SWEET11 and 15 as key players in seed filling in rice

Despite the relevance of seed filling mechanisms for crop yield, we still have only a rudimentary understanding of the pathways and transport processes for supplying the caryopsis with sugars. We hypothesized that the recently identified SWEET sucrose transporters may play important roles in nutrient import pathways in the rice caryopsis. We used a combination of mRNA quantification, histochemical analyses, translational promoter-reporter fusions and analysis of knock out mutants created by genomic editing to evaluate the contribution of SWEET transporters to seed filling. In rice caryopses, SWEET11 and 15 had the highest mRNA levels and proteins localized to four key sites: the nucellus proper at early stages, the nucellar projection close to the dorsal vein, the nucellar epidermis that surrounds the endosperm, and the aleurone. ossweet11;15 double knock-out lines accumulated starch in the pericarp while caryopses did not contain a functional endosperm. Jointly, SWEET11 and 15 show all hallmarks of being responsible for seed filling with sucrose efflux function at the nucellar projection and transfer across the nucellar epidermis/aleurone interface, delineating two major steps for apoplasmic seed filling, observations that are discussed in relation to observations made in rice and barely on the relative prevalence of these two potential import routes.


Introduction 51
Population growth is expected to lead to an increasing need for rice production, especially 52 in Africa (Sharma, 2014). A key question is thus how we can obtain maximal yield poten-53 tial. Rice grains are composed mainly of starch (over 90% in many cases; 54 www.knowledgebank.irri.org/ricebreedingcourse/Grain_quality.htm), which derives from 55 imported soluble carbohydrates. These carbohydrates are produced in the leaves with 56 the help of photosynthesis and are exported mainly as sucrose via the phloem, which 57 contains ~600 mM sucrose (Fukumorita & Chino, 1982). In order to fill the seeds, and 58 more specifically to generate starch, cell walls and provide energy, sucrose has to be 59 imported into developing caryopses. Phloem strands enter the seed coat, where sucrose 60 is unloaded and transferred into the developing caryopsis to supply cells with nutrients, 61 in particular sugars as sources of energy and as carbon skeletons for cell wall and starch 62

biosynthesis. 63
In plants, cell-to-cell transport of sugars is thought to be mediated by apoplasmic (export 64 from one cell by a plasma membrane transporter and subsequent import into the adjacent 65 cell by another transport protein) or by symplasmic transfer via plasmodesmata. Other 66 routes are conceivable, but evidence for vesicular transport processes are sparse (van 67 den Broek et al., 1997). Two of the key processes for long distance translocation are 68 phloem loading and seed filling. The transport processes that ultimately lead to cell wall 69 synthesis and storage product accumulation in seeds, in particular in cereal caryopses 70 are not fully understood. To delineate sym-and apoplasmic pathways in rice caryopses, 71 Oparka carried out a combination of ultrastructural and dye tracer studies in the early 80s 72 (Oparka & Gates, 1981a,b, 1982, 1984. The rice caryopsis is supplied by three vascular 73 bundles that pass through the pericarp. The dorsal vascular bundle is the major route for 74 sugar delivery to the developing caryopsis (Oparka & Gates, 1981a;Krishnan & Daya-75 nandan, 2003). He found symplasmic connections between the parenchyma of the dorsal 76 vascular bundle and the nucellar projection as well as a lipid barrier between the inner 77 TCACCAGTAGCAATGGCAGG-3') of OsSWEET11 was used with Cas9 for transfor-136 mation (see Fig. 2S). While method for TALEN-induced mutant lines (ossweet11-2, 137 ossweet15-1 and ossweet15-2) was also described (Li et al., 2014). One pair of TALENs 138 for OsSWEET11 and another pair of TALENs for OsSWEET15 were designed and engi-139 neered for rice transformation. For detailed sequence and location of TALEN targets, see 140 University Campus, CA, USA) or in greenhouses (Stanford and ISU) under long-day con-143 ditions (14h day/10h night, 28-30˚C). 144

Genotyping of rice plants 145
Rice genomic DNA was extracted using CTAB method. PCR was performed using ExTaq 146 DNA polymerase (Clontech, Mountain View, CA, USA) with a melting temperature of 56°C 147 and 53°C for OsSWEET11 and OsSWEET15, respectively (for primers, see Supplemen-148 tary Table 1). The PCR-amplicons from the mutant alleles has been sequenced. Chro-149 matograms were read and aligned using BioEdit. 150

