If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Branching enzymes (BEs) are essential in the biosynthesis of starch and glycogen and play critical roles in determining the fine structure of these polymers. The substrates of these BEs are long carbohydrate chains that interact with these enzymes via multiple binding sites on the enzyme’s surface. By controlling the branched-chain length distribution, BEs can mediate the physiological properties of starch and glycogen moieties; however, the mechanism and structural determinants of this specificity remain mysterious. In this study, we identify a large dodecaose binding surface on rice BE I (BEI) that reaches from the outside of the active site to the active site of the enzyme. Mutagenesis activity assays confirm the importance of this binding site in enzyme catalysis, from which we conclude that it is likely the acceptor chain binding site. Comparison of the structures of BE from Cyanothece and BE1 from rice allowed us to model the location of the donor-binding site. We also identified two loops that likely interact with the donor chain and whose sequences diverge between plant BE1, which tends to transfer longer chains, and BEIIb, which transfers exclusively much shorter chains. When the sequences of these loops were swapped with the BEIIb sequence, rice BE1 also became a short-chain transferring enzyme, demonstrating the key role these loops play in specificity. Taken together, these results provide a more complete picture of the structure, selectivity, and activity of BEs.
Starch, the primary energy storage molecule of plants, is a complex biomaterial built from large α-1,4- and α-1,6-branched glucose polymers packed into a well-ordered granule, with amorphous and semicrystalline layers throughout the granule (
) (with only 5–20 branches per molecule); and Amylopectin, a much larger polymer (65–85% starch weight, with average DP 11–15 for short chains, average DP 43–50 for long chains) with approximately 5 to 6% branching (
). The starch granule’s size, density, and fine structure are key to its function as the primary energy storage unit for all plants, leading to more or less bioavailability and robustness, depending on the plant’s energy requirements (
). The starch granule is biosynthesized by the interplay of several enzymes, including ADP-glucose pyrophosphorylase, which makes the ADP-glucose monomer building block, starch synthases, which convert ADP-glucose to α-1,4-linked glucan polymers, branching enzymes, which cleave α-1,4-linkages and transfer the resulting cleaved glucan to α-1,6 positions to form branches, debranching enzymes, which remove inappropriately placed branches, and kinases that phosphorylate the resulting polymer (
Comparison of starch-branching enzyme genes reveals evolutionary relationships among isoforms. Characterization of a gene for starch-branching enzyme IIa from the wheat D genome donor Aegilops tauschii.
). While BEI isoforms prefer amylose as a substrate and tend to transfer longer chains of 11 or more units, BEII isoforms favor amylopectin as a substrate and tend to transfer shorter chains of 6 to 7 glucose units, with BEIIa transferring a broader range of glucans while BEIIb isoforms transfer almost exclusively chains of 6 to 7 glucose units (
). They also differ in localization, with BEI and BEIIb almost exclusively found in the amyloplast. BEIIa is expressed throughout the plant, especially in the leaf, where it is exclusively responsible for branching the rapidly created and degraded starch (
). Herein we report the structure of maltododecaose (M12)-bound rice BEI (rBEI), which, we hypothesize, identifies the acceptor chain binding site on the surface of the enzyme for the first time. Mutagenesis experiments confirm the importance of this binding site for activity while also inferring that the site is not the donor chain binding site. Comparison of the rBEI structure with the M7-bound Cyanothece BE structure allows us to identify the likely donor chain binding site (
). Mutagenesis of two loops in the vicinity of this donor chain binding site converts the activity and specificity of rBEI to be very similar to that of rBEIIb, confirming the location of the donor chain binding site. Together, these results give the most complete picture to date regarding the mechanism of BE and identify the crucial regions for distinguishing the activity of the isoforms. These insights will enable the rational and systematic redesign of BEs to tailor their activities much more precisely, potentially leading to more precisely modified and optimized versions of this most critical biomaterial.
