Maternal effect contributes to grain-filling defects of Ospho1;2 rice mutants

In seeds, filial tissues, such as endosperm and embryo, and maternal tissues, such as the seed coat or caryopsis coat of cereal grains, are symplastically isolated (Krishnan & Dayanandan, 2003). Nutrient delivery through plasma membrane-localized transporters at the junction of filial and maternal tissues is indispensable for sustaining endosperm and embryo development. Phosphorus (P), an essential element, is transported primarily in the form of inorganic phosphate (Pi) across cell membranes (Chiou & Lin, 2011; Chiou, 2020). Successful P delivery is achieved by Pi efflux from maternal cells (unloading) followed by Pi influx into filial cells (uptake).

Recent studies have reported the importance of PHO1 Pi transporters in transferring Pi from maternal to filial tissues in developing seeds in addition to their previously identified role in root-to-shoot Pi translocation. In Arabidopsis seeds, PHO1 is explicitly expressed in the chalazal seed coat (CZSC) (Vogiatzaki et al., 2017). In rice, OsPHO1;2 is the closest homolog of Arabidopsis PHO1 (Secco et al., 2010). Loss-of-function Ospho1;2 mutants produced small and shrunken seeds with reduced starch content, resulting in grain chalkiness, highlighting the crucial role of OsPHO1;2 in determining grain quality and yield (Che et al., 2020; Ma et al., 2021). Although opposite Pi transport directions (influx or efflux activity) were concluded in these two reports that used different methods, both reports showed predominant OsPHO1;2 expression in the nucellar epidermis, the innermost cell layer of the caryopsis coat where nutrients are unloaded from maternal to filial tissues (Che et al., 2020; Ma et al., 2021). Thus, it is rational to envision that the grain-filling defects in Ospho1;2 mutants result from a reduced level of Pi in the starchy endosperm owing to impaired Pi unloading from maternal tissues. However, Ma et al. (2021) proposed a model in which an elevated Pi level in the Ospho1;2 starchy endosperm cells obstructs starch synthesis through inhibiting the activity of the first committed enzyme, ADP-glucose pyrophosphorylase (AGPase) because of reduced Pi efflux from the inner starchy endosperm cells. This model called into question the means by which the Ospho1;2 endosperm accumulates high Pi if its delivery from the maternal tissues is blocked. In the present study, we carefully examined the expression of OsPHO1;2 and total P and Pi distributions in the grains. We also analyzed the relevance of OsPHO1;2 in grain development through reciprocal crossing between Ospho1;2 mutants and wild-type (WT) rice. Our results demonstrate that the grain phenotypes of Ospho1;2 are caused by defective Pi unloading from maternal tissues. These results suggest that the model proposed by Ma et al. (2021) regarding the role of OsPHO1;2 in rice grain filling needs to be revised.

Because root-expressed OsPHO1;2 plays a crucial role in root-to-shoot Pi translocation (Secco et al., 2010) and its expression in node I participates in P distribution to seeds (Che et al., 2020), we first asked whether the grain-filling defects of Ospho1;2 mutants resulted from an overall low Pi level in aboveground tissues. We grew the Nipponbare WT rice plants under Pi sufficient (200 μm) or deficient (10 μm) conditions by hydroponic culture for 134 d until heading and grain maturation. We found that despite the reduction in panicle number and a remarkable decrease in Pi level in the flag leaf, the Pi concentrations in the caryopsis produced from high and low Pi growth conditions were not different, and no chalky seeds were observed in the low Pi-grown plants (Supporting Information Fig. S1). These results suggest that the general low Pi status reduces grain yield but does not cause the seed development defect as seen in Ospho1;2 mutants.

