Rice (Oryza sativa L.) is a staple grain crop. Six hundred million tons of rice are produced annually, and rice is widely cultivated throughout the world (IRRI, 2002). Low temperature (LT) is a major environmental factor causing reductions in yield (Boyer, 1982). In general, the mean critical temperature for rice is +4.7°C, and once the ambient temperature (AT) is below 5–10°C, irreversible LT-induced damage to seedling growth and development can occur (Sanchez et al., 2014). During different developmental stages, rice shows growth responses to LT stress. The reproductive stage in rice is a critical period. Farrell et al. (2001, 2006) reported that rice was susceptible to abundant LT-induced damage during the young microspore stage, and 12 h of exposure to LT can cause spikelet sterility. LT-induced spikelet sterility during the microsporogenesis stage is considered genotype dependent, which for cold-tolerant genotypes was a night-time air temperature of 15°C for 4 consecutive days and for cold-sensitive genotypes was a night-time air temperature between 17 and 19°C (Satake, 1969). Hence, developing and verifying cold-tolerant cultivars will be beneficial for overcoming the influence of LT stress on rice growth and development.

Low temperature-induced spikelet sterility, in particular during the reproductive period, could be related to the physiological response mechanisms in the anther, pollen, and stigma. Previous studies have indicated that short anther dehiscence, poor pollen grains, and low pollen germination on stigmas were all highly correlated with spikelet sterility under LT-induced stress (Matsui et al., 2001; Prasad et al., 2006). Moreover, Satake (1976) considered that the main reason for this could be attributed to LT-induced damage of male organs, such as pollen, not of female organs, such as the stigma. Hence, if pollen grains were deficient at anthesis, they could impede pollination, leading to spikelet sterility. These results might be similar to high temperature stress (HTs) in rice. A larger anther and longer stigma (Suzuki, 1981, 1982) were found to contribute to the increase in the basal pore length and found to improve LT tolerance during the reproductive stage, and larger anthers and longer stigmas also ultimately benefit successful pollination (Matsui and Kagata, 2003). In addition, panicle exsertion (PE) (Rang et al., 2011) and peduncle length (PeL) (He et al., 2009) have been reported to be responsible for enhanced spikelet sterility under water stress, as has HTs (Rang et al., 2011), but similar results have not been documented under LT stress.

In most rice genotypes, flowering and anthesis undergo a 5-day period with most spikelets at anthesis between 1000 and 1200 h; the similar patterns of flowering have been reported in indica and japonica subspecies (Nishiyama and Blanco, 1980; Prasad et al., 2006). Studies have shown that flowering (containing anthesis and fertilization), as a developmental stage in rice, in particular at the booting stage (microsporogenesis), is considered most susceptible to LT (Farrell et al., 2006; Maite and Ellis, 2015). The influences on spikelet flowering in rice have been verified under HTs (Rang et al., 2011). The climatic effects on the timing of day and the duration of flowering in rice also differ (Julia and Dingkuhn, 2012). Jagadish et al. (2007) analyzed the flowering patterns and the time of day of spikelets reaching anthesis under HTs in indica and japonica rice. Furthermore, spikelet fertility under stress conditions could be generally measured by marking individual spikelets during anthesis, which was adopted as an effective method. Indica–japonica hybrid cultivars (Yongyou series) have been successfully bred and have been widely used for rice production due to their exceptionally high grain yield potential. However, they are very susceptible to LT, and can succumb to spikelet sterility. Previous studies have not systematically investigated the effects of LT on spikelet flowering and spikelet fertility in indica–japonica hybrid cultivars (Satake, 1976; Nishiyama, 1982; Farrell et al., 2006; Julia and Dingkuhn, 2012; Maite and Ellis, 2015; Zeng et al., 2015).

The overall objectives in the present study were therefore to evaluate spikelet fertility of indica–japonica hybrid rice under LT stress and study the main traits of anthesis and pollen related to spikelet sterility. The experiments were conducted under AT (25°C) and LT (17°C) and involved the following: (1) recording the effect of LT on peduncle length (PeL) and PE in rice; (2) studying the influences on spikelet fertility, anther dehiscence, pollen grains, and pollen germination on stigmas under LT during anthesis; and (3) examining the relationship between spikelet flowering and spikelet fertility across cultivars.

