Starch Synthase 3 isoforms are essential for normal starch granule initiation in wheat endosperm
Jinjin Ding, Brendan Fahy, Ryo Matsushima, Qiantao Jiang, David Seung
TL;DR
This study shows that two Starch Synthase 3 isoforms in wheat are crucial for normal starch granule formation and can be used to increase resistant starch.
Contribution
The study reveals the essential roles of SS3a and SS3b in starch granule initiation and their potential for modifying wheat starch properties.
Findings
SS3a and SS3b are essential for normal A-type starch granule initiation in wheat endosperm.
The double mutant of SS3a and SS3b showed more severe starch granule defects and increased resistant starch.
SS3a and SS3b interact with each other and with BGC1, a known granule initiation component.
Abstract
Wheat grains have two distinct types of starch granules. Large, lenticular A‐type granules are formed from a single initiation per amyloplast during early grain development, while numerous small B‐type granules are initiated during later grain development.Here, we demonstrate that the two isoforms of Starch Synthase 3 in wheat (SS3a and SS3b) are essential for normal starch granule initiation in the endosperm.The ss3a mutant of durum wheat had deformed A‐type granules, while the ss3b mutant had no detectable differences in starch granule morphology. However, the ss3a ss3b double mutant had an enhanced phenotype compared to ss3a – with more aberrant A‐type granules, a stronger reduction in granule sizes, and an increased relative volume of B‐type granules. Interestingly, the A‐type granules with defective morphology in the ss3a and ss3a ss3b mutants were ‘semicompound’, resulting from…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9- —Chinese Scholarship Council
- —John Innes Foundation10.13039/501100004034
- —Sichuan Science and Technology Program of China
- —Biotechnology and Biological Sciences Research Council10.13039/501100000268
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsFood composition and properties · Polysaccharides and Plant Cell Walls · Wheat and Barley Genetics and Pathology
Introduction
Starch is the principal component of cereal grains, in terms of both mass and calories. It is composed of glucose polymers (amylopectin and amylose) that are organised into semicrystalline, insoluble starch granules (Smith & Zeeman, 2020; Apriyanto et al., 2022). The starch of Triticeae cereals (e.g. wheat, barley and rye) is unique in that it has a bimodal distribution of starch granules in the endosperm, containing large A‐type starch granules and small B‐type granules (Tetlow & Emes, 2017; Seung & Smith, 2019). This distribution results from a distinct pattern of starch granule initiation during grain development. During the early stages of grain development (between c. 4 and 15 d post anthesis – dpa), a single large A‐type granule is initiated and synthesised in each amyloplast (Parker, 1985; Esch et al., 2023). The A‐type granules undergo a morphogenesis program, starting as a spherical granule, then developing annular outgrowths around the equatorial plane to form lenticular morphology (Evers, 1971). B‐type granules are initiated between 15 and 20 dpa, and at least partially in amyloplast stromules (Parker, 1985; Bechtel et al., 1990; Langeveld et al., 2000; Chen et al., 2024). Although > 90% of the granules in the endosperm are B‐type granules, they comprise only 8–30% of the total volume of starch due to their small size (Lindeboom et al., 2004; Kamble et al., 2023; Chen et al., 2024; McNelly et al., 2025).
The control of this distinctive starch granule initiation and morphogenesis pattern in wheat is not fully understood. However, some important molecular components involved in the process have been identified. For example, STARCH SYNTHASE 4 (SS4) and B‐GRANULE CONTENT 1 (BGC1) are important for establishing the single initiation during early grain development that leads to an A‐type starch granule (Hawkins et al., 2021). SS4 is a glucosyltransferase, whereas BGC1 is the wheat ortholog of PROTEIN TARGETING TO STARCH 2 of Arabidopsis, a nonenzymatic protein with a coiled coil and a Carbohydrate Binding Module 48 (CBM48) domain (Seung et al., 2017). Mutations in either SS4 or BGC1 in wheat (or the BGC1 ortholog, FLO6, in barley) lead to multiple initiations in most amyloplasts, often packing into ‘compound’ granule structures (Chia et al., 2020; Hawkins et al., 2021; Matsushima et al., 2023, 2024). In the barley mutant Franubet defective in FLO6, some of these multiple initiations form ‘semicompound’ granules, in which the internal structure is made from multiple initiations but is covered by a smooth exterior surface (Suh et al., 2004). BGC1 is also critical for B‐type granule initiation later in grain development, as reduction in its gene dosage in wheat reduces the number of B‐type granules, which contrasts with the multiple initiations and compound granule formation caused by the complete loss of BGC1 function during early grain development (Chia et al., 2020). BGC1 acts in B‐type granule initiation with the plastidial phosphorylase, PHS1 – loss of PHS1 in wheat reduces the number of B‐type granules (Kamble et al., 2023). In addition, MYOSIN RESEMBLING CHLOROPLAST PROTEIN (MRC) plays an important role in the correct timing of granule initiation, possibly by regulating BGC1 (Chen et al., 2024; Fahy et al., 2025). Interestingly, all of these proteins have been implicated in starch granule initiation in chloroplasts of Arabidopsis leaves (Roldán et al., 2007; Malinova et al., 2017; Seung et al., 2017, 2018; Vandromme et al., 2019), although there are differences in the effect of the mutations between the species/tissues (Watson‐Lazowski et al., 2022; Uauy et al., 2025). For example, although ss4 or bgc1 mutations in wheat increase the frequency of initiations during early grain development (Hawkins et al., 2021), mutation of either ss4 or ptst2 in Arabidopsis reduces the number of granules per chloroplast (Roldán et al., 2007; Seung et al., 2017). However, in both contexts, these mutations consistently affect granule number. Overall, these findings suggest that the distinct spatiotemporal pattern of granule initiation in wheat amyloplasts was formed by adapting existing granule initiation proteins.
Plants contain multiple Starch Synthase (SS) isoforms, and SS3 isoforms have the most sequence homology to SS4 (Patron & Keeling, 2005). SS3 isoforms have a conserved SS (glucosyltransferase) domain at the C terminus, but they can be distinguished from other SS isoforms by their long N‐terminal extensions that contain three tandem carbohydrate‐binding modules (Pfister & Zeeman, 2016). Cereal species have two SS3 isoforms – SS3a and SS3b. SS3a is the major isoform in the endosperm and is distinct from SS3 sequences from dicots in that they have a longer N‐terminal region containing amino acid repeat motifs (Gao et al., 1998; Li et al., 2000; Pfister & Zeeman, 2016). SS3 has mainly been implicated in amylopectin synthesis. Arabidopsis ss3 mutants have altered amylopectin structure in leaf starch (Zhang et al., 2008). Likewise, mutants defective in SS3a of maize (referred to as dull1), rice and wheat have altered amylopectin chain length structure in endosperm starch (Wang et al., 1993; Gao et al., 1998; Fujita et al., 2007; Fahy et al., 2022). In rice, SS3b is the major isoform in leaves, and mutants defective in this isoform have altered amylopectin structure in leaf starch (Morita et al., 2023). SS3a mutants of cereals also have elevated amylose content (Wang et al., 1993, 2023; Fujita et al., 2007; Huang et al., 2024) – resulting in increased resistant starch (Fahy et al., 2022; Wang et al., 2023; Huang et al., 2024), which is a form of dietary fibre that provides health benefits (Li et al., 2019). These effects of SS3 mutations on polymer structure and composition strongly contrast with mutations in SS4, which do not cause major changes in amylopectin chain length structure or amylose content in Arabidopsis and wheat, despite causing strong alterations in granule number and morphology (Szydlowski et al., 2009; Hawkins et al., 2021).
