Genetic architecture of leaf photosynthesis in rice revealed by different types of reciprocal mapping populations

Several reliable QTLs for leaf photosynthesis were detected using reciprocal mapping populations derived from japonica and indica rice varieties with different photosynthetic capacities. 

Abstract

The improvement of leaf net photosynthetic rate (A_n) is a major challenge in enhancing crop productivity. However, the genetic control of A_n among natural genetic accessions is still poorly understood. The high-yielding indica cultivar Takanari has the highest A_n of all rice cultivars, 20–30% higher than that of the high-quality japonica cultivar Koshihikari. By using reciprocal backcross inbred lines and chromosome segment substitution lines derived from a cross between Takanari and Koshihikari, we identified three quantitative trait loci (QTLs) where the Takanari alleles enhanced A_n in plants with a Koshihikari genetic background and five QTLs where the Koshihikari alleles enhanced A_n in plants with a Takanari genetic background. Two QTLs were expressed in plants with both backgrounds (type I QTL). The expression of other QTLs depended strongly on genetic background (type II QTL). These beneficial alleles increased stomatal conductance, the initial slope of A_n versus intercellular CO₂ concentration, or A_n at CO₂ saturation. Pyramiding of these alleles consistently increased A_n. Some alleles positively affected biomass production and grain yield. These alleles associated with photosynthesis and yield can be a valuable tool in rice breeding programs via DNA marker-assisted selection.

Introduction

To meet the growing demand for rice (Oryza sativa L.) from the rapidly expanding world population, the world will need 85% more rice production by 2050 relative to 2013 (Ray et al., 2013; Long et al., 2015). Rice yield has increased remarkably since the release of modern semi-dwarf cultivars with high harvest indices in the mid-1960s (Khush, 1999). However, as harvest indices are now close to the theoretical maximum, increases in total biomass production are key to further increasing yields (Mann, 1999; Zhu et al., 2010; van Bezouw et al., 2019). Many genes relating to sink size have been identified through genetic studies (Ashikari et al., 2005; Huang et al., 2009; Ikeda-Kawakatsu et al., 2009; Miura et al., 2010; Yoshida et al., 2013). However, little yield enhancement with these genetic factors is found, indicating that enhanced source capacity is required (Ohsumi et al., 2011; Fukushima et al., 2017).

Dry matter production is determined by radiation use efficiency (RUE) that is affected by canopy architecture, individual leaf photosynthesis, and leaf area. Leaf area affects dry matter production through light interception at the early and late growth stages, when leaf area index (LAI) is less than the critical value (Gardner et al., 1985; Hay and Porter, 2006; Hirasawa, 2014). Canopy architecture affects dry matter production in the middle growth stages, when LAI is greater than the critical value (Hay and Porter, 2006; Hirasawa, 2014), whereas individual leaf photosynthesis affects dry matter production over the entire growing season (Hay and Porter, 2006; Hirasawa, 2014). Improving canopy architecture with short plant height and erect leaves contributed greatly to the yield increase in the last century (Kumura, 1995; Khush, 1999; Hay and Porter, 2006; Taylaran et al., 2009); however, it is unlikely to contribute to further large increases because it has already been subjected to a long breeding history (Kumura, 1995; Horton, 2000; Reynolds et al., 2011; Gu et al., 2014). Thus, improvements in the photosynthetic rates of individual leaves within the canopy have become the focus for increased rice grain production (Horton, 2000; Hubbart et al., 2007; Parry et al., 2011).

Increasing the rate of photosynthesis has been attempted by molecular engineering approaches such as modification of the kinetic properties and production of photosynthetic enzymes (Kebeish et al., 2007; Zhu et al., 2008; Raines, 2011; Sage and Zhu, 2011; Long et al., 2015; Ort et al., 2015; Simkin et al., 2015), and the introduction of the C₄ photosynthetic pathway into rice, a C₃ plant, to avoid wasteful photorespiration (Sage et al., 2017). Some trials have succeeded; for example, Kromdijk et al. (2016) optimized photoprotection recovery in tobacco leaves and increased dry matter production by 15%. However, these attempts still face challenges for production in terms of adaptation, ecological risks, and public perceptions of genetically modified crops. Another way to enhance photosynthesis is to exploit the natural variation among genetic resources through cross-breeding (Raines, 2011; Flood et al., 2011; Gu et al., 2014). This approach is expected to discover unknown regulatory and developmental genes that may not be identifiable by mutant screening (Sage et al., 2017). Mining existing genetic variation could be the most efficient method for short-term improvements (Parry et al., 2011). A wide range of varietal differences in net photosynthetic rate (A_n) among rice cultivars have been examined for the past half century (Murata, 1961; Cook and Evans, 1983; Sasaki and Ishii, 1992; Kanemura et al., 2007; Jahn et al., 2011). Several loci that enhance leaf photosynthesis have been detected on chromosomes 3, 4, 5, 6, 8, 9, and 11 (Teng et al., 2004; Adachi et al., 2011a; Gu et al., 2012); however, a very limited number of genes (GREEN FOR PHOTOSYNTHESIS and Carbon assimilation rate 8) responsible for photosynthesis have been identified (Takai et al., 2013; Adachi et al., 2017). More effort is needed to understand the genetic basis of leaf photosynthesis for developing breeding strategies.

Photosynthetic rate in C₃ species is limited by biochemical and diffusion capacities. Biochemical limitations can be subdivided into the capacity of Rubisco to consume ribulose-1,5-bisphosphate (RuBP) and the capacity for RuBP to regenerate (Farquhar et al., 1980). Diffusion limitations can be subdivided into stomatal and mesophyll limitations, termed stomatal conductance (g_s) and mesophyll conductance (g_m), respectively (Makino, 2011): g_s controls CO₂ flux into the intercellular spaces from the atmosphere, and g_m controls it from the intercellular spaces to the site of carboxylation in the chloroplast stroma (Sage et al., 2017). In rice, A_n is closely related to leaf nitrogen content (LNC), which affects the Rubisco content and electron transport supporting RuBP regeneration as well as g_s and g_m (Makino et al., 1994; Yamori et al., 2011). It is known that g_s is larger in cultivars with higher hydraulic conductance than in those with lower hydraulic conductance even at the same LNC (Adachi et al., 2011a; Taylaran et al., 2011). These differences underlie varietal differences in A_n (Kuroda and Kumura, 1990; Sasaki et al., 1996; Kanemura et al., 2007; Ohsumi et al., 2007; Adachi et al., 2011b; Taylaran et al., 2011). Therefore, these physiological factors should be examined along with A_n for a comprehensive understanding of the genetic control of leaf photosynthesis.

