STARCH SYNTHASE 4 is required for normal starch granule initiation in amyloplasts of wheat endosperm

Starch granule initiation is poorly understood at the molecular level. The glucosyltransferase, STARCH SYNTHASE 4 (SS4), plays a central role in granule initiation in Arabidopsis leaves, but its function in cereal endosperms is unknown. We investigated the role of SS4 in wheat, which has a distinct spatiotemporal pattern of granule initiation during grain development. We generated TILLING mutants in tetraploid wheat (Triticum turgidum) that are defective in both SS4 homoeologs. The morphology of endosperm starch was examined in developing and mature grains. SS4 deficiency led to severe alterations in endosperm starch granule morphology. During early grain development, while the wild type initiated single ‘A-type’ granules per amyloplast, most amyloplasts in the mutant formed compound granules due to multiple initiations. This phenotype was similar to mutants deficient in B-GRANULE CONTENT 1 (BGC1). SS4 deficiency also reduced starch content in leaves and pollen grains. We propose that SS4 and BGC1 are required for the proper control of granule initiation during early grain development that leads to a single A-type granule per amyloplast. The absence of either protein results in a variable number of initiations per amyloplast and compound granule formation.

amylose synthesis (Seung, 2020). In Arabidopsis leaves, SS4 is required for both normal 54 granule initiation and morphogenesis, but does not make a major contribution to 55 amylopectin structure (Roldán et al., 2007;  Transformants were selected in the T1 generation using the Basta resistance marker. Basta-220 resistant individuals from the T2 or T3 generation (heterozygous or homozygous for the 221 transgene; single or multiple insertions) with TaSS4 expression confirmed using 222 immunoblots were used for experiments. 223 224

Production of antibodies and immunoblotting 225
To produce TaSS4 and TaBGC1 antibodies, the coding sequence of the proteins (minus 226 transit peptide) were amplified using primers in Table S1, and TaSS4-1B:pDONR221 or 227 TaBGC1-4B:pDONR221 as templates. The amplicons were cloned into the pProExHTb vector 228 (Invitrogen) in frame with the N-terminal His6-tag using the Gibson assembly master mix 229 (New England Biolabs) for TaSS4-1B, or BamHI and XhoI sites for TaBGC1-4B. Proteins were 230 expressed in E. coli strain BL21 as described in Seung et al. (2015). Denaturing purification of 231 the protein with urea was carried out using the Ni-NTA Agarose (Qiagen). Immunisation of 232 rabbits was carried out at Eurogentec. Antibodies were enriched from antiserum using 233 protein A-agarose (Sigma-Aldrich). Affinity purification of TaBGC1 demonstrated that the antiserum recognised both homoeologs (Fig. S2). Protein extracts 293 from endosperms dissected from wild-type developing grains (10, 15 and 20 days post 294 anthesis (dpa)) were immunoblotted with the antiserum. A band that corresponded to the 295 predicted size of the mature polypeptide (98 kDa) was observed at all three timepoints, but 296 was most prominent at the 10 dpa timepoint (Fig. 2a). Several other bands at different 297 molecular weights were detected, but comparison of immunoblots from the wild-type and 298 mutant extracts showed that the 98 kDa band was missing in the latter while other bands 299 were unaffected (Fig. 2b). We conclude that the 98 kDa band represents TaSS4, and that the 300 other bands result from non-specific binding. levels were higher at the early stages of grain development (8-11 dpa) than later stages (16-305 22 dpa) (Fig. S3). These data are consistent with the observed decrease in TaSS4 protein 306 levels at later stages of grain development (Fig. 2a). 307

308
To assess the impact of the Tass4-1 mutations on endosperm starch, we purified starch 309 granules from mature grains and observed them using scanning electron microscopy. 310 Granules from control lines (AA BB) and the single homoeolog mutants (aa BB and AA bb) 311 had flattened A-type granules and round B-type granules typical of wheat starch (Fig. 3a). By 312 contrast, most starch granules from the double mutant (aa bb) had irregular, polyhedral 313 morphology. The irregular granules were highly variable in size, but rarely exceeded the size 314 of a typical A-type granule. A-type granules of normal appearance were also present in the 315 double mutant, but we rarely observed normal B-type granules. 316

