Plasma membrane H+-ATPase overexpression increases rice yield via simultaneous enhancement of nutrient uptake and photosynthesis

Nitrogen (N) and carbon (C) are essential elements for plant growth and crop yield. Thus, improved N and C utilisation contributes to agricultural productivity and reduces the need for fertilisation. In the present study, we find that overexpression of a single rice gene, Oryza sativa plasma membrane (PM) H+-ATPase 1 (OSA1), facilitates ammonium absorption and assimilation in roots and enhanced light-induced stomatal opening with higher photosynthesis rate in leaves. As a result, OSA1 overexpression in rice plants causes a 33% increase in grain yield and a 46% increase in N use efficiency overall. As PM H+-ATPase is highly conserved in plants, these findings indicate that the manipulation of PM H+-ATPase could cooperatively improve N and C utilisation, potentially providing a vital tool for food security and sustainable agriculture.

N itrogen (N) and carbon (C) are indispensable elements for plant growth and are required in large quantities for crop production 1 . Crops largely obtain N from the soil as NH 4 + and/or NO 3 − and C from the atmosphere as CO 2 . Synthetic N fertilisers are also applied in large amounts, with annual rates of >120 million tons worldwide 2 . Crops have a limited ability to utilise N 3 ; thus, excess N is continuously lost from agricultural systems, which pollutes the environment 4 . In addition, the productivity of C 3 plants such as rice and wheat is limited by inefficient CO 2 fixation by RuBisCO during photosynthesis, due to low CO 2 concentrations within the mesophyll cells of leaves. Plant biomass and crop production can be improved by the enhancement of intercellular CO 2 concentration, which creates an effect similar to CO 2 fertilisation 5 , but also emits excess CO 2 into the atmosphere 6 . Thus, it is critically important to determine how best to enhance N and CO 2 uptake by plants to improve crop production and environmental performance.
Plasma membrane (PM) H + -ATPase, a subfamily of P-type ATPases, generates a membrane potential and H + gradient across the PM, energising multiple ion channels and various H + -coupled transporters for diverse physiological processes 7,8 . In previous studies, we demonstrated that the PM H + -ATPase mediates light-induced stomatal opening 9,10 . Overexpression of the PM H + -ATPase in guard cells significantly enhances stomatal opening, photosynthesis and, subsequently, growth in Arabidopsis thaliana, a model plant 11 . It remains unknown if this manipulation would be efficient in crops, such as rice, which is the staple food for three billion people worldwide 12 .
Unlike most terrestrial plants, paddy rice grows in flooded soils where ammonium (NH 4 + ) ions constitute the dominant N source for root uptake 13 . To use NH 4 + as an N source, rice roots require efficient uptake ability and corresponding assimilation capacity for NH 4 + . Conversely, high tissue accumulation of unassimilated NH 4 + is usually negatively correlated with plant growth 14,15 . The assimilation of NH 4 + in root cells requires a C skeleton as the substrate for the synthesis of amino acids through the glutamine synthetase (GS)/glutamate synthase (GOGAT) cycle. The assimilation of one molecule of NH 4 + generates two molecules of H + in the cytoplasm 16 . PM H + -ATPase facilitates the transport of various nutrients, such as nitrate, phosphate and potassium (K + ) 17,18 , and maintains cytosolic H + homeostasis by pumping H + outside the cells 19 . In our previous study, NH 4 + nutrition was found to induce upregulation of PM H + -ATPase activity in rice roots 20 . Recently, we determined that enhanced PM H + -ATPase activity in rice roots ensures rice growth at high NH 4 + concentrations 21 . Therefore, we hypothesised that PM H + -ATPase may be involved in NH 4 + metabolism in rice plants.
In this study, we examined the involvement of PM H + -ATPase in NH 4 + uptake by rice roots and stomatal opening for CO 2 uptake and photosynthesis in rice leaves, with the aim of developing a new strategy to improve rice yield and N use efficiency (NUE) via the overexpression of a single gene, Oryza sativa PM H + -ATPase 1 (OSA1).

