Effects of free-air CO2 enrichment on heat-induced sterility and pollination in rice

ABSTRACT Global climate changes may cause heat-induced sterility in rice, threatening the global production of this important crop. Although little is currently known about the combined effects of the concentration of atmospheric CO2 ([CO2]) and temperature on heat-induced sterility, elevated [CO2] (E-[CO2]) will likely increase the panicle temperature and thereby exacerbate heat-induced sterility, but this was not tested in open fields. Therefore, we investigated the effect of E-[CO2] on heat-induced sterility and sterility-related traits in rice by increasing E-[CO2] by approximately 200 μmol mol−1 above ambient levels using a free-air CO2 enrichment (FACE) facility for six growing seasons with variable growing season temperatures. The percentage fertility was not significantly correlated with the air temperature (Ta) between 09:00 and 12:00 on each flowering day, but it did significantly vary among the years, with 2011 experiencing cool temperatures resulting in chilling-induced mild sterility. When data from 2011 were removed, there was a significant negative correlation between fertility and Ta between 09:00 and 12:00 on each flowering day under E-[CO2], whereas no such effect was seen under ambient [CO2]. E-[CO2] also significantly reduced the number of pollen grains deposited on the stigma by 10%, but it slightly increased the anther length by 1.3%, indicating that it had both negative and positive effects on heat-induced sterility. These findings suggest that E-[CO2] affects many traits related to heat-induced sterility and may sometimes exacerbate sterility by reducing pollen grain deposition. Abbreviations: A-[CO2]: ambient CO2; [CO2]: the concentration of carbon dioxide; E-[CO2]: elevated [CO2]; FACE: free-air CO2 enrichment; Ta: air temperature


Introduction
According to the 5th Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2013), the atmospheric concentrations of greenhouse gases have increased since 1750 as a result of human activities, with the concentration of carbon dioxide ([CO 2 ]) in 2011 (391 µmol mol −1 ) exceeding preindustrial levels by approximately 40%. This has led to a combined global average increase in land and ocean surface temperatures of 0.85°C (IPCC, 2013), which threatens the production of many crops, including rice, an important staple food in Asia and Africa.
Heat-induced sterility occurs when the air temperature (T a ) at anthesis is above 35°C in growth chambers (Jagadish et al., 2007;Satake & Yoshida, 1978;Weerakoon et al., 2008), temperature gradient chambers (Kim et al., 1996;Maruyama et al., 2013), open-top chambers (Chaturvedi et al., 2017), and the field (Ishimaru et al., 2016;Matsui et al., 2014;Tian et al., 2010). This is mainly caused by low pollination due to anther indehiscence (Kobayasi et al., 2011;Satake & Yoshida, 1978). Not only anther indehiscence but also the length of anther dehiscence at the basal part of the theca plays a critical role in stable pollination and fertilization under high temperature conditions (Matsui et al., 2005;Shah et al., 2011). For example, in a growth chamber experiment, Kobayasi et al. (2011) demonstrated that the relative contribution of anther indehiscence and the basal dehiscence length to pollination (percentage of sufficiently pollinated florets) changed according to T a , with the percentage of dehisced anthers becoming more important with increasing T a .
However, the combined effects of [CO 2 ] and heat stress on sterility, pollination, and anther dehiscence under openair field conditions remain unclear. Although some studies have shown that heat-induced sterility is exacerbated by elevated [CO 2 ] (E-[CO 2 ]) (Chaturvedi et al., 2017;Kim et al., 1996;Matsui et al., 1997), no such effect has been observed by others (Madan et al., 2012), and all of these experiments were conducted under enclosed conditions. E-[CO 2 ] increases the panicle temperature by causing the leaf stomata to close and by reducing leaf stomatal conductance Shimono et al., 2010). A model simulation showed that the panicle temperature increased by 0.5-1.0°C C on a flowering day as a result of E-[CO 2 ] alone, which is sufficient to exacerbate heat-induced sterility . Therefore, an assessment of the effects of E-[CO 2 ] on sterility, pollination, and anther dehiscence is important for accurately predicting the effect of heat stress on rice production and for evaluating the performance of heat-tolerant cultivars under heat stress conditions. The objective of this study was to reveal the effects of E-[CO 2 ] on floret sterility and sterility-related traits, such as pollination and anther dehiscence, in rice. To do this, we used the Tsukuba free-air CO 2 enrichment (FACE) facility (Hasegawa et al., 2013) because it represents the most reliable and feasible experimental platform for evaluating the effect of [CO 2 ] on the physiological and ecological traits of crops under open-air field conditions (Hasegawa et al., 2016;Kobayashi, 2001). We conducted the FACE experiments for six growing seasons with different temperatures, some of which exceeded the critical temperature for spikelet sterility.

