Yield responses to elevated CO2 concentration among Japanese rice cultivars released since 1882

ABSTRACT Atmospheric CO2 concentrations ([CO2]) have increased by more than 100 μmol mol−1 over the last century and are projected to rise further. Breeding cultivars with a greater response to elevated [CO2] (E-[CO2]) can be an effective adaptation to global climate change. We wondered whether E-[CO2]-responsive cultivars have been unintentionally selected through empirical breeding as [CO2] has increased. If so, modern cultivars may respond better to E-[CO2] than old ones. We conducted free-air CO2 enrichment (FACE) experiments in 2 years to examine whether rice cultivars bred in different eras differ in response to E-[CO2] and to determine any associated traits. We tested five Japanese cultivars: Aikoku (released in 1882), Norin 8 (1934), Koshihikari (1956), Akihikari (1976) and Akidawara (2009). The yields of Aikoku and Norin 8 increased by 19.3% and 30.3%, respectively, under E-[CO2], while those of Koshihikari and Akihikari increased by 15.9% and 3.4%, respectively. However, that of Akidawara, the newest cultivar, also increased by 19.0%. Norin 8’s strong response to E-[CO2] was associated with increases in both spikelet density and percentage of ripened grains, both of which were closely related to nitrogen uptake. These results suggest that breeding has not necessarily improved cultivars’ response to E-[CO2], and that selection for traits such as sink capacity and nitrogen uptake can be effective to improve rice productivity under E-[CO2]. Graphical Abstract


Introduction
Atmospheric CO 2 concentrations ([CO 2 ]) have increased by more than 100 μmol mol −1 from approximately 280 μmol mol −1 in pre-industrial times mainly through anthropogenic CO 2 emissions and are projected to reach as high as 936 μmol mol −1 by the end of this century (IPCC, 2013). Global warming and change of precipitation patterns resulting from this rise will have mainly negative effects on crop production (Semenov & Porter, 1995;Zhao et al., 2017). However, increasing [CO 2 ] will offer opportunities to increase crop productivity because CO 2 is a substrate for photosynthesis and increases photosynthesis rates. Significant intraspecific variation in the yield response to elevated [CO 2 ] (E-[CO 2 ]) was found in major food crops (Ziska, Manalo & Ordonez, 1996;Shimono et al., 2009;Hasegawa et al., 2013 for rice; Seneweera et al., 2010 for wheat;Bishop et al., 2015 for soybean). Therefore, screening for or breeding E-[CO 2 ]-responsive cultivars offers an effective way to adapt crops to global climate change (Ziska et al., 2012).
Rice (Oryza sativa L.), one of the most important staple food crops, feeds about half of the world's population (Maclean et al., 2002). Genetic improvement was one of the major technological advances in the last century. In Japan, nationally organized rice breeding started in 1893. Rice improvement had initially been based on the separation of pure lines from landraces until cross-breeding started in 1904. The major targets of the rice breeding programs were high yield and resistance to pests, diseases and cold injury. Continuing efforts in breeding, along with intensification of culture, resulted in rapid increases in yield (Chen, 2018;Horie et al., 2005) until the 1970s, when Japan achieved rice self-sufficiency. Since then, the main breeding objectives have shifted to high-eating quality (Horie et al., 2005). In the last few decades, breeders have returned their attention to productivity to reduce costs and have developed high-yielding cultivars by introducing indica genes (Yoshinaga et al., 2018).
Because the increase in [CO 2 ] over the last century has been steady, E-[CO 2 ]-responsive cultivars may have been selected unintentionally (Ziska et al., 2012). If so, modern cultivars may respond better to E- [CO 2 ]. Some crops have been evaluated for this response in glasshouses or open-top chambers. Old wheat cultivars showed greater yield responses to E-[CO 2 ] than modern cultivars, mainly owing to their inherent ability to form tillers and panicles (Manderscheid & Weigel, 1997;Ziska, 2008;Ziska, Morris & Goins, 2004). Oats showed similar results (Ziska & Blumenthal, 2007). But old and recent cultivars of barley showed no significant difference (Schmid et al., 2016). Thus, modern cultivars do not necessarily show a better response to E- [CO 2 ].