In addition to this canonical role in polymer biosynthesis, a potential role of SS3 in granule initiation is emerging, inferred mainly through conditional phenotypes. In Arabidopsis leaves, the ss3 mutant has no change in starch granule number per chloroplast (Vandromme et al., 2023), whereas the ss4 mutant has a strong reduction in the number of starch granules – with most chloroplasts containing no starch granules and a few chloroplasts containing one large granule (Roldán et al., 2007). However, the ss3 ss4 double mutant is almost completely starchless (Szydlowski et al., 2009). This demonstrates that SS3 is responsible for initiating the few granules in the ss4 mutant, but the severe phenotype of the ss4 mutant suggests that SS3 initiates granules with very poor efficiency (Szydlowski et al., 2009; Seung et al., 2016). In rice, the ss3a mutant accumulates granules that are smaller and more rounded than those of the wild‐type (WT), which are loosely packed in the endosperm (Fujita et al., 2007; Huang et al., 2024). The fact that rice produces compound starch granules makes it difficult to ascertain whether this is an initiation‐related phenotype. However, when mutated in combination with SS4b (in the ss3a ss4b double mutant), there is a strong reduction in starch content and a shift to making rounded, simple starch granules (Toyosawa et al., 2016). The wheat ss3a mutant has strong defects in A‐type granule morphology (Fahy et al., 2022), but it is not known whether this is due to an effect on granule initiation. Furthermore, the contribution of the SS3b isoform to wheat starch synthesis is not known.
In this study, we demonstrate that SS3a and SS3b isoforms play critical roles in granule initiation in developing wheat endosperm. We found that the defects in A‐type granule morphology in the ss3a mutant arise early during grain development due to supernumerary initiations, leading to the formation of semicompound starch granules. This phenotype is more pronounced in the ss3a ss3b mutant than in ss3a, suggesting overlapping functions among the two SS3 isoforms. The two isoforms interact with each other and with BGC1. Additionally, we demonstrate that eliminating both SS3 isoforms also leads to substantial increases in resistant starch in wheat.
Materials and Methods
Plant growth and materials
All durum wheat (Triticum turgidum L.) plants were grown in a climate‐controlled glasshouse fitted with supplemental LED lighting. The glasshouse was set to provide a minimum 16 h of light at 20°C, and 16°C during the dark, and constant relative humidity of 60%.
Mutant lines from the wheat in silico TILLING resource (http://www.wheat‐tilling.com) in durum wheat (cv Kronos) (Krasileva et al., 2017) were obtained from the Germplasm Resource Unit of the John Innes Centre. Genotyping of mutations was carried out using Kompetitive allele‐specific PCR (KASP) V4.0 genotyping (LGC, Teddington) with the primers in Supporting Information Table S1. Plants containing individual mutations in the A‐ or B‐homoeologs were crossed to combine the mutations. The WT segregant (AABB), single homeolog mutants (aaBB, AAbb) and the double homeolog mutant (aabb) were selected in the F_2_ generation using KASP genotyping. The bgc1‐3 mutant in Kronos was previously described in Hawkins et al. (2021).
Immunoblots and native gels
For immunoblot detection of SS3a, the endosperm was dissected from developing grains and homogenised with a pestle in extraction buffer [40 mM Tris–HCl, pH 6.8, 5 mM MgCl_2_, 2% (w/v) SDS, protease inhibitor cocktail (Roche)] at a ratio of 100 mg tissue per ml of buffer. The homogenate was heated at 95°C for 10 min, and insoluble material was pelleted at 20 000 ** g ** for 10 min. Proteins were collected in the supernatant, and 20 μl was loaded in each well. Blots were probed with a 1:20 000 dilution of the anti‐DU1N rabbit antiserum, which was raised against the N terminus of maize SS3a (Cao et al., 1999). For simultaneous detection of actin as a loading control, the anti‐actin antibody (Sigma‐Aldrich A0480) was used at 1:10 000 dilution.
For native gels, endosperm tissue was homogenised with a pestle in native protein extraction buffer [100 mM MOPS, pH 7.2, 1 mM EDTA, 5 mM DTT, 10% (v/v) glycerol, protease inhibitor cocktail (Roche)] at a ratio of 100 mg tissue per ml of buffer. Insoluble material was pelleted at 20 000 ** g ** for 10 min at 4°C. Soluble proteins were collected in the supernatant and mixed with Native PAGE loading buffer to a final concentration of 10% (v/v) glycerol, 0.01% (w/v) bromophenol blue. Samples were run on native PAGE gels containing 7.5% (w/v) acrylamide and 0.3% (w/v) glycogen, followed by activity staining in an adenosine diphosphate glucose‐containing buffer, exactly as described in Pfister et al. (2014).
Grain morphometrics and total/resistant starch contents of mature grains
Grain size was quantified using the MARViN seed analyser (Marvitech GmbH, Wittenburg). Total starch was quantified using the Total Starch HK Kit (Megazyme) according to the manufacturer's instructions, except the starting material was adjusted to 5–10 mg of flour, and all assay volumes were reduced by 6X. Resistant starch was quantified using the Resistant Starch Assay kit (Megazyme, Bray, Ireland), which was scaled down by reducing assay volumes by 10×.
Analysis of starch granule morphology and composition
Starch granules were purified from mature grains using a 70 μm nylon mesh and a Percoll cushion, exactly as described in Kamble et al. (2023). Granule size distributions were quantified with a Multisizer 4e Coulter Counter (Beckman Coulter, Brea, CA, USA) fitted with a 70‐μm aperture and using Isoton II electrolyte (Beckman Coulter). To derive values for mean A‐type and B‐type granule size and B‐type granule content (as a percentage of total starch volume), we used a curve fitting method as described in McNelly et al. (2025).
For scanning electron microscopy (SEM), purified starch was suspended in water, and a droplet of the suspension was air‐dried onto a glass coverslip attached to an SEM stub. Stubs were sputter coated with gold and observed using a Gemini 300 FEG SEM microscope (Zeiss, Cambourne, UK).
To observe starch granules in mature grains using light microscopy, c. 1‐mm cubic blocks were cut from the centre region of the endosperm and fixed in a solution containing 5% (v/v) formalin, 5% (v/v) acetic acid and 50% (v/v) ethanol for at least 12 h at room temperature. The procedures of resin embedding using Technovit 7100 resin (Kulzer, Tokyo, Japan) were described previously (Matsushima et al., 2023). Semithin sections (c. 1 μm) were prepared and stained with 40 times diluted Lugol solution and subsequently examined under a microscope.
Analysis of total starch content and granule morphology in developing grains
For analysis of total starch and starch granule properties in developing grains, we used the method in Kamble et al. (2023). Briefly, endosperms were dissected from developing grains collected at 8, 14, 18 and 22 dpa and homogenised in 0.7 M perchloric acid. Homogenates were spun at 10 000 ** g ** for 5 mins, and the pellet was resuspended in ddH_2_O and equally divided into two fractions.
One fraction was used to quantify starch content. The fraction was heated at 95°C for 15 mins to gelatinise the starch and digested to glucose with ⍺‐amylase (Megazyme) and amyloglucosidase (Roche). Glucose was quantified using the hexokinase/glucose 6‐phosphate dehydrogenase assay (Roche).