Experimental site, plant materials, and culture methods
This study was conducted at the Tsukuba FACE facility in Ibaraki Prefecture, Japan (35°58′N, 139°60′E; 10 m above sea level) from 2011 to 2016 using the standard japonica rice cultivar 'Koshihikari'. This area has a humid subtropical climate, with an average annual temperature of 13.8°C and annual precipitation of 1280 mm. Details of weather conditions and soil properties at the site are described by Hasegawa et al. (2013).
The crop cultivation practices that were used at Tsukuba FACE during the 2011-2016 growing seasons were essentially identical to those reported by Hasegawa et al. (2013). Briefly, all plots received equal amounts of P and K in early April, before ploughing, when PK compound fertilizer was applied at a rate of 4.36 g m -2 of P and 8.30 g m -2 of K. The fields were kept submerged after late April. Three kinds of N fertilizer were applied just prior to puddling: 2 g m -2 as urea, 4 g m -2 as one type of controlled-release fertilizercoated urea (type LP100, JCAM Agri. Co. Ltd, Tokyo, Japan), and 2 g m -2 as another type of controlled-release fertilizer (type LP140, JCAM Agri. Co. Ltd) in . In 2014, 2015, controlled-release fertilizer (type LP140) was changed to controlled-release fertilizer (type LP40).
Rice seeds were germinated in water and sown into seedling trays, each with 448 circular cells (16 mm in diameter and 25 mm in depth, Minoru Pot 448, Minoru Industrial Co. Ltd, Okayama, Japan). Each cell was filled with sterilized soil amended with fertilizer at the rate of 0.4 g N, 0.35 g P, and 0.5 g K per 1 kg of soil. Three seeds were sown in each cell. After seedling emergence, the seedling trays were transferred to the puddled open field and protected with a tunnel cloche or floating mulch for the first two weeks. The seedlings were then manually transplanted into hills (30 × 15 cm) on May 25, May 23, May 22, May 22, May 26 and May 25 in 2011, 2013, 2014, 2015. Planting density was 22.2 hills m -2 (three seedlings per hill). After the transplanting, the fields were kept flooded until August 20, about 3 weeks after heading, when the ponding water was drained for harvesting. We applied flush irrigations on several occasions to keep the soil moist.

CO 2 treatment
The method for controlling [CO 2 ] in an open field is described by Nakamura et al. (2012). Briefly, four blocks were established in paddy fields, with each block consisting of two octagonal plots (204 m 2 , 17 m across): an ambient CO 2 (A-[CO 2 ]) plot and an elevated CO 2 (E-[CO 2 ]) treatment plot. The E-[CO 2 ] plots were equipped with emission tubes on the perimeter, where CO 2 was released from the windward side to keep the [CO 2 ] measured at the central point at about 200 μmol mol −1 above the ambient control level.

Measurements of fertility, pollination, and anther dehiscence
Once plants reached the heading stage, we selected five panicles in each plot and marked the lemmas of all opened florets with an acrylic gouache paint 3-4 h after flower opening. We changed the paint color each day to distinguish the day of flowering and marked the florets on more than three flowering days per year, excluding rainy days. We then collected the painted panicles at maturity and examined their fertility by manually inspecting each floret for ovary development.
To examine pollination, we sampled more than five florets from the primary rachis-branches of different panicles in each plot at around 14:00 on more than three flowering days per year. The stigmata were extracted from the florets and stained with 0.1% cotton blue solution (0.1 g of cotton blue, 75 mL of 2% acetic acid, and 25 mL of glycerol), and the number of pollen grains deposited on the stigma was counted under an optical microscope (OPTIPHOTO; Nikon, Tokyo, Japan).
To examine the percentage of dehisced anthers and the length and width of the anther dehiscence at the basal part of the theca, we sampled more than five florets (different florets sampled for measuring pollination as mentioned before) from the primary rachisbranches of different panicles in each plot at around 14:00 on more than three flowering days per year (Matsui et al., 2005). The anthers were air-dried, and we counted the number of indehiscent anthers and measured anther length (both dehiscent and indehiscent anthers) under a digital USB microscope at 100 × magnification (Anyty 2.4; 3R System C., Fukuoka, Japan). The length and width of the basal anther dehiscence were measured under the same digital USB microscope (Matsui et al., 2005).

Measurement of weather conditions
We measured T a , relative humidity, wind speed, and solar radiation in the center at the Tsukuba FACE field. A temperature and humidity sensor (HMP45D; Vaisala Inc., Helsinki, Finland) was installed 2 m above the soil surface with a forced-ventilated radiation shield, and an anemometer and a wind vane (Model 03001-5 Wind Sentry; R. M. Young Co., Traverse City, MI, USA) were placed at a height of 2.5 m. Solar radiation was measured with a pyranometer (LI-200SA; LI-COR Inc., Lincoln, NE, USA). Each variable was measured every 10 s, and 10 min averages were recorded in a data logger (CR1000; Campbell Scientific Inc., Logan, UT, USA). T a , relative humidity, wind speed, and solar radiation are common in four replicates.

