The effects of HT during panicle initiation on yield formation
Most previous studies investigated damage induced by HT during the flowering stage [2, 13, 14, 28–30], and there have been few investigations of the effects of HT during panicle initiation on rice yield formation [2, 3]. Therefore, in this study, we assessed the detrimental effects of HT during panicle initiation in rice.
We found that HT treatment during panicle initiation seriously decreased yield (84%) and yield components, with the exception of the number of panicles per plant, in LYPJ plants; however, HT had a smaller effect on yield formation in SY63 plants in comparison with that observed in LYPJ plants (Table 1). In the HT-susceptible genotype LYPJ, the seed setting percentage showed the largest decline among yield components under HT treatment, indicating that the seed setting percentage was more vulnerable to HT in comparison with the other components, which was in agreement with previous reports [3, 31]. In the present study, the heat-tolerant variety SY63 had less yield damage in response to HT, which may be attributed to the high, stable seed setting percentage of this genotype under HT treatment (reduced by 20% in SY63 and 69% in LYPJ). It is important to note that the LYPJ plants had a relatively low seed setting percentage under the CK conditions in this study (Table 1), which may have been due to the natural high temperature (maximum temperature of more than 35℃ for five consecutive days) during early flowering (data not shown). It has been reported previously that the LYPJ genotype is susceptible to heat during flowering .
In the present study, LYPJ had very low spikelet fertility under HT (Fig. 2J), and we also found very few half-grains with filling initiation (0.5% in LYPJ and 1.4% in SY63). These findings showed that it was spikelet fertility, rather than the grain filling process, that was responsible for the reductions in seed setting percentage under HT during panicle initiation. These results are similar to those reported in a recent study by Cheabu et al. , in which spikelet fertility was the major restriction for observed reductions of grain yield under HT from booting to maturity.
The intrinsic factors responsible for low spikelet fertility under HT include low pollen productivity, low pollen viability, poor anther dehiscence and pollen reception, and poor pollen germination [1, 2, 26]. In this study, HT during panicle initiation significantly decreased the anther dehiscence rate (Fig. 2K). This finding is in accordance with reports by Jagadish et al.  and Kobayashi et al. , in which rice plants were exposed to HT during the flowering stage. However, Endo et al.  and Wu et al.  reported that the anther dehiscence rate was not substantially influenced by heat stress during panicle initiation. Moreover, HT treatment was performed for 15 days with a mean daytime maximum temperature of 36.1℃ by Wu et al. , whereas HT treatment was performed for 3 days with a mean daytime maximum temperature of 39℃ by Endo et al. . Therefore, these data suggest that the response of the anther dehiscence rate to HT depends on the developmental stage during which plants are exposed to HT and the intensity of the HT treatment. Significant decreases in pollen viability (reduced by 46%), the pollen shedding percentage of the anthers (reduced by 11%) and the anther dehiscence rate (reduced by 5%) were observed in LYPJ plants under HT treatment (Fig. 2C, G, I, K and L), indicating that pollen viability was more vulnerable to HT in comparison with the other components (Fig. 2I, K and L). Therefore, we considered that it was altered pollen viability, rather than changes in pollen shedding or anther dehiscence, that was mainly responsible for the lower spikelet fertility of HT-susceptible cultivar LYPJ under HT treatment during panicle initiation (Fig. 2I, J and K). However, subjecting SY63 plants to HT treatment had no significant effect on any of these traits (Fig. 2). In comparison with LYPJ, the process from anther dehiscence to complete dispersal of pollen grains from the anthers was relatively rapid in SY63 under both tested temperatures (9.0 min under CK and 16.2 min under HT for LYPJ, 2.2 min and 2.5 min for SY63, data not shown). Recently, Wu et al.  found that enclosed stigmas of SY63 plants contributed to high spikelet fertility under HT treatment. Therefore, higher spikelet fertility (67%) of HT-tolerant SY63 plants under HT treatment during panicle initiation may be attributed to higher pollen viability (85%), a higher anther dehiscence rate (98%), a shorter period of time required for complete dispersal of pollen grains from anthers, better pollen shedding from anthers (94%), and enclosed stigmas (Fig. 7).