The high-yielding indica cultivar Takanari has the highest A_n among rice cultivars, at 30–33 μmol m⁻² s⁻¹ (Kanemura et al., 2007; Hirasawa et al., 2010; Taylaran et al., 2011) during panicle formation to ripening. Its higher A_n is supported by higher Rubisco content due to greater LNC and by higher g_s due to larger hydraulic conductance (Hirasawa et al., 2010; Taylaran et al., 2009, 2011). Thus, Takanari seems to have favorable alleles that contribute to higher A_n. The high-quality japonica cultivar Koshihikari has a low A_n of 25–28 μmol m⁻² s⁻¹. Some backcross inbred lines (BILs) derived from a cross between Takanari and Koshihikari (Koshihikari/Takanari//Takanari) have 20–30% higher values of A_n than Takanari (Adachi et al., 2013). This indicates that Koshihikari might have some alleles that further increase A_n of indica cultivars.

In this study, we examined quantitative trait loci (QTLs) for A_n, using reciprocal backcross inbred lines (BILs) and chromosome segment substitution lines (CSSLs) derived from a cross between Takanari and Koshihikari in the field for 4 years (2 years for BILs and 2 years for CSSLs). Then we investigated the physiological properties of identified QTLs by using CSSLs grown in pots for 4 years (2 years for plants in Takanari genetic background and 2 years in Koshihikari genetic background). These investigations helped in understanding the genetic architecture of leaf photosynthesis controlled by genetic variation in the rice.

Materials and methods

Plant materials

We used BILs developed by Fukuda et al. (2018). F₁ plants from a cross between Koshihikari and Takanari were backcrossed to Koshihikari (for Koshihikari/Takanari//Koshihikari, BKT) or to Takanari (for Koshihikari/Takanari//Takanari, BTK) to produce BC₁F₁. Then BC₁F₆ (for the measurements in 2009) and BC₁F₇ (for the measurements in 2010) lines were generated from a single BC₁F₁ individual using 140 SSR markers distributed over the genomes of the BILs in Koshihikari background (BKTs) and BILs in Takanari background (BTKs). Eighty-two and 87 lines were used for measurements of BKTs and BTKs, respectively, in 2009, and 87 lines, for BKTs and BTKs in 2010 (Supplementary Table S1 at JXB online). Their genotypes are shown in Supplementary Table S1.

We also used 41 CSSLs covering most of the Takanari genome in the Koshihikari background (T-CSSLs, SL1201–1241) and 39 CSSLs covering the Koshihikari genome in the Takanari background (K-CSSLs, SL1301–1339) developed by Takai et al. (2014) as reciprocal CSSLs.

To detect the precise position of QTL regions, one T-CSSL (or K-CSSL) carrying a single QTL was crossed with Koshihikari (or Takanari), and the self-pollinated seeds were obtained. The progeny lines were selected based on their genotypes analysed with SSR markers. To develop QTL pyramiding lines, two T-CSSLs (or K-CSSLs) carrying a single QTL were crossed and a homozygous plant was selected from F₂ individuals with SSR markers.

Cultivation

Koshihikari, Takanari, BKTs and BTKs (in 2009 and 2010), and T-CSSLs and K-CSSLs (in 2011 and 2012) were grown in the paddy field of the University Farm located at Fuchu, Tokyo, Japan (35°41′N, 139°29′E). For comparisons of dry matter and grain production, plants were also grown in the paddy field in 2013 to 2016. Seedlings at the fourth-leaf stage were transplanted at 22.2 hills m⁻² (row width, 30 cm), with one plant per hill, and grown under submerged conditions. As a basal dressing, manure was applied at ~15 t ha⁻¹, and chemical fertilizer was applied at 30 kg N, 26 kg P, and 50 kg K ha⁻¹. One-third of the total N was applied as nitrogen sulfate, one-third as slow-release urea (LP-50), and one-third as very-slow-release urea (LPS-100; Chisso Asahi Fertilizer, Tokyo, Japan). No topdressing was applied. The experiments were arranged by randomized block design with two (for BILs) or three (for CSSLs) replicates and three or four replicates for biomass and yield determination. For BILs, one replication block consisted of one series of the lines and one row with eight hills for each line. For CSSLs, one replication block consisted of one series of the lines and four rows with nine hills per row for each line. For the determination of biomass and grain production, one replication block consisted of a set of comparable lines in an area of 10 m² or 4 m² per line for plants in the Koshihikari or Takanari genetic background, respectively.

CSSLs were also grown in pots with 12 liters of soil per pot. Seedlings at the fourth-leaf stage were transplanted into pots filled with a 1:1 (v/v) mixture of alluvial soil from the Tama River and Kanto diluvial soil at a density of three plants per hill, with three hills per pot, and grown under submerged conditions. Fertilizer was applied at 1.0 g N, 0.44 g P, and 0.83 g K per pot as a basal dressing, and 1.0 g N per pot as a topdressing a week before heading. The plants were grown outdoors with four to six replications arranged by randomized block design. One replication block consisted of a set of comparable lines with five pots per line.

Measurements of leaf photosynthesis rate and stomatal conductance

In the field in 2009–2012, we measured A_n and g_s with a portable photosynthesis system (LI-6400; LI-COR Inc., Lincoln, NE, USA) and a LED light source (LI-6400-02B; LI-COR Inc.) at the full heading stage in flag leaves attached to the main stem. We measured leaf color (SPAD value) of five or six leaves per replicate with a chlorophyll meter (SPAD-502; Konica Minolta, Osaka, Japan), and A_n and g_s of two or three leaves with average SPAD values.

The air temperature in the leaf chamber was controlled at 30 °C, and the photosynthetically active radiation at the leaf surface was controlled at 2000 μmol photon m⁻² s⁻¹. The ambient CO₂ concentration in the assimilation chamber (C_a) was adjusted to 370±1 μmol mol⁻¹ by controlling the CO₂ concentration of the air entering the chamber, and the leaf-to-air vapor pressure difference (VPD) was adjusted to 1.2–1.5 kPa by controlling the flow rate of air through the desiccant. Measurements were taken in sunlight between 08.30 and 11.30 h on clear days or at >400 μmol photon m⁻² s⁻¹ until 14.00 h on cloudy days. The leaf gas exchange rate (indicated by A_n and g_s) reached a steady state within a few minutes after the leaf was installed in the chamber. At least two measurements were taken on the same leaf on different days and averaged.

