose accumulates to high levels around mid-development, it disappears in the latter stages of grain maturation. The strong decline in cytosolic glucan activity, plus presumably a strong limitation on ADPGlc synthesis, can explain why maltose does not accumulate further, but its disappearance implies that it can be metabolized further by enzyme(s) in the maturing grain. The nature of these activities remains unknown.
Although the introduction of glucan synthase activity into the cytosol is highly deleterious for grain development, the enzyme is clearly capable of using the cytosolic ADPGlc pool to form glucans. Two further steps that build on this finding might have the desired effect of increasing the sink strength of the grain. First, introduction of a branching enzyme (BE) as well as a glucan synthase would be expected to result in a branched cytosolic glucan that could be degraded by the endogenous b-amylase to only a limited extent. The attempts to generate cytosolic glucans by transient expression in N. benthamiana leaves showed that the introduction of BEs as well as glucan synthases resulted in the accumulation of soluble, branched glucan polymers to levels several times greater than normal levels of starch. The branching of the product of glucan synthases also largely suppressed the accumulation of maltose and maltotriose. We suggest that the introduction of BE as well the GlgA-encoded glucan synthase into the cytosol might have a similar effect, and allow glucan accumulation without deleterious effects on starch. Second, given the dramatic accumulation of maltose and its likely negative effects, removal of the cytosolic b-amylase would provide a 'clean' background for expression of glucan synthase and BE. Loss of grain b-amylase has minimal impact on seed germination in rice (Saika et al., 2005) and barley (Kaneko et al., 2000), hence its removal in wheat is unlikely to have strong adverse effects. Its removal could be achieved through use of the TILLING population in the Cadenza cultivar of wheat, which contains multiple lines with predicted truncations in the A, B and D genome copies of this gene (www.wheat-tilling.com; Krasileva et al., 2017).

Plant growth and sample harvest
For initial analyses T1 plants were grown in a controlled environment room with 16 h light at approximately 400 lmol quanta m À2 sec À1 and 75% relative humidity and day/night temperatures of 20/15°C. Ears were harvested 30 days after anthesis (30 DAA: late soft dough stage under these growth conditions), immediately frozen in liquid nitrogen and then stored at À70°C. Grains for extractions were dissected from frozen ears rapidly on dry ice. Mature grains were harvested when plants had dried out. For analyses at different developmental stages, T3 plants were grown in natural light and temperature in a glasshouse in summer (July-September, day lengths 16.5 to 14 h). Ears were tagged at anthesis and harvested, stored and then dissected as above at the stated DAA. Each biological replicate was taken from a different plant.
Plant transformation was carried out as a service by the Rothamsted Research Cereal Transformation Group (https://www. rothamsted.ac.uk/cereal-transformation) using biolistics according to Sparks and Jones (2009). Construct pRRes14.3glgA1 was bombarded into immature embryos of the wheat cultivar Cadenza alongside a selectable marker plasmid pRRes1.111, which carries a Ubi-bar-nos cassette for plant selection. Genomic DNA from all putative transgenic plants was subjected to PCR analysis for GlgA1 (as in Crumpton-Taylor et al., 2013) and the bar selectable marker gene (bar). Genotyping and copy number assessments were carried out as a service by iDNA Genetics (idnagenetics.com), on leaves of seedlings grown from grains on T0 plants. Copy number was determined by TaqMan quantitative PCR, as described by Diaz et al. (2012). From 30 independent transgenic lines, three with low copy number and good levels of GlgA expression were selected for study. A line expressing the bar gene only was selected as a control.

Quantitative reverse transcription-PCR analysis
Analysis was performed on developing grain (30 DAA) from homozygous and heterozygous T1 plants grown in a controlled environment room. Outer floret grains from the central region of a spike (approximately 60% of spike length) were pooled, giving 10-14 grains. Each spike was from a different plant. RNA extraction and cDNA synthesis were performed as described in Connorton et al. (2017). Briefly, RNA was extracted with phenol/chloroform extraction and then treated with DNase using the TURBODNasefree kit (Thermo Fisher, www.thermofisher.com). cDNA was synthesized using oligo(dT) primers and SuperScript II reverse transcriptase (Thermo Fisher). The ADP-ribosylation factor gene Ta2291 was used as the reference gene for quantitative reverse transcription (qRT)-PCR normalization (Vicente et al., 2015).

Enzyme activities
Glucan (starch) synthase and AGPase activities were quantified on soluble extracts of 30 DAA grains sampled as for qRT-PCR analysis. Extraction (1:3 tissue to volume ratio) and assay procedures were as in Fahy et al. (2018). For native-gel assessment of glucan synthase activity during grain development, grains were homogenized in 100 mM MOPS, pH 7.2, 1 mM EDTA, 10% (v/v) glycerol, 1 mM dithiothreitol, and 19 Complete Protease Inhibitor cocktail (Roche, www.roche.com) at 1:3 tissue to volume ratio, then subjected to centrifugation. Soluble extract was loaded on 7.5% polyacrylamide gels containing 0.3% w/v oyster glycogen, at 4 mg tissue equivalent per lane. After electrophoresis, gels were stained with Lugol's iodine solution, according to Eicke et al. (2017).

Sugars, starch and glycogen
For sugar, starch and glycogen assays, frozen grains, sampled as for qRT-PCR analysis, were homogenized, extracted in perchloric acid, neutralized and subjected to centrifugation according to Fahy et al. (2018). Sugars were assayed on the soluble fraction, using enzymatic assays according to Delatte et al. (2005) and Martinis et al. (2014) and an enzymatic assay for maltose described by Smirnova et al. (2017). Starch was assayed on the insoluble pelleted material as glucose released by treatment with a-amyloglucosidase and a-amylase, according to Smith and Zeeman (2006). Glucan was assayed as for starch, on material precipitated from the soluble fraction by treatment with 75% methanol (Delatte et al., 2005). For measurements of the composition of the liquid in the pocket above the nucellar projection in GlgA grains, liquid was drawn out of five to seven grains with a fixed-needle syringe and pooled for each of the three transgenic lines. Sugars in the pooled liquid were assayed as above, with three technical replicates per line.

Dry weight and water content
Developing grains from the same ears used for metabolite measurements were weighed, dried completely at 60°C then reweighed. Water contents were calculated from these data. Mature grain size and weight were analysed with a Marvin seed analyser (www.marvitech.de).