The relationship of pollen viability with anther characteristics
Pollen sterility caused by heat stress has been associated with abnormal anther development in sorghum , wheat , tomato , cotton , dwarf bean , and rice [2, 33]. In our study, HT treatment disrupted the morphologic structures of the anther wall and spherical microspores in LYPJ (Figs. 3, 4 and 5). Specifically, HT treatment resulted in malformation of the pollen structure (obscure outline of the pollen exine, collapsed bacula, disordered tectum, and no nexine) in LYPJ (Fig. 5G and O) at stage 10. Additionally, we observed aborted pollens at stage 13 in LYPJ under HT treatment (Fig. 4O), manifesting as a shriveled and collapsed pollen surface with a hollow germinal aperture and uneven sporopollenin deposition. Previous studies found that the pollen surface severely shriveled under HT treatment in sorghum and maize, and this change was accompanied by poor pollen viability [19, 21]. In the present study, heat stress did not obviously alter anther development or anther structure in the heat-tolerant variety SY63 (Figs. 3, 4 and 5). These data indicate that abnormal pollen formation was responsible for low pollen viability under HT treatment during panicle initiation, and the heat tolerance of SY63 may be attributed to normal anther development (Fig. 7).
Additionally, we found that Ubisch bodies had blunt protrusions and an uneven distribution on the inner surface of the anther wall in LYPJ plants under HT treatment (Figs. 3N and 5O). HT also resulted in tight wrinkles of the knitted anther cuticle on the epidermis of LYPJ plants (Fig. 3V). Similarly, Uzair et al.  found marginal differences in the patterning of nano-ridges on the outer surface, as well as in the distribution of Ubisch bodies, on the inner surface of anthers in rice ptc2 mutants, which resulted in decreased pollen viability in comparison with that of wild-type plants. However, in SY63 plants, HT had no substantial effect on Ubisch bodies or anther cuticles (Fig. 3). Ubisch bodies carry a sporophytically produced structural protein that is essential for pollen development . The cuticle on the outer surface of the anther serves as a barrier and protects the microspore/pollen grain from various environmental stresses . These data suggest that well-developed Ubisch bodies and cuticles contribute to the heat tolerance of SY63 under HT during panicle initiation, whereas alterations in Ubisch bodies and cuticle formation may result in pollen sterility in LYPJ (Fig. 7).
In this study, the single-cell tapetum area of LYPJ was larger at stage 10 under HT treatment in comparison with that observed under CK conditions. In LYPJ plants exposed to HT treatment, the tapetum was still observed at stage 13 due to slow degradation, but the tapetum had completely disappeared at this stage in LYPJ plants under CK conditions and in SY63 plants under HT and CK conditions (Fig. 4). These results show that HT disrupted tapetum degradation in LYPJ. Similarly, halted and incomplete tapetum degradation was reported in rice ptc2 mutants  and rice plants under chilling , and as well in tomato and cotton plants under heat stress [25, 38]. Regarding causes for abnormal tapetum degradation, Mamun et al.  revealed that vacuolation and hypertrophy of the tapetum under chilling was caused by osmotic imbalance, which was triggered by the reabsorption of callose breakdown products in the absence of OsMST8 activity. Min et al.  found that delayed programmed cell death of the tapetum was mainly due to inactivation of starch synthase in cotton under HT treatment. These findings suggest that different regulatory mechanisms govern tapetum degradation in different organisms; however, the mechanism underlying tapetum degeneration retardation in rice under HT is not yet clear.
The tapetum, the innermost layer in the anther wall, serves as an active nutrient source for neighboring microspores [34, 39], and abnormal tapetum degeneration results in pollen sterility in photoperiod and thermosensitive genic male-sterile rice [40, 41] and rice mutants . These previous reports indicate that termination of secretory-type tapetum development and disruption of tapetal functions is partly responsible for pollen viability. In the present study, tapetum degradation in LYPJ was inhibited by HT treatment, whereas SY63 showed nearly tapetum degradation under the same conditions (Fig. 5R and T). Additionally, we observed several differences in the characteristics of the anther walls of LYPJ and SY63 plants following HT treatment. For example, the tapetum did not adhere to the endodermis (Fig. 4G) at stage 9 in LYPJ, at which point tapetum degradation was initiated. In addition, LYPJ plants had a tightly knitted anther cuticle on the epidermis (Fig. 3R and V). These heat-induced changes in anther development and tapetum degradation may partly explain the high pollen sterility rate of LYPJ plants. In contrast, the well-developed anthers of SY63 plants enhanced their heat tolerance in terms of pollen viability (Fig. 7).