The other fraction was used for starch purifications (as described previously for mature grains), followed by analysis on the Multisizer 4e Coulter counter (Beckman Coulter) fitted with a 70‐μm aperture tube and running in volumetric mode (analysing 2 ml of the suspension). This gave the number of granules in the suspension, which could be used to calculate the number of granules per starting fresh weight of endosperm. These data were also used to plot granule size distributions.
Amylose content
Purified starch (1 mg) was suspended in 200 μl water and solubilised with 200 μl of 2 M NaOH. Solubilisation was carried out at room temperature overnight. The solution was neutralised to pH 7 with 1 M HCl. The starch solution (5 μl) was diluted in 220 μl of water and 25 μl Lugol's iodine solution (Merck Life Science, Gillingham, UK) was added, followed by incubation at room temperature for 10 min to allow colour development. The absorbance at 535 and 620 nm was determined, and the apparent amylose content was estimated using the formula in Washington et al. (2000).
Phylogenetic analyses
Amino acid sequences of SS3 homologs in wheat were obtained from Ensembl plants using the bread wheat (Triticum aestivum L.; Refseq v1 – Appels et al. (2018)) and durum wheat (Svevo v1 – Maccaferri et al. (2019)) genomes, and homologs from other plant species were obtained using the Phytozome database (Goodstein et al., 2012). Sequences were aligned using Mafft (Katoh & Standley, 2013) and edited to only include regions with high‐quality alignment (Dataset S1). A maximum likelihood tree was built using raxML (Edler et al., 2020) using the BLOSUM62 model and 1000 bootstrap replicates.
Protein interaction assays
For cloning vectors for protein interaction studies, the TaSS3a‐B1 and TaSS3b‐A1 coding sequences (CDSs) (from TraesCS1B02G119300.1 and TraesCS2A02G468800.1, respectively) were synthesised into the pUC‐GW‐Kan vector by GeneWiz, flanked with attL sites. This vector was used directly in LR clonase reactions to recombine the full‐length CDS into the appropriate Gateway‐compatible destination vectors.
To clone the three SS3a fragments, the CDSs SS3a‐N (amino acids 1–725), SS3a‐CBM (amino acids 726–1153) and SS3a‐GT (amino acids 1154–1605) were amplified from the TaSS3a‐B1:pUC‐GW‐Kan vector using the attB‐flanked primers in Table S1. For cloning SS3b without the predicted transit peptide (amino acids 71–1307), the CDS was amplified from the TaSS3b‐A1:pUC‐GW‐Kan vector using the attB‐flanked primers in Table S1. These attB‐flanked coding sequences were recombined into the Gateway Entry vector pDONR221 using BP Clonase II (Life technologies, Paisley, UK), then subsequently recombined into Gateway‐compatible destination vectors using LR clonase II (Life technologies). Gateway entry vectors containing CDS for TaBGC1, TaMRC, TtPHS1 and TaSS4 sequences were from previous publications (Hawkins et al., 2021; Kamble et al., 2023; Chen et al., 2024).
For yeast‐2‐hybrid (Y2H) assays, CDSs were recombined directly into Gateway‐compatible pGADT7 (in frame with N‐terminal Activation Domain) or pGBKT7 (in frame with N‐terminal Binding domain). The assays were conducted as described in Chen et al. (2024).
For pairwise immunoprecipitation assays, CDSs were recombined directly into Gateway‐compatible pJCV52 [in frame with C‐terminal Hemagglutinin (HA)‐tag] or pB7RWG2 [in frame with C‐terminal Red Fluorescent Protein (RFP)]. Expression in Nicotiana benthamiana, protein extraction and immunoprecipitation was conducted as described in Kamble et al. (2023), except using μMACS magnetic beads conjugated to anti‐HA (Miltenyi Biotec, Bisley, UK). For chemiluminescence‐based immunoblot detection of fusion proteins, we used anti‐HA (Abcam plc; ab9110–1:5000) and anti‐RFP (Abcam plc; ab34771–1:2000) primary antibodies, and horseradish peroxidase–coupled secondary antibody anti‐rabbit HRP (Sigma; A0545), 1:15 000.
Results
SS3a but not SS3b impacts granule morphology in the wheat endosperm
To understand the contributions of SS3a and SS3b to endosperm starch synthesis in wheat, we identified the orthologs of both enzymes in bread wheat (Triticum aestivum) and durum wheat (Triticum turgidum) (Fig. S1a). Phylogenetic tree analyses confirmed that TaSS3a is encoded on Group 1 chromosomes (TraesCS1A02G091500, TraesCS1B02G119300, TraesCS1D02G100100) and TaSS3b is encoded on Group 2 chromosomes (TraesCS2A02G468800, TraesCS2B02G491700, TraesCS2D02G468900) (Fig. S1b). The equivalent loci in the durum wheat genome are TtSS3a (TRITD1Av1G036690, TRITD1Bv1G048610) and TtSS3b (TRITD2Av1G261450, TRITD2Bv1G224370), and these shared >99% amino acid identity to the bread wheat sequences (Table S2). All SS3a and SS3b gene models of wheat consisted of 16 exons, but SS3a genes had a longer exon 3 than SS3b (Fig. S1a), which encoded a longer N‐terminal extension in SS3a. Overall, wheat SS3a isoforms shared 57–73% identity with SS3b isoforms, depending on which homoeologs were used in the comparison (Table S2). Amino acid sequence alignment of these wheat isoforms with SS3 sequences from rice and maize showed high sequence homology in the C‐terminal glucosyltransferase domain and in the later part of the N‐terminal region that contained the CBMs (Fig. S2). However, the earlier part of the N‐terminal region aligned poorly between SS3a and SS3b sequences, since SS3b does not contain the longer N‐terminal extension. This extension is specific to SS3a isoforms (referred to as the ‘N‐terminal variable repeat region’ in Li et al. (2000), from the end of the transit peptide to near the start of the CBM domains). While some parts of this region aligned well among SS3a isoforms, other parts aligned poorly as there was variability in length and sequence between species, as previously noted (Li et al., 2000).
We then assessed the expression pattern of SS3a and SS3b. First, we used the RNAseq dataset from Ma et al. (2021) to compare transcript levels between different organs of bread wheat (Fig. S3a). TaSS3a transcripts were primarily detected in the spike and grains but were almost undetectable [Transcripts per Million (TPM) < 1] in roots, stems and leaves. By contrast, TaSS3b transcripts were detected in all tissues examined. Second, we used our own RNAseq dataset in durum wheat to assess transcript levels in the developing endosperm (Chen et al., 2023) (Fig. S3b). For TtSS3a, transcripts were present at all time points between 6 and 30 dpa, but peaked between 10 and 20 dpa. TtSS3b transcripts were also detected at all time points, but at levels that were c. 10‐fold lower than TtSS3a, and at relatively similar levels across the time points.
We explored the impact of SS3a and SS3b mutations on endosperm starch synthesis. We used the wheat TILLING mutant population in durum wheat Kronos (Krasileva et al., 2017) to identify mutant lines in SS3a and SS3b. All mutant lines carried premature stop mutations in exon 3 of SS3a and SS3b genes (Fig. 1a). For SS3a, we selected the K2505 and K3868 lines carrying mutations in the A and B genome copies, respectively. These parent lines were crossed to combine the mutations, then KASP genotyping was used to isolate the WT segregant control (AA BB), the single homeolog mutants (aa BB and AA bb) and the double mutant (aa bb). A similar procedure was done for SS3b, but selecting the K4276 and K2330 lines for mutations in the A and B‐genome copies.