With changing global climate, extreme cold conditions will be more frequent in the future, which make rice subject to adverse abiotic stresses. Improving the temperature stress tolerance in rice at anthesis during the susceptible reproductive stage is beneficial for adapting to highly dynamic climatic conditions in the future (Horie et al., 1996). In addition, increasing the absolute stress tolerance of rice could facilitate important physiological processes (e.g., anther dehiscence, pollen germination, pollination, and fertilization) being carried out with high spikelet fertility under stress (Jagadish et al., 2008, 2010; Rang et al., 2011). Rang et al. (2011) considered that spikelet fertility might be largely determined by morphological conditions at anthesis under different abiotic stresses. PE, which was primarily affected by PeL, could be described as affecting rice spikelet fertility when exposed to LT. However, in the present study, the panicles were partially exserted under LT; some spikelets within the panicles were still stuck in the leaf sheath of the four cultivars regardless of the degree of cold tolerance and EPE (Table 2). Interestingly, we found that PeL decreased significantly (P < 0.05) under LT in all cultivars apart from Yongyou9. These results were similar to the previous findings under HTs (Rang et al., 2011).

Although spikelets stuck in the leaf sheath could be reflective of spikelet fertility through altered anthesis and fertilization (O’Toole and Namuco, 1983; Cruz and O’Toole, 1984), the physiological process of spikelet fertility under LT did not invariably coincide with phenology. In general, anther dehiscence can affect the number of pollen grains on the stigma (Rang et al., 2011). In the present study, anther indehiscence under LT was higher in cold-sensitive cultivars than in cold-sensitive cultivars (Figure 2). It was mainly due to a sharply reduced FAD, resulting in decreased pollen grains on the stigma. Liu et al. (2006) reported that the rupturing of the septa, the expansion of locule walls, the swelling of pollen grains, and then the rupturing of the stomium were all highly related to anther dehiscence. However, in the present study, this phenomenon was not observed (Table 6). The reason was that anthers still dehisced under stress due to spikelet flowering; pollen grains could not swell, and they lost viability and did not shed, resulting in unfertilized pollen (Farrell et al., 2006; Liu et al., 2006; Jagadish et al., 2010). There was no relationship between greater anther dehiscence and more pollen grains on the stigma, as was described by Jagadish et al. (2010). However, spikelet fertility under cool temperature conditions was strongly correlated with large anthers (Suzuki, 1981), and anther length, width and area benefit increasingly large pollen grains with respect to fertilization (Farrell et al., 2006). Our findings also showed that anther volume, which was significantly affected in cold-sensitive cultivars, was more closely related to anther width (r = 0.96, P < 0.01; n = 17, data not shown) than anther length (r = 0.69, P < 0.01) under LT. In addition, spikelets scored for high fertility had at least 10 or more than 20 germinated pollen grains, which could be identified as the threshold number (Jagadish et al., 2010). It was therefore thought that cold-tolerant cultivars obtained more germinated pollen on the stigma (Figure 4A), and increased seed set compared with that of cold-sensitive cultivars under LT was ultimately confirmed (Table 4). It has been reported that pollen germination alters spikelet fertility, ultimately affecting rice productivity under high night-time temperatures (Mohammed and Tarpley, 2009) as well as under LT. Liu et al. (2006) reported that improved pollen stickiness, which could restrict pollen grains from shedding whether the anther pores dehisced or not, might affect pollen count on the stigma. Our results suggest that a small pollen count and germinated pollen grains under LT were observed in cold-sensitive cultivars (Figures 4, 5), mainly due to poor pollen fertility and shortened pollen diameters (Table 5). Moreover, there was strong positive relationship between spikelet fertility and pollen germination (Table 6). However, the mechanism controlling high pollen germination with low anther dehiscence in spikelets under LT stress is not clear.

