Genetic Improvements in Rice Yield and Concomitant Increases in Radiation- and Nitrogen-Use Efficiency in Middle Reaches of Yangtze River

The yield potential of rice (Oryza sativa L.) has experienced two significant growth periods that coincide with the introduction of semi-dwarfism and the utilization of heterosis. In present study, we determined the annual increase in the grain yield of rice varieties grown from 1936 to 2005 in Middle Reaches of Yangtze River and examined the contributions of RUE (radiation-use efficiency, the conversion efficiency of pre-anthesis intercepted global radiation to biomass) and NUE (nitrogen-use efficiency, the ratio of grain yield to aboveground N accumulation) to these improvements. An examination of the 70-year period showed that the annual gains of 61.9 and 75.3 kg ha−1 in 2013 and 2014, respectively, corresponded to an annual increase of 1.18 and 1.16% in grain yields, respectively. The improvements in grain yield resulted from increases in the harvest index and biomass, and the sink size (spikelets per panicle) was significantly enlarged because of breeding for larger panicles. Improvements were observed in RUE and NUE through advancements in breeding. Moreover, both RUE and NUE were significantly correlated with the grain yield. Thus, our study suggests that genetic improvements in rice grain yield are associated with increased RUE and NUE.

Genetic improvements in the grain yield and yield components. Fourteen varieties that were released and widely cultivated from 1936 to 2005 in Middle Reaches of Yangtze River were assessed in this study ( Table 1). The grain yield of newer varieties has been significantly increased compared with that of the older varieties. In 2013 and 2014, the grain yield increased to 61.9 and 75.3 kg ha −1 year −1 , respectively, which  (Table 2). A quadratic relationship between the grain yield and the year of release was observed in 2013, which demonstrated a decreasing trend in the genetic improvements in the grain yield over the last two decades ( Table 2; Fig. 2a). In 2014, the grain yield of varieties that had been released before 1990, with the exception of HHZ, the grain yield was higher. This trend resulted in a linear correlation between the grain yield and year of release in 2014 (Table 2; Fig. 2a). The genetic improvements in the grain yield since the 1930s have resulted from both an extended growth season and a higher daily grain yield (Fig. S1).
To dissect the causes of these improvements in grain yield, the yield components were measured ( Table 2; Fig. 2). Biomass production was significantly increased according to the year of release after the 1980s, and these changes were driven by the utilization of heterosis (Fig. 2). A quadratic relationship between the harvest index and the year of release was observed and had an R 2 value of 0.77 in 2013 and 0.52 in 2014. The harvest index of SLX, a tall variety, was 28.6 and 39.4% in 2013 and 2014, respectively, and these values were significantly lower than that of the semi-dwarf varieties (Table 3; Fig. 2c). Among the yield components, a significant increase in spikelets m −2 and grain filling percentage accounted for the genetic improvement in grain yield (Table 2; Fig. 2d,g). Newer varieties tended to present significantly fewer but larger panicles than the older varieties (Table 2 and Fig. 2e,f), and significant changes were not observed in grain weight during the breeding process (Table 2; Fig. 2h).
Genetic improvements in pre-anthesis radiation use efficiency. The total incident radiation from transplanting to heading was increased along with the year of release because of the extended growing season (Fig. S2a). The total incident radiation from transplanting to heading was higher in 2013 than in 2014 because of different weather conditions between the two years (Table 3; Fig. 1). Lodging, which could potentially reduce radiation interception, occurred during grain filling period for some varieties, including SLX, GCA, GC2, SY63, TQ and YLY1 in 2013 and SLX, AZZ, ZZA, SY63, YLY6 and YLY1 in 2014. However, bamboos and ropes were immediately used to help lodging plants stand up so as to minimize the reduction in radiation interception and biomass production. Therefore, light interception was measured from transplanting to heading, and significant changes were not observed in the pre-anthesis radiation interception efficiency, although the leaf area index (LAI) significantly increased along with the advances in breeding (Table 3; Fig. S2b and S3a). Significant correlations (p < 0.1) were observed between the grain yield and intercepted radiation from transplanting to heading in 2013 and 2014 (Fig. S4c).