317
We used cross-polarised light microscopy to examine the origins of the larger polyhedral 318 granules in the Tass4-1 double mutant endosperm. In the control line and the single 319 homoeolog mutants, there was one 'Maltese cross' per A-type or B-type granule, indicating 320 a single centre of organisation (Fig. 3b). The few normal A-type granules in the double 321 mutant also had single crosses. However, a complex birefringence pattern with faint or 322 multiple crosses were observed in most of the polyhedral granules, indicating multiple 323 initiation points. 324 325 Using a Coulter counter, we examined the granule size distribution in the endosperm 326 starches. As expected, starch from the control line and single mutants showed a bimodal 327 size distribution, with peaks at approx. 20 µm diameter for A-type and 7-8 µm diameter for 328 B-type granules (Fig. 3c). The size and relative proportion (by volume) of A-type and B-type 329 granules were quantified by fitting a mixed log-normal distribution (Table 1; Table  356 1). Thus, the altered granule morphology in the Tass4 can interact with each other. For two Tabgc1 aa bb double mutants (Fig. 7a) that 393 accumulate no detectable TaBGC1 protein (Fig. 2c), we established that starch granules had 394 morphologies like those described for Tabgc1 mutants in hexaploid wheat and barley: 395 mature grains contained A-type granules of normal appearance and small polyhedral 396 granules (Fig. 7b). As in the Tass4 mutant, these polyhedral granules were already present 397 during early grain development (8 dpa onwards) (Fig. S6). However, normal A-type granules 398 were more frequent in the Tabgc1 mutant ( Fig. 7b) than in the Tass4 mutant (Fig. 3a). 399 Coulter counter analysis also showed a prominent A-type granule peak in the Tabgc1 400 mutant at a similar diameter to that of the wild type (Fig. 7c, d). Such a distinct peak was not 401 observed in the Tass4-1 mutant (Fig. 3c). 402 403

Loss of TaSS4 affects starch synthesis in pollen 404
The Tass4 double mutant was indistinguishable from control lines in terms of plant growth 405 ( Fig. 8a). Most grains of the mutant appeared normal and the average weight of individual 406 grains was not significantly altered compared to the wild type, although we observed rare 407 examples of smaller, shrivelled grains in the mutant (Fig. 8b, S7). While the double mutant 408 produced comparable numbers of tillers to the control (Fig. 8c), the number of grains per 409 spike was significantly reduced in the mutant (Fig. 8d). This reduction in grain number was 410 most severe in the non-backcrossed Tass4-1 double mutant but was partly recovered after 411 backcrossing, suggesting that this phenotype was exacerbated by background mutations in 412 non-backcrossed lines (Fig. 8d). Since the fewer grains in the backcrossed mutant suggests 413 defective fertilisation, we examined starch accumulation in pollen grains of the mutant 414 using iodine staining. Less than a third of pollen grains from the double mutant contained 415 starch, contrasting those from control lines where almost all contained starch (Fig. 8e, S8a). 416 Cross-pollination experiments with the backcrossed Tass4-1 lines demonstrated that using 417 aa bb pollen to fertilise AA BB maternal plants resulted in significantly reduced fertilisation 418 rates compared to the reciprocal cross (Fig. S8b). These data suggest that TaSS4 is important  419 for normal pollen starch synthesis and viability. Interestingly, grains from low-yielding non-420 backcrossed and high-yielding backcrossed Tass4-1 lines had identical starch granule 421 morphology (Fig. S3), demonstrating that this phenotype is independent from the fertility 422 phenotype.

TaSS4 is necessary for normal granule initiation in the endosperm 452
In wheat endosperm, granule initiation is spatially and temporally coordinated such that 453 single A-type granules form in amyloplasts during early grain development and B-type 454 granules initiate later and at least partially in stroma-filled tubules (stromules) that emanate 455 from the amyloplast (Parker, 1985;Langeveld et al., 2000). This pattern is distinct from most 456 other grasses (e.g: rice), which form compound granules by initiating multiple granules per 457 amyloplast during early grain development (Matsushima et al., 2013(Matsushima et al., , 2015.  (Table 1), but resulted in the formation of compound 467 granules in the endosperm in place of most A-type granules (Fig. 3-6). A similar phenotype 468 was observed in mutants fully deficient in TaBGC1 in tetraploid wheat (Fig. 7), and in 469 hexaploid wheat (Chia et al., 2020), suggesting that the two proteins act in a similar process. 470 However, the Tass4 phenotype was more severe than the Tabgc1 phenotype as there were 471 substantially more normal A-type granules in the latter (Fig. 7). These observations parallel 472 those in Arabidopsis leaves, in which granule initiation is more compromised in the Atss4 473 mutant than in the Atptst2 mutant (Seung et al., 2017). The specific role of TaSS4 in B-type granule initiation must also be further explored. Very 526 few normal round B-type granules were observed in mature grains of the Tass4 mutant (Fig.  527 3). Also, at 15 dpa, we observed many compound granules in a linear arrangement in the 528 mutant, raising the possibility that they formed in stromules that normally enclose B-type 529 granules (Fig. 6)

