The tests were conducted in 2021 at the test site in Shatou, Yangzhou, with wheat being the previous crop in that area. The good-appearance soft rice varieties (Yangnongxiang 28 and Nanjing 5718), and the cloudy soft rice varieties (Wuxiangjing 113 and Yangjing 5118), used in 2020, were selected as test materials. For notations, Y28 represents Yangnongxiang 28; N5718 represents Nanjing 5718; W113 represents Wuxiangjing 113; Y5118 represents Yangjing 5118.

The test site was selected from plots with medium to good ground strength and balance (basic ground strength), yields above 400 kg and easy access to irrigation. A normalized group design was used.

The cultivation method was machine inserting blanket seedlings, in rows 30cm × 11.7cm apart, and with 4-5 plants per hole.

Nitrogen fertilizer level was 270kg /hm², based on base fertilizer: tiller fertilizer: spike fertilizer = 3.5:3.5:3. Tiller fertilizer was applied seven days after transplanting, and spike fertilizer was applied at the inverted four-leaf stage. The ratio of N: P: K was 2:1:2, with phosphate as a one-time base fertilizer and potash applied in equal amounts before plowing and at the nodulation stage respectively. Water management, pest and weed control and other related cultivation measures were implemented following high yield cultivation requirements.

Spikes with uniform growth and flowering on the same day were tagged at the beginning of flowering for each variety. Seed samples were taken 5, 10, 15 and 20 days after flowering (DAY) and at maturity.

The chalkiness of the seeds at maturity was determined according to GB/T 17891-2017. The chalkiness, chalk area and chalky rate were calculated (WS-SC-E, China).

When the rice sample was placed in a 1 cm thick cuvette, the transparency of the rice can be expressed by its transmission using a colorimeter with a D65 light source (CM-5 Japan). The transparency of normal seeds at maturity and non-chalky seeds. were measured.

Representative rice was arranged neatly in a dark room. The appearance of the white rice was photographed with a stereo microscope (Leica S8AP0, Germany).

The seed samples were fixed in pre-chilled 2.5% glutaraldehyde solution immediately after leaving the spike at the filling stage, and washed three times with 0.1 mol/L (pH 7.2) phosphate buffer after three hours. Then ethanol gradient dehydration, isoamyl acetate replacement, and CO2 critical point drying were conducted. The final sample was prepared using ion beam sputtering coating and then observed using a scanning electron microscope (Philips XL-30 ES-EM, Netherlands).

Mature seeds were frozen in liquid nitrogen for 3~5 min. Then the external force was applied to the rice grains with forceps to make them brittle. The sample was pasted, sputter-coated and observed in the same way as described above.

The particle size distribution of starch was investigated using a laser diffraction particle size analyzer (Mastersizer 2000, Malvern, UK). About 0.2 g starch samples were immersed in 500 mL absolute ethyl alcohol and stirred at 2000 rpm. The instrument was calibrated to measure starch particle sizes ranging from 0.1 to 2000 µm.

T2 (transverse relaxation time) values of cooked rice were measured using a nuclear magnetic resonance analyzer (NMI-20Analyst, Shanghai Niumag Electronic Technology Co., Shanghai, China) with the Carr-PurcellMeiboom-Gill (CPMG) sequence. A subsample of each rice sample used to determine taste was weighed to 5 g in glass vials and placed in a nuclear magnetic tube with a diameter of 25 mm. The center frequency was first calibrated with a free induction (FID) decay pulse train using the NMR spectrum analysis software. The parameters of the CPMG pulse train for each rice sample were set as follows: main frequency SF1=20MHz; sampling frequency SW=200 kHz; 90° hard pulse RF pulse width P1=6 μs; 180° hard pulse RF pulse width P2=13.04 μs; signal sampling points TD=600010; waiting time for repeat sampling, TW=2500 ms; the number of replicates, NS=4; echo number NECH=15000. This technique was studied and explained in depth by Klitsadee et al. (Klitsadee et al., 2019).

After obtaining the images by scanning electron microscopy, 200 starch grains were selected for each sample according to the magnification-corrected scale, and the number of starch grain cavities was calculated. The area of the cavities was measured using the PS software, and their mean and standard deviation were obtained using Excel.

Each experiment was repeated at least three times independently. All of the reported date were calculated into averages and analyzed using one-way analysis of variance (ANOVA) and Tukey’s test in SPSS (version 16.0).

The appearance of Y28 and N5718 differed significantly from that of W113 and Y5118 ( Figure 1 ). Y28 and N5718 had transparent texture, and small chalky areas could be seen in the core and base of individual grains ( Figures 1A, B ). The endosperm of W113 and Y5118 generally showed a cloudy opaque phenotype.