The RUE for pre-anthesis intercepted global radiation (pre-anthesis RUE) was designated as the amount of biomass produced using intercepted global radiation from transplanting to heading, and the values ranged from 1.05 to 1.55 g MJ −1 in 2013 and from 1.35 to 1.85 g MJ −1 in 2014. Significant correlations (p < 0.1) between the pre-anthesis RUE and the year of release were observed in both 2013 (R 2 = 0.22) and 2014 (R 2 = 0.32). The genetic improvements in the pre-anthesis RUE were 0.0032 and 0.0047 g MJ −1 year −1 on an absolute basis, and 0.45 and 0.40% per year on a relative basis for 2013 and 2014, respectively. The AZZ and SLX varieties released in the 1930s and 1940s had the lowest pre-anthesis RUE, whereas the YLY1 and HHZ varieties released in the 2000s had the highest values (Fig. 3). Significant correlations were observed between the grain yield and pre-anthesis RUE at p < 0.1 in both 2013 and 2014 (Fig. S4d).
Genetic improvement in nitrogen uptake and use efficiency. The total N uptake significantly increased (p < 0.01) along with the year of release, and the values ranged from 138 to 208 kg ha −1 in 2013 and from 150 to 248 kg ha −1 in 2014. In both years, the SLX variety released in the 1930s accumulated the lowest amount of N, whereas the YLY6 and YLY1 varieties, both released in the 2000s, had the highest N uptake ( the correlation between the NHI and the year of release was significant at p < 0.01 (R 2 = 0.48), whereas in 2014, the correlation was significant at p < 0.1 (R 2 = 0.22) (Fig. 4b).
The nitrogen use efficiency for grain production (NUEg) reflects the physiological efficiency of assimilated N for grain yield production, and this value was significantly increased along with the year of release (p < 0.01) (Fig. 4c). Genetic improvements in the NUEg were 0.21 kg kg −1 year −1 (0.84% per year) in 2013 and 0.20 kg kg −1 year −1 (0.57% per year) in 2014. The partial factor productivity of N fertilizer (PFP) reflects the use efficiency of applied N fertilizer in grain yield production, and the value was linearly correlated (p < 0.01) with the year of release with R 2 of 0.79 in 2013 and 0.87 in 2014 (Fig. 4d). The PFP ranged from 26.9 kg kg −1 for SLX to 63.0 kg kg −1 for LYPJ in 2013 and from 28.9 kg kg −1 for SLX to 69.1 kg kg −1 for YLY1 in 2014 ( Table 4). The genetic improvements in PFP were 0.41 and 0.50 kg kg −1 year −1 on an absolute basis, and 1.18 and 1.16% on a relative basis for 2013 and 2014, respectively. Both the total N uptake and NUEg significantly contributed to the increased grain yield at p < 0.01 in 2013 and 2014 (Fig. S5a,b).

Discussion
Genetic improvements in the grain yield. Taking the whole period of 70 years (from 1936 to 2005), annual gains of 61.9 and 75.3 kg ha −1 in grain yield were observed for rice varieties grown in Middle Reaches of Yangtze River in 2013 and 2014, respectively (Fig. 2a). The rate of genetic improvements in the grain yield showed in present study falls within the range reported by similar studies on rice at IRRI and in Texas 16,18 . However, the increasing trend was not linear in 2013 (Table 2; Fig. 2a), which is consistent with the recent yield stagnation observed in 79% of rice planting areas in China 4 . Despite the relatively lower rate of yield increases in the last two decades, breeding efforts have significantly improved the stability of grain yield for varieties since the 1990s. Compared with 2013, grain yield of varieties released prior to 1990s (except for SLX and NJ11) was reduced by

Table 2. Grain yield and yield components of rice varieties grown in different years since 1930s in Middle
Reaches of Yangtze River in 2013 and 2014.