Structure of M12-bound rBEI
The structures of both apo and maltopentaose-bound rBEI were previously determined (
). Three maltopentaose binding sites were identified, all distal from the active site. In the present work, a new rBEI crystal form was identified and the M12-bound rBEI structure obtained by soaking these crystals. This new crystal form was obtained using a tetra-mutant rBE1 that was truncated by 60 residues on its N-terminus (see Experimental procedures for details). Both activity and chain specificity assays showed this construct to behave identically to that of wild-type rBE1. Of the two M12 molecules bound to rBE1 (Fig. 1, A and B), one occupies the sites identified in the previously described M5-bound rBE1 structure (labeled “site 1”) (
). Four glucose units are ordered in site 1 and the interactions between glucan and protein are similar to that seen in the previous structure (Fig. 2A). The conformation and curvature of the oligosaccharide are similar to that seen in an amylose single helix (Fig. 2B). The second molecule occupies a binding site heretofore not identified in BE (“site 4”) and has all 12 glucose units ordered (Fig. 3A). It begins at a region on the catalytic domain quite far from the active site and advances toward the active site traversing the width of the catalytic domain. The glucan adopts a helical conformation (Fig. 3B), with six glucose units per turn, similar to one chain of a glycogen double helix (
), but it deviates from this conformation as it approaches the active site. The surface of rBE1 is predisposed toward binding the helical conformation, with aromatic stacking interactions (Tyr487 and Trp535) that serve to project the glucan away from the surface and hydrogen-bonding interactions with the glucose units that directly contact the surface (Fig. 3A). The residues that directly interact with M12 in this binding site are highly conserved in plant and animal BEs, with 11 of the 19 residues that directly interact with M12 identical in virtually all plant and animal BEs (Fig. 3C). However, little conservation is seen in bacterial BEs for this surface, indicating this binding site to be common to the eukaryotic BEs and distinct from the bacterial enzymes, including the starch-making cyanobacterial enzymes with bacteria-like BEs such as Cyanothece (GH13_9) (
When the structures of M12-bound BE1 are compared with BE1 structures with and without M5, little structural change is observed, apart from the flexible loop between residues 468 and 474 (This loop is either disordered or found in two distinct conformations in the other BEI structures). M12 binding causes this loop to adopt a conformation not seen previously. Numerous residues in the loop make interactions with M12, necessitating the loop to adopt the orientation seen in the M12-bound structure (Fig. 4A). This loop interacts with the glucose moiety found closest to the active site and may act as a “door” into the active site. No other large conformational changes are seen in the M12 binding site when all three rBEI structures are compared.
The strong conservation of the interacting residues and the relative proximity to the active site suggest that the M12 binding site may play a direct role in the reaction of the enzyme. This possibility was evaluated by site-directed mutagenesis, activity, and transfer chain specificity assays. Unlike the activity assays, transfer chain specificity for this enzyme is rather a complicated experiment, and it needs to be carefully designed. The duration of the transfer chain specificity assay is a critical parameter since the observed specificity is affected by the time that BE is branching a substrate. As shown in Figure 5A, longer reaction times invariably produce more short-branched chains and fewer longer chains. The reason for this observation is that the longer chains produced by rBEI are also substrates for secondary transfers. Secondary transfers will be limited to the branches that are still long enough to be substrates of the enzyme. Mutated enzymes with lower activity must therefore be calibrated with enzymes that have higher activity by varying the reaction time in the assay. This calibration was accomplished by terminating the transfer chain specificity assay for each mutant only when the iodine assay absorption reached 50% of the initial absorption at 660 nm.
Table 1 summarizes the activities of various rBEI mutants relative to the wild type. One unit was defined as the amount of the enzyme needed to decrease the absorbance of the substrate–iodine complex by 1% per min. The wild-type rBEI specific activity on amylose substrate is 10,100 U/mg, which is comparable to 12,300 U/mg, reported previously (
). Though several point mutants in the M12 binding site (W535A, Y487A, and D483A) showed significant loss of activity, none of these mutants significantly impacted the branch chain specificity (Figs. 5B and S2–S9). We also identified a large insertion of 11 residues in the 525 to 553 loop (Loop 541) found in all BEII enzymes (
) (Fig. 4B). This loop is proximal to the M12 binding site, and five conserved residues in this loop make interactions with M12, suggesting that it plays an important role in M12 binding. To study differences between BEI and BEII isoforms, an 11-residue insertion found in rBEIIb was introduced to the 525 to 553 loop in BEI (loop 541 mutant). Though a significant loss of activity was observed, no change in branch chain specificity was identified (Fig. S9); one of the most significant differences in BEI versus BEII activity is the preference of BEII enzymes for the transfer of shorter (6–7 units) chains relative to BEI enzymes (
A second loop, encompassing residues 146 to 156 (Loop 143 mutant), found proximal to the 525 to 553 loop, also had significant sequence deviation between BEI and BEII enzymes (Fig. 4B). When both this loop and the 525 to 553 loop were replaced in BEI with the sequence found in rBEIIb (Fig. 4C), dramatic differences were seen in the activity. First the overall activity was significantly decreased when amylose was used as substrate (Table 1). Second, the branch chain specificity of BEI was converted from a preference for longer (11–12 glucose units) chains to an almost exclusive preference for chains of 6 to 7 glucose units (Fig. 5, C and D), similar to that seen for rBEIIb. We therefore conclude that these two loops work together to control the branch chain specificity in rice, and likely other, plant branching enzymes.
