To study the functional role of SBEI on molecular structure of amylose and amylopectin, Wt Sbe1a starch and mutant sbe1a starch were fractionated from mature kernels. The maize sbe1a mutant contains a Mu transposon in the 14th exon of the Sbe1a gene, and was previously shown to be null for the expression of SBEI transcript and protein 37 . The proportions, iodine binding properties, and size-exclusion chromatograms for the amylopectin and amylose fractions were similar for Wt and sbe1a starch (data not shown). To study the molecular fine structure, β-amylolysis and subsequent isoamylase and pullulanase debranching were applied to both the amylopectin and amylose fractions from Wt and sbe1a. Despite a similar chain length (CL) profile observed for both fractions from the two genotypes (see Additional File 1 online), the CL distribution after various extents of β-amylolysis showed differences for Wt and sbe1a (Figure 1A; see Additional File 2 online).

For the amylopectin fraction from both genotypes, hydrolysis with β-amylase caused a dramatic change in CL distribution within the first 10 min (Figure 1A): A major increase was observed below degree of polymerization (DP) ~10. In this region for Wt, the change in the CL distribution from 10 min to 24 h of β-amylolysis was primarily a reduction of the DP 4 stubs to DP 2 stubs; however, for the sbe1a sample no further reduction in DP 4 was observed after 10 min (Figure 1A). After 24 h of β-amylolysis, conditions necessary to produce β-limit dextrin (β-LD) 39 40 , the sbe1a sample had a much smaller proportion of the DP 2 chains and a much larger proportion of DP 4 chains than the Wt sample (Figure 1A; see Additional File 2 & 3 online).

For the amylose fraction from both genotypes, β-LD was produced. Analysis of the CL distribution of isoamylase-debranched β-LDs showed a higher proportion of chains of DP ≥ 100 and lower proportions of other chains (DP < 100), before and after pullulanase addition (Figure 1B; see Additional File 4 online). The subsequent pullulanase debranching led to an increase in both the DP 3 and DP 2 areas for both genotypes, and this increase was greater in sbe1a (Figure 1B; see Additional File 4 online). The subsequent pullulanase debranching also led to a decrease in chains of approximately DP 8-9 for both genotypes (Figure 1B).

Starch hydrolysis is an important feature of starch function both in the plant and when the plant is used for human food. Hydrolysis of starch ingested as food can vary both with respect to the rate and the extent of digestion by pancreatic α-amylase. In the human digestive tract, the undigested starch that reaches the colon is termed RS; the level of RS is a measure of the extent of digestion by this enzyme. An official in vitro method (AOAC 2002.02) is used for determination of the RS level. This method was modified to allow study of both the digestion rate and the extent of digestion 41 42 . F-tests performed for a fully nested analysis of variance (ANOVA) showed an effect of genotype (p = 0.000), but no effect of biological replication (p = 0.334). The RS value was higher in the sbe1a mutant starch (13.2%) than in the Wt starch (1.6%) from measures of 3 biological replications (p < 0.05).

The digestion pattern was similar among the three biological replications for each genotype (data not shown). For graphic illustration of the digestion time-course, curves for one biological replication for each genotype are shown in Figure 2. The kinetics of digestion were analyzed using a five-parameter, double-exponential decay model (see "Materials and Methods"), and the calculated parameters are presented in Table 1. A higher y₀ (the limit of digestion as determined using the model) was found in sbe1a than for Wt (Table 1), consistent with the higher limit of digestion given by the RS value for this genotype.

Scanning electron microscopy was used to image starch granules from Wt and sbe1a mutant plants in native form and after the 16 h in vitro digestion with pancreatic α-amylase for determination of the RS value. Prior to digestion, native starch granules from Wt and sbe1a had similar morphology (Figure 3A) with an average diameter of 10.2 μm for Wt and 9.8 μm for sbe1a. However, after digestion, differences were observed between the two genotypes (Figure 3B; see Additional File 5 & 6 online). Samples of the sbe1a RS contained many residual granules with distinct holes in the surface and hollow interiors, whereas for Wt only small fragments of residual granules were seen (Figure 3B). The Wt fragments also showed evident alternating layers on the edge of the pieces, which was less evidently present in sbe1a samples (Figure 3B; see Additional File 6 online).