To clarify the controversy about whether the shrunken and chalky grains of Ospho1;2 mutants are caused by impaired Pi unloading from the maternal caryopsis coat or by excessive Pi accumulated in the inner starchy endosperm, we re-analyzed the expression pattern of OsPHO1;2 in the seeds during the grain-filling stage at 9 days after pollination (DAP). We excised the embryo and gently squeezed the milky inner starchy endosperm from the caryopsis coat. The outer endosperm containing aleurone and subaleurone layers (Krishnan & Dayanandan, 2003; Iwai et al., 2012; Wu et al., 2016), likely attached to the caryopsis coat, were pooled as one fraction. To verify the enrichment of our isolation, we measured the transcript level of SUGAR WILL EVENTUALLY BE EXPORTED TRANSPORTER 11 (SWEET11), known to be expressed specifically in the nucellar projection, nucellar epidermis and aleurone (Yang et al., 2018), and starch branching enzyme 1 (SBE1), which is mainly expressed in the scutellum between the endosperm and embryo (Qu & Takaiwa, 2004) by reverse transcription quantitative polymerase chain reaction. Consistent with previous reports, the expression of SWEET11 was detected explicitly in the caryopsis coat and aleurone fraction, whereas SBE1 was preferentially expressed in the inner endosperm (Fig. 1a,b), indicating the enrichment of specific tissues in these fractions. We then examined the expression of OsPHO1;2 and found that its mRNA level was much more abundant (about sixfold higher) in the caryopsis coat and outer endosperm fraction than in the inner endosperm of the WT (Fig. 1c). OsPHO1;2 mRNA was barely detected or significantly reduced in the Ospho1;2-1 (NC0015) and Ospho1;2-2 (NE2044) mutants generated by Tos17 insertion (Fig. 1c), two loss-of-function allelic mutants characterized previously (Secco et al., 2010). Interestingly, the expression of SBE1 was reduced substantially in the Ospho1;2 endosperms, even though its tissue preference was retained. The reduction in SEB1 expression reflects the defective starch biosynthesis in the Ospho1;2 endosperms.

We next separated the grains into husk and caryopsis. We detected their total P and Pi concentrations along with the fractions of caryopsis coat and outer endosperm and starchy endosperm at three different seed development stages, 9 DAP (S3), 16 DAP (S4), and mature stage (S5, for husk and caryopsis only). Compared with the WT, both Ospho1;2 mutants had a low total P but high Pi concentration in caryopsis, particularly at the S5 stage (Fig. 1d,e). A similar trend of changes was observed in the caryopsis coat and aleurone fraction. However, Ospho1;2 mutants accumulated excessive total P (twofold that of the WT) and Pi (four- to fivefold WT) in the S5-stage husk but much less total P and Pi in the S3- and S4-stage starchy endosperms (Fig. 1d,e). The reduced total P and Pi levels in the Ospho1;2 starchy endosperms were further supported by the upregulation of several Pi starvation-induced genes, such as INDUCED BY PHOSPHATE STARVATION1 (OsIPS1) (Hou et al., 2005), Phosphate Transporter Protein 8 (OsPT8) (Jia et al., 2011), and Purple acid phosphatase 10c (OsPAP10c) (Lu et al., 2016) (Fig. 1f–h). Total P in grains is stored as phytic acid (PA), also known as inositol hexakisphosphate (InsP₆), primarily in the bran (caryopsis coat and outer endosperm) and embryo of rice (Iwai et al., 2012; Perera et al., 2018). We did indeed find that Ospho1;2 caryopsis contains only 50–60% of the WT level of PA (Fig. 1i). These observations suggest an impairment in delivering Pi from the maternal into the filial tissues of Ospho1;2 grains for PA synthesis.

To validate the reduced P/Pi accumulation in the endosperm of Ospho1;2 mutants, we applied laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) to analyze the ³¹P distribution in the cross-section of mature caryopsis. The images were quantified and normalized with ¹³C⁺ as an internal standard element. We observed diminished P signals in the endosperms of both Ospho1;2 mutants, especially in the inner endosperm, corresponding to the chalkiness area (Fig. 1j). Furthermore, we sectioned caryopsis of the WT and Ospho1;2-1 mutant at 16 DAP and then stained Pi using a recently published method, inorganic orthophosphate staining assay (IOSA) (Guo et al., 2024), which offers a semi-quantitative image of Pi distribution. Under the same condition, we found that Pi staining in the Ospho1;2-1 mutant was increased in the caryopsis coat (dorsal vascular bundle, cross cell and nucellar epidermis) and subaleurone layers but not in the inner starchy endosperm compared with the WT (Fig. 1k). The Pi staining in the WT caryopsis was generally very faint except for a weak signal at the lateral subaleurone layers (block arrowheads in the upper left panel of Fig. 1k), likely as a result of an active synthesis of PA from Pi. These observations agree with the results of PA and LA-ICP-MS (Fig. 1i,j) analyses, which taken together, support the notion that Ospho1;2-1 mutants are defective in Pi transfer from maternal to filial tissues and impaired conversion of Pi into PA in the bran. Despite there being no apparent differences in Pi staining in the inner starchy endosperm between WT and mutants (Fig. 1k), the results of both physical separation (Fig. 1d–f) and LA-ICP-MS image analyses (Fig. 1j) clearly showed reduced P and Pi accumulation in the inner starchy endosperm of Ospho1;2 mutants.