Statistical analysis
We conducted an analysis of variance by applying a randomized block design, in which year, [CO 2 ] treatment, and block are the factors, and year × [CO 2 ] is the interaction factor with four replications, using R (version 3.2.5, R Development Core Team, 2016). We analyzed our data by means of correlation procedures using R. In all tests, statistical significance is indicated for p < 0.001, p < 0.01, or p < 0.05, and the actual p-values are presented where 0.05 < p < 0.1.

Weather conditions during the flowering periods
The daily mean T a during the flowering periods of 2011-2016 ranged between 24.8 and 29.2°C (Table 1), being the highest in 2015. The daily maximum T a sometimes reached 35°C in 2014, 2015 and 2016, which is the temperature at which heat-induced floret sterility occurs. By contrast, in 2011, the average daily mean T a was <25°C and the daily minimum T a did not exceed 25°C during the flowering period.
The daily mean wind speed was low during all six flowering periods, remaining below 3 m s −1 ( Table 2). The average daily mean relative humidity varied from 73.2% in 2014 to 85.8% in 2013, whereas the daily mean solar radiation ranged from 10.0 MJ m −2 in 2011 to 23.6 MJ m −2 in 2014.

Fertility and pollination
The percentage fertility significantly differed among the years (p < 0.001), but it was not significantly affected by [CO 2 ] or the interaction between the treatment and the year (Table 3). Fertility was below 90% in 2011 and higher than 90% in all other years in the control and treatment plots.
The number of pollen grains deposited on the stigma also significantly differed among the years (p < 0.001) and was significantly affected by [CO 2 ] (p < 0.05), with no significant interaction between the year and [CO 2 ]. The number of deposited pollen grains was the highest in 2015 and the lowest in 2011, when T a was the lowest, and decreased by approximately 10% under E-[CO 2 ].

Anther dehiscence and length
Anther indehiscence was rarely observed in this experiment, so data on the percentage anther dehiscence are Values are ranges, with means in parentheses. not shown. Both the length and the width of the anther dehiscence at the basal part of the theca significantly differed among the years (p < 0.001) but were not significantly affected by [CO 2 ] or the interaction term (Table 4). The basal dehiscence length was the greatest in 2012 and was also high in 2011, when T a was the lowest, and was smaller in 2014, 2015, and 2016, when the daily mean T a was relatively high. The width of the anther dehiscence zone showed similar trends to the length. The anther length significantly differed among the years (p < 0.001) and with [CO 2 ] (p < 0.01), and the interaction between the year and [CO 2 ] was also significant (p < 0.001). Anthers were the longest in 2011, when T a was the lowest, and were particularly short in 2016. E-[CO 2 ] averaged across the six years slightly increased the anther length by approximately 1.3% although E-[CO 2 ] slightly decreased the anther length in 2011 and 2012.

Relationship between weather factors and traits related to fertility and pollination
There was no significant correlation between the percentage fertility and T a between 09:00 and 12:00 on each flowering day. However, there was a significant difference in fertility among the years, with relatively high sterility occurring in 2011 despite the moderate T a during the flowering period. This may have been due to the extraordinarily low minimum T a that was experienced approximately 10 days before heading (<17.5°C), at which stage cool temperatures may cause mild sterility (Nishiyama, 1995). When the 2011 data were removed, there was a significant negative correlation between fertility and T a between 09:00 and 12:00 under E-[CO 2 ] but not under A-[CO 2 ] (Figure 1).
No significant correlation was observed between percentage fertility and relative humidity between 9:00 and 12:00 on each flowering day (data not shown). Solar radiation between 9:00 and 12:00 was not also significantly correlated with the percentage fertility (data not shown).
The number of pollen grains deposited on the stigma was not correlated with T a between 09:00 and 12:00 on each flowering day under E-[CO 2 ] or A-[CO 2 ] (Figure 2). However, the number of deposited pollen grains was significantly positively correlated with wind speed between 09:00 and 12:00 in both the treatment and the control plots (Figure 3), with a wind speed of >1.5 m s −1 resulting in more than 100 pollen grains being deposited.
The basal dehiscence length was not significantly correlated with T a between 09:00 and 12:00 on each flowering day (Figure 4), but it did significantly vary among years.