FRET sucrose sensor analysis in HEK293T cells 211
The OsSWEET11 and OsSWEET15 coding sequences were cloned into the Gateway 212 entry vector pDONR221f1, and then cloned into vector pcDNA3.2V5 by LR recombination 213 reaction for expression in HEK293T cells. HEK293T cells were co-transfected with a plas-214 mid carrying the OsSWEET11 or OsSWEET15 and the sucrose sensor FLIPsuc90μ-215 sCsA, using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). For FRET imaging, 216 HBSS medium was used to perfuse HEK293T/ FLIPsuc90μ-sCsA cells pulsed with 20 217 Results 220

Identification of OsSWEET11 and OsSWEET15 in rice caryopses 221
To identify SWEETs that are expressed specifically in rice caryopses, we analyzed public 222 microarray data from RiceXPro (ricexpro.dna.affrc.go.jp). We focused on members of 223 clade 3 since they had been shown to function as plasma membrane sucrose transport-224 ers. Among the five clade 3 SWEETs analyzed, OsSWEET11 and 15 had the highest 225 mRNA levels in the endosperm between 7 and 14 DAP (Fig. S1). To validate the micro-226 array data, we harvested immature seeds at different developmental stages from green-227 house-grown plants and re-analyzed mRNA levels of OsSWEET11 and 15 by qRT-PCR. 228 For comparison, we analyzed OsSWEET4, which had been shown to play an important 229 role as a hexose transporter in seed development (Sosso et al., 2015)( Fig. 1). At 1 DAP, 230 SWEET4 had the highest mRNA levels, but already at 3 DAP, levels had declined about 231 ~3-fold. In comparison, OsSWEET11 was low at DAP1, but equal to OsSWEET4 at 3 232 DAP. OsSWEET11 gradually increased throughout seed development. While in the mi-233 croarrays OsSWEET15 was only 2-3x lower compared to OsSWEET11, our qRT-PCR 234 analysis indicated a much lower relative level however the developmental pattern of 235 OsSWEET15 mRNA levels was similar to that of OsSWEET11. 236

OsSWEET11 plays a key role in seed filling 237
OsSWEET11 had previously been shown to function as a plasma membrane sucrose 238 transporter (Chen et al., 2012). Since OsSWEET11 was by far the most highly expressed 239 SWEET gene, we hypothesized that knock-out mutants might be affected in seed filling. 240 Two independent ossweet11 mutants, one carrying a single nucleotide deletion leading 241 to a frameshift was created by CRISPR-Cas9 and a second TALEN-derived mutant car-242 rying a 489 bp deletion were characterized phenotypically (Fig. S2). In the greenhouse, 243 both mutants had incompletely filled seeds at maturity (Fig. 2a). Depending on the growth 244 conditions, the phenotype was more or less severe (see for example Fig. 2 and Fig. S3). 245 The effects became more severe in paddy field conditions (single field experiment in 246 2016; see also parallel study (Ma et al., 2017). Moreover, panicle development of 247 ossweet11 mutants only much later (>60 DAP). As a result, mutants had a significantly 250 reduced yield (both percentage of mature seeds after harvest and 1000-grain weight; 251 Fig. S3a,b). Of note however, plant height, spikelet number and panicle length were sim-252 ilar as in wild-type also in paddy field conditions (Figs. 2c,S3c,d). 253