To further confirm that the shrunken and chalky grains seen in Ospho1;2 mutants can be attributed to impaired Pi unloading from maternal tissues, we carried out reciprocal crosses between Ospho1;2-1 mutants and WT. Notably, we found that only the heterozygous F1 progeny generated from the maternal Ospho1;2-1 (homozygous pho1;2-1pho1;2-1 allele) yielded the shrunken and chalky grains, as shown by less transparency from backlighting (Fig. 2a) despite the presence of a functional OsPHO1;2 allele in the endosperm. Moreover, the F2 seeds collected from self-pollination of the F1 (heterozygous PHO1;2pho1;2-1 allele), either from WT × Ospho1;2-1 or Ospho1;2-1 × WT crossing, all displayed the WT-like phenotype (Fig. 2a). Indeed, the segregation of the F2 offspring genotypes fit the Mendelian ratio of 1 (PHO1;2PHO1;2) : 2 (PHO1;2pho1;2-1) : 1 (pho1;2-1pho1;2-1) (Fig. 2b). To reinforce this observation, we measured the F1 and F2 caryopsis dry weight along with the parental lines. We found that the heterozygous F1 seeds generated from the two maternal Ospho1;2 mutants showed > 50% reduced dry weight (Fig. 2c), but all F2 seeds, no matter whether the initial maternal came from WT or Ospho1;2-1, had a similar dry weight to the WT (Fig. 2d) and without chalkiness (Fig. S2). It is worth noting that the F2 seeds with homozygous pho1;2-1pho1;2-1 null allele display the WT phenotype because their heterozygous F1 lines carry a functional OsPHO1;2 allele in the maternal tissues. The WT-like phenotype of F2 seeds with the pho1;2-1pho1;2-1 null allele also implies that OsPHO1;2 has a dispensable role in endosperm cells. The result of reciprocal crosses provides unequivocal evidence that a maternal effect contributes to the shrunken and chalky grains of Ospho1;2 mutants.

Ma et al. reported that considerable expression of OsPHO1;2 mRNA was detected in the WT endosperm (c. 65% of which was reported to be expressed in the nucellar epidermis) (Extended Data, fig. 4c,d in Ma et al., 2021) and c. 20% increased Pi relative to the WT control was observed in the Ospho1;2 endosperms (Extended Data, fig. 7e in Ma et al., 2021), which contradicts our results (Fig. 1). How the endosperm isolated from the pericarp and nucellar epidermis was not described in the experiments reported in Ma et al., and the purity of each fraction and whether the endosperm fraction contains aleurone or subaleurone are unknown. Furthermore, no OsPHO1;2 expression signal was detected in the endosperm according to the immunohistochemistry and promoter-reporter analyses conducted in their study (Fig. 2c; Extended Data, fig. 4e in Ma et al., 2021), which argues against the conclusion and model proposed by the authors.

In summary, we conclude that loss-of-function of OsPHO1;2 in the nucellar epidermis of the caryopsis coat rather than in the endosperm, resulting in defective unloading of Pi from maternal tissues, is the primary cause of the grain-filling defects observed in Ospho1;2 mutants. A revised working model is presented in Fig. 2(e). Additionally, our results point to an alternative mechanism inhibiting the starch biosynthesis in the endosperm of Ospho1;2 mutants. Although the activity of AGPase is usually allosterically inhibited by Pi (Ballicora et al., 2004), interestingly, the RNA levels of many genes involved in starch biosynthesis, including AGPL2 and AGPS2b, are downregulated in Ospho1;2 mutants during the grain-filling stage (fig. 4a and supplementary fig. 6 in Ma et al., 2021), suggesting the alteration of starch biosynthesis in Ospho1;2 grains may occur at the transcriptional or post-transcriptional level. How reduced Pi in the Ospho1;2 endosperm affects starch biosynthesis awaits future investigation.

None declared.

W-CL, J-CH, H-FC and M-JL performed the experiments and analyzed the data. S-SK, K-CY and T-JC contributed to the study design and supervision. S-SK and T-JC wrote and finalized the manuscript. S-SK, W-CL and J-CH contributed equally to this work.