We previously reported that the yield response of rice to E-[CO 2 ] varied widely among cultivars including some old and new cultivars in a free-air CO 2 enrichment (FACE) experiment under open-field conditions in a single season (Hasegawa et al., 2013). In that study, we showed the intraspecific variation in the yield responses to E-[CO 2 ] from a single-season, but detailed analyses of the associated yield traits particularly with regard to the time cultivars developed remain undone. In this study, we repeated the experiment to compare Japanese cultivars released since 1882 with additional measurements of key growth traits. The objectives were to examine the effects of breeding era on yield sensitivity to [CO 2 ] and associated yield traits.

Study sites
FACE experiments were conducted in farmers' fields in Tsukubamirai City, Ibaraki Prefecture, Japan (35°58ʹN, 139°60ʹE, 10 m a.s.l.), during the 2010 and 2011 growing seasons. The soil is a Fluvisol, which is typical of alluvial areas with 23% clay and 40% silt contents. The total carbon and nitrogen concentrations were 21.4 and 1.97 mg g −1 , respectively.

Cultural practices
Germinated seeds were sown in seedling trays on 26 April 2010 and 25 April 2011. After emergence, seedlings were raised in puddled open fields, with a tunnel cloche for the first 2 weeks. They were then transplanted by hand at a spacing of 30 cm × 15 cm (22.2 hills m −2 ), with three seedlings per hill, on 26 May 2010 and 25 May 2011.
Equal amounts of phosphorus (P) and potassium (K) were given to all the plots in early April, before plowing; compound P-K fertilizer was applied at a rate of 4.36 g P m −2 and 8.30 g K m −2 . Fields were kept submerged after late April. Three kinds of nitrogen (N) fertilizer were applied just before puddling: 2 g N m −2 as urea, 4 g N m −2 as another coated controlled-release fertilizer (LP100, JCAM Agri. Co. Ltd., Tokyo, Japan), and 2 g N m −2 as another type of controlled-release fertilizer (LP140). The two types of coated urea differ in the rate of release: LP100 releases 80% of its total N over 100 days at 25°C and LP140 does so over 140 days.
The field was kept flooded until late August, when the surface water was drained for harvesting. We applied flush irrigations on several occasions to keep the soil moist. Each cultivar was harvested at maturity in mid-to-late September.

Growth and yield measurements
In each plot, we monitored panicle emergence of at least four hills of each cultivar every other day and defined the heading date as the date when 50% of the productive tillers reached panicle-tip emergence.
At heading and maturity, four hills of each cultivar (six hills of Koshihikari) per plot were sampled. We also dug up a block of soil 30 cm wide, 15 cm long and 15 cm deep around one of those hills (two of Koshihikari) and carefully washed the soil from the roots with running water. We separated plants into living and dead leaf blades, stems (including leaf sheaths), roots and panicles, and measured the living leaf area with a leaf area meter (AAM-8; Hayashi denko, Tokyo, Japan). The plant parts were then ovendried at 80°C for 72 h and weighed. The samples were ground, and the N concentrations were determined by NC analyzer (Sumigraph NC-22F, Sumika Chemical Analysis Service, Tokyo, Japan). Total N contents were calculated from dry weight and N concentration of each organ. Here, we regarded the total N contents as the amount of N uptake till then.
At physiological maturity, eight hills per cultivar (21 of Koshihikari) were sampled to measure grain yield and yield components. After the plants were air-dried under a rain shelter, we counted panicles. After threshing, we measured the total weight of the spikelets. Each spikelet sample was then split into three subsamples. One subsample (half of the spikelets) was dehulled to determine the brown rice weight. We measured the moisture content of the grains with a grain moisture tester (Riceter f, Kett Electric Laboratory, Tokyo, Japan). Brown rice yield and single-grain mass were expressed on a 15% moisture content basis. Another subsample (a quarter of the spikelets) was used to determine the proportion of ripened spikelets by sorting in an ammonium sulfate solution with specific gravity of 1.06. The third subsample was stored for future chemical analyses.