Isolation of durum wheat mutants defective in SS3a or SS3b. (a) Gene models of TtSS3a and TtSS3b homeologs in durum wheat. Boxes represent exons. The position of premature stop codon mutations in the TILLING mutants are marked with red arrowheads. Mutants in individual homeologs were crossed, and we isolated the single homeolog mutants (aaBB and AAbb), the double homeolog mutant (aabb) and the wild‐type (WT) segregant control (AABB) for both SS3a and SS3b. (b) Plant photograph of SS3a and SS3b mutants compared to the Kronos WT. (c) Photograph of mature grains from SS3a and SS3b mutants. Bar, 1 cm.
We then analysed plant growth and grain characteristics of the ss3a and ss3b mutants. Both ss3a and ss3b mutant plants were indistinguishable from the WT (Fig. 1b). The grains were also indistinguishable (Fig. 1c), and there were no significant differences in Thousand Grain Weight (TGW) or grain size (area, width and length) (Fig. S4).
Next, we characterised endosperm starch in the mutants. There were no statistically significant differences in total starch content in any of the mutant lines (Fig. 2a). We then purified starch granules from mature grains and examined starch granule morphology using SEM. Consistent with the findings of Fahy et al. (2022), there were strong alterations in granule morphology in ss3a aabb, particularly the formation of lobed A‐type granules (Figs 2b, S5). However, no alterations in granule morphology were apparent in the ss3b aabb mutant or in any of the single homeolog mutants. Quantitative analysis of starch granule size distributions on the Coulter counter showed that all genotypes had a bimodal distribution of starch granule size. However, in the ss3a aabb starch, the position of the A‐type granule peak was shifted to the smaller size range (Fig. 2c), whereas the size distribution was not altered in ss3b aabb (Fig. 2d). We used a curve fitting method to derive A‐ and B‐type granule parameters from the distribution curves (McNelly et al., 2025). The B‐type granule content of ss3a aabb (as relative percentage of total starch volume) was significantly higher than those of the controls (Fig. 2e). Indeed, ss3a aabb had a significant decrease in A‐type granule diameter (Fig. 2f), while the B‐type granule size remained unchanged (Fig. 2g). No differences in any of the granule size parameters were observed in the ss3b aabb mutants relative to its WT segregant control.
Durum wheat SS3a mutant has highly defective starch granule morphology. For all analyses, we analysed the single homeolog mutants (aaBB and AAbb), the double homeolog mutant (aabb), the wild‐type (WT) segregant control (AABB) and the Kronos WT. Panels (a, c–g) show the mean ± SE for n = 3 replicates, where each biological replicate represents grains harvested from a different plant. Statistical groupings are from a one‐way ANOVA and Tukey's HSD test, where groups with different letters are significantly different (P < 0.05) (a) Total starch content of mature grains. (b) Granule morphology observed using scanning electron microscopy. Bars, 10 μm. (c, d) Volumetric granule size distributions quantified with a Coulter counter. Panel (c) shows the mutants in SS3a, while panel (d) shows mutants in SS3b. Note that the single homeolog mutants are not shown for clarity, as they were identical to the WT controls. (e–g) A‐ and B‐type granule size parameters calculated using curve fitting on the size distribution data from (c, d).
The defects in granule morphology in the ss3a mutant stem from aberrant granule initiation
The defects in starch granule morphology observed in the ss3a mutant were largely consistent with what was observed in ss3a mutants of hexaploid bread wheat (Fahy et al., 2022). However, the mechanism by which the aberrant morphology forms was not known. We therefore examined starch granule morphology and number during grain development, analysing 8, 14, 18 and 22 dpa to cover both A‐ and B‐type granule initiation and growth. Total starch content during grain development appeared to be similar between the ss3a aabb and its controls at 8 dpa, but a small significant decrease in starch content in this mutant became apparent from 14 dpa onwards (Fig. 3a). The starch content of the ss3b aabb mutant was identical to the controls at all time points. We then used the Coulter counter to calculate the number of granules per unit of endosperm. There were no significant differences in the total number of starch granules in either ss3a aabb or ss3b aabb mutants at any timepoint (Fig. 3b).
Semicompound starch granule formation during early grain development in the durum wheat ss3a aabb mutant. Developing grains were harvested at 8, 14, 18 and 22 d post anthesis (dpa) from the Kronos wild‐type (WT), WT segregant control (AABB) and the double homeolog mutant (aabb) for both SS3a and SS3b. The endosperm was dissected before all analyses. (a) Total starch content of the endosperm. (b) Number of starch granules per unit endosperm weight. Panels (a, b) show the mean ± SE for n = 3 replicates, where each biological replicate represents grains harvested from a separate plant. Statistical groupings are from a one‐way ANOVA and Tukey's HSD test within each time point, where groups with different letters are significantly different (P < 0.05). (c) Starch granule morphology viewed under scanning electron microscopy and polarised light microscopy. Bars, 10 μm. (d) Close‐up images of semicompound starch granules of the ss3a aabb mutant and WT segregant at 8 dpa, viewed under polarised light microscopy. Red arrows mark examples of multiple crosses within granules. Bars, 10 μm. (e) As for D, but for starch purified from mature grains. Bars, 10 μm.
We then examined granule morphology in the developing grains. Consistent with our findings from mature grains, only the ss3a aabb mutant had altered granule morphology (Fig. S6). Strikingly, the lobed granules of the ss3a aabb mutant were already present at 8 dpa (Fig. 3c), suggesting that the altered granule morphology is independent from the formation of B‐type granules, which were initiated in the WT between 14 and 18 dpa (Figs S6, S7). Since the lobed granule morphology could arise from the aggregation or fusion of multiple granules, we examined the granules in developing (8 dpa) and mature grains from the ss3a mutant using polarised light. Indeed, multiple initiation centres were observed in almost all of the lobed granules in ss3a in both immature and mature grains (Fig. 3d,e). When multiple granules aggregate within the amyloplast, they most commonly form ‘compound’ granules – which have a distinct tessellated appearance and fall apart into individual polygonal granulae during extraction (e.g. the compound granules of rice, or the wheat ss4 mutant; Hawkins et al. (2021)). However, ss3a granules were distinct in that they had a smooth exterior surface where no tessellation pattern was apparent, and the multiple initiation centres were only visible under polarised light. The multiple initiations also remained joined during starch extraction. This type of granule was reported before in the barley ‘Franubet’ mutant defective in PTST2/FLO6/BGC1 and is referred to as ‘semicompound’ (Suh et al., 2004).
It therefore appears that SS3a plays an essential role in controlling the number of initiations during early grain development, facilitating proper A‐type granule formation. Loss of SS3a results in the formation of semicompound granule morphology from supernumerary initiations per amyloplast.