For A_n measurements in the pots in 2013–2016, C_a and leaf temperature were controlled at 400 μmol mol⁻¹ and 30 °C. The photosynthetically active radiation was 2000 μmol photon m⁻² s⁻¹. The VPD was adjusted to 1.2–1.5 kPa. The A_n–C_i curve was obtained at 13 values in the order of 330, 400, 500, 600, 800, 1000, 1500, 300, 230, 160, 120, 80, and 40 μmol mol⁻¹ of the intercellular CO₂ concentration (C_i) by changing C_a, and the initial slope of the curve and photosynthetic rate at CO₂ saturation (A_sat) were estimated. A_n was measured at 400 μmol CO₂ mol⁻¹ between 08.30 and 11.30 h on clear days and 14.00 h on cloudy days, and A_n–C_i curves were drawn. SPAD values of five to six leaves per replicate were measured, and one or two leaves with average SPAD values were selected for measurements of A_n and g_s.

Quantification of leaf nitrogen

LNC of leaves in which photosynthesis was measured was quantified with a CN analyser (MT700 Mark II; Yanako, Tokyo, Japan).

Measurements of dry weight, yield, and yield components

Before sampling for dry weight measurements, we counted panicle numbers in 25 successive hills per replicate. Then we harvested 7–10 hills with an average number of panicles from each replicate and removed roots after washing stem and roots with tap water. The plants were then dried in a ventilated oven at 80 °C for >4 d to a constant dry weight. Plants in an area of 1–2 m² in each replicate were harvested to determine grain yield per unit area. We determined spikelet number per panicle, ripening percentage, and 1000-grain weight in three or four hills with an average number of panicles in each replicate. Panicle number per m² was determined as the average of 25 hills per replicate. Grains thicker than 1.8 mm were selected by sieving as fully ripened.

QTL analyses

Data for each parameter were averaged for each replication and used in QTL analyses using genetic linkage maps of the BKT and BTK populations. QTL analyses were performed by composite interval mapping using the Zmapqtl program (model 6) in QTL Cartographer v. 2.5 software (Wang et al., 2005). Genome-wide threshold values (α=0.05) were used to detect putative QTLs based on the results of 1000 permutations.

We conducted analysis of variance (ANOVA) to examine the effects of CSSL on A_n and traits relevant to A_n. Significant CSSL effects were explored with Dunnett’s test, using Koshihikari as the control for the Koshihikari background and Takanari as the control for the Takanari background in JMP v. 13 software (SAS Institute, Cary, NC, USA). To delineate candidate QTL regions, we conducted substitution mapping by comparing overlapping segments among the CSSLs according to Takai et al. (2014).

Statistical analysis

Differences and interactions between rice lines and year effects on variance were tested by two-way ANOVA. Differences were also tested by the Tukey–Kramer test. Pearson’s correlation coefficient was calculated to test the significance of the relationships between A_n, LNC, and g_s and between days to heading from sowing (DTH) and A_n. All statistics were analysed in JMP v. 13 software.

Results

Detection of QTLs

The heading time was 7–9 d earlier in Koshihikari than in Takanari (Supplementary Table S2). Takanari consistently had 20–30% higher A_n than Koshihikari (Figs 1–3). A_n of BKTs and BTKs showed a continuous distribution; the means in BKTs and BTKs were close to those of Koshihikari and Takanari, respectively (Fig. 1).

Fig. 1.

Frequency distributions of photosynthetic rate (A_n) of BILs with Koshihikari background (A, C) and Takanari background (B, D) grown in the paddy field. Arrowheads indicate average values: black (K), Koshihikari; white (T), Takanari; gray, inbred lines.

Fig. 3.

Photosynthetic rates (A_n) of flag leaves at full heading stage of K-CSSLs (Takanari background) grown in the paddy field. Error bars represent SD (n=3). Asterisks indicate significant differences from Takanari at the *5%, **1%, and ***0.1% α level (Dunnett’s test). NA, not available. (This figure is available in color at JXB online.)

The phenotype data are shown in Supplementary Table S1. QTL analysis for A_n using the reciprocal BILs was conducted separately for populations and years. The analysis of BKTs estimated two QTLs with significant logarithm of odds (LOD) values on chromosomes (Chrs) 3 and 4 in both 2009 and 2010 (Table 1). The Koshihikari allele on Chr. 3 and the Takanari allele on Chr. 4 increased A_n. The analysis of BTKs estimated six QTLs on Chrs 1, 2, 3, 4, 5, and 7 (Table 1). Only one QTL, at RM3534 on Chr. 4, was detected in both years. Koshihikari alleles on Chrs 1, 2, and 7 and Takanari alleles on Chrs 3, 4, and 5 increased A_n. The QTL on Chr. 4 was detected in both populations across the years.

Table 1.

QTLs for enhancing flag leaf photosynthesis estimated by using BILs

Line	Year	Chr	Nearest marker or interval	LOD	Confidence interval (Mb)	PVE (%)a	Additive effectb	QTL
BKT	2009	3	RM1371	4.02	6.8	13.4	1.42	qHP3a
		4	RM3534	3.59	4.7	12.1	−1.71	qHP4
	2010	3	RM1371	3.85	6.8	11.7	1.25	qHP3a
		4	RM3534	5.56	4.7	17.8	−2.08	qHP4
BTK	2009	1	RM7594	4.03	7.8	13.5	1.69	qHP1b
		3	RM3204	3.84	5.1	16.9	−3.03	qHP3b
		4	RM3534	3.97	4.7	13.9	-1.7	qHP4
		5	RM3838–RM1386	3.54	5.8	13.6	−1.46	qHP5
	2010	2	RM5521	3.54	11.7	10.4	0.98	—
		4	RM3534	5.28	4.7	16.3	−1.45	qHP4
		7	RM6420	2.61	10.6	8.3	0.98	qHP7b

^a Phenotypic variance explained.

^b Koshihikari allele increases.

In the analysis of T-CSSLs (Koshihikari genetic background), lines SL1208, SL1217, and SL1235 had significantly higher A_n than Koshihikari in 2011 and 2012 (Fig. 2A, B). Additionally, SL1222 and SL1224 had significantly higher A_n than Koshihikari in 2011, while SL1227 had significantly lower A_n (Fig. 2A). In the analysis of K-CSSLs (Takanari genetic background), SL1301, SL1311, SL1324, and SL1325 had consistently and significantly higher A_n than Takanari in both years (Fig. 3A, B). Additionally, SL1303 had significantly higher A_n than Takanari in 2011, and some lines (SL1315, SL1335, and SL1336) had significantly lower A_n than Takanari (Fig. 3A); SL1336 had significantly lower A_n than Takanari in both years. We also found SL1324 and SL1325 carried different sets of QTLs for A_n from the fine mapping experiment of SL1324 (see Supplementary Fig. S1; Supplementary Table S3) although they have an overlapping segment of Koshihikari. Confidence intervals of QTLs for A_n estimated by using CSSLs and by narrowing down are shown in Supplementary Table S3.