The relationship among anther structures, anther dehiscence and pollen shedding
It has been reported that pollen reception (pollen numbers on stigma) influences spikelet fertility under HT treatment [1, 13, 29]. Wu et al.  also found that HT treatment at the heading stage led to poor pollen shedding in heat-susceptible cultivars. In this study, we observed that HT treatment during panicle initiation had a negative effect on anther dehiscence and the pollen shedding percentage of the anthers in LYPJ plants; however, SY63 plants showed stable anther dehiscence and pollen shedding under HT and CK conditions (Fig. 2K and L). It has been reported that poor pollen shedding may be a disadvantage for successful reproduction under HT treatment .
The existence of the tapetum at flowering may halt anther dehiscence . Our microscopic observations demonstrated that the tapetum did not degenerate until stage 13 (anther maturity stage) in LYPJ plants under HT treatment (Table 2, Figs. 4O, 5C, K and 6C), and the anther wall (i.e., more remaining cell layers) remained between the locule and the lacuna. Similarly, Matsui et al.  reported that locules were kept closed by parenchyma and endothecium cells at anthesis due to the remaining anther wall cell layers in rice, which subsequently led to poor pollen shedding under HT treatment. However, Bagha  reported that the cell layers of the anther wall in rice were not affected by HT treatment during panicle initiation, while failure in lysis of the septum cell wall inhibited anther locule opening. In our study, lysis failure of the septum cell wall was also observed in LYPJ plants under HT treatment at anthesis (Fig. 4O), and the septum cell wall degraded in LYPJ plants under CK conditions (Fig. 4M), as well as in heat-tolerant SY63 plants under both CK and HT conditions (Fig. 4N and P). Therefore, our study suggests that both anther wall degradation and septum cell wall lysis together regulate anther dehiscence under HT conditions during panicle initiation. HT treatment during panicle initiation inhibited anther wall degradation and septum cell wall lysis in heat-susceptible LYPJ plants and subsequently resulted in low anther dehiscence in comparison with that of heat-tolerant genotype SY63 (Fig. 7).
Anther locules are opened via the stomium splitting. In our study, HT treatment did not affect the number of stomium cells in LYPJ or SY63 plants, and the two varieties had similar numbers of stomium cells under CK and HT (Table 2 and Fig. 6). Similarly, Bagha  did not find differences in stomium cell abundance between various genotypes grown under CK and HT conditions. Therefore, these results suggest that inhibition of anther dehiscence and pollen shedding at stage 13 under HT treatment may be attributed to septum cell lysis rather than stomium splitting.
In this study, we also found that the number of septum cells in SY63 plants was lower than that of LYPJ plants under both CK and HT conditions, and HT treatment resulted in a slight increase in the abundance of septum cells in both varieties (Table 2). This result was consistent with the results of Bagha , who reported that the number of septum cells was not affected by HT in heat-susceptible or heat-tolerant varieties; moreover, the heat-tolerant variety had fewer septum cells. Therefore, a lower number of septum cells may be a favorable characteristic for anther dehiscence in heat tolerant varieties, and this factor may have contributed to the high heat tolerance of SY63 (Fig. 7). Our study did not investigate the cause for the cessation of septum splitting in LYPJ plants under HT treatment. Inhibition of septum cell wall lysis may be attributed to low cell wall invertase activity caused by HT treatment . The underlying physiological mechanism for regulation of septum splitting under HT merits further investigation.
Our study found that HT inhibited pollen shedding of the anthers (Fig. 2G and L). There are two likely reasons for our observation of inhibited shedding under HT conditions. First, Ubisch bodies on the inner surface of the anther wall show non-wettability due to the distribution of hydrophobic substances, and a continuous hydrophobic layer appears on the inner surface due to locule shrinkage after dehiscence . Therefore, the occurrence of Ubisch bodies decreased the sticking properties of the pollen to the locule wall , and decreased sticking favors pollen dispersal . HT induced uneven distribution of Ubisch bodies in heat-susceptible LYPJ plants (Fig. 3N), which may be one of the reasons for the reduction in pollen shedding of the anthers (Figs. 2L and 7). Second, Pacini et al.  indicated that anther stomata may play a crucial role in anther dehydration in rice, which may favor anther dehiscence and pollen shedding. Recently, in ICE1 (ice1-2) Arabidopsis mutants, Wei et al.  found that decreased stomata density and abnormal stomata development were not advantageous for anther dehydration and anther water movement, so the epidermis did not shrink to dehisce, regardless of the existence of stomium cells in the anther; these changes resulted in low pollen dehiscence, low pollen viability and a low germination percentage. Therefore, to elucidate the mechanism underlying poor pollen shedding, more investigations, such as studies of anther stomata development and the physiological changes associated with anther dehydration in dehisced anthers, should be performed under HT conditions.