Light micrographs of iodine-stained native granules are shown in Figure 3C. For both genotypes, all native starch granules were stained blue and produced a characteristic Maltese Cross when viewed in the polarized light microscope; however, sbe1a native starch showed more heterogeneity in staining as compared to Wt, as there were more relatively dark-stained granules in sbe1a than in Wt native starch (24.3% and 8.7%).

As endosperm starch from the sbe1a mutant has a lower susceptibility to pancreatic α-amylase, we suspected that the sbe1a endosperm starch might be less readily utilized during kernel germination. To study the effect of sbe1a on kernel germination, starch utilization and coleoptile growth during germination of Wt and sbe1a mutant kernels were examined.

All the kernels from three different ears of both Wt and sbe1a genotypes were germinated, demonstrating no differences in germination rate. The coleoptile length of each genotype was measured daily over 11 days (Figure 4). The average length of sbe1a coleoptiles was shorter than Wt from Day 7 onward (Figure 4). For both genotypes the endosperm starch content decreased over time (Figure 4). On Days 6, 8, and 11, the starch content was higher in sbe1a germinating endosperm as compared to Wt, suggesting less utilization of starch. This trend is consistent with the reduced growth of sbe1a coleoptiles after Day 6.

In the present study, rapid degradation of chains DP ≥18 and DP 8-17 were observed for both Wt and sbe1a samples in the first 10 min of β-amylolysis (Figure 1A). As β-amylase cannot bypass branch points to hydrolyze starch chains, a plausible interpretation for the less extensive degradation of DP 8-17 in sbe1a would be that the B chains (those chains with other chains attached) 43 would have slightly longer internal segments and shorter external chains. For the second stage of β-amylolysis 44 , a slow reduction in the amount of DP 4 chains was observed in Wt samples over the period of 10 min to 24 h but not in sbe1a samples (Figure 1A), suggesting differences in the proportion of branch points that would differentially limit access of the enzyme to glycosidic linkages 40 .

Amylopectin branching pattern models for both sbe1a and Wt are presented to account for this difference in β-amylase action on DP 4 stubs (Figure 5A). In the model for sbe1a, DP 4 stubs would be difficult for β-amylase to hydrolyze to DP 2 when closely associated branch points present a steric barrier to binding of β-amylase. Although most of the DP 4 is from residual A chains 43 , some DP 4 chains from residual B chains would result from short B chains with short internal segments. The incomplete hydrolysis of DP 4 in sbe1a suggests that A chains are preferentially localized near another branch point, leading to 1) hindered hydrolysis of residual A chains of DP 4 to DP 2 due to steric constraint, and 2) more residual B chains with DP 4 due to incidence of short internal segments (Figure 5A). In the model for Wt, the DP 4 stubs would be slowly hydrolyzed to DP 2, as there is less steric hindrance from proximal branch points. According to the two models, sbe1a amylopectin contains a higher proportion of closely associated branch points than Wt. Furthermore, based on CL profiles (see Additional File 1 online), the calculated overall average branching density is similar in the two amylopectins. Thus, we suggest that the effect of the sbe1a mutation is to increase the local concentration of branch points but not to influence the overall amount of branch points in amylopectin.

In the debranched β-LDs from the amylose fraction (but not in intact amylose), a higher proportion of long chains of DP ≥ 100 was observed in sbe1a (Figure 1B and Additional File 1 online). The higher proportion of longer chains in β-LDs of amylose from sbe1a can be explained by branch points that tend to be closer to the non-reducing ends, so that longer internal chains result.