Fig. 2.

Photosynthetic rates (A_n) of flag leaves at full heading stage of T-CSSLs (Koshihikari background) grown in the paddy field. Error bars represent SD (n=3). Symbols indicate significant differences from Koshihikari at the †10%, *5%, **1%, and ***0.1% α level (Dunnett’s test). (This figure is available in color at JXB online.)

All QTLs for A_n detected in BILs and CSSLs were mapped on the rice genome and visualized by graphical genotyping (Fig. 4A, B). Most were detected more than twice among the years; we named them ‘qHP’ (high photosynthesis) and focused on them for further analysis. Four QTLs where Takanari alleles enhanced A_n were identified in two or more years in CSSLs or BILs on Chrs 2, 4, 5, and 10. Among them, only qHP4 was detected in both backgrounds. qHP5 was detected in BTKs and K-CSSL populations (Takanari background), even though the Takanari allele increased A_n. Six QTLs where Koshihikari alleles enhanced A_n were identified in two or more years in CSSLs or BILs on Chrs 1, 3, and 7. Most were identified in the populations of the Takanari background, and only qHP3a was detected in the BKTs and T-CSSL populations (Koshihikari background).

Fig. 4.

Summary of identified QTLs. (A) Graphical genotypes of mapping populations. Orange, homozygous for Koshihikari; blue, homozygous for Takanari. BILs (BKTs and BTKs) were used in 2009 and 2010, and CSSLs (T-CSSLs and K-CSSLs) were used in 2011 and 2012. (B) Locations of QTLs for photosynthetic rate (A_n). (C) Locations of QTLs for stomatal conductance (g_s). (D) Locations of QTLs for leaf nitrogen content (LNC). Circles represent the locations of QTLs for A_n that were repeatedly identified in different years.

In the field in 2009, Takanari had 91% higher g_s and 22% higher LNC than Koshihikari, and the mean LNCs in BKTs and BTKs were close to those of Koshihikari and Takanari, respectively (see Supplementary Fig. S2). These trends were the same in 2010 (data not shown). There were close correlations between g_s and A_n and between LNC and A_n in BKTs and BTKs (Supplementary Fig. S3) and in CSSLs, although those between LNC and A_n were lower in K-CSSLs than in T-CSSLs (Supplementary Fig. S4).

We mapped the QTLs associated with increased g_s and increased LNC as we did for A_n from the results of the BILs (see Supplementary Tables S4, S5) and CSSLs (Supplementary Figs S5–S8). Four QTLs where Takanari alleles enhanced g_s were identified in two or more years in CSSLs or BILs on Chrs 3, 4, 10, and 11, and two of them overlapped with qHP4 and qHP10 (Fig. 4C). No QTL where the Koshihikari allele enhanced g_s was identified in any year in CSSLs or BILs. Four QTLs where Takanari alleles increased LNC were identified in two or more years in CSSLs or BILs on Chrs 1, 2, 4, and 9, and two of them overlapped with qHP2 and qHP4 (Fig. 4D). One QTL where Koshihikari alleles increased LNC was identified in two or more years in CSSLs or BILs on Chr. 7, and overlapped with qHP7b (Fig. 4D). We also identified several loci for g_s and LNC without any association with the QTLs for A_n. These results indicate that QTLs for increased g_s or LNC don’t always correspond with enhanced A_n.

Validation of QTLs using pot-grown rice

We grew the parental lines and CSSLs that carry QTLs for A_n in the background of either Koshihikari or Takanari in pots and examined the cause of the higher A_n. Takanari had a significantly larger A_n and g_s than Koshihikari, reflecting its larger C_i/C_a (Fig. 5A, B, E). LNC was significantly higher in Takanari than in Koshihikari (Fig. 5F), reflecting the significantly larger initial slope of the A_n–C_i curve (Fig. 5C), but there was no significant difference in A_sat between cultivars (Fig. 5D). All CSSLs had significantly higher A_n than the respective parental cultivars across the years: SL1217 and NIL10 (developed from SL1235 and carrying a Takanari chromosome segment from RM6128 to RM25840) had significantly higher A_n than Koshihikari, and SL1301, SL1303, SL1311, SL1324, and SL1325 had significantly higher A_n than Takanari (Table 2). These results clearly indicate that the QTLs identified in the field also had positive effects on A_n even in pots. We did not use SL1208 for measurements because the plant size was very small (Takai et al., 2014).

Fig. 5.

Comparisons of photosynthetic rate (A_n) (A), stomatal conductance (g_s) (B), initial slope of the A_n–C_i curve (C), A_n at CO₂ saturation (A_sat) (D), ratio of intercellular CO₂ concentration (C_i) to ambient CO₂ concentration (C_a) (C_i/C_a) (E), and leaf nitrogen content (LNC) (F) between Koshihikari and Takanari grown in pots. Error bars represent SD (n=6). Asterisks indicate significant differences between Koshihikari and Takanari at the *5% and **1% α level (Student’s t-test). ns, not significant.

Table 2.

Comparisons of photosynthetic rate (A_n), stomatal conductance (g_s), initial slope of the A_n–C_i curve, photosynthetic rate at CO₂ saturation (A_sat), and leaf nitrogen content (LNC) of Koshihikari, Takanari, and CSSLs carrying QTLs for enhancing A_n; all plants were grown in pots