Gas exchange measurements
To measure leaf gas exchange at heading, we measured light-saturated net photosynthesis rate (A sat ) of the uppermost fully expanded leaf of one plant of each cultivar per replicate with a portable photosynthesis system (LI-6400, Li-Cor, Lincoln, NE, USA). During measurements, the photosynthetic photon flux density was maintained at 1800 μmol m −2 s −1 , the leaf temperature at 32°C, [CO 2 ] at the growth [CO 2 ] (380 μmol mol −1 for the A-[CO 2 ] plots, 580 μmol mol −1 for the E-[CO 2 ] plots) and relative humidity at >60%.

Statistical analyses
We conducted an analysis of variance (ANOVA), using a split-split-plot design, with year as the whole factor, [CO 2 ] as the split-plot factor and cultivar as the splitsplit-plot factor; we used the general linear model procedure of SPSS 17.0 for Windows (SPSS Japan Inc.; now IBM, Tokyo, Japan). We tested the significance of the response ratio (E-[CO 2 ]: A-[CO 2 ]) by ANOVA after natural-logarithm (ln) transformation to avoid heterogeneity in the error variance (Hedges, Gurevitch & Curtis, 1999). We examined the relative importance of yield components by multiple regression analysis based on the response ratio after ln transformation because the relationships between yield and its components were multiplicative. For simplicity, we used the three components for the analysis: spikelet density (panicle number × spikelets per panicle), percentage of ripened grain, and single grain mass.

Meteorological conditions
Meteorological conditions in the 2 years were shown in Table 1. The mean air temperature over the growing season was 25.0°C in 2010 and 24.1°C in 2011, higher than that of the long-term  average of~23°C , and was nearly the same among the cultivars. During 30 days before heading (DBH) and after heading (DAH), which correspond to panicle formation and grain-filling periods, respectively, the mean air temperature differed among the cultivars by up to 1.1°C and 0.9°C, respectively. The mean solar radiation showed the similar trend as the air temperature, with a lager difference among the cultivars.
Growth and development E-[CO 2 ] shortened days to heading (DTH) by 1 day, when averaged across the years and cultivars (P < 0.001, Table 2).  Since the mean temperature during the growing season was higher in 2010 than in 2011, the average DTH across the cultivars was 4 days shorter in 2010 than in 2011 (P < 0.001). DTH differed significantly among the cultivars (P < 0.001), ranging from an average across the years, [CO 2 ] of 57 days for Akihikari to 82 days for Norin 8. DTH of the earliest maturing cultivar, Akihikari, did not differ between years, resulting in a significant year × cultivar interaction (P < 0.001).
Neither plant height nor leaf area index (LAI) at heading was affected by E-[CO 2 ], when averaged across the years and cultivars (Table 2). Maximum tiller number was marginally greater in E-[CO 2 ] (P > 0.05). Plant height and LAI averaged across the cultivars were greater in 2011 than in 2010 (P < 0.05). Maximum tiller number and LAI differed significantly among cultivars (P < 0.05). Averaged across the years, they were highest in Norin 8 and lowest in Akidawara and Akihikari. Plant height showed a significant year × cultivar and CO 2 × cultivar interaction (P < 0.05).
Total dry weight across the years and cultivars was increased by E-[CO 2 ] by 12.8% at heading and 15.9% at maturity (P < 0.001, Table 3). Total dry weight of each cultivar was generally lower in 2010 than in 2011 at both stages owing to shortened growth duration at the higher temperature, though that of Akihikari was higher in 2010 at heading, resulting in a significant cultivar × year interaction (P < 0.01). It also differed significantly among cultivars at both stages (P < 0.001). Norin 8 had the largest total dry weight, followed by Akidawara and Koshihikari, and Akihikari had the smallest in both years. Although there was no significant cultivar × CO 2 interaction, Aikoku had the highest response to E-[CO 2 ] at maturity in both years. E-[CO 2 ] increased shoot (living and dead leaf blade, leaf sheath and stem) and root dry weights at the maturity (P < 0.01, Table 3). Both differed significantly among cultivars at both stages (P < 0.001). Means across years were highest in Norin 8 at both stages and lowest in Koshihikari at heading and in Akihikari at maturity. The responses of shoot and root dry weights at maturity to E-[CO 2 ] differed significantly among cultivars (cultivar × CO 2 : P < 0.01). The oldest cultivar, Aikoku, had the highest response of shoot dry weight over years, as in total dry weight. However, Norin 8 had the highest response of root dry weight, as well as the largest root mass among cultivars.