The unique function and specificity of branching enzymes, their role in synthesizing and modifying growing polymeric substrates, their relatively imprecise, though widely divergent transfer chain specificities, depending on species or even isoform, and their ability to properly space branch chains, make them relatively unusual enzymes, given the high specificity for substrate and product of most enzymes. Though a number of branching enzyme structures are known (
), many of which are bound to malto-oligosaccharides, the structural details that give rise to the unique characteristics of BEs remained mysterious.
M12 is the largest oligosaccharide to be observed at atomic resolution bound to a BE and elaborates a binding surface starting at a point distal from the active site, and stretching nearby, but not entering, the active site. Numerous mutations confirm the importance of the binding surface for the catalytic activity of the enzyme, and the residues interacting with M12 are highly conserved in eukaryotic (GH13 sub-family 8) BEs. Together this leads to the hypothesis that the M12 binding surface defines a part of either the donor or acceptor chain binding site. Notably, the Y487A mutation, though quite far from the active site, displays very low activity (0.61% ± 0.05 of wild-type rBEI). If we suppose that M12 represents the donor chain binding site, this tyrosine can be involved in interactions between the donor chain and the enzyme only if rBEI is transferring chains longer than ten glucose units. The fact that the fraction of transferred chains smaller than 11 glucose units still accounts for almost 15% of all transferred chains is inconsistent with only 0.61% ± 0.05 activity. In addition, the fact that none of the mutations to this binding surface resulted in any noticeable change in transfer chain length also argues against its involvement in donor chain binding (
) shows for the first time a donor chain bound in the active site of a BE. Many of the residues that define the donor chain binding in the M7-bound Cyanothece BE structure are conserved in bacterial and eukaryotic BEs, including rBEI, suggesting that all BEs use a similar donor chain binding surface. The M7 binding surface follows a trajectory distinct from that of the rBEI M12 surface (Fig. 6). This leads naturally to the conclusion that the M12 binding surface represents part of the acceptor chain binding site. Using the M7-bound Cyanothece BE structure as a guide, an M7 was modeled into the putative donor chain binding surface of M12-bound rBEI (Fig. 6). As shown, the donor chain runs between the 525 to 553 and 146 to 157 loops, in a crevice that links the donor strand binding surface and the CBM48 domain. On the other hand, there is no overlap between the putative donor and M12 acceptor chain binding sites, though they are proximal. Interestingly, the flexible 525 to 553 loop is located between the putative donor and acceptor chain binding sites. This is the same loop that is substantially extended in BEIIb isoforms. As previously discussed, replacement of the rBEI loop sequence with that of rBEIIb reduced the activity of the enzyme and resulted in a significant change in branch chain specificity toward shorter chains, more similar to that of rBEIIb (Fig. 5). Simultaneously exchanging both this loop and the 146 to 156 loop substantially altered branch chain specificity, essentially converting rBEI into an rBEIIb in its product specificity. Thus rBEI, which is the isoform that prefers transfer of the longest chains of any of the isoforms (preferring chains of M11 to M20–M30), is converted to an rBEIIb-like enzyme, that transfers almost exclusively only the shortest, M6 and M7 chains. Together, these results pinpoint the region of plant BEs that is responsible for the specificity of the enzyme for the first time. Notably the 146 to 156 loop represents the end of the N-terminal domain. A previous publication, using maize BEI and BEIIb chimeras, implicated the N-terminal domain in branch chain transfer specificity, consistent with our observation regarding the importance of this loop in the same activity in rBEI and rBEIIb (