SS3b only partially compensates for the loss of SS3a in granule initiation
Given that SS3a and SS3b are both expressed in the endosperm (Fig. S3), we wondered whether the presence of SS3b can partly compensate for the loss of SS3a. We therefore generated a ss3a ss3b double mutant by crossing the ss3a aabb and ss3b aabb lines. The double mutant (ss3a ss3b), single mutant segregants (ss3a and ss3b) and WT segregant (WT seg.) were isolated in the F_2_ generation. The ss3a ss3b double mutants were phenotypically indistinguishable from the control lines (Fig. 4a) and produced grains of normal appearance (Fig. 4b). There were no significant alterations in TGW or grain size parameters (Fig. S8).
Isolation of durum wheat defective in both SS3a and SS3b. The cross between the ss3a aabb and ss3b aabb mutant was used to segregate out the ss3a ss3b double mutant defective in all homeoalleles of both genes, as well as the single mutants defective in all homeoalleles of either gene (ss3a or ss3b) and the wild‐type segregant (WT seg.). (a) Plant photograph of mutants. Bar, 10 cm. (b) Photograph of mature grains. Bar, 1 cm.
We then characterised endosperm starch in the double mutant. Total starch content was significantly reduced in both the ss3a and ss3a ss3b mutants compared with the WT control, but there was no significant difference between ss3a and ss3a ss3b mutants (Fig. 5a). The reduction in starch content in the ss3a mutant was statistically significant in this experiment, although it was not in our previous experiment (Fig. 2a). This suggests that the reduction in starch content in ss3a is modest and close to the limit of reliable detection. Coulter counter analysis of purified starch granules followed by curve fitting analyses revealed that ss3a ss3b had a significantly stronger reduction in A‐type granule size and increase in B‐type granule content compared with ss3a (Fig. 5b–d). The size of B‐type granules in the single and double mutants was significantly smaller than the WT segregant (Fig. 5e). Similar to the ss3a mutant, we observed the presence of aberrant, semicompound granules in the ss3a ss3b double mutant under polarised light microscopy (Fig. 5f). We scored the presence of such aberrant granules with multiple crosses and found that > 70% of the large A‐type granules in the ss3a single mutant had semicompound morphology, but this rose to almost 90% in the ss3a ss3b double mutant (Fig. 5g).
Starch granule morphology in the durum wheat ss3a ss3b double mutant – Part I. The cross between the ss3a aabb and ss3b aabb mutant was used to isolate the ss3a ss3b double mutant defective in all homeoalleles of both genes, as well as the single mutants defective in ss3a or ss3b, and the wild‐type segregant (WT seg.). These were compared with the Kronos WT. Panels (a–e) and (g) show the mean ± SE for n = 3 replicates, where each biological replicate represents grains harvested from a different plant. Statistical groupings are from a one‐way ANOVA and Tukey's HSD test, where groups with different letters are significantly different (P < 0.05) (a) Total starch content of mature grains. (b) Volumetric granule size distributions quantified with a Coulter counter. (c–e) A‐ and B‐type granule size parameters calculated using curve fitting on the size distribution data from (b). (f) Close‐up images of semicompound starch granules of the ss3a aabb mutant and WT segregant at 8 dpa, viewed under polarised light microscopy. Red arrows mark examples of multiple crosses within granules. Bars, 10 μm. (g) Frequency of semicompound starch granules. The first 100 starch granules observed that were greater than 10 μm were assessed with polarised light microscopy. Those that had multiple initiation centres visible were counted as semicompound.
The strong prevalence of these lobed granules was also evident under SEM, where most of the granules in the ss3a ss3b double mutant had multiple lobes and were smaller than those of the ss3a single mutant (Fig. 6a). These changes in granule size and morphology were also evident in thin sections of mature grains (Fig. 6b). Overall, the occurrence of aberrant granule morphology from semicompound granule formation was much more pronounced in the ss3a ss3b double mutant than in the ss3a single mutant, suggesting that both SS3a and SS3b can influence the control of granule initiation – but SS3a plays the dominant role, and the presence of SS3b only partially compensates for the loss of SS3a.
Starch granule morphology in the durum wheat ss3a ss3b double mutant – Part II. The cross between the ss3a aabb and ss3b aabb mutant was used to segregate out the ss3a ss3b double mutant defective in all homeoalleles of both genes, as well as the single mutants defective in ss3a or ss3b, and the wild‐type segregant (WT seg.). The bgc1‐3 mutant was included for comparison. (a) Scanning electron microscope observation of purified starch granules. Bars, 10 μm. (b) Light micrograph observation of iodine‐stained thin sections of endosperm, made from mature grains. For all panels, examples of lobed, semicompound starch granules are marked with red arrows, while examples of compound granules in bgc1‐3 are marked with yellow arrows. Bars, 20 μm.
SS3 isoforms interact directly with granule initiation proteins, including BGC1
Since our genetic analyses suggested that SS3a and SS3b share a function in granule initiation, we explored the biochemical interactions of the two isoforms. We first performed immunoblots of SS3a in developing endosperm extracts from the mutants using an antibody raised against the N terminus of maize SS3a (anti‐DU1N; Cao et al. (1999)). This maize antigen shares some homologous regions with wheat SS3a (Fig. S2) and has been shown to react with wheat SS3a in Fahy et al. (2022). As observed in that study, the antibody recognised a double band corresponding to SS3a, present in the WT controls and absent in the ss3a aabb mutant (Fig. 7a). Interestingly, the intensity of the bands decreased substantially in the AAbb single mutant, but not in the aaBB single mutant, suggesting that TtSS3a‐B1 may be the dominant homeolog in durum wheat. Although the predicted molecular weight (MW) of the mature TtSS3a‐A1 and TtSS3a‐B1 homeologs are 177.5 and 175.6 kDa respectively, both homeologs migrated above the 250 kDa marker. The reason for the anomalous migration is not known but was previously observed for SS3a in wheat (Fahy et al., 2022) and maize (Cao et al., 1999). We then analysed the genotypes isolated from the cross between ss3a and ss3b. SS3a could not be detected in the ss3a single or ss3a ss3b double mutant but was detectable in the ss3b mutant at a comparable level to the WT, suggesting that there is no change in SS3a abundance following the loss of SS3b (Fig. 7b).
Immunoblots and native gel analyses of durum wheat ss3a ss3b mutants. (a) Immunoblot of wheat endosperm extracts from the SS3a single homeolog mutants (aaBB and AAbb), the double homeolog mutant (aabb) and the wild‐type (WT) segregant control (AABB). The Kronos WT is included for reference. Blots were probed with the anti‐DUL1N antibody (raised against the N terminal of maize SS3a) or anti‐actin as a loading control. The location of molecular mass markers in kDa is marked on the left. Gels were loaded on an equal fresh weight basis (1.5 mg per lane). (b) Same as (a), but with genotypes from the ss3a × ss3b cross. (c) Native PAGE analysis of endosperm extracts. Bands corresponding to SS3a and SS3b are indicated. Gels were loaded on an equal fresh weight basis (6.4 mg per lane). (d) Same as (c), but with genotypes from the ss3a x ss3b cross.
We then conducted native PAGE analysis to observe SS3 activity. Notably, a band with low mobility (‘Upper’ band) was present in all WT controls, but absent in both the ss3a aabb and ss3b aabb mutants (Fig. 7c). The fact that this band disappears in both mutants suggested that it could represent a heteromer containing both SS3a and SS3b subunits. A faster migrating ‘Lower’ band was also observed in the WT controls and both ss3a and ss3b single mutants but was abolished in the ss3a ss3b double mutant (Fig. 7d). This lower band therefore could represent SS3a and SS3b subunits that are not engaged in heteromer formation, and thus in a lower molecular weight configuration.