Results
PM H + -ATPase mediates NH 4 + absorption. We first investigated the relationship between PM H + -ATPase and NH 4 + uptake by rice roots. We treated rice roots with the fungal toxin fusicoccin (FC), a stimulator for PM H + -ATPase activity 22 , and found that the rate of 15 NH 4 + absorption increased by 17% in darkness and by 11% under illumination, compared to the corresponding controls (mock) (Fig. 1a). These results clearly indicate that PM H + -ATPase is involved in NH 4 + uptake in rice roots. Under illumination, we also observed an additional increase in the 15 NH 4 + absorption rate (Fig. 1a) and induction of leaf stomatal opening for transpiration (Fig. 1b). These results suggest that enhanced transpiration in leaves also contributes to NH 4 + uptake by roots. Therefore, we inferred that the overexpression of PM H + -ATPase in rice roots and/or stomatal guard cells would efficiently improve NH 4 + absorption.
Overexpression of PM H + -ATPase enhanced NH 4 + uptake. To understand the effects of PM H + -ATPase overexpression on N uptake, we compared the isotopic 15 N ( 15 NH 4 + ) absorption rate between WT and OSA1-oxs (or osa1 mutants), and determined the absorption rate of 15 NH 4 + within 5 min by roots. 15 NH 4 + concentrations ranging from 0.5 to 8 mM were used to test 15 NH 4 + uptake via different NH 4 + transport systems in rice roots. Interestingly, the 15 NH 4 + absorption rate in all OSA1-oxs was significantly higher than that in the WT, under both low (≤1 mM) and high (≥1 mM) NH 4 + concentration conditions involving the high-and low-affinity transport systems, respectively 24 (Fig. 2f). By contrast, all osa1 mutants exhibited lower 15 NH 4 + absorption rates under all NH 4 + concentration conditions (Fig. 3f). We also examined the 15 NH 4 + absorption rate within 30 min under 2 mM 15 NH 4 + . The rate of 15 NH 4 + absorption was 20-30% higher in OSA1-oxs than in the WT, but was markedly lower in osa1 mutants ( Supplementary Fig. 5). In all rice lines, 15 NH 4 + absorption rates were significantly repressed by treatment with 0.35 µM vanadate, an inhibitor for PM H + -ATPase ( Supplementary Fig. 5). These results confirmed that PM H + -ATPase modification in rice roots regulated NH 4 + absorption. Consequently, under laboratory hydroponic conditions, total N accumulation was found to be 16-57% higher in OSA1oxs ( Fig. 2g) but lower in osa1 mutants (Fig. 3g) compared to the WT. In addition, the contents of other nutrients such as K, P, Ca, S, Fe, and Zn were also increased in OSA1-oxs and decreased in osa1 mutants compared to the WT (Supplementary Fig. 6a-c, f, j). Interestingly, total C accumulation was 21-47% higher in OSA1oxs but lower in osa1 mutants compared to the WT (Figs. 2h and 3h). Because C is not taken up by plant roots, these results suggest that OSA1 modification influenced CO 2 uptake and/or fixation in rice leaves.