). This represented one of the few studies, previous to the present one, that suggested which regions of the enzyme were responsible for this specificity. This second loop lies on the opposite side of the donor chain binding site (Fig. 6) such that the two loops surround the nonreducing end of the donor chain, exactly where they would be expected to be to play a role in branch chain specificity. This serves to confirm that the donor chain trajectory is very similar in rBEs to that of Cyanothece BE, making it likely that all eukaryotic BEs share a common donor chain binding surface. Further, it seems that loops on both sides of the donor chain are required for controlling donor chain length. We hypothesize that the longer 525 loop found in BEII enzymes reaches over the donor chain binding surface, interacts with the 146 to 156 loop and the end of an M6 or M7 donor chain to select for shorter donor chains. It is interesting to note that a different loop in Cyanothece BE occupies the space of the 525 loop, interacts with the nonreducing end of the M7 glucose unit, and likely provides some of the specificity for shorter glucan chains seen in the Cyanothece enzyme. This loop is not conserved in other bacterial enzymes, many of which have branch chain specificities distinct from that of Cyanothece BE. The proximity of the 525 loop to both donor and acceptor chains and the fact that residues in this loop make direct interactions with the putative M12 acceptor chain in the M12-bound rBEI structure suggest the possibility that there is allosteric communication between donor- and acceptor-binding sites such that binding of one does not inhibit the binding of the other in the active site. Their proximity also suggests that there may be interaction between donor and acceptor chains when both are bound, as suggested for pullulanases (
). Together, these results suggest that interaction between donor and acceptor chains may be required for acceptor chain binding within the active site. Such a requirement would prevent acceptor chain binding from inhibiting the donor chain from occupying the active site. This would explain why the M12 chain, though making numerous interactions with the surface of the enzyme, nonetheless does not reach into the active site.
Other questions arising from the structure regard the size and shape of the acceptor glucan chain–enzyme interface. Since many of the interactions occur in the groove of the helical glucan, it would appear to preclude binding of a double helical strand in this binding site. Further, while the 6-hydroxyl groups of the four glucoses closest to the active site (Fig. 7) are at least partially buried in the binding site, with many making direct interactions with rBE1, the last eight 6-hydroxyls of the malto-oligosaccharide point away from the protein surface and are not blocked by protein interactions, indicating that while the first four positions would appear to have difficulty accommodating a branch, the last eight positions would appear to easily accommodate a branched oligosaccharide. This arrangement should inhibit the binding of an oligosaccharide that contains a branch on a sugar within 4 to 6 glucan units of the acceptor site, therefore preventing rBE1 from attaching branches too close together, while still allowing reaction with an acceptor chain branched at a sugar position more than 6 units from the acceptor site, consistent with previous studies demonstrating rBE1 reactivity with branched acceptor chains (
In conclusion, with the insights gained from the recent donor-chain bound Cyanothece BE structure and the acceptor-chain bound rBEI structure described here, combined with the mutagenesis results that define the regions responsible for controlling donor chain specificity in plant isoforms, an atomic resolution picture of the critical branching process in the dynamic biosynthesis of the starch granule finally begins to emerge.
Full-length rice BEI gene was obtained from the National Institute of Agrobiological Sciences in Japan (Sciences, National Institute of Agrobiological, J. Rice Genome Resource Center, n.d.). However, the sequence of the gene delivered differed by four residues (L40M-V280M-S443P-T669A) from the complete DNA sequence found on the Rice Genome Resource Center’s website http://ricexpro.dna.affrc.go.jp/GGEP/, using the keyword AK119436. The pet-28b vector and the BL21 codon plus expression cells were purchased from Novagen and Agilent Technologies, respectively. Crystallization solutions (Crystal Screen 1 & 2, PEG/Ion 1 & 2, Salt Rx 1 & 2 and Index) were commercially available from Hampton Research. Amylose substrate used in assays was purchased from Sigma-Aldrich (CAS number: 9005-82-7, Amylose from potato). Maltododecaose samples were provided by Dr Park’s lab. (See supporting information for more details, section 5).