When debranching of β-LDs from amylose was performed with isoamylase without subsequent pullulanase digestion, there were fewer DP 2 than DP 3 chains (Figure 1B; see Additional File 4 online). For β-LD from amylopectin, all of the DP 3 and some of the DP 2 chains are known to be debranched by isoamylase 40 . However, our results of β-LD from amylose for both genotypes suggest that even some DP 3 chains are not debranched by isoamylase. Comparing sbe1a to Wt, more of DP 2 and DP 3 chains are not debranched by isoamylase in β-LD from sbe1a amylose (Figure 1B). As the structures escaping isoamylase debranching may have closely associated branch points and those structures can be debranched by pullulanase 40 , a greater increase in both DP 2 and DP 3 by subsequent pullulanase treatment suggests that a higher proportion of these structures are resistant to isoamylase in amylose from sbe1a. Amylose branching pattern models are presented in Figure 5B to account for the difference in isoamylase action. In the model for sbe1a, A chains are preferentially attached by branch points close to each other whereas in Wt, A chains are not, leading to less hindered isoamylase debranching.

Our data suggest that amylose of sbe1a mutant starch has 1) longer internal chains and 2) more A chains attached by branch points close to each other. This evidence can be used to create an overall model for amylose branching patterns of sbe1a and Wt (Figure 5C). The models are drawn taking into account similar CL profiles (see Additional File 1 online) and assuming that ~50% of amylose molecules are branched, with ~5-6 branches per molecule 45 . According to the proposed model, for sbe1a, A chains are closer to each other, and the location of the chains tends to be more towards non-reducing end. For Wt, A chains are farther away from each other, and the location of the chains is more random and thus more distributed.

Kinetic analysis shows that the y₀ value for Wt starch is effectively zero (Table 1), in agreement with the RS value for Wt starch (1.6%), and the y₀ and RS values for sbe1a starch are also in good agreement.

The kinetic model is based on the presence of two general types of starch substrate: a rapidly-digested substrate (S ₁ ), and a slowly-digested substrate (S ₂ ) 41 42 . The two genotypes differ both in the proportions of S ₁ and S ₂ and the reaction rate constants for these two components. The S ₁ components of Wt and sbe1a starch were 85.9% and 59.8% respectively. This suggests that the sbe1a mutation altered the starch structure and this resulted in less rapidly-digested component. Consistent with our results, Ao et al. 46 found that increased branch density led to a decreased proportion of RDS (analogous to our S ₁ ) and an increased proportion of SDS (analogous to our S ₂ ).

Two microscopic techniques, scanning electron microscopy (SEM) and light microscopy (LM), were employed to observe granular structure before and after RS digestion by pancreatic α-amylase. Native starch granules from Wt and sbe1a appear similar in size, shape, degree of birefringence, and morphology, as described in a previous report for wx and sbe1a wx granules 47 . Polarized light microscopy (see Additional File 5 online) shows that almost all of the digested Wt granules had lost their birefringence, while for sbe1a, many digested granules had maintained some birefringence in the peripheral area of the granules, which indicates that the center of the digested sbe1a granules is either gone or no longer crystalline enough to show birefringence. The presence of a hollow interior in the digested sbe1a granules was confirmed by SEM (Figure 3B), indicating a relatively greater resistance to digestion for the exterior portion of the sbe1a granule.

Most of the recovered RS from Wt were represented by small granule fragments. However, the sbe1a RS showed variations in morphology, from small fragments to hollow granules. The difference in digestion of individual granules may be due to differences in heterogeneity in granule structure, as a higher proportion of relatively dark-stained granules were observed in sbe1a than in Wt native starch (Figure 3C). SEM revealed the presence of alternating layers in the Wt residual fragments (Figure 3B), which probably reflect the residual growth rings after digestion.

By observing the sbe1a RS by SEM (Figure 3B), one may roughly estimate that, for the recovered granules, approximately 40% of granule content has escaped digestion. However the RS value for sbe1a starch is approximately 13%. Therefore, some of the sbe1a granules were likely to have been digested completely. The heterogeneity found among sbe1a granules (Figure 3C) may account for different degree of digestion of individual granules. Thus, it can be reasoned that the micrographs of the sbe1a RS may disproportionately represent the more resistant granules.