Rice transparency is measured as the overall transparency of sampled fine rice grains in previous studies, which refers to light transmission performance of sampled fine rice under transmitted light. It can be affected by both chalky and non-chalky parts of rice grains. In order to further investigate the causes of the differences in soft rice transparency, the non-chalky grain transparency of four soft rice varieties was also measured in this study. This transparency can represent original chalk-free transparency of rice grains without the interference of chalkiness ( Table 1 ).

Table 1 shows no differences in the moisture content of rice grains ( Table 1 ). The good-appearance soft rice varieties had significantly higher chalk-free seed transparency than the cloudy soft rice varieties, indicating that the grains of the good-appearance soft rice varieties were originally more transparent. The transparency was 0.9%-1.32% and 2.97%-3.08% lower than their chalk-free seed transparency for the good-appearance soft rice varieties and the cloudy soft rice varieties, respectively. This indicates that rice grains of the cloudy soft rice varieties had lower transparency, higher chalkiness and larger differences in transparency from their chalk-free seed transparency. This should be influenced by the different chalky phenotypes, because chalky area, chalky rate and chalkiness of the good-appearance soft rice varieties were significantly lower than those of the cloudy soft rice varieties. This indicates that the good-appearance soft rice varieties had higher chalk-free seed transparency and lower chalkiness.

The above analysis shows that the appearance of these four varieties was significantly different regarding chalk-free seed transparency and chalkiness. Normal and non-chalky grains of these four varieties were selected to further investigate the endosperm structure that caused differences in these two types of appearance.

The differential endosperm structure of different soft rice varieties is closely related to the grain-filling process. What characteristics does endosperm exhibit during this process? Therefore, the endosperm structure of different soft rice varieties at different stages after flowering was systematically investigated.

Better appearance quality has always been an effective means for rice resources to enhance their market competitiveness. However, there is no industry-accepted standard for determining good-appearance soft rice varieties. In China, the appearance quality of ordinary rice is mainly measured by chalkiness (Li and Zhen, 2022). It indicates that lower chalkiness represents better rice appearance quality in the existing industry standards. In this experiment, it was found that similar to non-soft rice japonica rice, different appearance types of soft rice differed significantly in chalkiness. It shows that chalky differences are common to both soft and non-soft rice, and that less chalky soft rice also means having better appearance quality. However, differ from non-soft rice, soft rice tends to have a low transparency phenotype, which seriously affects the appearance quality of soft rice (Lin et al., 2016; Li et al., 2018; Zhang et al., 2019a). Li et al. found that soft rice varieties Nanjing 46 and Guangdong 194 showed a significant decrease in grain endosperm transparency at a storage moisture content below 12% (Li et al., 2018). Zhang et al. studied the transparency of dried brown rice with the same genotype but different AC and found that the transparency of rice with the AC of soft rice was significantly lower than that of normal rice (Zhang et al., 2019a). Wu et al. found that the endosperm of all low AC materials (including early indica rice varieties (lines), restorer lines and sterile male lines of hybrid rice) showed easily identifiable cloudy traits under indoor storage conditions for three months. However, the appearance of the endosperm of the varieties with medium or high AC did not change significantly and remained transparent (Wu et al., 2001). The transparency of different appearance types of soft rice was also found to be significantly different in this study, indicating that transparency is not only a trait that distinguishes soft rice from non-soft rice, but also varies greatly between different soft rice. The chalk-free seed transparency and chalkiness differed significantly between different soft rice varieties. Compared with the cloudy soft rice varieties, the good-appearance soft rice varieties had lower chalkiness but higher transparency. Therefore, the appearance quality of soft rice should be measured with chalkiness and transparency-related indexes.

The endosperm structure is a direct cause of the rice appearance quality. In many previous studies, the endosperm structure of highly chalky rice was compared with that of low chalky rice. It was found that chalkiness was mainly caused by the insufficient filling of endosperm, insufficient fullness of starch grains and loose arrangement (Dong et al., 2006; Zhang et al., 2006; Liu et al., 2007; Deng et al., 2021; Wang et al., 2021). In some transgenic rice studies, by comparing the endosperm structure of the chalky mutant with that of the wild type, it was found that the starch arrangement of the chalky mutant endosperm was more lax (Wang et al., 2010; Zhu et al., 2018; Can et al., 2020; Wu et al., 2022). In this study, we found that the chalky phenotype of soft rice was also due to the loose arrangement of endosperm starch grains, and the low chalky soft rice had fuller starch grains and a tighter starch grain arrangement structure. This was consistent with previous studies. In the starch granules size and distribution of rice grains test, it was also found that the average starch grain size was larger in the low chalky soft rice, and this result also confirmed the phenomenon that fuller starch grains contribute to the reduction of chalkiness. This may be due to the fact that fuller starch grains can easily compress to form a tight endosperm structure, and gaps between starch grains are less likely to exist.