The full-length rBEI∗ gene (with four mutations, L40M, V280M, S443P, T669A) was originally cloned into an adjusted pet-28a vector encompassing histidine and Sumo tags upstream of the gene of interest, giving rise to the following sequence: pET-28-His-Sumo-BamHI-RBEI∗-XhoI. Initially, all crystallization and activity studies were carried out with rBEI∗. After fixing all the mutations, activity assays (kinetics and chain length distribution assays) were repeated with the wild-type rBEI and mutants. The crystallization of wild-type rBEI did not produce diffracting crystals.
Starting from pET-28-His-Sumo-BamHI-RBEI-XhoI, looping out the N-terminal leader 65-residue sequence from the full-length cDNA generated the mature rBEI. In order to cleave the C-terminal 60 residues, a TEV-protease cut site was introduced. The TEV-protease strategy was employed because attempts at expressing the C-terminally truncated protein were unsuccessful, and as described below, crystallization of the full-length protein proved problematic. Mutations on the mature wild-type rBEI were implemented in the study. The cells were grown at 37 °C to 0.7 OD600, and the expression was induced with 1 mM IPTG at 16 °C. Overnight cultures were collected at 50,000g at 4 °C, sonicated for 6 min (15 s on, 45 s off, 50%) at 4 °C. The supernatant was collected at 45,000g for 30 min at 4 °C. The protein was purified with Ni-NTA affinity resin using wash buffer [100 mM NaCl, 1 mM BME, 20 mM imidazole and 50 mM Tris (pH 8.0)] and elution buffer [100 mM NaCl, 1 mM BME, 200 mM imidazole, and 50 mM Tris (pH 8.0)]. The purified proteins were used for all activity and chain length specificity assays. Protein prepared for crystallization contained four mutations (L40M, V280M, S443P, T669A). After purification by Ni-NTA affinity chromatography, the protein to be crystallized was immediately cleaved with TEV-protease and sumo-protease (0.5 mg protease for 50 mg protein, overnight, at 4 °C) sequentially, leading to a 702-residue protein, truncated by 60 residues at its c-terminus, a final construct almost identical to that previously crystallized. This construct was found to have essentially identical activity to that of full length, wild-type rBE1. The protein was further purified by size-exclusion chromatography (Superdex 200 16/60 column from GE healthcare), in a buffer of 100 mM NaCl and 50 mM Tris-Cl (pH 8.0). If there were any impurities after this step (usually His-Sumo-protease and His-TEV-protease), they were removed by trapping them in an Ni-NTA affinity column. The PCR primers for mutagenesis are listed in Table S1.
The final purified proteins were concentrated to 3.0 mg/ml (Nano-drop, MW = 82,000 g/mol, ε = 133,000 M−1 cm−1) in the size-exclusion chromatography buffer. Attempts to crystallize the full-length 820-residue protein using a multitude of conditions (from Hampton Research mentioned in the Experimental procedures section above) at 4 °C and 25 °C employing the hanging drop method failed to produce diffracting crystals. The mature BEI produced crystals that either did not survive or gave poor resolution upon soaking in dodecaose. On the other hand, soaking BEIΔC crystals in the polymer did no damage to these and the best crystals grew in 30% PEG8K, 550 mM sodium acetate, and 100 mM sodium cacodylate (pH 6.9) in 1 week at room temperature. A maximum concentration of 45 mM dodecaose dissolved in the same buffer; and crystals were soaked in the polymer for 330 min, which was the longest soak time attempted before crystals were damaged. The crystals were then cryoprotected in conditions consisting of the growth solution in addition to 9.55% glycerol and flash frozen in liquid nitrogen.
The crystal structure of rBEI, soaked with M12 was determined by molecular replacement using the rice BEIΔC structure previously determined (PDB ID: 3AMK) as a search model. The data was refined to 2.35 Å, complete to 95.24%, and refined to an Rwork of 17% and Rfree of 23%. The X-ray diffraction data were collected from a single crystal at the Advanced Photon Source, LS-CAT 21-ID-G beamline. The diffraction images were processed using HKL2000. The structure was refined using PHENIX. The structure shows one molecule in the asymmetric unit and belongs to space group P212121 with unit cell parameters a = 47.67, b = 80.11, and c = 182.72 Å. Table 2 shows detailed data statistics. X-ray structure and coordinates were deposited in the PDB as 7ML5.