PM H + -ATPase overexpression enhanced stomatal conductance and photosynthetic activity. Stomata are crucial for gas exchange, particularly for CO 2 diffusion into the leaf 12 . Light, the most effective environmental signal for stomatal opening, then activates PM H + -ATPase 10,11,[25][26][27][28] . PM H + -ATPase-induced hyperpolarisation in the PM of guard cells enables K + uptake through inward-rectifying K + channels. The accumulation of K + and its counter ions in guard cells prompts guard-cell swelling and stomatal opening 29 . Therefore, we investigated stomatal phenotypes in OSA1-oxs. Representative closed and open stomata in a WT rice leaf are shown in Fig. 4a. In darkness, the level of stomatal closure in OSA1-oxs was similar to that in the WT, whereas under light, the ratio of open to closed stomata was significantly higher in OSA1-oxs (Fig. 4b). Conversely, in osa1 mutants, the ratio of open to closed stomata was significantly lower than in the WT under light treatment ( Supplementary  Fig. 7a). In all rice lines, stomatal opening was suppressed by the plant hormone abscisic acid (ABA) ( Fig. 4b and Supplementary  Fig. 7a), suggesting that ABA action was unaffected in guard cells of both OSA1-ox and osa1 mutant plants. Stomatal density, size, and shape in OSA1-oxs and osa1 mutants were comparable to those of the WT ( Supplementary Fig. 8), suggesting that overexpression or mutation of PM H + -ATPase in rice had no effect on stomatal morphology or development; these results were similar to our observations in Arabidopsis thaliana 12 .
Given that stomatal aperture is a limiting factor for photosynthesis 12,30 , we examined the photosynthetic properties of OSA1-ox plants. Under saturated white light (WL) conditions, stomatal conductance in OSA1-oxs was almost double that in the WT ( Fig. 4c and Supplementary Table 2), and photosynthetic rates in OSA1-oxs were 26-28% higher than in the WT ( Fig. 4d and Supplementary Table 2), indicating that enhanced lightinduced stomatal opening in OSA1-oxs conferred higher photosynthesis rates. By contrast, osa1 mutants exhibited 22-37% lower stomatal conductance and 27-35% lower photosynthetic rates ( Supplementary Fig. 7b, c). Next, we examined photosynthetic light response curves in detail. Along with increased stomatal conductance (Fig. 4e), the photosynthetic rates of OSA1-ox plants were 15-34% higher than those of the WT (Fig. 4f), particularly under high-intensity light (500-1500 µmol m −2 s −1 ). Photosynthetic CO 2 response curves (A-Ci curves) were also higher for OSA1-oxs than for the WT (Fig. 4g), indicating a higher photosynthetic capacity among OSA1-ox plants. The water use efficiency of OSA1-oxs was 13-21% lower than that of the WT (Supplementary Table 2).
Genome-wide effect of OSA1 on gene expression. To identify differentially expressed genes (DEGs) and associated pathways that may provide a molecular basis for the described OSA1-ox and osa1 mutant phenotypes, we analysed the comprehensive gene expression profiles in the leaves and roots of 4-week-old WT, OSA1-ox (OSA1#2) and osa1-2 mutant plants using RNAsequencing (RNA-seq) analysis. Among the DEGs, 1373 and 1124 transcripts were upregulated in the leaves and roots of the OSA1ox line, and 347 and 3295 transcripts were downregulated in the leaves and roots of the osa1-2 mutant, respectively (Fig. 5a). By contrast, 1895 and 1304 transcripts were downregulated in the leaves and roots of the OSA1-ox line, and 1859 and 2913 transcripts were upregulated in the leaves and roots of the osa1-2 mutant, respectively (Supplementary Fig. 9a-c). Consistent with OSA1 expression levels, we detected 59 and 82 genes in the leaves and roots, respectively, that were upregulated in the OSA1-ox line but downregulated in the osa1-2 mutant (Fig. 5a).