Although a uniform level of PE, PeL, and WPL between treatments was observed among the cultivars, rice plants exposed to LT did have lower spikelet fertility across cold-sensitive cultivars (Table 4). This was caused by different spikelet positions within the panicle (P < 0.05); the upper parts of the panicle under LT stress and LT alone could especially aggravate poor spikelet fertility. However, under stress conditions, the grains at the upper parts of the panicle are usually filled primarily (Stansel, 1975). The upper parts are also exposed to LT first, which can enhance the extent of cold-induced damage and lead to spikelet sterility, even when the panicles are exserted. However, the lower and middle parts of the panicle were correspondingly subjected to less damage due to the site of PE. Similar results on the grain characteristics of rice under high night-time temperatures and spikelet position can be obtained from the report of Mohammed and Tarpley (2010). Furthermore, the effect of spikelet position within the panicle on spikelet fertility across cultivars was also related to the exsertion of the panicle.

The cumulative negative influence of extended LT stress on consecutive flowering days was examined in terms of spikelet fertility (seed set) using the individual spikelets marked. In the study, we observed a significant reduction in spikelet fertility with the duration of flowering in cold-sensitive cultivars but not in cold-tolerant cultivars (Figure 3). This result was in accordance with that in the studies of Jagadish et al. (2007) and Rang et al. (2011) on the effects of HTs. Flowering under AT peaked within a 5-day period earlier in indica–japonica hybrid cultivars than in japonica rice (Figure 6). The former cultivars contain the indica gene besides the japonica gene (Prasad et al., 2006), resulting in early flowering, even though the pattern of flowering, which reached peaked between 1030 and 1230 h, on the third day was similar among cultivars (Figure 7). Compared with cold-sensitive cultivars, cold-tolerant cultivars still had fewer numbers of flowering spikelets under LT (Figures 6, 7), allowing floral organs to avoid damage from LT. Jagadish et al. (2007) proposed that more spikelet anthesis is correlated with a higher spikelet fertility under HTs. However, this phenomenon was not verified in our study, because spikelet flowering did not significantly affect spikelet fertility (Table 6). Hence, our results stated that spikelet fertility was not the consequence of spikelet flowering. It had been understood that the normal action of dehisced pollen sacs in anthers and pollen vitality might be affected by LT stress as well as by HTs a day ahead flowering (Matsui and Omasa, 2002). It is possible that in cold-susceptible cultivars such as Yongyou9 and Yongyou17, other processes that occur before fertilization were affected, such as pollen shrinkage and anther volume reduction (Farrell et al., 2006) as well as effects on pollen tube growth rate and pollen germination (Kakani et al., 2002; Coast et al., 2015). This could confirm that high pollen germination was observed in Yongyou538 and Zhejing88. Nonetheless, it is unlikely that ovary development was unaffected after fertilization, although spikelets subjected to LT after opening were fertile in both cold-sensitive cultivars and cold-tolerant cultivars. This phenomenon could also result in spikelet sterility.

In general, indica species are more tolerant to HTs; conversely, japonica species are more tolerant to LT stress (Farrell et al., 2006). Moreover, indica–japonica hybrid rice, which simultaneously contains indica and japonica genes, was sensitive to temperature fluctuations, resulting in spikelet sterility under LT, despite the neutral alleles at the two loci being used to solve pollen sterility in indica–japonica hybrid rice (Lu et al., 2000). In this study, indica–japonica hybrid rice cultivars, which were tested for the gene frequencies of indica and japonica using SSR molecular markers (data not shown), indeed had different responses to LT. Our study may provide evidence of indica–japonica hybrid rice adapting to extreme climatic changes for breeding purposes.

Our study has shown that LT induces spikelet sterility by shrinking anther size and decreasing pollen function (the number of pollen grains, pollen fertility, and pollen germination on stigmas) in cold-sensitive cultivars during anthesis. We found that spikelet fertility was strongly correlated with pollen germination rather than spikelet flowering across cultivars and treatments. Spikelet fertility was not attributed to the patterns of flowering and the number of spikelets reaching anthesis. Therefore, this aspect should be considered for breeding indica–japonica hybrid rice varieties to tolerance during flowering.

DZ conceived and designed the experiments. YPZ, JX, NU, and XP provided experimental opinions and assistance. YPZ and YHZ performed the experiments. YHZ analyzed the data and wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.