OsSWEET11 accumulation in nucellar projection, nucellar epidermis and aleurone 254
Oparka had predicted symplasmic diffusion of sugar in the pericarp and apoplasmic 255 transport at the nucellar epidermis-aleurone interface all around the endosperm, which 256 contrasts the sucrose import patterns found in developing barely seeds, where the main 257 import route was through the nucellar projection (Oparka & Gates, 1984;Melkus et al., 258 2011). To determine whether OsSWEET11 exports sucrose at the nucellar projection or 259 the nucellar epidermis/aleurone interface, we analyzed transgenic rice plants expressing 260 translational GUS fusions containing a 2kb promoter fragment and the whole coding re-261 gion including all introns. Crude histochemical GUS analysis of caryopses showed com-262 parable GUS staining in seeds in 8 out of 18 independent transformants. Two independ-263 ent lines were used for a more detailed analysis. In early stages (up to 3 DAP), we ob-264 serve GUS activity in maternal tissues including the ovular vascular trace and the nucel-265 lus, possibly indicating a role in remobilization of carbohydrates during nucellar degrada-266 tion ( Fig. S4). At 5 DAP, GUS activity was detected in the ovular vascular trace, the nu-267 cellar projection, the nucellar epidermis surrounding the developing endosperm, the re-268 maining nucellar proper and also in the aleurone layer of the endosperm (Fig. 3a-c). To 269 our surprise, we find OsSWEET11 expression in nucellar projection as well as the nucel-270 lar epidermis, and in addition also in the outermost endosperm cell layer, the aleurone, 271 providing a potential path for sucrose export out of the nucellar projection into the endo-272 sperm, and a parallel pathway for export from the circumferential nucellar epidermis and 273 then subsequently a potential import via OsSWEET11 into the aleurone. A parallel study 274 observed a similar expression pattern using a transcriptional GUS fusion (Ma et al., 2017). 275 Potential compensation for ossweet11 deficiency by other SWEETs 276 the expression levels of clade 3 SWEET genes in the ossweet11 mutant showed that 279 OsSWEET13 mRNA levels were slightly increased, however the absolute levels were 280 extremely low. The mRNA levels of OsSWEET15 were about twofold higher in ossweet11 281 seeds compared to wild-type (Fig. 4). Since depending on the experiment, OsSWEET15 282 was expressed at only slightly lower levels compared to OsSWEET11 and was further-283 more candidate that may contribute to compensation in the mutant, we first tested 284 whether it functions as a sucrose transporter, determined its expression pattern in devel-285 oping caryopses and then analyzed knock-out mutants. As one may have predicted, 286 OsSWEET15 also functioned as a sucrose transporter when co-expressed with a sucrose 287 sensor in HEK293T cells (Fig. S5). Translational GUS fusions of the OsSWEET15 gene 288 driven by their native promoter showed similar tissue specificity as compared to 289 OsSWEET11 (13 independent lines): i.e. GUS activity was detected at early stages in the 290 nucellus proper, and later in the ovular vascular trace, the nucellar projection and the 291 aleurone (Fig. 3d,e). At 9 DAP, OsSWEET15 GUS activity was also detected in nucellar 292 epidermis (Fig. 3f). The two SWEET transporters exhibit similar expression pattern in the 293 developing seeds, especially ovular vascular and interface between the nucellar epider-294 mis and the aleurone layer, assuming redundant roles during seed development. How-295 ever, on their own, two independent ossweet15 knock-out mutants generated via 296 CRISPR-Cas9 (frameshift mutations that prevent production of a functional OsSWEET15 297 protein, Fig. S2) did not show any detectable phenotypic differences compared to wild-298 type in 4 independent experiments (Fig. 2b). 299

OsSWEET11 and 15 are essentially for seed filling 300
Since the seed filling of ossweet11 mutants was only partially affected relative to 301 ossweet4 mutants (Sosso et al., 2015), and OsSWEET15 appeared to be expressed in 302 the same cell types to substantial levels, and even possibly compensates in part for 303 OsSWEET11 deficiency in the mutant, we generated ossweet11;15 double mutants for 304 both alleles of the two loci. In greenhouse conditions both at ISU and Stanford, the double 305 mutant phenotype was very severe, much more than the single ossweet11 mutant (Fig.  306 5). The differences were even more severe in the ISU greenhouses, where ossweet11 307 came apparent at ~5 DAP (Fig. 5a). Differences became much stronger at 7 DAP, a time 310 point at which ossweet11 mutants already started to develop a wrinkled grain morphol-311 ogy, while ossweet11;15 was characterized by grains that were flattened with a smaller 312 diameter (Fig. 5a,b). Sections through the grain showed that the mutants were endo-313 sperm-deficient and either had only remnants of the endosperm or lost the endosperm 314 completely (Fig. 5c). Cytohistological analyses of resin-embedded sections showed that 315 cellularization of the endosperm started at ~3 DAP in both wild-type and ossweet11, while 316 cellularization was not observed in ossweet11;15 mutant (Fig. S7). The cellularization of 317 the endosperm was completed and nucellus were degenerated in the wild-type from 5 to 318 7 DAP (Fig. S7a). In the ossweet11 mutant, endosperm was not fully cellularized and 319 degradation of nucellus were delayed (Fig. S7b). In the ossweet11;15 mutant, endosperm 320 cells defected cellularization and nucellus remained until 7 DAP (Fig. S7c). 321