We then performed Gene Ontology (GO) term enrichment analysis of the DEGs upregulated in the OSA1-ox line and downregulated in the osa1-2 mutant to investigate the molecular mechanisms underlying OSA1-mediated biological processes ( Supplementary Fig. 9d, e and Supplementary Data 1). The results indicated that 12 biological processes were significantly enriched, including photosynthesis, NH 4 + assimilation, glutamate biosynthesis, amino acid metabolism, carbohydrate transmembrane transport, various ion transport, and N utilisation ( Supplementary Fig. 9d, e and Supplementary Data 1). Genes associated with transmembrane transporter activity, ion transport, substrate-specific transmembrane transporter activity, cation transmembrane transporter activity, carbohydrate transmembrane transporter activity, and PM part were also significantly enriched in overlapping genes that were upregulated in OSA1-ox roots and leaves, but downregulated in the osa1-2 mutant (Supplementary Data 2). In addition, we compared leaf and root transcriptomes between the WT, OSA1-ox line, and osa1-2 mutant, and found that genes associated with nucleic acid binding transcription factor activity, response to chitin, response to organonitrogen compound, regulation of N compound metabolic process, regulation of nucleobase-containing compound metabolic process, and RNA biosynthetic process were significantly enriched in the overlapping genes downregulated in OSA1-ox leaves and upregulated in osa1 mutant leaves (false discovery rate [FDR] < 0.05) (Supplementary Data 3). However, no GO terms were found to be significantly enriched in overlapping genes downregulated in OSA1-ox roots and upregulated in osa1-2 mutant roots (Supplementary Data 3).
We examined the expression levels of NH 4 + -responsive genes, including AMT1;1, GS1;2, NADH-GOGAT1, NADH-GOGAT2 and GS2 using quantitative reverse-transcription polymerase chain reaction (PCR) (Fig. 5d-h). The expression of all investigated genes increased significantly in the OSA1-ox lines. Notably, GRF4, a key transcription factor in N metabolism and C fixation in rice 31 , was highly expressed in response to OSA1 overexpression (Fig. 5c).

Roots
Leaves Relative expression of  Fig. 6a, b. At all three locations, grain yield of the OSA1-ox lines was 27-39% (mean, 33%) higher than that of the WT (Fig. 6e and Supplementary  Tables 3-5). Conversely, in osa1 mutants, grain yield was significantly lower than that of the WT at all three locations (Supplementary Tables 3-5). In OSA1-oxs, the higher yield was correlated with higher panicle weight (18-42%) (Fig. 6f), which was attributed to increased numbers of panicles per hill (15-20%) (Fig. 6c, g) and spikelets per panicle (8-16%) (Fig. 6d, h). Plant height, panicle length, filled grain rate and 1000-grain weight were nearly identical between OSA1-ox, osa1 mutant and WT plants (Supplementary Tables 3-5). Similar patterns were observed across fertilisation levels (Fig. 6j, Supplementary Fig. 11 and Supplementary Tables 3-5). Notably, under N-N conditions, grain yield was 12-20% higher in OSA1-oxs than in the WT at all test locations ( Fig. 6j and Supplementary Tables 3-5). The NUE of OSA1-oxs was~46% higher than that of the WT at all N fertilisation levels (Fig. 6i). Even when treated with only half the amount of N fertiliser (L-N, 100 kg ha −1 ), the grain yield of the OSA1-ox lines was significantly higher than that of the WT grown under M-N conditions (200 kg ha −1 ) (Fig. 6j). Thus, the same grain yield was attained using only half the amount of N fertiliser when the WT was replaced with OSA1-oxs.
To further verify the practical outcome of OSA1 overexpression in rice, we conducted an independent field trial in Hainan, southern China, which has a tropical climate and short-day conditions, and therefore produces lower yield than the  Table 6).

Discussion
Increasing crop yield by improving NUE and C fixation is important for sustainable agriculture and environment performance. In this study, we demonstrated the critical role of the PM H + -ATPase gene OSA1 in controlling both NUE and photosynthesis in paddy rice production. Overexpression of OSA1 in rice plants increased the activity of PM H + -ATPase (Fig. 2e), promoted NH 4 + uptake and assimilation in roots (Fig. 2f, g) and enhanced light-induced stomatal opening and stomatal conductance and photosynthetic rate under saturated WL in leaves (Fig. 4b-d and Supplementary Table 2), leading to higher NUE and grain yield (Fig. 6). Our results demonstrate the cooperative enhancement of NH 4 + metabolism, photosynthesis rate and grain yield through the expression modulation of a single PM H + -ATPase gene in rice plants.
