The field experiments were conducted in 2016 and 2017 in the subtropical environment of Danyang City, Jiangsu Province, China (32°0′ N, 119°70′ E, 51 m altitude). Weeds, pests, and diseases were intensively controlled to avoid yield loss. The climate data regarding daily radiation and air temperature were measured at a meteorological station located within 1 km of the experimental site. The daily solar radiation and temperature were measured by a silicon pyranometer (LI-200, LI–COR Inc., Lincoln, NE, USA) and a temperature/RH probe (HMP45C, Vaisala Inc., Helsinki, Finland), respectively. The meteorological data of the two rice growing seasons were shown in Table 1.

The experiments were randomized in a complete block design with three replications. Two conventional japonica varieties, namely, Nanjing 9108th (N9108) and Ningjing 7th (N7), with two grain filling temperature, namely, LT (implement by setting late sowing date, average temperature = 20.3°C) and CK (implement by setting early sowing date, average temperature = 24.7°C), were used. Two different nitrogen panicle fertilizer (NPF) levels were adopted in this study, namely, no NPF (N0), and 150 kg N ha⁻¹ at panicle initiation (N150), and were applied in panicle initiation. The whole experimental field was applied with the same amount of 450 kg P ha⁻¹ (Calcium superphosphate) + 150 kg K ha⁻¹ at transplanting, and 150 kg K ha–¹ at panicle initiation stage (Potassium chloride) + 330 kg N ha⁻¹ as carbamide at transplanting in 2016 and 2017. The soil properties of the topsoil layer (0–20 cm) before transplanting were measured in both years as follows: 1 kg soil contains 1.25 g total N, 6.8 mg NH4+, 0.9 mg NO3-, 27.9 mg Olsens-P and 168 mg NH₄OAc-K, and pH = 6.3. Seedlings were sowed on May 30 and July 1, and machine-transplanted on June 20 (CK) and July 20 (LT) in 2016 and 2017, respectively, with the hill spacing for 14 cm × 30 cm. Machine-transplantation was performed using a rice trans-planter (PZ640, Iseki Agricultural Machinery Co., Ltd., Japan).

Fifty milligrams of plant sample was weighed into a 2 ml plastic microtube and frozen in liquid nitrogen, dissolved in 1 ml methanol/water/formic acid solution (15:4:1, v/v/v). Ten microliters of internal standard mixed solution (100 ng/ml) was added into the extract as internal standards for the quantification. The mixture was vortexed for 10 min, and centrifuged for 5 min (12,000 rpm/min, and 4°C), and the supernatant was then transferred to clean plastic microtubes, followed by evaporation to dryness and dissolved in 100 μl 80% methanol (v/v), and filtered through a 0.22 μm membrane filter for further LC-MS/MS analysis. ABA and ACC contents were detected by MetWare (http://www.metware.cn/) based on the AB Sciex QTRAP 6500 LC-MS/MS platform (Wuhan Metware Biotechnology Co., Ltd., Wuhan, China). The quantification of ACC and ABA were analyzed using an UPLC-ESI-MS/MS system (UPLC'ExionLC™ AD' https://sciex.com.cn/; MS'Applied Biosystems 6500 Triple Quadrupole, https://sciex.com.cn/). The content of ACC and ABA was determined using the external standard method and is expressed as ng/g fresh weight (FW). Three biological replications were performed.

Analysis of variance was performed using SPSS version 20.0 (SPSS Statistics, SPSS Inc., Chicago, USA), and the results are expressed as means (±SD) of three biological replicates. The treatment means were compared based on the least significant difference (LSD) at a 0.05 level of probability.

The present study was carried out to determine the effect of different temperature and sowing dates on the yield and yield related attributes of rice during the rice-growing seasons 2017–18 as presented in Table 1. To accomplish the filling stage temperature treatments in the field, two different sowing date treatments were adopted. Lower Mean temperature (Tmean), Day temperature (Tday), and Night temperature (Tnight) and shorter growth duration were found in LT treatment compared to that in CK treatment (Tables 1, 2). The LT and NPF treatments showed a significant effect on total spikelets number, seed setting rate, and 1,000-seed weight that determine the yield of the rice. The total spikelet number could be enhanced by both NPF and LT treatments while seed setting rate and 1,000-seed weight reduced. Moreover, their balancing eventually altered the grain yield (Table 3). However, the grain yield responds differentially to NPF between CK and LT treatments. Under CK condition, the grain yields of N9108 and N7 could be significantly improved by NPF for about 12.7 and 9.4%, respectively, mainly due to enhanced population size (total spikelet number). However, under LT conditions, the grain yield of both varieties showed no significant change if NPF was applied or not since significantly lower seed-setting rate and 1,000-seed weight were observed, although their total spikelet number increased at the same time. The higher seed-setting rate (12.3 ± 2.4%) and 1,000-seed weight (7.4 ± 1.3%) of both varieties were found in CK treatment compared to those in LT treatment. Similarly, a lower seed-setting rate (9.1 ± 1.7%) of both varieties could be induced by applying NPF.