Total N uptake was stimulated by E-[CO 2 ] by 4.6% at heading (P > 0.05) and 9.2% at maturity over years and cultivars (P < 0.01, Table 3). Year had a significant effect: total N uptake over cultivars was larger in 2011 than in 2010, by 13.6% at heading and by 8.7% at maturity (P < 0.01). Total N uptake over years differed significantly among cultivars at heading (P < 0.001) but not at maturity. There was no significant cultivar × CO 2 interaction at either stage, though two old cultivars (Aikoku, Norin 8) tended to respond better to E-[CO 2 ] at maturity than the others.
E-[CO 2 ] significantly increased all of the yield components except single-grain weight across the years and cultivars (P < 0.05, Table 4). The overall effect of E-[CO 2 ] was most apparent in spikelet density (10.6%). All yield components differed significantly among cultivars across the years and [CO 2 ] (P < 0.001). E-[CO 2 ] effects differed between yield components: In panicle number, Koshihikari had the largest response over years (12.9%), resulting in a significant cultivar × CO 2 interaction (P < 0.05). In spikelets per panicle, Akidawara had Table 3. Effects of elevated [CO 2 ] on total dry weight, shoot and root dry weight, and total N uptake at heading and maturity of five cultivars tested under FACE conditions in 2 years.   the largest response (8.4%), but there was no significant cultivar × CO 2 interaction. It is worth noting that the three older cultivars (Aikoku, Norin 8, Koshihikari) responded to E-[CO 2 ] more strongly in panicle number than in spikelets per panicle, while the newest one (Akidawara) responded conversely. As a result, cultivar × CO 2 interaction was not significant in spikelet density (panicle number × spikelets per panicle; Figure 3(a)). The effect of E-[CO 2 ] on the percentage of ripened grain was highest in Norin 8, followed by Akidawara and Aikoku (Figure 3(b)). Harvest index was increased by E-[CO 2 ] over the years and cultivars (P < 0.001, Table 3). Harvest index over years and [CO 2 ] was highest in Akihikari and lowest in Norin 8 (P < 0.001). E-[CO 2 ] increased that of Norin 8 by 9.3% over years. On the other hand, Koshihikari and Akihikari showed negligible effects, resulting in a significant cultivar × CO 2 interaction (P < 0.01). E-[CO 2 ], overall, increased brown rice yield mainly through spikelet density, followed by percentage of ripened grain, but not through single-grain weight (Table 5).

Relationships between yield components and N uptake
There was a linear and positive relationship between root mass and N uptake at both heading and maturity (Figure 4). Spikelet density was linearly and positively correlated with total N uptake at heading irrespective of year, CO 2 treatment or cultivar when Norin 8 was excluded ( Figure  5). Norin 8 had a lower spikelet density than the other cultivars at the same N uptake. Overall, the percentage of ripened grain was positively correlated with N uptake from heading to maturity ( Figure 6).

Discussion
We grew five rice cultivars released since 1882 under ambient and elevated [CO 2 ] to examine the effects of past breeding on rice [CO 2 ] sensitivity and associated traits.
In contrast, E-[CO 2 ] sensitivity was not responsible. Old cultivars (Aikoku, Norin 8) responded better to E-[CO 2 ] than modern ones (Koshihikari, Akihikari), except the most recent, Akidawara (Figure 2, Table 4). Similar results were found in wheat (Manderscheid & Weigel, 1997;Ziska, 2008;Ziska et al., 2004) and oats (Ziska & Blumenthal, 2007), in which higher responsivity of older cultivars to E-[CO 2 ] was associated with increased tiller and thus panicle numbers. Here, older cultivars also tended to respond better in panicle number (Table 4). On the other hand, newer cultivars responded relatively well in spikelets per panicle (Table 4), canceling out any significant differences between old and new cultivars (Figure 3(a)). In addition to responses in panicle number, old cultivars also responded better in percentage of ripened grain (Figure 3(b)). The percentage of ripened grain was the second major contributor to E-[CO 2 ] sensitivity (Table 5). These results suggest that simultaneous responses of panicle number (spikelet density) and percentage of ripened grain (Table 4) led to the higher yield responses of older cultivars to E-[CO 2 ].