Group	Year	Cultivar, line	QTL	A n (μmol m−2 s−1)	g s (mol m−2 s−1)	Initial slope (mol m−2 s−1)	A sat (μmol m−2 s−1)	LNC (g m−2)
A	2015	Koshihikari		25.9±1.7	0.54±0.06	0.128±0.011	41.4±4.8	1.98±0.03
		SL1217	qHP4	32.2±2.5**	0.69±0.05**	0.144±0.020	49.6±1.5**	2.27±0.19**
		NIL10	qHP10	29.0±1.9*	0.72±0.11**	0.134±0.006	44.4±4.9	2.04±0.09
	2016	Koshihikari		24.1±0.6	0.47±0.02	0.108±0.005	36.0±3.1	1.60±0.18
		SL1217	qHP4	26.9±1.7**	0.48±0.06	0.127±0.005**	40.7±2.6**	1.82±0.06*
		NIL10	qHP10	26.1±1.1*	0.52±0.03*	0.116±0.005*	38.9±1.1	1.59±0.19
	ANOVA	Line		<0.001	<0.001	<0.001	<0.001	<0.001
		Year		<0.001	<0.001	<0.001	<0.001	<0.001
		Line×Year		<0.05	<0.05	ns	ns	ns
B	2013	Takanari		31.2±0.4	1.16±0.15	0.132±0.007	42.0±4.4	1.54±0.04
		SL1301	qHP1a	33.8±1.6**	1.31±0.11	0.150±0.006	44.2±1.6	1.59±0.09
		SL1303	qHP1b	—	—	—	—	—
		SL1311	qHP3b	35.0±1.0**	1.27±0.12	0.161±0.011**	51.0±0.3**	1.72±0.10*
		SL1324	qHP7a	33.0±0.5*	1.37±0.29	0.151±0.006	44.2±1.9	1.58±0.08
		SL1325	qHP7b	33.7±0.9**	1.20±0.35	0.152±0.010*	49.1±3.7**	1.72±0.08**
	2014	Takanari		29.6±1.6	0.85±0.11	0.141±0.008	38.8±3.4	1.82±0.19
		SL1301	qHP1a	34.8±1.1**	1.25±0.17**	0.154±0.006**	44.3±2.3**	1.92±0.05
		SL1303	qHP1b	34.1±0.6**	0.87±0.05	0.154±0.004*	44.1±3.2*	2.16±0.07*
		SL1311	qHP3b	32.2±0.7**	0.90±0.11	0.144±0.006	42.0±2.5	2.06±0.05
		SL1324	qHP7a	33.5±1.4**	1.00±0.09	0.149±0.005	44.8±1.5**	1.92±0.27
		SL1325	qHP7b	34.8±1.2**	1.10±0.12**	0.154±0.004**	46.0±0.8**	2.10±0.14**
	ANOVA	Line		<0.001	<0.01	<0.001	<0.001	<0.01
		Year		ns	<0.001	ns	<0.001	<0.001
		Line×Year		<0.001	ns	ns	<0.01	ns

Groups A and B represent plants in the Koshihikari and Takanari genetic background, respectively. NIL10 was developed from SL1235 and carries a Takanari chromosome segment from RM6128 to RM25840. Plants were sown on 11 April 2015, 18 April 2016, 6 June 2013, and 6 June 2014. Asterisks indicate significant difference from Koshihikari in A group, or from Takanari in B group at the *5% and **1% α level (n=4–6, Dunnett’s test). ns, not significant.

In the CSSLs with the Koshihikari background, SL1217, which harbors qHP4, had significantly higher values of all photosynthetic parameters measured than Koshihikari (Table 2) in one or both years. In contrast, NIL10, which harbors qHP10, had significantly higher g_s but similar initial slope of the A_n–C_i curve (2015 only), and similar A_sat and LNC to Koshihikari (Table 2). These QTL effects are consistent with the field results, in which qHP4 and qHP10 affected g_s and qHP4 affected LNC (Fig. 4C, D). All CSSLs with the Takanari background had significantly higher A_sat than Takanari in one or both years, and all except for SL1324 had significantly higher initial slope of the A_n–C_i curve in one or both years, while only SL1301 and SL1325 had significantly higher g_s in one year, and SL1303, SL1311, and SL1325 had significantly higher LNC in one or both years (Table 2). These QTL effects were not observed in the field, except that qHP7b (associated with SL1325) affected LNC (Fig. 4D).

Plant or culm length, dry matter and grain production, and yield components of the plants carrying QTLs

In field experiments, SL1217 showed significantly shorter plant length and significantly lower dry weight of above-ground parts at harvest and grain yield than Koshihikari (Table 3); the lower grain yield was caused by the very small number of spikelets per panicle (Table 4). In contrast, in NIL10, plant length did not differ significantly from that of Koshihikari, but dry weight at harvest and grain yield were significantly larger than those of Koshihikari in both years (Table 3); the higher grain yield was caused by the larger number of panicles and spikelets per panicle (2013 only) (Table 4).

Table 3.

Comparisons of plant (culm) length, dry weight of above-ground parts at harvest (biomass), and grain yield of Koshihikari, Takanari, and CSSLs carrying QTLs for enhancing A_n; all plants were grown in the paddy field

Group	Year	Cultivar, CSSL	QTL	Plant length (cm)	Culm length (cm)	Biomass (g m−2)	Grain yield (g m−2)
A	2016	Koshihikari		114.2±3.4	—	1421±39	585±18
		SL1217	qHP4	106.4±0.4**	—	1321±42*	542±20*
	2013	Koshihikari		112.3±0.4	—	1445±101	517±11
		NIL10	qHP10	112.8±20.8	—	1804 ±183*	664±13***
	2014	Koshihikari		104.1±1.0	—	1547±31	564±35
		NIL10	qHP10	103.9±0.8	—	1645±17***	642±12*
	ANOVA	Line		ns	—	<0.01	<0.001
		Year		<0.001	—	ns	ns
		Line×Year		ns	—	ns	<0.05
B	2015	Takanari		—	73.9±3.8	1329±129	513±30
		SL1301	qHP1a	—	73.5±2.8	1455±113	647±49**
		SL1303	qHP1b	—	91.5±14.5*	1633±293†	536±58
		SL1311	qHP3b	—	89.2±3.9*	1359±114	534±49
		SL1324	qHP7a	—	72.6±1.8	1371±110	589±27
		SL1325	qHP7b	—	75.1±3.2	1461±133	612±54*
	2016	Takanari		—	76.2±2.2	1461±94	829±54
		SL1301	qHP1a	—	82.3±1.7	1502±26	735±24
		SL1303	qHP1b	—	97.4±3.4***	1640±95†	709±39
		SL1311	qHP3b	—	86.5±2.9***	1268±119*	631±113*
		SL1324	qHP7a	—	75.0±1.4	1398±114	861±109
		SL1325	qHP7b	—	79.4±2.6	1373±100	847±82
	ANOVA	Line		—	<0.05	<0.001	<0.001
		Year		—	<0.001	ns	<0.001
		Line×Year		—	ns	ns	<0.01

Groups A and B represent plants in the Koshihikari and Takanari genetic background, respectively. NIL10 was developed from SL1235 and carries a Takanari chromosome segment from RM6128 to RM25840. Plants were sown on 15 April 2013, 15 April 2014, 5 May 2015, and 6 May 2016. Grain yield was deterrmined at 14.5% moisture content. Symbols indicate significant differences from Koshihikari in Group A or Takanari in Group B at the ^†10%, *5%, or **1% α level (n=3 or 4, Dunnett’s test); ns, not significant.

Table 4.