. Interestingly, Chia et al. (2020) reported that reducing gene dosage of 530
TaBGC1 in hexaploid wheat can almost eliminate B-type granules while retaining normal A-531 type granule morphology. By contrast, B-type granule volume was not affected in either of 532 the single homoeolog mutants in TaSS4, but it is possible that a further reduction in gene 533 dosage is required to see an effect. However, we noted that while TaSS4 protein levels are 534 highest during early grain development and decrease at the later developmental stages, 535 TaBGC1 transcript and protein levels increase and are highest during the period of B-type 536 granule initiation ( Fig. 2; Fig. S3). Thus, it is possible that TaBGC1 has a specific role during B-537 type granule initiation that is independent of TaSS4. 538 539 While other members of the Triticeae (e.g., barley and rye) also have A-and B-type 540 granules, most other grasses produce compound granules in the endosperm (Matsushima et 541 al., 2013(Matsushima et 541 al., , 2015. The fact that loss of SS4 or BGC1 gives rise to some compound granules in 542 wheat makes it tempting to speculate that differences in the extent and timing of SS4 543 and/or BGC1 expression between species could determine whether a given species 544 possesses compound granules. However, the difference between compound and other 545 patterns of granule initiation is unlikely to be so simple. Thus, the formation of compound granules in rice is likely to involve multiple genes that 549 control starch synthesis and amyloplast morphogenesis. 550 551

TaSS4 is required for proper granule initiation in leaves and pollen 552
In leaves of the Arabidopsis Atss4 mutant, over 75% of chloroplasts had no visible starch 553 granule, and the majority of remaining chloroplasts contained one large granule (Roldán et 554 al., 2007;Seung et al., 2017). Leaves of the Tass4 mutant had a percentage of starchless 555 chloroplasts that was comparable to the Arabidopsis mutant, but the remaining chloroplasts 556 mostly contained multiple granules (Fig. 9). The reason for this difference between the 557 Arabidopsis and wheat phenotypes is unknown, but could reflect differences in the 558 compensation mechanism following loss of SS4. The few granules present in the Arabidopsis 559 Despite a reduction in gene dosage to 50% in our single homoeolog wheat mutants, we did 576 not observe an effect on granule number in leaf chloroplasts. On first glance, this is in 577 contrast to a previous report that hexaploid wheat mutants deficient in only TaSS4-1D have 578 reduced numbers of granules per chloroplast (Guo et al., 2017). However, we showed that 579 some hexaploid cultivars, including the reference cultivar Chinese Spring, have a natural 580 polymorphism that leads to a premature stop codon in TaSS4-1A (Fig. 1). It is possible that 581 TaSS4-1B is the only functional homoeolog in the TaSS4 TaSS4 also appears to be required for normal starch synthesis in wheat pollen. Publicly 585 available gene expression data for hexaploid wheat suggests that TaSS4 is expressed in 586 microspores in addition to leaves, stems, roots and grains (Fig. S3b); and most pollen grains 587 from our Tass4 mutants were starchless (Fig. 8e, S8). In rice, starch synthesis in pollen 588 appears to be essential for viability, as rice pgm mutants lacking pollen starch are sterile 589 (Lee et al., 2016). Consistent with this, the pollen from the Tass4 mutant had significantly 590 reduced fertilisation success in cross-fertilisation experiments (Fig. S8b), and the mutants 591 produced fewer grains per spike (Fig. 8d, S8c)  Starch content was determined as glucose equivalents and is expressed as a percentage of 931 the flour weight. Amylose content of starch was determined by HPLC-SEC. The mean 932 diameters of A-type and B-type granules and the relative volume of B-type granules were 933 determined using a Coulter counter. All values are mean ± S.E from n = 3 biological 934 replicates, defined as grains harvested from three different plants. There were no significant 935 differences between any of the lines in any of these parameters under a one-way ANOVA at 936 p<0.05.    (a), but granules were observed using cross-polarised light microscopy. The multiple hila in the large polyhedral granule are indicated with red arrows. Bars = 10 µm. (c) Size distribution of endosperm starch granules. The volume of granules at each diameter relative to the total granule volume was quantified using a Coulter counter. Values represent mean (solid line) ± SEM (shading) of three replicate starch extractions from grains of three different plants.