A distinct feature of the recovered sbe1a RS is the presence of holes on the surface of the peripheral portion of the granules. These holes are possibly from the enlargement of the surface pores in native granules by α-amylase hydrolysis 48 . The presence of these holes on the shell is consistent with previous studies demonstrating that digestion of normal granules starts with surface pores and proceeds through deeper hydrolysis in channels 49 50 51 52 , followed by fragmentation 48 . In the current study, the presence of remaining shells with holes in the sbe1a RS indicates continuing difficulty in digestion by α-amylase. Neither holes nor shells were observed in the Wt RS, indicating a more complete digestion.

As observed under microscopy, the RS from Wt consists mostly of portions of residual growth rings, while the RS of sbe1a is mostly residual peripheral regions. The kinetic analysis shows that the digestion of sbe1a starch reached a plateau by 16 h, suggesting the RS from sbe1a is not further digested. When the RS is observed by SEM, one can conclude that some of the peripheral regions in sbe1a starch granules cannot be further digested. Enrichment of amylose has been found by some to exist toward the granule peripheral region 53 54 . SEM showed that the peripheral regions were more resistant to α-amylase digestion in sbe1a granules. It is possible that these differences may be preferentially localized in the peripheral region of the granules, where starch synthesis may be more influenced by deficiency of SBEI 10 . The CL distribution of residual starch collected after α-amylase digestion showed some small differences between Wt and sbe1a (see Additional File 7 online). However, no direct evidence was obtained in the current study about whether the molecular structure in the peripheral regions was different in sbe1a.

An endogenous α-amylase is considered to be responsible for attacking the starch granule and initiating starch hydrolysis in germinating cereal endosperm 55 . Starch hydrolysis continues by the action of limit dextrinase, α-amylase, β-amylase, and α-glucosidase to produce maltose and glucose for plant utilization 55 . The observed reduction in starch hydrolysis during the later stages of germination raises the possibility that continued hydrolysis of α-amylase-hydrolyzed glucans is hindered in the sbe1a mutant. The altered carbon metabolism could then cause a deficiency in general plant growth characteristics such as coleoptile length 23 . The structural analysis of sbe1a starch suggests that the decreased starch utilization of sbe1a seeds is due to an altered starch branching pattern.

Differences in SBE activity in sbe mutants could be simply due to the amount of a remaining SBE isoform or to biochemical or physical interactions that modulate the activities of an isoform; for the latter possibility SBEI may be regulated through complex interactions with other starch synthetic enzymes. Colleoni et al 21 showed that two migratory forms of SBEI are missing in maize endosperm of the maize ae mutant, indicating a possible interaction of SBEI and SBEIIb. Seo et al. 24 found that when SBEs were heterologously expressed in a yeast system, SBEIIa and/or SBEIIb appear to act before SBEI on synthesizing glucan structure. The studies of Yao et al. 25 26 suggest that in the absence of SBEIIb, a reciprocal inhibition exists between SBEI and SBEIIa, and that the presence of either SBEI or SBEIIa increases amylopectin branching as opposed to the presence of SBEI and SBEIIa together.

Direct evidence for protein-protein interactions between SBEs and different members of all the proteins involved in starch biosynthesis has also been reported by several groups, based on co-immunoprecipitation and affinity purification methods. Tetlow et al. 27 reported that SBEI from wheat amyloplasts was present in a high molecular weight complex with starch phosphorylase and SBEIIb. A separate study 56 using maize amyloplasts showed that eliminating SBEIIb caused significant increases in the abundance of SBEI, BEIIa, SSIII, and starch phosphorylase in the granule, without affecting SSI or SSIIa. Hennen-Bierwagen 30 reported that SBEI and SSI were shown to interact in one of three independent methods tested; SBEI did not interact with any of the other proteins in their study (SSIIa, SSIII, SBEIIa, SBEIIb), and unlike the other five proteins in their study, SBEI was the only protein to exist as a monomer in gel permeation chromatography.