Next, we used Y2H to screen whether the SS3 isoforms could interact with each other and granule initiation proteins (BGC1, MRC, PHS1 and SS4). Screening with SS3a revealed that it could interact with itself, as well as with BGC1 and MRC (Fig. 8a). A similar screen using SS3b revealed that it could also interact with BGC1, but not MRC (Fig. 8b). Importantly, SS3b could also interact with both itself and SS3a, consistent with the observation of a potential SS3a‐SS3b heteromer on native PAGE gels (Fig. 7c,d).
Interaction between wheat SS3 and BGC1. For all protein–protein interaction experiments, we used amino acid sequences from bread wheat, except PHS1, where the durum wheat sequence was used. (a–c) Yeast two‐hybrid experiments. Proteins fused to either the Activation Domain (AD) or Binding Domain (BD) are indicated to the left of each panel. The large T‐antigen (T) and tumour suppressor p53 pair represents the positive control, whereas the T and lamin (Lam) pair represents the negative control. The double dropout medium (–LW) allows growth of all yeast strains carrying the plasmids, whereas the quadruple dropout medium (–LWAH) selects for yeast strains in which the AD and BD domains functionally interact. All panels show yeast colonies spotted at 1/1, 1/10, 1/100 and 1/500 dilutions, from left to right. (a) Interactions between SS3a and granule initiation proteins. (b) Interactions between SS3b and granule initiation proteins. (c) Interactions between SS3a fragments and BGC1. (d) Pairwise immunoprecipitation (IP) using anti‐HA beads of SS3a‐N‐HA and BGC1‐RFP coexpressed in Nicotiana benthamiana leaves. Input and IP samples were blotted with HA and RFP antibodies. Migration of molecular weight (MW) markers in kDa is shown on the left of each panel. Bands of lower MW likely representing a degradation product of the fusion protein are marked with ‘Deg.’. (e) Close‐up images of semicompound starch granules of the bgc1‐3 mutant at 8 d post anthesis, viewed under polarised light microscopy. Red arrows mark examples of multiple crosses within granules. Bars, 10 μm.
We further focussed on the interaction between SS3 and BGC1, since BGC1 is known to affect granule initiation at the early stage of grain development (Suh et al., 2004; Hawkins et al., 2021), and it was the only common protein interacting with both SS3a and SS3b in the Y2H. First, to locate the regions of the SS3a protein important for the interaction with BGC1, we made constructs encoding SS3a fragments, dividing the protein into three parts: the N‐terminal domain (SS3a‐N), carbohydrate binding module (CBM) domains (SS3a‐CBM) and glucosyltransferase (GT) domains (SS3a‐GT). Y2H showed that all three fragments could maintain some degree of interaction with BGC1, suggesting that the interaction could occur over multiple sites (Fig. 8c). To confirm an interaction between SS3a and BGC1 in planta, we used a co‐immunoprecipitation assay with epitope‐tagged proteins transiently expressed in N. benthamiana leaves. Despite multiple attempts, we were not able to express the full‐length SS3a protein in Nicotiana, likely due to its large size. However, we could express the SS3a‐N with a HA tag (SS3a‐N‐HA), together with BGC1 tagged with RFP (BGC1‐RFP). Immunoprecipitation with anti‐HA beads showed that BGC1‐RFP copurified with SS3a‐N‐HA (Fig. 8d), suggesting that the two proteins interact.
Taken together, we propose that SS3a controls granule initiation during early grain development by interacting with BGC1. We therefore compared the phenotypes of our ss3 mutants with the previously reported bgc1 mutant in the same durum wheat cultivar, Kronos, and we noted several similarities and differences. Previous work had reported semicompound granules in the barley mutant, Franubet, defective in the BGC1 ortholog, FLO6 (Suh et al., 2004). Indeed, we also observed them in the wheat bgc1 mutant (Fig. 8e). However, compared with ss3a or ss3a ss3b, semicompound granules were rare. Rather, granules in bgc1 were more often fully compound with typical polygonal morphology, resulting from tessellated packing in the endosperm (Fig. 6). Thus, although ss3a, ss3a ss3b and bgc1 mutations all result in supernumerary initiations per amyloplast, a key difference is that the initiations merge into semicompound granules with a contiguous surface in ss3a and ss3a ss3b, while in bgc1, the initiations mostly remain as separate granulae in tessellated compound granules.
Eliminating both SS3 isoforms results in high resistant starch in wheat
Given the partial overlap in function between SS3a and SS3b for starch granule initiation in wheat endosperm, we also tested whether their roles also overlap in determining starch polymer structure and composition. We first analysed amylopectin chain length structure using High Performance Anion Exchange Chromatography with Pulsed Amperometric Detection (HPAEC‐PAD). Both ss3a and ss3b single mutants had only minor alterations in amylopectin chain length distributions (< 0.5% difference at each degree of polymerisation (DP)) (Fig. 9a). However, in the ss3a ss3b double mutant, we observed a stronger change in the chain length distribution, with an increase in DP 6, a decrease in short chains from DP 8 to 11, and an increase in intermediate chains from DP 12 to 19. Consistent with Fahy et al. (2022), the ss3a mutant had elevated levels of amylose and resistant starch (Fig. 9b,c). Although the ss3b single mutant had no change in amylose and resistant starch, levels of both were significantly higher in ss3a ss3b than in ss3a. Thus, knocking out both isoforms of SS3 can significantly boost the amount of resistant starch, relative to the previously proposed strategy of knocking out SS3a alone (Fahy et al., 2022). This parallels recent success in rice, achieving high resistant starch by targeting both SS3a and SS3b (Wang et al., 2023; Huang et al., 2024). These results suggest that the two isoforms partially overlap in function in starch polymer biosynthesis, in addition to the functional overlap observed in starch granule initiation.
Starch polymer characteristics of the durum wheat ss3a ss3b mutant. The cross between the ss3a aabb and ss3b aabb mutant was used to segregate out the ss3a ss3b double mutant defective in all homeoalleles of both genes, as well as the single mutants defective in ss3a or ss3b, and the wild‐type (WT) segregant (WT seg.). These were compared with the Kronos WT (WT). All panels show the mean ± SE for n = 3 replicates, where each biological replicate represents grains harvested from a different plant. Statistical groupings are from a one‐way ANOVA and Tukey's HSD test, where groups with different letters are significantly different (P < 0.05). (a) Amylopectin chain length distribution was quantified using High‐Performance Anion Exchange Chromatography with Pulsed Amperometric Detection. Values represent the difference in the relative chain lengths at each Degree of polymerisation for each genotype relative to the Kronos WT. (b) Amylose content (% of total starch) measured by iodine colourimetry. (c) Resistant starch content (% of total starch) measured using an in vitro assay.