ability under E-[CO 2 ] is discussed in the next section.) As a result, larger sink formation ability and higher source function of Akidawara might be the main factors in its E-[CO 2 ] sensitivity. In addition to Akidawara, recent high-yielding indica or indica-derived cultivars showed higher E-[CO 2 ] sensitivity, for example, 'Takanari' (Hasegawa et al., 2013). Since continuous inbreeding has reduced genetic diversity (Yamamoto et al., 2010;Yonemaru et al., 2012), the introduction of indica cultivars or indica traits may improve E-[CO 2 ] sensitivity.  Table 4 and are 2-year averages with SD (n = 4). Bars with different letters are significantly different at 5% level by Tukey's multiple comparison test. P value indicates the statistical significance between old (Aikoku, Norin 8) and modern (Koshihikari, Akihikari) cultivars.  Table 4 and are 2-year averages with SD (n = 4). Bars with different italic letters are significantly different at 10% level by Tukey's multiple comparison test.    In this study, we applied N fertilizers at the current standard level. Since N application level was low in the era when old cultivars had been bred, old cultivars were expected to show higher yields than modern cultivars did. However, previous studies reported that modern and new high-yielding cultivars had higher yields than old ones did even under low or no N application conditions (Hasegawa, 2003;Taylaran et al., 2009;Zhang & Kokubun, 2004). As for [CO 2 ] responsivity, high-yielding cultivar Takanari showed larger yield responses than modern cultivar Koshihikari did under no N application condition, resulting in a significant cultivar × CO 2 × N interaction (Hasegawa et al., 2019). However, it remains still unclear that old cultivars show a CO 2 × N interaction. Therefore, the order in the [CO 2 ] responsivity among the five cultivars tested in this study may change depending on N application levels.
A 2-year field trial with limited cultivars cannot, of course, clarify all of the past breeding improvements associated with [CO 2 ] sensitivity of rice. However, our results suggest that past breeding had not necessarily been selecting E-[CO 2 ]-responsive cultivars.
Traits associated with [CO 2 ] sensitivity E-[CO 2 ] increased the yield of the old Japanese cultivar Norin 8 by 30.3% on average under open-field conditions. This response is comparable to those of Chinese hybrid and indica inbred cultivars: increases of 34.1% in the hybrid cultivar 'Shanyou 63ʹ and 30.1% in 'Liangyoupeijiu' (Liu et al., 2008;Yang et al., 2009), and of 32.9% in the indica inbred cultivar 'Yangdao 6 Hao' by FACE (Zhu et al., 2015). Ours is the first report of a yield increase of >30% in a japonica inbred cultivar in a multi-year FACE experiment.
Among the yield components, spikelet density contributed most strongly to the yield increase by E-[CO 2 ] ( Table 5). Similar results were reported in previous FACE studies using different rice cultivars under different N regimes and environments (Hasegawa et al., 2013;Kim et al., 2003;Liu et al., 2008;Shimono et al., 2009;Yang et al., 2006Yang et al., , 2009Zhu et al., 2015). Therefore, the responsivity of spikelet density is the most important factor in [CO 2 ] sensitivity in rice. It is well known that spikelet density is closely and positively correlated with N uptake by rice before heading, irrespective of [CO 2 ] (Hasegawa et al., 2019;Kim et al., 2001;Shimono et al., 2009). We found a similar relationship ( Figure 5). These results suggest that the responsivity of spikelet density under E-[CO 2 ] was closely related to the increased N acquisition before heading. Interestingly, the [CO 2 ]responsive Norin 8 had a much lower spikelet production efficiency than the other cultivars ( Figure 5), somewhat mitigated by its large N uptake by heading stage (Table 3).