In present study, the sbe1a mutant line is nearly isogenic with the Wt control. Most if not all mutant phenotypes are likely the result of many effects, direct and indirect, on the overall growth, development and physiology of the plant, so it is impossible to truly isolate a primary effect of the mutation when looking at a whole plant level phenotype, even the starch structure phenotype. Modifying SBE activity may induce modifications in the distribution of phosphate groups within amylopectin such as in potato 57 58 . This may alter accessibility of amylase (α or β) to its substrate and may induce differences in digestibility. Nonetheless, there is value in observing and characterizing the phenotype of these plants, both at the macro and molecular levels as we have presented. We have a sister paper 36 which does investigate the effect of various SBE mutations on leaf starch which further sheds light on the SBEI function in the context of pleiotropic consequences.

This work for the first time reports a specific and unique function for SBEI during the life cycle of maize. Molecular structure analysis suggests an important function of SBEI in modulating the branching pattern in normal starch by decreasing local clustering of amylopectin branch points. Thompson 59 emphasized the non-random nature of the distribution of branch points in starch. A specific type of non-random branching pattern may be required to optimize both storage and hydrolysis. It is reasonable to hypothesize that alteration in the specific non-random branching pattern could lead to an altered granule organization, rendering it more or less favorable to the plant for storage and/or for enzyme hydrolysis during utilization. Our data from in vitro starch digestibility and from plant germination analysis support this hypothesis.

Gene duplication and neo-functionalization are well known mechanisms by which specific genes can evolve to express different isoforms of enzymes with slightly specialized expression patterns or different enzymatic activities 60 61 62 . With the evidence from current and previous work, we can infer that an ancestral Sbe gene has duplicated at least twice during the evolution of maize, and these evolved to express three different SBE isoforms with highly specific functions in starch biosynthesis. A detailed phylogenetic analysis of the branching enzymes was published by Deschamps et al. 14 . This work demonstrated that genes belonging to both the SBEI and SBEII families can be identified in the green alga, which supports the theory that these two families of genes evolved approximately a billion years ago based on phylogenetic estimates of the divergence between the Chlorophyta and Magnolippyta lineages (estimates range from 729-1210 million years ago) 63 64 . This example of extreme evolutionary conservation is strong evidence for a specific and vital role for each enzyme isoform in starch biosynthesis. While most plant species studied retain genes representing each subfamily of SBE, Arabidopsis does not, suggesting that somewhere in the lineage leading to Arabidopsis, the gene was lost with minimal consequences to the species 65 .

The evidence presented in this work strongly supports the hypothesis that SBEI is required to synthesize endosperm starch granules that allow normal hydrolysis and utilization during germination. Considering plant survival in the wild, optimal seedling vigor would be a strong evolutionary force to select for genotypes of plants with starch granules optimized for molecular structure that would lead to efficient storage and utilization. The reduced seedling vigor of sbe1a mutant seeds observed in this work provides powerful evidence for a specialized and important role of SBEI in plant development, consistent with the evolutionary conservation of SBEI in all higher plants.

Maize plants of Wt and sbe1a mutant were grown during summer, 2007 at Penn State Horticultural Research Farm (Rock Springs, PA). In order to compare starch material within a highly similar genetic background, homozygous Sbe1a/Sbe1a (i.e. Wt) and sbe1a/sbe1a mutant siblings were identified from a single segregating population derived from seeds of selfed Sbe1a/sbe1a plants to obtain ears for endosperm analysis. Genotyping of Wt and sbe1a mutant plants followed Blauth et al. 37 . The detected homozygous Wt and sbe1a mutant plants were self-pollinated to produce ears for endosperm analysis, and are segregated from a BC₄F₃ population backcrossed by Blauth et al. 11 37 . Starch extraction from three different ears, considered as three biological replications, for each genotype, was according to Yao et al. 66 . Starch fractionation followed Klucinec and Thompson 67 .