Discussion
SS3 isoforms are essential for normal granule initiation in early wheat grain development
We demonstrate that the two SS3 isoforms of wheat play an important role in starch granule initiation during grain development. Although the important role of SS3 in amylopectin synthesis has been established in various species (Arabidopsis, rice, maize and wheat) (Wang et al., 1993; Fujita et al., 2007; Zhang et al., 2008; Fahy et al., 2022), a role for SS3 in starch granule initiation is less established, since mutant phenotypes related to granule number have been conditional in all species examined thus far. In Arabidopsis leaves, loss of SS3 only affects starch granule number when SS4 is also absent (Szydlowski et al., 2009; Vandromme et al., 2023). In rice endosperm, ss3a mutants produce smaller starch granules that are rounded in shape compared with the WT, although it is unclear whether this is due to a defect in granule initiation (Fujita et al., 2007). Nevertheless, a similar genetic relationship with SS4 has been demonstrated in rice, since the effect of the rice ss3a mutation on granule morphology becomes stronger after the simultaneous elimination of SS4b (Toyosawa et al., 2016). Despite these conditional phenotypes, a recent bioinformatics study proposed SS3a to be a strong candidate gene for determining differences in granule initiation patterns among grass species, based on interspecies differences in SS3a gene expression patterns that correlate with the occurrence of simple or compound granule morphology (Watson‐Lazowski et al., 2026).
Eliminating SS3a alone in wheat leads to strong granule initiation‐ and morphology‐related phenotypes. First, the striking change in the A‐type granule morphology of ss3a mutants is due to supernumerary granule initiations, leading to semicompound starch granule formation. Defects in A‐type granule morphology were already observed in an ss3a mutant of hexaploid bread wheat, although the mechanism of their formation was not known (Fahy et al., 2022). In our study, polarised light microscopy showed that > 70% of the large granules in the ss3a single mutant were composed of multiple initiations (Fig. 5). Furthermore, we generated an ss3a ss3b double mutant, which to our knowledge has not been previously reported in any Triticeae species, allowing the assessment of redundancy between the two paralogs in relation to A‐ and B‐type granule formation. Although the ss3b single mutant had no defects in granule morphology, the ss3a ss3b double mutant had stronger effects on granule formation, with > 90% of the large granules having semicompound morphology. This demonstrates that both SS3a and SS3b can contribute to granule initiation, but SS3a is the major activity that is indispensable, while SS3b partially complements in the absence of SS3a.
The loss of SS3 appears to affect A‐type granule formation more than B‐type granule formation. Semicompound A‐type granules were already visible at 8 dpa, when only A‐type granules are present (Fig. 3). The B‐type granules of ss3a and ss3a ss3b had normal spherical morphology (Figs 2, 6). In some experiments, we measured a decrease in B‐type granule size in the mutants, but this was not consistent and more likely to reflect the decrease in total starch content (Fig. 6). B‐type granule content (by relative volume) increased in the ss3a and ss3a ss3b mutants, but as this is expressed relative to the total granule volume, it more likely reflects decreased A‐type granule volume – especially since the mean size of A‐type granules was significantly smaller in the mutants than the WT controls.
These overlapping roles of SS3a and SS3b in granule initiation are consistent with the protein–protein interactions we have observed. First, we demonstrated using Y2H assays that SS3a and SS3b can interact with themselves and with each other. Such interaction between isoforms strongly suggests functional overlap. Consistent with this, a band that could represent a high‐molecular‐weight heteromer was observed on native PAGE gels (Fig. 7). While such oligomerisation of SS3 has not previously been reported, SS3a has been observed to be in high‐molecular‐weight complexes in gel filtration experiments using endosperm extracts from rice (Crofts et al., 2015) and maize (Hennen‐Bierwagen et al., 2008). Second, we demonstrated that both SS3a and SS3b can interact with BGC1. Such a direct link between SS3 and the granule initiation machinery via protein–protein interaction has not been demonstrated in any species. Additionally, we observed an interaction with the granule initiation protein MRC for SS3a but not for SS3b (Fig. 8a,b). We did not focus on this interaction in this study because MRC plays a defined role in controlling the timing of B‐type granule initiation (Chen et al., 2024; Fahy et al., 2025), but the primary effect of ss3a mutations was on A‐type granules. Nevertheless, in Arabidopsis, combining mutations in both SS3 and MRC leads to a dramatic effect on granule morphology (Vandromme et al., 2023), providing rationale for future investigations into the interaction in wheat. Overall, our data expand our knowledge of SS3 interactions, building upon the reported interactions between SS3a and other biosynthetic proteins (such as other starch synthase and branching enzyme isoforms) in maize and rice (Hennen‐Bierwagen et al., 2008; Crofts et al., 2015). More broadly, our work builds on previous knowledge of interactions between other key starch biosynthesis proteins in wheat endosperm (Tetlow et al., 2004). However, one limitation is that the interactions between SS3 isoforms and BGC1, and between SS3a and MRC, have only been observed in heterologous systems (i.e. yeast and Nicotiana). Further work using endosperm extracts could shed further light on the stability of these interactions, and how the numerous interactions of these proteins are regulated in vivo.
Semicompound granule formation in the ss3 mutant
Whereas a normal A‐type granule forms from a single granule initiation per amyloplast, the formation of semicompound starch granules requires supernumerary initiations within amyloplasts. Despite that, the total number of granules measured using the Coulter counter was not increased in the ss3a and ss3a ss3b mutant compared with the WT, even at 8 dpa (Fig. 3). Since the Coulter counter would count each semicompound granule as a single particle, the merging of multiple initiations into semicompound granules likely occurs with great efficiency in the mutant, such that almost all initiations within each amyloplast are incorporated. Thus, the final number of A‐type granules (regardless of whether they are normal or semicompound) remains the same between the WT and mutants and is likely determined by the number of amyloplasts. Likewise, the fact that total granule number is not altered in the mutants is strong evidence that there are almost no true compound granules in the mutant, which would each fall apart into multiple granules, and greatly increase the total number of countable granules. Consistent with this, we have not observed any granules with the typical polygonal morphology of compound granules (Fig. 6).
The formation of semicompound granules makes the ss3 phenotype in wheat distinct from existing granule initiation mutants (Fig. 6). Loss of function in SS3, SS4 or BGC1 in wheat all lead to supernumerary initiations per amyloplast during early grain development, but these initiations lead to different fates – with semicompound granule formation in ss3a and ss3a ss3b mutants, and true compound granule formation in ss4 (Hawkins et al., 2021). The only report of semicompound granules in the Triticeae is in the Franubet mutant of barley, which is defective in the barley ortholog of BGC1 (FLO6) (Suh et al., 2004). This is particularly striking given that we discovered SS3 interacts with BGC1 (Fig. 8). Although this interaction with a known granule initiation protein further implicates SS3 in initiation, it cannot fully explain the semicompound phenotype – especially since in the wheat bgc1 mutant, semicompound granules were a rare occurrence (Fig. 8) and the vast majority of granules had true compound morphology (Fig. 6).