As shown in Table 4, the grain filling rate of N9108 is significantly higher than that of N7, and it is more significant under the LT condition (Figure 1). Under CK condition, the inferior and superior mean grain filling rate (GR_mean) of N9108 were 0.504 ± 0.073 and 0.811 ± 0.053 (mg·grain⁻¹ · D⁻¹), respectively, was about 25 and 43% higher than those of N7 [0.372 ± 0.079 and 0.617 ± 0.033 (mg·grain⁻¹ · D⁻¹)]. However, under LT condition, the GR_mean of inferior and superior grains in N9108 were about 0.416 ± 0.075 and 0.803 ± 0.03 (mg·grain⁻¹ · D⁻¹), which were significantly higher than those of N7 [0.324 ± 0.069 and 0.595 ± 0.011 (mg·grain⁻¹ · D⁻¹)], with an increment of about 40 and 67% compared with those of CK condition. In this study, both NPF and LT treatments reduced the grain filling rate of the two varieties, delayed the time to reach the maximum grain filling rate, and increased the initial grain filling potential (Table 4). When NPF was applied, the inferior grain filling rate of N9108 and N7 decreased by about 16 ± 2.7% and 27 ± 8.3%, respectively. The time to reach the maximum filling rate was delayed by about 13 ± 3.3% and 21 ± 7.1%, and the initial grain filling potential decreased by about 13 ± 5.4% and 15 ± 7.8%. In addition, the covariance (CV) of filling rates significantly differed by varieties and seed positions, indicating their differential stability. The CV of GR_mean, max grain filling rate (GR_max), and max grain weight (W_max) in inferior seeds of N7 were 0.33, 0.21, and 0.31, respectively. The CV of GR_mean, GR_max, and W_max in superior seeds of N7 were 0.12, 0.09, and 0.13, respectively. Differently, the CV of GR_mean, GR_max, and W_max in inferior seeds of N9108 were 0.22, 0.14, and 0.21, respectively. The CV of GR_mean, GR_max, and W_max in superior seeds of N9108 were 0.07, 0.04, and 0.04, respectively. It is obvious that the stability of superior and inferior grain filling levels of N9108 was significantly higher than those of N7. The stability of the superior grain filling level of both varieties was higher than that of inferior grain filling, which indicated that grain filling of inferior seeds could be more regulated by NPF and LT treatments.

The starch accumulation pattern was consistent with the grain filling rates results as shown in Figure 2. In Figure 2, the starch accumulation of both inferior and superior seeds in N9018 is significantly higher than that in N7. Compared to superior seeds, the starch accumulation and filling rates were significantly higher than that of inferior seeds (Figures 1, 2). Unlike inferior seeds, the starch accumulation of superior seeds was not significantly regulated by temperature and fertilizer treatments. Under CK conditions, the total amount of starch accumulation in inferior seeds was suppressed by NPF. The LT treatment showed a lower inferior seeds starch accumulation compared to that of CK, and the application of NPF had no significant effect under LT conditions. The suppression effect of NPF on starch accumulation in CK condition was higher than that in the LT condition, which was consistent with grain filling results (Figure 2). On the other hand, the effect of temperature and fertilizer treatment on the starch accumulation of N7 was significantly higher than those of N9108 (Figure 2).

To further provide metabolism level clues about changing the biosynthesis of starch, we measured the key intermediate metabolite, amylose, and amylopectin content. We found that, similar to the changing pattern of total starch accumulation, the amylose and amylopectin contents of both varieties were increased rapidly in all measured periods (Figure 5). The content of amylose and amylopectin was significantly reduced by LT and NPF treatments in both varieties. On the other hand, the dynamics of amylose and amylopectin were lower in N7 compared to that in N9108, which is consistent with gene expression data (Figure 3).

Since the grain filling difference of superior seeds in both varieties among different treatments was not significant, it could be concluded that the difference of grain weight was mainly derived from inferior seeds starch biosynthesis among all treatments. Thus, the relative expression levels of genes related to starch accumulation in inferior seeds were verified in this study (Figure 3). Due to starch, synthesis could be controlled by several key enzymes including starch synthase (SSS, encoded by FLO4), starch branching enzyme (SBE, encoded by SBE I), and starch phosphorylase (SPS, encoded by PHO-l) (Yoshida and Hara, 1977; Yang et al., 2001a; Zhao et al., 2004; Zhu et al., 2004; Duan and Sun, 2005). The expression pattern of the above genes was measured by qRT-PCR and was consistent with the physiological results (Figure 2). FLO4 gene expression was high and kept stable from 5 DAA (Figure 3). FLO4 expression, in both varieties under CK condition, was higher than those under LT condition, but there was no significant difference among different NPF application treatments. The expression of SBE1 in superior and inferior grains increased continuously, while the expression of PHO-l in superior and inferior grains increased continuously from 0 to 15 DAA and then decreased slightly since 15 DAA. The expression of SBE1 and PHO-l were significantly suppressed by LT and NPF treatments. Overall, the expression levels of N9108 were significantly higher than those in N7, which indicated a higher starch biosynthesis ability (Figure 3). The gene expression data was consistent with the enzymatic activities data in Figure 4. The encoded gene expression levels and activities of SBE, SPS, and SS were significantly suppressed by LT and NPF treatments, which also indicated suppressed starch biosynthesis ability.