5.8-21.9% in 2014 due to the lower radiation and temperature, whereas grain yield of varieties released after 1990s (except for HHZ) was increased by 4.3-10.6% (Table 2). These results suggest that maintenance breeding has improved the adaptation of newer varieties to the environmental conditions of low radiation and low temperature that have a negative impact on older varieties 36 . Grain yield is the product of biomass and harvest index 37 . Harvest index (HI) was increased significantly when the sd1 gene was utilized in rice breeding in the 1950s 38 , and this improvement was demonstrated in this study by the significantly lower HI of the tall variety SLX (Table 2; Fig. 2c). Biomass has significantly increased since the end of the 1970s in present study due to the use of heterosis 13 . On the other hand, the grain yield could be divided into several components, namely spikelet m −2 comprised of panicles m −2 and spikelets panicle −1 , grain filling percentage, and grain weight 37 . Among the yield components, the enlarged sink size caused by the heavier panicles largely accounted for the genetic improvements in grain yield of rice (Table 2; Fig. 2e), and this result was similar to the results found in an analysis of 21 indica hybrid varieties released since 1976 22 and 12 indica inbred and hybrid varieties released since 1940s 35 in China. A moderate number of tillers and large panicles have been the target traits in many rice breeding programs such as the new plant type (NPT) breeding program at IRRI 11 and the "super" hybrid rice breeding program in China 39 . To maintain a high grain filling percentage for the large sink size, other morphological traits must be simultaneously improved to increase biomass production through a combination of the ideotype approach and the utilization of intersubspecific heterosis 11 . These traits are mostly related to the leaf morphology, such as the leaf length, angle and thickness and LAI of the top three leaves 39 . Modifications in these plant morphological traits result in the improvements in the light distribution and ventilation within the canopy, which could lead to increases in canopy photosynthesis 40,41 . Genetic improvements in RUE. From a physiological perspective, biomass at maturity is the product of total incident radiation during the rice growing period, efficiency with which radiation is intercepted by the crop (radiation interception efficiency), and efficiency with which intercepted radiation converted into biomass 42 . In    the advance of breeding, the total incident radiation during the rice growth period has increased because the growing season has been extended ( Fig. S1a and S2a). RUE is one of the most promising traits for further improvements in the grain yield of rice 43 , since the efficiency of radiation interception has been significantly improved because plant stature and canopy architecture have been optimized 17,24 . A question remains as to whether RUE has been increased during the crop breeding process. Many lines of evidence indicate that improvements in RUE have contributed to genetic improvements in the grain yield of many crops, although some has argued that there has been little or no improvement in the conversion efficiency of radiation into biomass 6,17,30 . In present study, grain yield was significantly correlated    with pre-anthesis biomass accumulation and pre-anthesis RUE, but not with post-anthesis biomass accumulation (Fig. S4a,b). RUE was significantly increased along with the year of release of the varieties, and a 27.8% improvement in pre-anthesis RUE was observed over the 70 years covered in present study (approximately 0.43% per year). Similar results have been reported in other crops. In soybean, the efficiency of light conversion to biomass has increased by approximately 36% over 84 years, and together with the improvement in light interception efficiency, these changes are responsible for the observed yield increases in the advance of breeding 17 . For wheat varieties developed from 1972 to 1995 in the UK, increases in pre-anthesis RUE drove increases in the number of grains and the accumulation of soluble carbohydrates for grain filling, which led to significant improvements in grain yield 30 . On one hand, the increase in RUE discussed above may have resulted from improvements in the intrinsic photosynthetic rate. In rice, the higher biomass of newer varieties released at IRRI has led to an increased light saturated photosynthetic rate per unit leaf area 44 . A similar trend has also been found in other studies in rice 45,46 . On the other hand, the optimization of plant architecture might have decreased the photoinhibition of the top leaves and increased the assimilation rate of lower leaves through an optimized distribution of radiation within the canopy 47,48,49 . Moreover, higher photosynthetic rates have been observed in rice during the grain filling period in newer varieties 45,46 , and such changes could contribute to improvements of RUE during the grain filling period because the photosynthetic rate in the flag leaf after heading was positively correlated with grain yield 50 .
The pre-anthesis RUE in present study ranged from 1.05 to 1.85 g MJ −1 (Table 3), and these values are similar to that of two super-hybrid, two ordinary hybrid and two inbred rice varieties (1.08-1.66 g MJ −1 ) grown in Hunan and Guangdong provinces in China 51 , and to that of seven high-yielding rice varieties (1.29-1.72 g MJ −1 ) grown in Yunnan province of China and Kyoto of Japan 52 . Here, a significant difference was observed in pre-anthesis RUE values between 2013 and 2014 (Table 3). Significant variation of the RUE values in two consecutive experimental years was also found in soybean 14 , and this may have resulted from a significant negative relationship between RUE and intercepted (or incident) radiation 42 . Recently, a meta-analysis found that RUE was 18% higher in shaded conditions compared with that under full sunlight 53 .

Genetic improvements in NUE.
Several studies demonstrated that breeding for new varieties in rice significantly increased the response of grain yield to N applications 18,46 , which has led to an overuse of N fertilizers and a reduction in NUE in rice production 54 . In present study, we demonstrated that N uptake was significantly higher in newer varieties, and NUEg was also significantly increased (Fig. 4). These results are consistent with the results of a recent study in which yield improvements were accompanied by increases in N uptake and NUEg for varieties widely grown in Jiangsu Province over the past 70 years 35 . Improvements in N uptake and NUEg in the advance of breeding were also found in wheat 32 , maize 33 and cotton 55 . These results indicate that the empirical viewpoint that newer varieties developed under conditions of ample N application have lower NUE is unauthentic, and show that the lower NUE frequently observed in rice production is mainly caused by inappropriate N management 35,54 .