The endosperm structure of soft rice was also compared with that of ordinary rice in previous studies (Lin et al., 2018; Shun et al., 2019; Yu et al., 2019). Zhang et al. analyzed the relationship between transparency and endosperm structure of starch grains of transgenic rice that was constructed based on glutinous rice and had different AC (Zhang et al., 2019a). They found many cavities between endosperm starch grains of glutinous rice. The pore number decreased with increasing AC, while rice transparency was enhanced. This indicates that the opacity of glutinous rice and the translucency of soft japonica rice were attributed to the cavities between starch grains. These cavities caused light scattering, thus affecting rice transparency (Zhang et al., 2013; Zhang et al., 2017). Hao et al. found that the number of cavities in individual starch grains was significantly higher in the low AC rice group compared to the medium and high AC rice groups (Hao, 2021). In the study of the Wx^IV gene, the presence of starch granule cavities in the endosperm of low transparency was found (Zhang et al., 2019b). However, these studies ignored the effect of chalk on transparency or only compared the endosperm transparency of individual seeds. In this study, the effect of chalkiness on transparency was well excluded by using chalk-free seed transparency as the focus indicator. It was found that starch grain cavities were responsible for the transparency of soft rice without chalky seeds. Combined with the previous studies, it is suggested that the fewer and smaller starch grain cavities contributed to the higher transparency of soft rice. Previous studies have found that the effect of moisture content on transparency were significantly and positively correlated with the size of starch grain cavities (Zhang et al., 2019a; Wang et al., 2021; Zhu et al., 2021; Hao, 2021). There are fewer studies on the moisture distribution state of rice grains on rice appearance. The relaxation time (T2) is closely related to the type and state of hydrogen protons, and the value of the transverse relaxation time reflects the water freedom inside the specimen (Song et al., 2016; Klitsadee et al., 2019; Yifu et al., 2019). This study found that the content and proportion of free water in good-appearance soft rice varieties were significantly higher than that in the cloudy soft rice varieties at the same moisture content between 14%-15%. The authors suggest that during rice grain drying to reduce moisture content, the water in the starch grain cavities was more easily precipitated than the semi-free water tightly bound to the starch. The cloudy soft rice lost more semi-free water due to more and larger starch grain cavities. The cavities that lost water scattered light and thus reduced endosperm transparency. This may be the pathway by which starch grain cavities affect the transparency of soft rice.

Many scholars have concluded that the development of rice chalkiness was closely related to the dynamic variation of dry matter accumulation, filling rate and moisture content of rice grains during rice filling (Yang et al, 2003; Liu et al., 2007; Wenting et al., 2021). The presented study found that at DAF 10, there were significant differences in the endosperm back and core. This indicates that differences in the endosperm structure of different soft rice varieties started to develop in the early filling stage. The differences were mainly in the back and core, possibly because from DAF 5 to 10, the filling material of all soft rice varieties in the early stage of filling was delivered more to the back and core parts and less to the belly part. This also shows that the good-appearance soft rice varieties had a higher filling rate than the cloudy soft rice varieties. At DAF 15, the difference in endosperm structure was mainly in the endosperm belly. This indicates that during DAF 10-15, more filling material was delivered to the belly and the belly structure rapidly increased in densitye after the endosperm cells at the back and core were filled. Thus, the difference in the endosperm structure between good-appearance soft rice varieties and cloudy soft rice varieties was further increased. At DAF 20, the differences between the endosperm of different rice varieties were evident in all endosperm parts. Possibly, in the middle and late filling stages, the filling efficiency of the good-appearance soft rice varieties was still higher than cloudy soft rice, and all parts of the endosperm were further filled. This suggests that the endosperm structure responsible for the chalky differences in soft rice starts to form at least ten days after filling and that the differences increase dynamically.

Yang et al. reported that the formation of starch grain cavities may be due to gradual water loss in the transition to the yellow ripeness stage after the number and shape of starch grains are set and filled with endosperm cells in the milky ripeness stage (Yang, 2015). Hao et al. conducted a structural analysis of the dynamic development cross-section of starch grains. They found that the cavity area of the single starch grain in the endosperm was formed at about DAF 4. It generally increased in size from DAF 4 to 20. At DAF 25-30, the cavity area of the single starch grain was stabilized (Hao, 2021). In this study, the cavities were not observed at DAF 5. However, at DAF 10, cavities occurred in the starch grains which was observed later than that of Hao et al (Hao, 2021). This may be due to the difference in the test varieties or the cell wall hindering ang observation. The results at least indicate that starch grain cavities occur before yellow ripeness stage. During DAF 10-20, these cavities showed a dynamic increase, which is consistent with the findings of Hao et al (Hao, 2021). This suggests that the endosperm structure, which causes differences in soft rice transparency, also begins to form at least ten days after filling. In conclusion, in addition to the selection of soft rice varieties with good appearance, the structural differences affecting the appearance quality of soft rice have already begun to form in the early filling stage. Therefore, reasonable fertilizer and water management should be ensured from the filling early stage in order to achieve a good filling rate. Thus, the endosperm can develop closely and ultimately promote the formation of good-appearance quality of soft rice.