The reason why ss3 mutants form semicompound granules as opposed to true compound granules is not known and will require further investigation. We speculate two possibilities: First, the presence of functional SS4 and/or BGC1 in the ss3 mutants could influence the phenotype. For example, SS4 or BGC1 might play an additional role in granule morphogenesis that allows multiple initiations to merge and undergo contiguous surface growth. Indeed, orthologs of both proteins have been implicated in granule morphogenesis. In Arabidopsis leaves, SS4 is required for proper granule shape, as the ss4 mutant produces spherical granules (Burgy et al., 2021). In potato tubers, a specialised PTST2/BGC1 isoform, PTST2b, also plays a similar role in granule morphogenesis (Hochmuth et al., 2025). A second possibility is that the alterations in polymer structure and composition in the ss3a and ss3a ss3b mutants facilitate semicompound granule formation. Altered amylopectin structure and amylose content are a key feature of these mutants (Fig. 9), which sets them apart from ss4 and bgc1 mutants that do not have these alterations (Hawkins et al., 2021). Although semicompound granules have never been reported following ss3 mutations in other species (will be discussed later), they have been reported in starch branching enzyme (SBE) mutants of other species, particularly maize and potato (Boyer et al., 1977; Jiang et al., 2010; Tuncel et al., 2019). It was suggested that in the maize ae (sbe2b) mutant, long chains exposed on the outside individual granules could interact, bringing together granules and allowing fusion (Jiang et al., 2010). In barley, reductions in SBE2 activity cause the formation of elongated starch granules (Matsushima et al., 2024). In wheat, stronger reductions in SBEII activity cause a ‘deflated’, ‘sickle‐shaped’ appearance in A‐type granules while the B‐type granules adopt an elongated appearance (Regina et al., 2006; Sestili et al., 2010). This is distinct from ss3a and ss3a ss3b in which A‐type granules become semicompound, while B‐type granules retain normal shape. Thus, the link between the semicompound granule morphology phenotype and high amylose content (or the presence of long amylopectin chains in branching enzyme mutants) is clearly not simple. This relationship could be tested by using mutations in Granule Bound Starch Synthase (GBSS) to reduce the amylose content in the ss3a or ss3a ss3b background.
Comparison to other cereal species and biotechnological applications
The severe effects on starch granule morphology following loss of SS3 in wheat, particularly in the ss3a ss3b mutant, contrast the relatively subtle effects on granule morphology in other cereals. The SS3a mutant of maize (dull1) is reported to have a decrease in starch granule size, but relatively normal granule shapes (Gao et al., 1998; Zhu et al., 2016). The SS3a mutant of rice also has subtle effects on granule morphology, with a decrease in granule size, and some granules that appear more rounded (Fujita et al., 2007; Watson‐Lazowski et al., 2026). Thus, SS3a does not appear to have a major influence over granule morphology in maize and rice that makes simple and compound granules, respectively, and the highly defective semicompound granule morphology appears to be an effect that is so far specific to wheat. Beyond cereal species, the antisense repression of SS3 in potato led to some compound granule formation, but there was no report of semicompound granule formation (Abel et al., 1996).
However, compared with other cereals, the influence of SS3 isoforms on polymer composition appears more conserved, making both isoforms an ideal target for wheat breeding for higher resistant starch. To our knowledge, we demonstrated for the first time in wheat the full potential of SS3 as a target for engineering high resistant starch. Although it is known that eliminating SS3a results in increased amylose and resistant starch content (Fahy et al., 2022), we showed that eliminating both SS3a and SS3b can approximately double amylose and resistant starch compared to eliminating SS3a alone (Fig. 9). This parallels recent results in rice, in which eliminating both isoforms had a stronger effect on amylose and resistant starch content than eliminating single isoforms (Wang et al., 2023; Huang et al., 2024). The amylose content values in wheat ss3a ss3b mutants were c. 50% (Fig. 9), which is lower than values of c. 75% generated in wheat by targeting SBE isoforms (Regina et al., 2015). However, one potential advantage is that ss3a ss3b mutants had no significant effect on grain weight under controlled experiments, and starch content was either not significantly affected or slightly reduced depending on the experiment (Figs 2, 5, S4, S8). This contrasts with strong reductions in both these traits after targeting SBE isoforms (Regina et al., 2015). SS3 isoforms are therefore promising targets for engineering high resistant starch. To realise this potential, further work must focus on (1) evaluating field performance in comparison with existing strategies based on targeting SBEs, and (2) evaluating how food quality and cooking properties are impacted by the altered granule morphology in these mutants, which accompanies the high resistant starch phenotype.
Competing interests
None declared.
Author contributions
DS conceived and designed the study. JD conducted most of the experimental work, analysed data and produced the figures. DS and QJ provided supervision to JD. BF produced the SS3a and SS3b wheat mutant lines. RM performed the light microscope imaging of starch granule morphology in sections. DS wrote the manuscript with input from all authors.
Disclaimer
The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.
Supporting information
Dataset S1 Amino acid sequence alignment used for phylogenetic analysis in Fig. S1.
Fig. S1 Identification of SS3a and SS3b gene models. Fig. S2 Amino acid alignment of cereal SS3a and SS3b sequences. Fig. S3 Transcript levels of SS3a and SS3b homeologs. Fig. S4 Grain size parameters for ss3a and ss3b single mutants of durum wheat. Fig. S5 Starch granule morphology in the durum wheat ss3a and ss3b single mutants. Fig. S6 Starch granule morphology in developing grains of durum wheat ss3a and ss3b single mutants. Fig. S7 Starch granule size distributions in developing grains of durum wheat ss3a and ss3b single mutants. Fig. S8 Grain size parameters for the durum wheat ss3a ss3b mutant. Table S1 Oligonucleotide primers used in this study. Table S2 Amino acid identity and similarity values between wheat SS3 isoforms.Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Abel GJW , Springer F , Willmitzer L , Kossmann J . 1996. Cloning and functional analysis of a c DNA encoding a novel 139 k Da starch synthase. The Plant Journal 10: 981–991.9011082 10.1046/j.1365-313x.1996.10060981.x · doi ↗ · pubmed ↗
- 2Appels R , Eversole K , Stein N , Feuillet C , Keller B , Rogers J , Pozniak CJ , Choulet F , Distelfeld A , Poland J et al. 2018. Shifting the limits in wheat research and breeding using a fully annotated reference genome. Science 361: eaar 7191.30115783 10.1126/science.aar 7191 · doi ↗ · pubmed ↗
- 3Apriyanto A , Compart J , Fettke J . 2022. A review of starch, a unique biopolymer – structure, metabolism and in planta modifications. Plant Science 318: 111223.35351303 10.1016/j.plantsci.2022.111223 · doi ↗ · pubmed ↗
- 4Bechtel DB , Zayas I , Kaleikau L , Pomeranz Y . 1990. Size‐distribution of wheat starch granules during endosperm development. Cereal Chemistry 67: 59–63.
- 5Boyer CD , Daniels RR , Shannon JC . 1977. Starch granule (amyloplast) development in endosperm of several Zea mays L genotypes affecting kernel polysaccharides. American Journal of Botany 64: 50–56.
- 6Burgy L , Eicke S , Kopp C , Jenny C , Lu KJ , Escrig S , Meibom A , Zeeman SC . 2021. Coalescence and directed anisotropic growth of starch granule initials in subdomains of Arabidopsis thaliana chloroplasts. Nature Communications 12: 6944.10.1038/s 41467-021-27151-5PMC 862648734836943 · doi ↗ · pubmed ↗
- 7Cao H , Imparl‐Radosevich J , Guan H , Keeling PL , James MG , Myers AM . 1999. Identification of the soluble starch synthase activities of maize endosperm. Plant Physiology 120: 205–216.10318698 10.1104/pp.120.1.205PMC 59252 · doi ↗ · pubmed ↗
- 8Chen J , Chen Y , Watson‐Lazowski A , Hawkins E , Barclay JE , Fahy B , Denley Bowers R , Corbin K , Warren FJ , Blennow A et al. 2024. Wheat MYOSIN‐RESEMBLING CHLOROPLAST PROTEIN controls B‐type starch granule initiation timing during endosperm development. Plant Physiology 196: 1980–1996.39158075 10.1093/plphys/kiae 429PMC 11531834 · doi ↗ · pubmed ↗