Discussion
We found that an increased T a between 09:00 and 12:00 on each flowering day slightly decreased the percentage fertility of rice plants under E-[CO 2 ] but had no effect under A-[CO 2 ] (Figure 1). Previous experiments conducted under enclosed conditions have similarly shown that heatinduced sterility is often exacerbated by E-[CO 2 ] (Chaturvedi et al., 2017;Kim et al., 1996;Matsui et al., 1997), which increases the panicle temperature by inducing leaf stomatal closure and reducing leaf stomatal      conductance. Although we did not measure the panicle temperature here, if the 0.5-1.0°C increase estimated by Yoshimoto et al. (2005) under E-[CO 2 ] is valid, the panicle temperature will sometimes exceed 35°C, where heatinduced sterility occurs. E-[CO 2 ] had varied effects on traits related to heatinduced sterility and pollination, decreasing the number of pollen grains deposited on the stigma (Table 3) but increasing the anther length (Table 4) and having no effect on anther dehiscence in the basal part of the theca. Matsui et al. (1997) similarly observed a reduction in pollen grain deposition under E- [CO 2 ] conditions in an open-top chamber experiment. In terms of the other environmental conditions measured, the number of deposited pollen grains was not significantly affected by T a between 09:00 and 12:00 on each flowering day under E-[CO 2 ] or A-[CO 2 ] (Figure 2), but it was significantly positively correlated with wind speed (Figure 3), with a wind speed of >1.5 m s −1 resulting in more than 100 pollen grains being deposited. Similarly, Tian et al. (2010) previously observed significant losses in seed set due to sterility under a high relative humidity and a low wind speed of <1 m s −1 , although the maximum T a was approximately 35°C. In our experiment, the correlation coefficient between daily mean T a and wind speed during the flowering periods was 0.372 (p < 0.001), indicating that high temperatures usually accompany higher wind speeds. Therefore, it is likely that only mild sterility was observed at E-[CO 2 ] in our experiment because the negative effect of increased panicle temperature was offset by the stronger wind under increased T a , highlighting the importance of considering wind speed when evaluating heat-induced sterility.
The effects of E-[CO 2 ] on anther dehiscence and length were not clear in our study. Although E-[CO 2 ] significantly increased the anther length (Table 4), it did not improve pollination (Table 3). Cool-weathertolerant rice varieties have longer and wider anthers (Tanno et al., 1999), and Matsui and Omasa (2002) argued that longer anthers can hold more pollen grains, making it an effective trait for tolerating heatinduced sterility. Zhao et al. (2010) also reported that the percentage of sufficiently pollinated florets was correlated with the basal dehiscence length and the percentage seed set. However, we only used one cultivar ('Koshihikari') in our study, so variations in the anther length and dehiscence in response to changes in the environment may have been small. Since the anther length and dehiscence are highly heritable traits (Tanno et al., 1999;Zhao et al., 2016), genetic improvement would improve the tolerance of this cultivar to heat-induced sterility.
Fertility was less than 90% in 2011, likely due to the minimum T a being below 17.5°C approximately 10 days before heading, when cool temperatures may cause slight sterility (Nishiyama, 1995), and E-[CO 2 ] further increased sterility in this year (Table 3). Okada et al., (2005) similarly reported that E-[CO 2 ] increased sterility under low temperatures indirectly by increasing the spikelet number. However, we did not observe any difference in the number of spikelets per panicle between A-[CO 2 ] and E-[CO 2 ] in 2011, and we found a little difference in the number of deposited pollen grains (Table 3). Therefore, a possible explanation for the mild sterility that was observed in this year is that the cool temperature affected pollen maturation during the booting stage, resulting in pollen sterility, unstable pollen germination and pollen tube elongation after pollination.
In our study, E-[CO 2 ] had only a small effect on sterility. In addition to the effect of wind speed mentioned above, this was likely due to only a slight increase in panicle temperature occurring because of decreased transpiration during the flower opening periods, when the sensitivity to heatinduced sterility is highest (Satake & Yoshida, 1978). During the vegetative stage, leaf stomatal conductance under E-[CO 2 ] in Shizukuishi FACE was low at 09:00 but became higher around noon (Shimono et al., 2010). The flower opening time in our study was around 10:00-11:00, so there may have been a little difference in T a between E-[CO 2 ] and A-[CO 2 ] just before this time.
Although we only detected small changes in sterility and traits related to heat-induced sterility and pollination under E-[CO 2 ], this treatment had both negative and positive effects on heat-induced sterility, suggesting that the evaluation and prediction of E-[CO 2 ] effects require data obtained from open-field experiments as well as under controlled conditions. However, now that Tsukuba FACE project has finished, it will be difficult to conduct additional experiments under E- [CO 2 ] in the open field. Therefore, the development of a model to estimate heat-induced sterility under E-[CO 2 ] using various datasets, including the data obtained here for six growing seasons with variable growing season temperatures, and application of the micrometeorological model that has been developed for simulating panicle temperature (Yoshimoto et al., 2011) will be useful for predicting the effects of E-[CO 2 ] on the sterility and pollination of rice.