Table 2Crystallographic data collection and structure refinement statistics
Maltododecaose-bound rice branching enzyme I
P 21 21 21
a = 47.67, b = 80.107, c = 182.716, α = 90, β = 90, γ = 90
). As the amylose substrate is branched by the enzyme, the absorption of the glucan–iodine complex decreases. One unit of activity is defined as a decrease in absorbance of 1.0 per min at 30 °C at 660 nm and is measured in U/mg. Iodine/Iodate stock solution was made by dissolving 2.6 g of KI and 0.26 g of I2 in 10 ml water, and later it was diluted for running assays (1.95 ml Iodine/Iodate stock solution to 50 ml). Amylose stock solution was made by dissolving 50 mg of amylose in 2 ml water and 0.5 ml 10% NaOH (aq). In order to dissolve the amylose completely, the stock solution was heated by microwave for 30 s. Amylose working solution was made by taking one part freshly made amylose stock solution, one part 1 M sodium citrate buffer (pH 7.0), and eight parts waters with the final pH adjusted to 8.0 by adding HCl. After centrifugation at 11,000g for 5 min to separate undissolved solids, 900 μl amylose working solution was equilibrated at 30 °C. In total, 100 μl of enzymes stock solution (50 μg/ml) was added to the amylose working solution, resulting in a final enzyme concentration of 5 μg/ml, and 15 μl samples were taken into 0.985 ml Iodine/Iodate solution at 30 s intervals for 5 min and then at 6, 7, 10, 15, 20, and 60 min (and 180 min for some mutations). Activities were calculated based on the initial linear slope (1 min for relatively active mutations and the wild type, 60 min for less active mutations) and relative to the wild type.
Chain length specificity assay
To evaluate the branch chain specificity, amylose was initially branched by the rBEI enzymes as described in activity assay section and immediately the product of branching reaction was debranched by iso-amylase. The debranched mixture was analyzed with LC-TOF-MS to quantify the abundance of different sized oligosaccharides. For the chain length specificity assays, reaction was performed in the same way as activity assays. All reactions were stopped by adding 20 μl concentrated HCl when the absorption of glucan–iodine complex at 660 nm was 50% of its initial absorption (reaction times were estimated from the result of activity assays). After heating the reaction mixture in boiling water to denature the branching enzyme, the pH was adjusted to 3.5 to 4.0 by adding HCl. The debranching reaction was carried out overnight in 40 °C by adding 3 units of isoamylase HP (glucan 6-glucanohydolase) purchased from Megazyme Inc. The debranching reaction was stopped by adding 3× volume Methanol and centrifugation at 11,000g for 5 min to remove the precipitates. Solvent from the supernatant was removed using a speed-Vac. The resulting solids were dissolved in 50% methanol and desalted using HyperSep Hypercarb SPE cartridges and 25% acetonitrile as eluent. Chain length distribution was analyzed by LC-TOF-MS previously described by Vismeh et al. (
The authors declare that they have no conflicts of interest with the contents of this article.
Generous support was provided by the NIH (R01GM101353) and DOE (DE-FG02-06ER15822). All crystallographic data were collected at the Advanced Photon Source, an Office of Science User Facility operated for the US Department of Energy (DOE) Office of Science by Argonne National Laboratory, supported by the US DOE under contract no. DE-AC02-06CH11357. Use of the LS-CAT Sector 21 was supported by the Michigan Economic Development Corporation, the Michigan Technology Tri- Corridor (grant 085P1000817), and the MSU office of the Vice President for Research. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
H. N. G. and J. H. G. conceptualization; H. N. G., Remie Fawaz, and J. H. G. data curation; H. N. G., Remie Fawaz, and J. H. G. formal analysis; J. H. G. funding acquisition; H. N. G., Remie Fawaz, N. E., D. W., K. P., Raoul Fulgos, S. P., Z. A., A. G., and J. H. G. investigation; H. N. G., Remie Fawaz, N. E., S. P., and J. H. G. methodology; J. H. G. project administration; J. H. G. resources; H. N. G. and J. H. G. supervision; H. N. G., Remie Fawaz, and J. H. G. validation; H. N. G., Remie Fawaz, and J. H. G. visualization; H. N. G., Remie Fawaz, and J. H. G. writing–original draft; H. N. G. and J. H. G. writing–review and editing.
Comparison of starch-branching enzyme genes reveals evolutionary relationships among isoforms. Characterization of a gene for starch-branching enzyme IIa from the wheat D genome donor Aegilops tauschii.