Comparisons of yield components of Koshihikari, Takanari, and CSSLs carrying QTLs for enhanced A_n; all plants were grown in the paddy field

Group	Year	Cultivar, CSSL	QTL	Panicle number (m−2)	Spikelet number per panicle	Ripening percentage (%)	1000-grain weight (g)
A	2016	Koshihikari		297±12	103±2	90.8±1.4	21.1±0.1
		SL1217	qHP4	326±11*	84±6***	94.2±1.9*	21.1±0.6
	2013	Koshihikari		281±8	107±6	81.0±3.2	21.3±0.5
		NIL10	qHP10	332±8**	119±3*	83.7±2.5	20.0±0.5*
	2014	Koshihikari		290±11	104±2	87.9±3.8	21.4±0.3
		NIL10	qHP10	338±6**	104±1	90.3±2.7	20.2±0.1**
	ANOVA	Line		<0.001	<0.05	ns	<0.001
		Year		ns	<0.01	<0.01	ns
		Line×Year		ns	<0.05	ns	ns
B	2015	Takanari		166±46	262±58	66.3±2.5	18.7±0.3
		SL1301	qHP1a	228±29	175±9†	82.4±4.6*	19.8±0.2**
		SL1303	qHP1b	182±58	269±45	60.0±13.9	19.5±0.7†
		SL1311	qHP3b	186±58	248±73	71.5±6.2	17.5±0.3**
		SL1324	qHP7a	173±27	213±24	84.9±2.3**	19.1±0.5
		SL1325	qHP7b	180 ±24	266±42	67.5±3.1	19.2±0.1
	2016	Takanari		229±19	201±11	85.7±4.2	21.1±0.2
		SL1301	qHP1a	242±7	164±13***	85.8±5.3	21.6±0.3*
		SL1303	qHP1b	183±16†	215±9	86.2±5.3	21.0±0.3
		SL1311	qHP3b	228±44	204±11	71.8±2.1***	18.9±0.2***
		SL1324	qHP7a	269±33	174±5*	90.2±0.6	20.4±0.2***
		SL1325	qHP7b	257±28	196±2	78.9±1.7†	21.3±0.1
	ANOVA	Line		<0.10	<0.001	<0.001	<0.001
		Year		<0.001	<0.001	<0.001	<0.001
		Line×Year		<0.10	ns	<0.001	<0.05

Groups A and B represent plants in the Koshihikari and Takanari genetic background, respectively. NIL10 was developed from SL1235 and carries a Takanari chromosome segment from RM6128 to RM25840. Ripening percentage represents the ratio of the number of grains thicker than 1.8 mm to the total number of spikelets. Grain weight was deterrmined at 14.5% moisture content. Symbols indicate significant difference from Koshihikari in Group A, or from Takanari in Group B, at the ^†10%, *5%, and **1% α level (n=3 or 4, Dunnett’s test).

In SL1301 and SL1325, culm length and dry weight of above-ground parts at harvest did not differ from those of Takanari, and grain yield was significantly higher than that of Takanari in 2015 but not in 2016 (Table 3). One-thousand-grain weight was significantly larger in SL1301 than in Takanari in both years (Table 4). Panicle number per m² tended to be larger in SL1325 than in Takanari in both years, although the differences were not significant (Table 4). In SL1311 and SL1324, dry weight of above ground parts at harvest and grain yield were similar to those of Takanari except for SL1311 in 2016, when ripening percentage and 1000-grain weight were also significantly smaller in SL1311. SL1303 showed large culm length and dry weight compared with Takanari in both years, with no significant difference in grain yield or yield components except for 1000-grain weight in 2015 and panicle number in 2016 (Tables 3, 4).

Effects of QTL pyramiding on A_n

We tested the effects of some QTL combinations. All combinations tested showed clear enhancements of A_n relative to plants carrying one QTL in the Koshihikari (Fig. 6A) or Takanari (Fig. 6B–D) background. The line carrying qHP4 and qHP10 in the Koshihikari background had 30% higher A_n than Koshihikari, comparable to the Takanari value (Fig. 6A).

Fig. 6.

Photosynthetic rate (A_n) of plants carrying Takanari allele(s) in the Koshihikari genetic background in 2016 (A) and Koshihikari allele(s) in the Takanari genetic background in 2015 (B–D) in the paddy field. C_a was controlled at 400±1 μmol mol⁻¹ and other conditions were the same as in the field measurements in 2009–2012. Error bars represent SD (n=4). Bars with the same letter are not significantly different among the cultivar(s) and lines in each figure at the 10% α level (Tukey–Kramer test).

Discussion

We identified QTLs for leaf photosynthesis in rice by directly measuring leaf gas exchange and other relevant traits in the field, and then evaluated their effects using plants grown in pots. The use of two elite mapping populations that were recently developed by DNA marker-assisted selection (MAS) and of repeated trials over different years enabled us to identify reliable QTLs controlling the natural genetic variation of A_n. We examined the effects of pyramiding the various QTLs on A_n, and explored the associations between photosynthesis-related parameters and A_n and the associations between the A_n-enhancing QTLs and dry matter and grain production by using plants grown in the field and in pots.

Expression and pyramiding effects of QTLs for leaf photosynthesis

We detected 10 QTLs for A_n—qHP1a, qHP1b, qHP2, qHP3a, qHP3b, qHP4, qHP5, qHP7a, qHP7b, and qHP10—in at least two years by using BILs and CSSLs in the field (Fig. 4B). Among them, four QTLs (qHP1b, qHP4, qHP5, and qHP7b) were detected by using BILs as well as CSSLs, five QTLs (qHP1a, qHP2, qHP3b, qHP7a, and qHP10) were detected only by using CSSLs, and one QTL (qHP3a) was detected only by using BILs. By using CSSLs grown in pots, we confirmed that Takanari alleles at qHP4 and qHP10 increased A_n in plants with a Koshihikari background and that Koshihikari alleles at qHP1a, qHP1b, qHP3b, qHP7a, and qHP7b increased A_n in plants with a Takanari background (Table 2). These results indicate that reliable QTLs for enhanced A_n exist in rice. It is likely that qHP3a is expressed only following allelic interaction with other Koshihikari alleles, because the Koshihikari allele at this QTL was identified only in the populations with the Koshihikari background. There were two patterns of QTL expression: expression in plants with both genetic backgrounds (type I) and expression dependent on genetic background (type II). Both qHP4 and qHP7b are type I QTLs. qHP4 is located in a chromosomal region containing the GPS (GREEN FOR PHOTOSYNTHESIS) gene that Takai et al. (2013) identified by using the same BIL mapping population as we used. qHP4 is very effective at increasing A_n as the Takanari allele or decreasing A_n as the Koshihikari allele. On the other hand, qHP7b is as effective at increasing A_n as the Koshihikari allele or decreasing A_n as the Takanari allele. The other detected QTLs are type II QTLs. Effects of the Koshihikari allele at qHP3a and the Takanari alleles at qHP2 and qHP10 on A_n were consistently detected only in plants with a Koshihikari background, and effects of the Koshihikari alleles at qHP1a, qHP1b, qHP3b, qHP5, and qHP7a were consistently detected only in plants with a Takanari genetic background (Fig. 4B).