Starch accumulation in the pericarp of ossweet11;15 double mutants 322
Based on the localization of OsSWEET11 and 15, we predicted that inhibition of the trans-323 porters would lead to starch accumulation in cells that export sucrose and cells outside 324 the endosperm. In wild-type, starch is stored transiently in the pericarp until 7 -9 DAP. 325 Rapid starch degradation in the pericarp correlated with starch accumulation in the endo-326 sperm starting ∼7 DAP (Wu et al., 2016). We found starch in the endosperm, but only 327 residual amounts in the pericarp of wild-type caryopses (Fig. 6a,d). By contrast, starch 328 accumulated to high levels in the pericarp of the ossweet11 mutant (Fig. 6b,e). In 329 ossweet11;15 double mutants, starch accumulated to even higher levels in the pericarp, 330 while the endosperm did not show substantial starch levels (Fig. 6c,f). The accumulation 331 of starch in the pericarp of ossweet11 and ossweet11;15 double mutants supports the 332 critical roles of OsSWEET11 and OsSWEET15 in sugar translocation and mobilization 333 towards the developing endosperm. 334 expression, translational reporters to map tissue-specific protein accumulation, nuclease-337 induced knock out mutants: (i) OsSWEET11 and 15 are the most highly expressed su-338 crose transporting clade 3 SWEETs in the rice caryopsis, (ii) if we assume that they mark 339 apoplasmic import routes, sucrose can enter both directly below the vein via the nucellar 340 projection as well as the circumferential nucellar epidermis; (iii) they may not only play 341 roles in cellular efflux at these two sites, but also be responsible for importing sucrose 342 into the aleurone cells; (iv) OsSWEET11 and 15 are both contribute to seed filling with 343 redundant roles, but since the single ossweet15 mutant alone has no apparent phenotypic 344 differences to wild-type, OsSWEET11 appears to function as the dominant transporter,  (Fig. 7). There appear to be four locations that re-359 quire sucrose transporters: 1. Parenchymatic cells in the vascular bundle: a localiza-360 tion that may be expected based on the OsCIN2 localization which requires efflux of su-361 crose (and subsequent import of hexoses for providing sucrose as a substrate (Wang et 362 al., 2008). Of note, the oscin2/gif1 cwINV mutant has a clearly distinct phenotypes with 363 markedly more grain chalkiness and is thus not similar to ossweet11 (Wang et al., 2008). 364 influx and is not expected to efflux sucrose. However, SWEETs appear to function as 371 uniporters, thus the sucrose gradient simply determines the direction of flux. We had pre-372 viously found that SWEET4 in maize and rice were likely key to the import of cwINV-373 derived hexoses in seeds (Sosso et al., 2015). A sucrose gradient across the two cell 374 types (nucellar epidermis and aleurone) driven by a high rate of delivery from the maternal 375 side and rapid conversion in the endosperm would allow the use of the same transporters 376 on both cell types. This situation is remotely similar as in the human intestine, where 377 transcellular transport across the intestinal epithelia is mediated by GLUT2 on both the 378 apical and basal membrane under conditions where the glucose concentrations in the 379 lumen exceed those of the blood stream (Kellett et al., 2008). In addition to the two 380 SWEETs, SUTs, which are expressed in the aleurone, may contribute to secondary active 381 sucrose import into the aleurone (Ishimaru et al., 2001;Scofield et al., 2002;Bai et al., 382 2016). 4. Nucellar projection: The presence of OsSWEET11 and 15 in the cells of the 383 nucellar projection may appear as the most surprising site, since plasma membrane su-384 crose transport is not in line with radiotracer import studies, which indicated that in rice, 385 the import of sugars occurs exclusively via the nucellar epidermis-aleurone pathway 386 (Oparka & Gates, 1981b). However, others have suggested that the nucellar projection 387 may help to transporting sugars to the developing endosperm also in rice (Krishnan & 388 Dayanandan, 2003). Importantly, the nucellar projection pathway appears to be the main 389 pathway for sugar import in barely caryopses as shown by magnetic resonance imaging 390 (MRI) (Melkus et al., 2011). Of note, we are aware that by contrast to Oparka's radiotracer 391 studies, we do not measure actual translocation of assimilates, but rather the presence 392 of a protein, and we do not know whether the two SWEETs are active at the plasma 393 membrane of these cells. Nevertheless we suggest that it may be useful to reassess 394 sugar entry pathways, for example by MRI at different stages and in different varieties. 395 for sucrose efflux from the seed coat also showed very complex changes in cellular ex-402 pression during seed development (Chen et al., 2015b). 403