β-Dextrins were prepared by the method of Xia and Thompson 40 with slight modifications in sample size. Amylopectin samples (48 mg) were dispersed in 480 μL of 90% dimethyl sulfoxide (DMSO) by heating in a boiling water bath for 10 min. To the dispersion, warm sodium acetate buffer (3.52 mL, 50°C 0.02M, pH 6.0) was added. The mixture was heated in a boiling water bath for 10 min and cooled to 50°C. A 200-μL aliquot of a β-amylase (from barley, Cat.No. E-BARBL; Megazyme International Ireland, Ltd.) solution (250 U/mL, 0.02M sodium acetate, pH 6.0) was added, and the samples were incubated at 50°C with constant agitation (200 strokes/min). At approximately 10 min, 30 min, 1 h, 2 h, 6 h, and 24 h, a 0.5-mL aliquot of sample was removed and heated in a boiling water bath for 10 min to stop the reaction. The procedures for precipitating β-dextrins and debranching β-dextrins by successive action of isoamylase (from Pseudomonas sp., Cat.No. E-ISAMY; Megazyme) and pullulanase (from Klebsiella planticola, Cat.No. E-PULKP; Megazyme) were the same as used previously for β-LDs) 39 40 .

The preparation and debranching of β-LDs followed the procedures in Klucinec and Thompson 39 with slight modifications in sample size. After the β-LDs were debranched with isoamylase for 24 h, a 30-μL aliquot of the digested solution was added to 270 μL of DMSO and reserved for analysis by high-performance size-exclusion chromatography (HPSEC). Then the β-LDs were further debranched with pullulanase for 24 h, afterwards another 30-μL aliquot of the digested solution was added to 270 μL of DMSO for HPSEC analysis 40 .

The official method for in vitro RS determination (AOAC 2002.02, AACC 32-40) was employed, which was scaled-down and modified for direct analysis of the digestion supernatant for total carbohydrate 41 . The modification allowed analysis of digestion time-course for small starch samples (~20 mg). For RS determination, after the 16 h digestion step at 37°C with porcine pancreatic α-amylase and amyloglucosidase (enzymes from RS Assay Kit, Cat.No. K-RSTAR, Megazyme), the sample tube was removed from the water bath and to an aliquot of each sample was added 1 volume of 95% (v/v) ethanol with 0.5% (w/v) EDTA. After centrifugation (1,500 × g, 10 min), the supernatant was analyzed in duplicate for total carbohydrate using the phenol sulfuric acid method 68 . The percent non-digested starch (% NDS) was calculated from this data and was the basis for the calculation of the RS value. Starch isolated from Wt and sbe1a mutant endospermes from three separate plants (triplicate biological replications) were subjected to triplicate pancreatic α-amylase digestion, for determining the RS values.

For determination of digestion time-course, the starch samples were digested as described above. An aliquot was removed at approximately 30 sec, 3 min, 6 min, 15 min, 30 min, 1 h, 2 h, 3 h, 4 h, 5 h, 7 h, 10 h, 13 h, and twice at 16 h, and added to 1 volume of ethanol/EDTA solution to ensure immediate deactivation of the enzymes. After centrifugation the supernatants were analyzed for total carbohydrate as described above.

Digestion time-course was analyzed following the method developed by Rees 42 to obtain kinetic data. A "Double, 5 parameter" regression model in SigmaPlot (Systat Software, Inc.) was selected to fit the data using the double exponential decay equation:

where y is % NDS, x is the time, y ₀ is the y-value that the model asymptotically approaches, S ₁ and S ₂ are the concentrations of the two different substrate components, and k ₁ and k ₂ are the reaction rate constants for the decay of the two different components. The units for y ₀ , S ₁ , and S ₂ were % of initial starch, and the units for the rate constants were min^-1. After running the regression program, the software gives three possible completion status messages depending on how well the model fits the data:

(1) Converged, tolerance satisfied.

(2) Converged, tolerance satisfied. Parameter may not be valid. Array numerically singular on final iteration.

(3) Didn't converge, exceeded maximum number of iterations.

The data were kept for further regression analysis if message 1 or 2 resulted, and were discarded if message 3 resulted.