By using CSSLs derived from a cross between Koshihikari and Habataki, a sister cultivar of Takanari, Adachi et al. (2011b, 2014) detected four QTLs where Habataki alleles enhanced A_n in plants with a Koshihikari background. But despite using CSSLs derived from related materials, we detected only qHP4 (detected as qCAR4 by Adachi et al., 2014). The Habataki allele at qCAR5 on Chr. 5 enhanced A_n in plants with a Koshihikari background (Adachi et al., 2014); although qHP5 lies in the same region as qCAR5, in the current study the Takanari allele at qHP5 did not enhance A_n in the plants with a Koshihikari background, but the Koshihikari allele at qHP5 decreased A_n in the plants with a Takanari background. Thus, the expression of a type II QTL may be affected also by the combination of parents of the mapping population, although it remains to be seen whether qCAR5 and qHP5 are identical.

The value of A_n in a plant carrying only one Takanari allele (qHP2, qHP4, or qHP10) did not reach to the level of Takanari (Fig. 2). But the value of A_n in a plant carrying Takanari alleles of both qHP4 and qHP10 in the Koshihikari background increased to the level of Takanari (Fig. 6A), as it did in a plant carrying Habataki alleles of both qCAR4 and qCAR8 (Adachi et al., 2014). Similar pyramiding effects of Koshihikari alleles on A_n were observed in plants with the Takanari background (Fig. 6B–D). These results suggest that QTL effects on A_n are additive in introgression lines with both the Koshihikari background and the Takanari background, and in QTLs of type I as well as type II. Adachi et al. (2013) identified two rice lines with extremely high A_n (BTK-a and BTK-b) among the same BILs (Koshihikari/Takanari//Takanari). From the genotypes and our results, it is clear that BTK-a carries Koshihikari alleles of qHP1b, qHP3b, qHP7a, and qHP7b (Fig. 7). This observation suggests that A_n of plants with a Takanari background could be increased by more than 20% above that of Takanari by introgression of Koshihikari alleles. The genetic basis of both types I and II QTL expression should be a target for future research to enhance A_n.

Fig. 7.

Graphical genotype of line BTK-a, with extremely high A_n, which was identified previously among the BILs (Koshihikari/Takanari//Takanari) by Adachi et al. (2013). Black, homozygous for Koshihikari; white, homozygous for Takanari. Boxes: QTLs for A_n located on the Koshihikari homozygous region in BTK-a. A_n, initial slope of the A_n–C_i curve, and A_sat of BTK-a were ~25–30% higher than those of Takanari (Adachi et al., 2013).

Causes of the increase in A_n by Koshihikari and Takanari alleles

The rate of photosynthesis in rice is controlled by g_s, g_m, the capacity of Rubisco to consume RuBP, and the capacity of RuBP regeneration (Farquhar et al., 1980; Terashima et al., 2011). Takanari had a higher g_s, higher initial slope of the A_n–C_i curve, and higher LNC than Koshihikari (Fig. 5). The initial slope of the A_n–C_i curve is associated with the combination of Rubisco capacity and g_m (Sharkey et al., 2007; Adachi et al., 2013). We suggest that Takanari has a higher A_n than Koshihikari owing to its higher g_s and its higher capacity of Rubisco to consume RuBP (Hirasawa et al., 2010), although we did not examine g_m. The Takanari alleles at qHP4 were associated with the increased g_s, initial slope of the A_n–C_i curve, and A_sat in SL1217, and that at qHP10 was associated with the increased g_s in NIL10 (Table 2). The increases in g_s and initial slope of the A_n–C_i curve are likely due to the Takanari alleles. However, the higher A_sat might be due to the combination of the Takanari allele of qHP4 with the Koshihikari genetic background, because Takanari did not have a significantly higher A_sat than Koshihikari (Fig. 5).

Although g_s, the initial slope of the A_n–C_i curve, and LNC were smaller in Koshihikari than in Takanari, and there was no difference in A_sat between Koshihikari and Takanari (Fig. 5), some Koshihikari alleles increased g_s (at qHP1a and qHP7b), the initial slope of the A_n–C_i curve (at qHP1a, qHP1b, qHP3b, and qHP7b), and A_sat (at qHP1a, qHP1b, qHP3b, qHP7a, and qHP7b) in the plants with a Takanari genetic background grown in pots (Table 2). Some of the increase might be caused by the increase in LNC, as reported previously (Makino et al., 1994; Yamori et al., 2011). It is interesting to note that the correlation between LNC and A_n was lower in K-CSSLs than in T-CSSLs, or was not significant (Supplementary Fig. S4). The combination of a Koshihikari allele with the Takanari background may enhance the initial slope of the A_n–C_i curve or A_sat without increasing LNC (Table 2). Increased g_m was a cause of the much higher A_n in BTK-a and BTK-b than in Takanari (Adachi et al., 2013).

Phenology may affect A_n through weather conditions and soil properties. However, in the current study, the difference in days to heading from sowing (DTH) between Koshihikari and plants carrying QTLs for enhancing A_n in the Koshihikari background and between Takanari and plants carrying QTLs enhancing A_n in Takanari background was within 11 d (see Supplementary Figs S9, S10). SL1208 showed heading 10–11 d later than Koshihikari and, on the contrary, SL1311 and SL1324 showed heading 7–10 d earlier than Takanari in 2011 and 2012 (Supplementary Figs S9, S10). No significant correlations were observed between DTH and A_n except for T-CSSLs in 2012 (Supplementary Fig. S11). These results suggest that A_n measured in the current research was affected not by differences in phenology, but by genetic properties.

Our findings suggest that A_n and traits relevant to A_n (g_s, g_m, Rubisco capacity, or RuBP regeneration) can be improved beyond the level of the parent cultivar by combining a Takanari allele with the Koshihikari genetic background or a Koshihikari allele with the Takanari genetic background. The genetic basis of the effects of the QTL alleles and genetic background on diffusion and biochemical capacities should be a target in future research.