PM H + -ATPase was found to regulate NH 4 + uptake in rice ( Fig. 1 and Supplementary Fig. 5). Furthermore, genetic evidence based on OSA1 overexpression/knockout showed that OSA1 modulation regulated the rate of NH 4 + absorption by rice roots across a wide range of rhizosphere NH 4 + concentrations (Figs. 2f  and 3f). RNA-seq analyses revealed the upregulation of at least six NH 4 + transporter genes (AMT3;3, AMT3;1, AMT2;3, AMT2;1, AMT1;2 and AMT1.1) that encode both high-and low-affinity NH 4 + transporters in OSA1-overexpressing lines (OSA1-oxs), and the downregulation of these genes in the osa1 mutant (Fig. 5). These results indicate that there is a close relationship between PM H + -ATPase and NH 4 + transporters in rice root cells. These coordinated expression pattern of different genes is also the fundamental mechanisms that enable OSA1-oxs rice roots to efficiently take up NH 4 + in the field soils with frequently fluctuated NH 4 + concentration. Our results also indicate that OSA1 overexpression may enhance NH 4 + assimilation capacity. Genes responsible for NH 4 + assimilation such as glutamine synthetase (GS1;2 and GS2) and glutamate synthase (NADH-GOGAT1, NADH-GOGAT2 and Fd-GOGAT) were upregulated in OSA1-oxs (Fig. 5b, e-h). Because NH 4 + uptake and assimilation are closely synchronised in plant roots 32 , enhanced GS and GOGAT activity can transfer root-absorbed NH 4 + to amino acids for the synthesis of various N-containing compounds during plant growth and development, which in turn prevent NH 4 + overloading in the root cytoplasm due to the acceleration of NH 4 + uptake in OSA1-oxs (Fig. 2f). However, the process of NH 4 + assimilation also generates H + , which is toxic if excessively  We also observed that OSA1 is involved in C fixation through the regulation of stomatal opening (Supplementary Fig. 12). Stomatal conductance and photosynthetic rates were enhanced in OSA1-oxs due to increased stomatal aperture opening (Fig. 3), compared with rates in the WT and osa1 mutants (Supplementary Fig. 7). This result is consistent with the finding that genes related to photosynthesis were upregulated by OSA1 overexpression and downregulated by OSA1 knockout (Fig. 5b), for example, Psb28, PsbQ, PsaH, PFPA2, PsbR1, GLO4 and RbcS [38][39][40] . Enhanced photosynthesis in OSA1-oxs might also have contributed to the uptake and assimilation of NH 4 + in rice roots by providing more C skeletons and energy for NH 4 + metabolism processes 41 . Therefore, N acquisition and photosynthetic activity are intrinsically linked through overall N and C status in rice plants 42 , resulting in a globally coordinated increase in C and N accumulation in rice plants through OSA1 overexpression (Supplementary Fig. 13). Together, these results show that OSA1 overexpression promotes both C and N uptake and assimilation, which further regulate the expression of genes involved in C and N metabolism and contribute to plant growth and grain yield.