In addition to spikelet density, the percentage of ripened spikelets contributed to yield responses to E-[CO 2 ] ( Table 5). As discussed above, all cultivars with high [CO 2 ] sensitivity in this study, especially Norin 8, responded to E-[CO 2 ] in both spikelet density and percentage of ripened grain (Figure 3, Table 4). On the other hand, Koshihikari had the highest response in spikelet density with a lower response in percentage of ripened grain, resulting in a moderate increase in grain yield (Figure 3, Table 4). These results are consistent with those of highly responsive Chinese hybrid and inbred cultivars -Shanyou 63 (Liu et al., 2008), Liangyoupeijiu (Yang et al., 2009) and Yangdao 6 Hao (Zhu et al., 2015) in which yield increases of >30% were not only attributed to stronger sink generation but also enhanced grain filling capacity. It may be possible to achieve a >30% yield increase by increasing spikelet density by >30% at the same grain filling percentage, although previous studies did not (Liu et al., 2008;Yang et al., 2009;Zhu et al., 2015). Therefore, for large yield responsivity (say, >30%) to E-[CO 2 ], enhanced grain filling ability can be effective in addition to increased spikelet production by E-[CO 2 ].
What traits are associated with the responsivity in percentage of ripened spikelets to E-[CO 2 ]? We found a close relationship across the cultivars between percentage of ripened spikelets and N uptake during grain filling ( Figure 6). All cultivars with a high response in percentage of ripened spikelets increased N uptake under E-[CO 2 ] during grain filling (Table 3). Increased N uptake during grain filling will maintain leaf N concentration and hence source ability by avoiding N translocation from leaf to grain and leaf senescence. Although we did not measure A sat during grain filling, Chen et al. (2014) reported that the [CO 2 ]-sensitive Takanari could retain a higher leaf N concentration until late grain filling, resulting in a high grain filling ability. Thus, we suggest that enhanced N uptake during grain filling could allow a response in percentage of ripened spikelets under E-[CO 2 ] mainly through enhanced or maintained source ability.
Meteorological conditions during panicle formation and grain filling periods differed between the cultivars because of the maturity differences (Table 1, 2). Since the clear relationship was found between spikelet density and N uptake before heading in this and previous studies under different meteorological regimes (Hasegawa et al., 2019;Kim et al., 2001;Shimono et al., 2009), the differences of mean air temperature and solar radiation seemed not to have strong effects on the extent of [CO 2 ] responsivity of the cultivars. In contrast, grain filling has been known to be affected by air temperature and solar radiation (Yoshida, 1981). In this study, the differences of both air temperature and solar radiation between the cultivars during the grain filling periods were relatively small in 2010 (up to 0.3°C, 1.1 MJ m −2 day −1 , Table 1). In 2011, however, the mean air temperature and solar radiation ranged 25.2°C to 26.1°C and 15.5 MJ m −2 day −1 to 17.8 MJ m −2 day −1 , respectively, among the cultivars. Though the mean air temperatures were slightly higher than the optimum range for grain filling (20-25°C: Matsushima & Manaka, 1957;Morita, 2000a;Yoshida & Hara, 1977), the negative effect of higher temperature on grain weight was counteracted by solar radiation (Morita, 2000b), mainly because of the positive relationships between both under normal weather conditions. Therefore, the differences of meteorological conditions in this study might have limited effects on the differences of [CO 2 ] responsivity between the five cultivars.
Under A-[CO 2 ], percentage of ripened spikelets was relatively lower in the old and newest cultivars than the modern ones (Table 4). The similar result was found in the previous studies (Zhang & Kokubun, 2004). These results may suggest that the old cultivars are source-limited and there is a room to be improved by enhanced N uptake and source ability by E-[CO 2 ]. The cultivars with a high percentage of ripened spikelets will hardly respond to E-[CO 2 ] without increase in spikelet density. In other words, such cultivars may be already well adapted to current [CO 2 ].
Our results show that both spikelet density and percentage of ripened spikelets were associated with high [CO 2 ] sensitivity and were closely related to the response in N uptake. We found a positive relationship between root mass and N uptake at heading and maturity (Figure 4), as did Kim et al. (2001). As detailed study of the response of root mass and its functions to E-[CO 2 ] are limited, especially under open-field conditions , further studies to reveal their genotypic variation in E-[CO 2 ] response are needed for the improvement of E-[CO 2 ] sensitivity.