In present study, both the increased N uptake and NUEg contributed to the increases in grain yield (Fig. S5a,b), because N significantly affects grain yield through its effect on both the source and sink. Increases in N uptake may have contributed to the improvement in RUE because RUE is dependent on photosynthesis and respiration 43 , and these metabolic processes are affected by plant N uptake 31,56 . On the other hand, larger N accumulation in vegetative and early reproductive growth stage is necessary for producing large number of spikelets 57 , and N top-dressing at the panicle initiation stage was most efficient in increasing spikelet number 58 . In present study, breeding for high yield is accompanied by a significant increase in the number of spikelets per panicle and N uptake and NUEg. Coincidences of QTLs for yield and its components with genes encoding cytosolic GS and the corresponding enzyme activity were detected in maize 59 and rice 60 . Recently, genetic link between number of spikelets per panicle and nitrogen use efficiency in rice was demonstrated by one gene locus "DEP1" 61,62 . The gene is first cloned to reduce length of inflorescence internode, and increase number of grains per panicle and grain yield in rice 61 . Afterwards, a major rice NUE quantitative trait locus (qNGR9) is cloned, and interestingly this gene locus is synonymous with DEP1 62 .

Conclusions
Yield potential has been the main target in rice breeding program under the pressure of increasing population. In this study, the grain yield was significantly increased with the year of release. The genetic improvements in the grain yield partially resulted from the significant increase in RUE. In addition, both nitrogen uptake capacity and nitrogen use efficiency for grain production in newer varieties was improved significantly, which contributed to the increase in grain yield. Currently, the world is facing challenges from environmental pollution, depletion of natural resources, climate change, and growing population, so crop yield potential has to be increased together with resource use efficiency. Compared with the theoretical maxima, there is still room for further improvement in RUE which is an important target in future breeding program. Recent progress in identification of NUE-related genes in rice may facilitate the breeding for high NUE. Overall, RUE and NUE should be concomitantly increased in the future.

Experiment design and plant materials. The experiments were conducted in a farmers' field at Dajin
Township (29°51′ N, 115°53′ E), Wuxue County, Hubei Province, China, during the rice-growing season from May to October in 2013 and 2014. The soil from experiment field had a texture of clay loam with pH 5.47, organic matter 29.10 g kg −1 , total N 2.2 g kg −1 , available P 12.14 mg kg −1 and available K 92.2 mg kg −1 .
Fourteen historical indica mega varieties that were released from 1936 to 2005 were used in this study, and they were all cultivated as middle-season rice in a large-scale area at that time in the Middle Reaches of Yangtze River of China during the last 70 years. Among them, Shenglixian is a tall variety, Aizizhan, Guangchang'ai, Scientific RepoRts | 6:21049 | DOI: 10.1038/srep21049 Zhenzhu'ai, Nanjing11, Ezhong2, Guichao2, Teqing and Huanghuazhan are inbred rice, Shanyou63 is an ordinary hybrid rice, and IIYou725, Liangyoupeijiu, Yangliangyou6, Yliangyou1 are super high yielding rice varieties that were certified by China's Ministry of Agriculture (Table 1).
All cultivars were arranged in a randomized complete block design with four replications and plot size of 5 × 6 m. Pre-germinated seeds were sown in seedbed. Seedlings (25 d old) were transplanted on 9 June in 2013 and 6 June in 2014. The planting density was 25 hills m −2 at a hill spacing of 30.0 cm × 13.3 cm with three seedlings per hill. Fertilizers included urea for N, single superphosphate for P and potassium chloride for K, and they were applied at the rates of 150 kg N ha −1 , 40 kg P ha −1 and 100 kg K ha −1 . N fertilizer was split-applied at a ratio of 4:2:4 at basal (1 day before transplanting), tillering (7 days after transplanting), and panicle initiation. P and K were all applied at basal. The experimental field was flooded from transplanting until 7 days before maturity. Pests and weeds were intensively controlled using chemicals to avoid yield loss.