Digestion time-course analysis was performed for three biological replications per genotype. For each biological replication, two technical replications were performed. If both sets of data "converged" using the model (message 1 or 2), no further analyses were performed. If message 3 appeared, a new technical replication was done until the data "converged." The data from the two "converged" technical replications for each biological replication were combined, and the software program was run on the combined data. For all samples, the regression model fit for the combined data completed with convergence (Message 1), and generated valid parameters for analysis. Using the combined data, values for five parameters in the equation were determined for each biological replication. A mean and standard deviation of the five parameters for each genotype was then calculated, and comparisons among genotypes were made by one-way ANOVA analysis.

Bright field and polarized light microscopy were performed using a light microscope (BX51; Olympus) with an attached digital camera (Spot II RT; Diagnostic Instruments). 5 mg of native starch sample was mixed with 0.5 mL of deionized water in a micro-centrifuge tube. For the resistant starch samples, the supernatant was removed after centrifugation of digestion solution and 20 μL of deionized water was added to the pellets to disperse the sample. To examine the sample under the microscope, 20 μL of the dispersed sample was added to a glass slide, and a cover slip was fixed over the sample with fingernail polish. Examination of iodine-stained starch followed the method in Evans and Thompson 54 . 20 μL of iodine solution (0.08% I₂, 0.12% KI) was placed onto 20 μL of the dispersed sample to give a final I₂ concentration of 0.04%. In order to compare birefringence between granules, the camera's automatic exposure function was turned off, and the exposure was set the same for all samples. The same sample field was examined under bright field and polarized light.

Heterogeneity of iodine staining was evaluated quantitatively by a volunteer panel. Differentially iodine-stained starch granules were classified into two categories, dark or light stained granules, and were sorted visually by five individual evaluators who were not otherwise involved in the research. The evaluators were trained to understand the difference between dark and light stained granules, by observing granules in a portion of a micrograph for sbe1a granules. Four micrographs for each genotype (Wt or sbe1a) were used for sorting. The evaluators were then given those eight micrographs, unlabeled and in randomized order, and asked to sort the granules into two categories. The proportion of dark granules for each micrograph was calculated based on the sorting results from all five evaluators, and a mean proportion was obtained for each micrograph. For each genotype, a mean was calculated from the means of the four micrographs. Comparison between two genotypes was made by one-way ANOVA.

A thin layer of starch sample was applied to double-sided sticky carbon tape on a specimen stub, and sputter-coated with 10 nm Au/Pd (BAL-TEC SCD 050; US-TechnoTrade). Samples were then examined using a scanning electron microscope (JSM-5400; JEOL Ltd.) at an accelerating voltage of 20 keV and at different magnification levels (1,500 ×, 3,500 ×, and 10,000 ×). For image collection, lower magnification was first employed to examine the whole view of samples, and higher magnification was then used to focus on sample areas that were representative overall.

A kernel germination assay was performed according to the method in Dinges et al. 23 with slight modifications. Mature, dried maize kernels were surface-sterilized by immersion in 15 mL of 1% sodium hypochlorite for 5 min and then washed three times with deionized water. 15 kernels from each of three ears for each genotype were placed in Petri dishes containing three layers of moist Whatman paper and incubated at 30°C in the dark. The length of each coleoptile was measured by a ruler on successive days throughout the 11-day incubation period. To measure the amount of endosperm starch remaining, the roots, coleoptiles, embryo, and pericarps were removed from 2 kernels at days 1, 6, 8, and 11. The remaining endosperm was ground with a mortar and pestle on ice. The powered tissue was washed into a tube with deionized water and homogenized with a Tissumizer (Model SDT 1810; Tekmar) at 20,000 rpm for 1 min. The ground tissue was washed with deionized water, centrifuged at 1500 × g for 10 min, and suspended in 3 mL of deionized water. For calculating the dry weight of samples, 1 mL of this suspension was dried at 70°C overnight and weighed. The remaining 2 mL of the suspension was boiled for 30 min, and the total glucan polysaccharide in the solubilized solution was quantified in triplicates, using a commercial assay kit that measures glucose released after digestion with α-amylase and amyloglucosidase (Cat.No. K-TSTA; Megazyme). The quantified starch content was normalized against the dry weight for comparison between genotypes.