PM H + -ATPase is also involved in the uptake of various nutrients from plant roots by providing proton motive force 17,18,43 . In this study, the stronger acidification in OSA1-ox rice roots ( Supplementary Fig. 2) could provide higher proton motive force in the rhizosphere for the uptake of nutrients. This is consistent with the enhanced NH 4 + uptake rate in OSA1-oxs as compared with WT plants (Fig. 2f-g and Supplementary Fig. 6), and also consistent with upregulating the expression of various nutrient transporter genes, such as ammonium transporters AMT1;1/1;2/2;1/3;1/3;3, phosphate transporter PT1/PT8/PHO1.1 and potassium transporter HAK1 in roots of OSA1-oxs (Fig. 5b). These results coincided with the increased contents of N, P and K in OSA1-oxs as compared with WT plants (Fig. 2g and Supplementary Fig. 6). Recently, GRF4, a transcription factor in rice, was reported to integrate N assimilation, C fixation and plant growth; multiple N metabolism genes, such as AMT1;1, GS1;2, GS2 and NADH-GOGAT2, are positively regulated by GRF4 31 . Here, GRF4 was found to be upregulated by OSA1 overexpression (Fig. 5b, c). It is possible that some master regulators of nutrient uptake and metabolism, such as GRF4, could be activated by OSA1 overexpression. The enhanced C fixation and N metabolism could also have a feedback on the expression of nutrient transporter genes in order to ensure sufficient supply of nutrients for the promotion of the plant growth. Further study is deserved to investigate the underlying molecular mechanisms responsible for the signal transduction initiated by OSA1 overexpression.
In contrast to CO 2 , which is taken up from the atmosphere, N is derived from fertilisers for most non-legume crops. Thus, cultivars with improved NUE are in urgent demand for the sustainable development of agriculture. The green revolution has boosted crop yields; however, the resulting cereal varieties are associated with reduced NUE 44 . Even precision crop management has led to only a slight improvement in NUE 3 . In this study, OSA1-ox rice exhibited both higher yield and higher NUE than the WT under a wide range of N fertilisation rates, from 0 to 300 kg N ha −1 (Fig. 6i, j and Supplementary Tables 3-5). Higher NUE in OSA1-ox rice leads to a lower demand for N fertilisers to produce similar yields of rice grain. To obtain the same yield as WT rice, OSA1-ox rice requires only half the amount of N fertiliser (Fig. 6j). This benefit will drastically reduce the cost of rice production as well as the environmental load produced by excess N accumulation due to rice production.
Given that the molecular mechanisms of nutrient uptake 17,43 and light-induced stomatal opening 27 are conserved in most plant species, this manipulation strategy could be applicable to many valuable crops. Therefore, we suggest designating plants overexpressing PM H + -ATPase as promotion and upregulation of plasma membrane proton-ATPase (PUMP) plants. If PM H + -ATPase overexpression can be realised using non-transgenic methods such as genome editing, these crops could have great potential for commercial use, conferring greater yields and potentially critical environmental benefits. For the gas-exchange experiment, seedlings of the WT, OSA1-overexpressing lines and osa1 mutants were grown in ½ IRRI nutrient solution for 1 week, followed by 5 more weeks of growth in modified IRRI nutrient solution (pH 5.5) containing 2 mM NH 4 Cl. The solution in the containers was replaced every 3 days. Plants for most of the laboratory experiments were grown in a greenhouse at 30°C/24°C (day/night) and 60-80% relative humidity. Plants for stomatal aperture and gasexchange measurements (Fig. 4 and Supplementary Figs. 4 and 7) were grown in a growth chamber (NC-410HC, Nippon Medical & Chemical Instruments Co., Ltd, Osaka, Japan) under~150 μmol m −2 s −1 fluorescent light at 30°C/24°C (12 h/12 h) and 60-80% relative humidity. The rice seeds for these experiments were of the same age.