In this study, we measured the content of ABA and ACC of inferior seeds among different treatments at three grain-filling stages. The ABA concentration was low at the early grain filling stage, increased from 10 DAA, reached a peak at 15 DAA, and declined at 20 DAA. In all treatments, the concentration of ABA was significantly decreased by LT and NPF treatments compared to CK and no NPF treatments, respectively (Figure 5). The reduction was more severe in additive LT and NPF treatment compared to other treatments. In contrast to ABA, the concentration of ACC was high at the early grain filling stage (Figure 6) but continuously declined until 20 DAA. Throughout the grain filling period, the concentration of ACC was largely enhanced by LT and NPF treatments compared to CK and no NPF treatments, respectively. Similar to ABA changing pattern, we found a large increment of ACC concentration in additive LT and NPF treatment compared to other treatments. The variance of ACC responding to NPF was more pronounced than that of LT. The ABA concentration of N9108 among all duration was significantly higher than that of N7, while ACC concentration showed no significant difference.

Nitrogen is one of the most important elements for plant growth, and the application of nitrogen fertilizers plays an important role in increasing rice yield due to their increment on total spikelet number (Fu et al., 2019; Wang et al., 2019). However, previous studies also proved that inappropriate nitrogen panicle fertilizer could reduce the grain filling of rice in the field (Ohnishi et al., 1999; Yang A. et al., 2004; Singh et al., 2011; Jiang et al., 2016; Zhang et al., 2021). The grain yield of both varieties could be enhanced by applying NPF since their larger population size. NPF increased grain yield for 11.7 ± 3.3% and 1.6 ± 0.9% under CK and LT conditions, respectively. This meant that the effect of NPF on grain yield was not significant under LT condition (Table 3). Under LT condition, the NPF induced enhancement of the population size and could exaggerate the filling issue similar to how a previous study (Fu et al., 2019) reported. This study explained that under abiotic stress in the filling stage, the grain yield of high nitrogen level was even lower than that of low nitrogen level due to insufficient grain filling. Unlike the responding pattern of the grain yield, both NPF and LT suppressed the grain filling while NPF enhanced grain yield as in previous studies since the increment on total spikelet number (Ding et al., 2003; Wang et al., 2015; Peng, 2016; Fu et al., 2019). Since 20 ± 4.9% higher total spikelet number and 9.7 ± 2.1% lower grain yield was found in N7 compared to those in N9108, it could be concluded that the major difference of the grain yield of the two varieties was mainly derived from the grain weight which was closely related to their filling level instead of their population size. The variance of grain weight of N9108 was significantly lower than that of N7 indicated their higher filling stability (Table 3). To further investigate the underlying mechanism of the different sensitivities of grain weight, the filling rate was examined in this study, and the results were consistent with the grain yield (Table 4).

Abiotic stress, like temperature, suppressed the grain filling rate by slowing enzymes activities, which further reduce the starch biosynthesis (Wang et al., 2021). In the present study, grain filling rate was suppressed by LT but, on the other hand, the late sowing of rice increases the spikelet numbers that are in line with the previous study, which in turn enhances the vegetative stage temperature and rice population size (Wang et al., 2021). As shown in the result, both NPF and LT suppressed the grain filling rate while they both enhanced the spikelet number (Table 3). It could be concluded that increasing grain filling could be a more effective method to increase grain yield in the field rather than increasing population size, since the rice population size has been already enlarged a lot due to the popularization of “super rice” varieties. The grain yield was determined by the balancing between grain filling reduction and spikelet number enhancement. Therefore, the variety selection of late sowing conditions in the middle and lower reaches of the Yangtze River could be guided by strong sink activity rather than sink capacity. The varieties with higher sink activities showed higher LT and NPF adaptivity than low sink activities varieties in this study. The reduction of grain filling level in response to NPF and LT was further proved by lower sugar-to-starch conversion and starch biosynthesis level. Low starch content was found to be produced under LT and NPF conditions, mainly due to lower expression levels of starch biosynthesis correlated with gene expression (FLO4, PHO-l, and SBE1). A previous study also showed that abiotic stresses affected the grain filling level by controlling starch biosynthesis (Wang et al., 2019). Compared to LT treatment, grain yield could be more significantly regulated by the NPF application in CK condition due to their insufficient filling rate (Table 3). The grain filling rate was reduced by individual LT treatment (15 ± 2.9%). Moreover, the grain filling rate was further reduced (23 ± 4.6%) by applying NPF simultaneously (Table 4 and Figures 2, 3). It could be concluded that higher NPF's yield increasing effect was found in CK condition compared to that in LT condition, which indicated that the appropriate NPF should be selected in LT condition.