Measurements. Plant sampling at heading. Plants from 12 hills in each plot were sampled at heading, and then separated into leaves, stems and panicles. The area of green leaf blades was measured with a Li-Cor area meter (Li-Cor Model 3100, Li-Cor Inc., Lincoln, NE, USA), and expressed as leaf area index (LAI). The dry weights of leaves, stems and panicles were measured after oven-drying at 70 °C to constant weight. The specific leaf weight was calculated as the ratio of leaf weight to leaf area. The aboveground total dry weight was the summation of dry weights in different plant parts.
Yield and yield components. At maturity, plants of 5 m 2 in the center of each plot (to avoid border effect) were harvested to determine the grain yield which was adjusted to 14% moisture content. Grain moisture content was measured with a digital moisture tester (DMC-700, Seedburo, Chicago, IL, USA). Grain yield per day was calculated as the ratio of grain yield to total growth duration. Destructive sampling of 12 hills from each plot was conducted to determine the yield components. After the panicle number was counted, the plants were separated into straw and panicles. The straw dry weight was determined after oven drying at 80 °C to a constant weight. The panicles were hand threshed, and the filled spikelets were separated from the unfilled spikelets by submerging the spikelets in tap water, then half-filled and empty spikelets were separated by seed wind machine (FJ-1, China). Subsequently, three 30 g subsamples of filled spikelets, 15 g subsamples of half-filled spikelets, and 2 g subsamples of empty spikelets were collected to count the number of spikelets. The dry weights of the rachis and filled, half-filled and empty spikelets (unfilled spikelets) were determined after oven drying at 80 °C to a constant weight. The panicles m −2 , spikelets per panicle, total spikelets m −2 , 1000-grain weight, and harvest index were all calculated. Total dry weight (TDW) was calculated by summing the total dry matter of straw, rachis, filled and unfilled spikelets. Harvest index was calculated as the percentage of grain yield to the aboveground total biomass.
Radiation interception and use efficiency. The climate data (temperature and solar radiation) were collected from the weather station located within 2 km from the experimental site. A datalogger (CR800, Campbell Scientific Inc., Logan, Utah, USA) was used as the measurement and control module. A silicon pyranometer (LI-200, LI-COR Inc., Lincoln, NE, USA) and temperature/RH probe (HMP45C, Vaisala Inc., Helsinki, Finland) were used to measure total solar radiation and temperature, respectively. The total incident radiation were summation of daily global solar radiations from transplanting to maturity.
The canopy radiation interception (LI) was measured from transplant to heading in 2013 and 2014. LI was not measured during grain filling because of the occurrence of lodging. The measurements were performed between 1100 and 1300 h on clear-sky days at an interval of 7-15 days during the growing season with a line ceptometer (AccuPAR LP-80, Decagon Devices Inc., Pullman, WA, USA). In each plot, the light intensity inside the canopy was measured by placing the light bar in the middle of two rows and at approximately 5 cm above the water surface. The light intensity was then recorded above the canopy. In total, six measurements were performed in each plot, with three measurements performed in wider rows and three performed in narrower rows. The LI was calculated as the percentage of light intercepted by the canopy (light intensity above the canopy-light intensity below the canopy)/light intensity above the canopy 39 . The intercepted radiation during each growing period was calculated using the average LI and accumulated global radiation during the growing period. The intercepted global radiation from transplanting to heading (Ri) was the summation of intercepted global radiation during each growing periods as in Equation 1: where n represents the time when LI was measured and R represents total global solar radiation during the period between two consecutive measurements of LI. LI at transplanting (LI 0 ) was assumed to be zero. Radiation use efficiency for intercepted global radiation from transplanting to heading (pre-anthesis RUE) was calculated as the ratio of aboveground total dry weight at heading relative to total intercepted global radiation based on Equation 2 63 . Nitrogen uptake and use efficiency. At maturity, after measurement of yield components and biomass from 12-hill samples, dry matter of each plant part [stem plus leaf (straw), filled grains, and unfilled grains plus rachis] was ground to powder to measure N concentration with Elementar vario MAX CNS/CN (Elementar Trading Co., Ltd, Germany). The total aboveground N uptake was then calculated as the product of the N concentration and dry weight of each aboveground part. Nitrogen harvest index (NHI) was calculated as the ratio of grain N content to total aboveground N uptake according to Equation 3. Nitrogen use efficiency for grain production (NUEg) was calculated as the ratio of grain yield over total aboveground N uptake according to Equation 4. Partial factor productivity of applied nitrogen fertilizer (PFP) was calculated as the ratio of grain yield to the fertilizer N input according to Equation 5.