Effects of the QTLs on dry matter and grain production

The enhanced A_n is expected to increase dry matter and grain production through increased carbon gain of plants. In fact, NIL10, a line carrying a qHP10 region in the Koshihikari background, produced heavier dry matter of above-ground parts at harvest and showed higher grain yield than in Koshihikari across two years (Table 3). However, if other traits relevant to dry matter production and grain yield are affected by the QTL for A_n, known as pleiotropic effects, dry matter production and grain yield also might be affected by them. It was shown in the previous report that the A_n-enhancing QTL, qCAR8, promoted the earlier heading resulting in shorter growth period and as a result, produced small dry matter at harvest (Adachi et al., 2017). Similarly, in this research, SL1311 and SL1324 displayed earlier heading than Takanari and showed no enhancement of dry matter production. Plant height affects dry matter production through canopy architecture (Ookawa et al., 2010). SL1217 with shorter canopy produced smaller dry matter than Koshihikari and, on the other hand, SL1303 with taller canopy produced heavier dry matter than Takanari (Table 3). Dry matter production is a highly complex trait, which is determined not only by A_n as measured in the current research, but also by phenology, leaf area and architecture of the canopy (Gardner et al., 1985; Hay and Porter, 2006). To elucidate how each QTL allele identified here is associated with dry matter production and grain yield, detailed investigation on leaf area development, which affects light interception of the canopy, and canopy architecture, which affects light penetration and CO₂ diffusion into the canopy as well as individual leaf photosynthesis in the canopy, are required (Hirasawa, 2014).

Toward a comprehensive understanding of the genetic control of rice A_n

Several QTLs for enhanced A_n have been reported previously (Teng et al., 2004; Hu et al., 2009; Adachi et al., 2011a; Gu et al., 2012). Except for QTLs (qHP4 and qHP5) identified by using relevant mapping populations (Adachi et al., 2011a), no QTLs for A_n or traits relevant to A_n that have been identified using different mapping populations to those used here (Ishimaru et al., 2001; Ogawa et al., 2016) share the same position as those identified here. This is probably because of the strong dependence on the combination of allele and genetic background. The comprehensive information on the genetic components of A_n in Koshihikari and Takanari that we have identified will be useful for understanding the genetic basis of rice A_n. The first step will be to clone the genes underlying the QTLs to understand the molecular mechanisms of their functions. The QTLs detected repeatedly over several years (qHP1a, qHP1b, qHP3b, qHP7a, qHP7b, and qHP10) would be good targets for fine mapping by MAS. The second step will be to develop predictive genetic models for selecting plants with appropriate combinations of genes. Genomic selection, a newly developed strategy for crop selection on the basis of haplotype, relies on a training model developed from phenotypic and genotypic data (Heffner et al., 2009; Jannink et al., 2010). QTLs for g_s and LNC, which did not always correspond to QTLs for high A_n (Fig. 4C, D), could be incorporated in connection to the higher A_n in the model. QTL information presented here could be used to improve prediction models. Furthermore, ‘omics’ approaches, including transcriptome and metabolome analyses, can bridge the gap between genomes and phenotypes (Kuroha et al., 2017). Information on the causal genes will help to explain how the genome controls A_n in field-grown plants. For a comprehensive understanding of the genetic components of rice A_n, using mapping populations from other combinations of cultivars and genome-wide association study are promising. Our study, which demonstrated several types of QTL for A_n based on genetic and environmental background, will provide a guide for future studies evaluating the applicability of estimated QTLs to rice breeding programs.

Conclusions

We found several reliable QTLs for enhanced A_n in rice. Pyramiding the QTLs greatly enhanced A_n. A_n was increased in the high-quality japonica cultivar Koshihikari with a low A_n by introducing alleles from the high-yielding indica cultivar Takanari with a high A_n; furthermore, the A_n of the high-A_n cultivar Takanari could be increased further by introducing alleles of the low-A_n cultivar Koshihikari. However, the effects of many QTLs for enhanced A_n were limited to specific combinations of allele and genetic background. Prediction modeling and gene mapping would contribute to the selection of plants or genomes for further increasing A_n. By characterizing the effects of each QTL on A_n and other traits associated with rice production, such as leaf area development, canopy structure, and earliness of heading, breeders could use alleles that enhance A_n in programs to improve rice productivity.

Supplementary data

Supplementary data are available at JXB online.

Fig. S1. Genotype and photosynthetic rate of lines used for narrowing down QTL regions on chromosomes 7 and 10.

Fig. S2. Frequency distributions of stomatal conductance and leaf nitrogen content of BILs with Koshihikari background and Takanari background in 2009.

Fig. S3. Relationships between stomatal conductance and photosynthetic rate (A_n), and between leaf nitrogen content and A_n, in BKTs and BTKs.

Fig. S4. Relationships between stomatal conductance and photosynthetic rate (A_n), and between leaf nitrogen content and A_n, in T-CSSLs and K-CSSLs.

Fig. S5. Stomatal conductance of flag leaves at full heading stage of T-CSSLs (Koshihikari background).

Fig. S6. Leaf nitrogen content of flag leaves at full heading stage of T-CSSLs (Koshihikari background).

Fig. S7. Stomatal conductance of flag leaves at full heading stage of K-CSSLs (Takanari background).

Fig. S8. Leaf nitrogen content of flag leaves at full heading stage of K-CSSLs (Takanari background).

Fig. S9. Days to heading from sowing of T-CSSLs (Koshihikari background).

Fig. S10. Days to heading from sowing of K-CSSLs (Takanari background).

Fig. S11. Relationships between days to heading from sowing and photosynthetic rate in T-CSSLs and K-CSSLs in 2011 and 2012.

Table S1. Averaged value of photosynthetic rate, stomatal conductance and leaf nitrogen content and genotypes for the BILs.

Table S2. Heading date of Koshihikari and Takanari.

Table S3. Genomic locations of QTLs for enhanced flag leaf photosynthesis estimated by using CSSLs and by narrowing down with further crosses.

Table S4. QTLs for stomatal conductance estimated by using BILs.

Table S5. QTLs for leaf nitrogen content estimated by using BILs.

Glossary

Abbreviations

A_n

net photosynthetic rate

A_sat

net photosynthetic rate at CO₂ saturation

C_a

ambient CO₂ concentration in the assimilation chamber

C_i

intercellular CO₂ concentration

DTH

days to heading from sowing

g_s

stomatal conductance

g_m

mesophyll conductance

LAI

leaf area index

LNC

leaf nitrogen content

MAS

marker-assisted selection

QTL

quantitative trait locus

RuBP

ribulose-1,5-bisphosphate.