Starch branching enzymes (SBEs) play a pivotal role in amylopectin biosynthesis by catalysing chain transfer by cleavage of an α-1,4 linkage following a condensation of an α-1,6 linkage . In cereal species, such as rice, maize, barley and wheat, there are three classes of starch branching enzymes (SBE I, SBE IIa and SBE IIb). Barley (cv. Golden Promise) was genetically transformed to increase the starch amylose content by blocking amylopectin biosynthesis through silencing of all SBE genes. A single, multifunctional DNA construct was designed with the intention to simultaneously target the expression of the three genes encoding isoforms of SBEs in barley by RNA interference (RNAi) (Figure 1a). We found that expression of all three SBE genes was simultaneously reduced in grains of transgenic plants (Figure 1b). There was very little sequence similarity among the target sequences of the three different SBE genes (Additional file 1), which suggest that the simultaneous silencing was not an effect of homologous inhibition to corresponding genes by one of the segments in the hairpin. This conclusion is in agreement with a similar approach in rice , where it was demonstrated that up to three members of a gene family could be specifically silenced by a single chimeric hairpin construct of non-homologous segments. In this work the authors also conclude that transitive RNA silencing where formation of siRNA extent beyond the target sequence does not occur for endogenous genes in rice. We did not study that in this work, however it is unlikely that this would have an effect in the SBE gene family because of low sequence similarity among SBEI, SBEIIa and SBEIIb. In line with the observed reduction of gene expression of the SBE genes we found that starch branching enzyme activity was reduced by 82% in the SBE RNAi4.1 line when compared with a wild type control line grown under similar conditions (Figure 1c). This shows that the reduction in gene expression similarly reduced the level of enzyme activity.
Using size exclusion chromatography (SEC), we found that the amylose fraction in control starch was 29.9% of the total starch. In contrast, the amylopectin constituted less than 1% of total starch in SBE RNAi4.1, where instead a major double peak characteristic for amylose was identified (Figure 1d). The λ-max for these fractions are all above 580 nm supporting that this starch fraction is amylose. Some residual starch branching enzyme activity was observed in the SBE RNAi4.1 line suggesting that biosynthesis of amylopectin requires a certain threshold (above 18%) of SBE activity. SBE activity below 18% of control is not capable of synthesising amylopectin and the possibility that the amylose deposited in the SBE RNAi4.1 line contains some degree of branching not detectable by the iodine staining cannot be excluded. However, the SBE RNAi4.1 line also had significant increased expression of some of the starch synthases (Figure 3d), increasing the capacity for biosynthesis of non-branched starch.
Amylopectin is a semi-crystalline material with distinct thermal characteristics . The thermal and solubility properties of the starch of the amylose-only line was analyzed and compared to starch from control barley. The control starch had as expected a peak gelatinization endotherm typical for amylopectin when using differential scanning calorimetry (DSC) (Figure 2a and Additional file 3). This endotherm was completely absent in the SBE RNAi4.1 starch confirming the absence of normal amylopectin in this starch. The endotherm seen at approx. 95°C in SBE RNAi 4.1 starch (data not shown) is characteristic of the amylose Vh crystal polymorph and supports the presence of normal amylose. Together with the SEC data these results demonstrated that the phenotype was amylose-only with a characteristic molecular fingerprint of amylose. Silencing of two SBE genes SBEIIa and SBEIIb in barley increases amylose content to a certain degree (70%) . Our data show that amylose-only barley can be obtained when simultaneous suppression of all of the three SBE genes is performed (Figure 1b). This effect underlines the important role played by SBEI in barley endosperm starch biosynthesis in contrast to the apparent non-functionality suggested for SBEI in Arabidopsis leaves  and wheat endosperm .
Swelling and solubilisation of starch in aqueous systems, e.g. during cooking, are crucial for efficient enzymatic starch digestion leading to glycemic response. Heating of granular starch in excess water disrupts the crystalline structure as an effect of breakage of the extensive hydrogen bonding network between water molecules and the hydroxyl groups of the starch. This causes granule swelling and gelatinization  and the branched amylopectin, but not amylose is primarily responsible for this effect . In control line the start point of swelling coincided with melting of amylopectin at 66°C. Starch extracted from SBE RNAi4.1 did not show any visible swelling (Figure 2a).
Major suppression of enzymatic degradation rates and a major increase in the RS content fraction were found for the SBE RNAi4.1 starch as compared to control starch as evaluated by the Englyst method for determination of RS. The amount of RS in the amylose-only starch was 90%, 65% and 68% respectively for native, gelatinized and retrograded starches (Table 5). For comparison cooked banana and potato starches, which is considered very high in RS, do not exceed 30% RS. These data demonstrate important health-associated features of this novel all-native resistant starch. It also provides the last link to complete the compositional range of starch produced in the cell from 0% amylose, the so called waxy type starch , to 100% amylose to generate the entire range of amylose:amylopectin ratios in plants important for completing our understanding of starch bioengineering.