Starch in the pericarp as a transient buffer 404
In rice caryopsis development, large amount of starch grains accumulated in pericarp at 405 6 DAP, followed by those starch grains degraded from 7 to 9 DAP (Wu et al., 2016). This 406 type of starch accumulation and degradation has been observed in pericarp of barely and 407 wheat (Radchuk et al., 2009;Xiong et al., 2013). Starch accumulation occurred in the 408 ossweet11 and ossweet11;15 of pericarp at 9 DAP (Fig. 6), suggesting the presence of 409 one route for delivery of sucrose in the pericarp. timing of mRNA accumulation, tissue specificity and the combined effect observed in dou-417 ble knock out mutants that OsSWEET15 can compensate for OsSWEET11 deficiency. 418

Developmental control of seed filling 419
Rice caryopsis occurred dynamical changes in cell expansion, cell wall thickening, and 420 starch grain accumulation (Wu et al., 2016). Expression levels of OsSWEET11 and 421 OsSWEET15 are gradually increased during caryopsis development. In contrast 422 OsSWEET4, which is expressed mainly at the base of the caryopsis, and mutation strong 423 seed filling defect, however timing very different, very high early and then declining (Fig.  424 1). Early seed development, imported sucrose released from maternal tissue is cleaved 425 expressed early stage (Hirose et al., 2002;Cho et al., 2005). Since OsSWEET4 belongs 427 to clade 1 SWEET and shown as a glucose transporter (Sosso et al., 2015), role of clade 428

Relevance for pathogen susceptibility 431
The finding that OsSWEET11 and 15 play important roles in seed filling is also relevant 432 in the context of the fact that OsSWEET11 serves as a blight susceptibility gene (Yang et 433 al., 2006;Yuan et al., 2010;Antony et al., 2010;Chen et al., 2010). Ectopic expression 434 of OsSWEET11 is activated by pathovar-specific effectors of the blight pathogen Xan-435 thomonas oryzae pv oryzae, and mutations in the effector binding sites in the 436 OsSWEET11 promoter lead to resistance to Xoo (Yang et al., 2006;Yuan et al., 2010;437 Antony et al., 2010). Therefore, it will be important to ensure that engineering of the 438

OsSWEET11 promoter in resistant lines retains proper OsSWEET11 expression in seeds 439
to ensure that resistant lines do not carry a yield penalty. This goal appears feasible since 440 apparently mutants (xa13) that are used by breeders do not show yield deficiencies (Laha 441 et al., 2016). 442

Conclusions 443
The analysis of SWEET gene expression in rice caryopses together with the characteri- wild-type greenhouse-grown seeds. Data was shown as mean ± s.e.m., n=3; expression 577 levels were normalized to rice Ubiquitin1 levels. was measured with quantitative RT-PCR (qRT-PCR) in wild-type and sweet11. Expres-600 sion of OsSWEET13 and OsSWEET15 were increased in the ossweet11-1 mutant com-601 pared to wild-type (*p<0.05). Data was shown as mean ± s.e.m., n=3; expression levels 602 were normalized to rice Ubiquitin1 levels. OsSWEE12 transcripts were not detected.  The model was made based morphological observations from this study and previous 620 studies (Oparka & Gates, 1981a,b;Wu et al., 2016). Sucrose may move from the 621 phloem to parenchyma cells in the ovular vascular bundle and then to nucellar projec-622 tion and the nucellar epidermis through symplasmic pathway via plasmadesmata (a). 623 We surmise that OsSWEET11 and OsSWEET15 mediates sucrose export from xylem 624 parenchyma cells into the apoplasmic space (b). In addition, OsSWEET11 and 625 OsSWEEET15 may be involved in sucrose export out of cells at the nucellar projection 626 and the nucellar epidermis to apoplasm. followed by import into the aleurone in endo-627 sperm (c). Transfer across the nucellar epidermis/aleurone would require a sucrose gra-628 dient across both cell types. al, aleurone; ne, nucella epidermis; np, nucellar projection; 629 pa, parenchyma. 630