The transgenic grains had a characteristic wrinkled phenotype (Figure 3a) and the SBE RNAi4.1 endosperm cavity appeared elongated and enlarged. Interestingly, the wrinkled seed is a phenocopy of the pea phenotype rugosus described by Gregor Mendel in his study on the laws of inheritance published in 1865 [45–47], which is also due to a loss-of-function in SBE activity . The easily recognizable phenotype allowed us to score segregation (Table 1). The phenotype segregated 3:1. The fact that the SBE RNAi construct permits simultaneous targeting of three independent SBE genes, is of particular practical importance in breeding. That is because segregation in a single locus is practically more feasible as compared to the traditional alternative of differential suppression by independent RNAi constructs targeting each of the SBE genes  or crossing of multiple individual SBE loss-of-function genes, which each segregates independently. The strategy has been presented previously by . However this is to our knowledge the first time that the method has been applied in a situation where silencing of multiple independently segregating genes is necessary for achieving a particular biosynthetic product, which in our case is amylose-only starch. For higher plants this is especially important where many metabolic pathways are highly redundant due to presence of isoenzymes and gene families in metabolic networks [48, 49] and single gene loss-of-function is therefore often phenotypically silent.
Increased amylose content in cereal grains has been demonstrated to be correlated with altered accumulation of others grain constituents like β-glucan and water content . Similarly in SBE RNAi4.1 and SBE RNAi4.9 wrinkled grains the β-glucan content was significatively higher than in control barley grains (Additional file 5). Cereal grain β-glucan has been shown to be associated with important dietary health benefits .
Simultaneous suppression of the only two classes of starch branching enzymes, SBE I and SBE II present in dicotyledonous plants such as pea and potato using a single  or a sequential [30, 33] round of transformation in potato led only to a partial suppression of the amylopectin content and a dramatic increase of starch phosphate. Here we found that SBE suppression in barley had no significant effects on the content of starch bound phosphate (data not shown). Hence, the starch generated in this study provides for the first time an amylose-only model with no effects on starch phosphate.
Starch granule morphology and structure were severely altered in the amylose-only chemotype (Figure 3b, Figure 4 and Additional file 6). Normal starch granule morphology and crystallinity arises from repeated amylopectin double-helical lamellae. The disordered morphology of the SBE RNAi granules therefore reflects the lack of ordered amylopectin and suggests the presence of abnormal multiple granule initiations typical for high amylose chemotypes . These novel granules are expected to compose new combinations of crystal polymorphic packing. There are two main starch crystalline polymorphs: the A polymorph present in cereal seed starch and the B polymorph typically found in tuberous storage starch, transitory leaf starch and amylose-rich starch. A third single helical Vh polymorph is typical for amylose, especially in complexation with lipids, iodine or alcohols . We found a shift from A-type in the control starch to a mixed B/Vh-type polymorph in the SBE RNAi line, typical for high-amylose starch . Such starch is also associated with resistance to enzymatic hydrolysis and dietary fiber-like properties [52, 53].
Yield and germination were investigated in the T2 generation of SBE RNAi 4.1 plants grown under semi-field conditions. An analysis of the individual components contributing to overall yield: spike number, grains per spike and grain weight showed that the yield penalty in the amylose-only barley is mainly due to fewer spikes per plant, and to a lesser extent lower grain mass. All together, the overall yield was 22% below that of control plants grown under identical conditions, which is much less dramatic as compared to other high-amylose systems . Hence, the cereal system all-together has excellent potential for large-scale production of pure amylose. The wrinkled phenotype may indicate decreased starch content, however the starch content of the amylose-only grains was 47.2%, which is only slightly lower than in the control starch (52.8%).
No difference was found in the germination frequency of SBE RNAi4.1 and control grains (Additional file 8). However, SBE RNAi4.1 line plants exhibited slower growth of the young plantlet as compared to control but the difference disappeared throughout later development (Figure 5). We hypothesized that this effect indicates an impediment of endosperm starch remobilization, during early development when the plant is dependent on the starch as a carbon source. The in vitro dark germination showing less grain mass for the control than for the SBE RNAi4.1after germination confirmed this hypothesis. Less dry biomass had been redistributed to the coleoptile and radicle in the SBE RNAi4.1 grains (21%) compared with control grains (49%). This demonstrates the physiological importance of amylopectin in the structural ordering of carbohydrate to allow a more rapid energy remobilization.
The endosperm of developing SBE RNAi grains had increased expression levels of a number of key starch biosynthetic enzymes. The most prominent increases were found for SSIV, and for GBSSIb which was previously reported to be specifically expressed in pericarp rather than in endosperm . In durum wheat where SBEIIa was silenced, a similar up-regulation of the genes encoding GBSSI, SSIII, Limit Dextrinase (LD) and Isoamylase 1 (ISAI) has been detected . This general up-regulation of the starch synthases in cereals may explain how our amylose-only barley line can compensate starch synthesis, preventing severe yield loss as seen in e.g. high-amylose potato .