Methods
For field experiments, plants were grown in the summer of 2016 and 2017 at four well-controlled biological experimental stations in Hainan in 2016 (N18°67′, E108°76′), southern Nanjing in 2016 (N32°01′, E118°51′), northern Nanjing in 2017 (N32°11′, E118°46′) and Fengyang in 2017 (N32°52′, E117°33′). Hainan is in a tropical monsoon zone with sandy soil, whereas the experimental sites in Nanjing and Fengyang are in a subtropical monsoon climate zone with yellow-brown soil. For each field experiment (Fig. 6, Supplementary Fig. 11 and Supplementary  Tables 3-6), four levels of N (urea) fertiliser were applied: 0, 100, 200 and 300 kg N ha −1 (N-N, L-N, M-N and H-N). Seeds were germinated and seedlings were grown in a greenhouse for~1 month at the beginning of May. The rice seedlings were hand-transplanted in a flooded field with regular hill spacing. Each fertilisation treatment was performed in one plot (6.5 m × 4 m). Rice seedlings were planted in 14 rows with 20 hills per row, for a spacing of 25 and 20 cm, respectively. Each OSA1-ox, osa1 mutant and WT line was planted in three rows (excluding border hills). At the edge of each plot, the same rice line of inside neighbour was also planted as the border hills (red box) to avoid the margin effects on the rice growth inside the plot. Each plot contained 480 hills, for a total of 1920 hills. Each field experiment consisted of four plots with different N fertilisation levels. Prior to Fig. 6 Overexpression of OSA1 increases grain yield and N use efficiency (NUE) in the field. a-d Photographs of 100-day-old WT and OSA1-ox plants in the field (a), in pots in the field (b) and harvested panicles (c) and spikelets (d) in 2017 in northern Nanjing under 200 kg N ha −1 (M-N) fertilisation. e Grain yield, f panicle weight per plant, g panicles per hill and h spikelets per panicle of WT and OSA1-ox plants in field tests at three locations (n ≥ 6). i Relative agronomic NUE in WT and OSA1-ox plants in field tests under low (L-N; 100 kg N ha −1 ), moderate (M-N; 200 kg N ha −1 ) or high (H-N; 300 kg N ha −1 ) levels of N fertilisation. Columns and error bars represent the means ± SEs (n = 3). j Grain yield of WT and OSA1-ox plants in field tests under different N conditions. Black asterisks represent significant differences between WT and OSA1-ox plants under the same N fertilisation level; small circles in e-h, j represent data points of collected samples in individual experiments (n = 6 in 2017 Nanjing-N and n = 8 in 2016 Nanjing-S and 2017 Fengyang). Centre line indicates the median, upper and lower bounds represent the 75th and the 25th percentile, respectively. Whiskers indicate the minimum and the maximum in the box plots (e-h, j). Differences were evaluated using the two-tailed Student's t test (*P < 0.05; **P < 0.01; n.s., not significant). The exact P values are provided in the Source Data file. seedling transplantation, the paddy field was fertilised with 80 kg P ha −1 as Ca (H 2 PO 4 ) 2 and 110 kg K ha −1 as K 2 SO 4 . The first N fertilisation was carried out at 2 days before transplantation using 33.3% of the total amount of N fertiliser, which was mixed into the soil. At the tillering stage (~1 week after transplanting), the second N fertilisation (33.3% of the total N) was carried out. The final N application (33.3% of the total N) was conducted 4 weeks later. The plant growth period (transplantation to harvest) differed among rice lines and N levels, as follows. At 0 or 100 kg N ha −1 , the growth period was 109 ± 3 days for the WT and osa1 mutant and 102 ± 2 days for the overexpression lines; at 200 or 300 kg N ha −1 , the growth period was 119 ± 2 days for the WT and osa1 mutant and 112 ± 2 days for the overexpression lines.
Grain yield was determined at harvest in October. At maturity, 6-8 hills of plants from each rice line were randomly selected at the centre of the plot from among a 22 × 18 array of hills (excluding the border hills) and harvested. Yield and its components were determined 45,46 with minor modifications. The samples were divided into grain and straw for nutrient content analysis. Agronomic NUE was defined as the yield increase per kg N fertiliser in the field experiment. Relative agronomic NUE (Fig. 6i) was calculated as the ratio to WT rice under L-N treatment in each field trial. Statistical analyses were performed using two-tailed Student's t tests and one-way analysis of variance followed by Tukey's test.

Construction of the overexpression vector and transgenic plants.
The openreading frame of OSA1 was amplified using gene-specific primers (Supplementary Table 7). The fragment was treated with restriction enzymes, inserted into vectors and sequenced before transformation. Embryonic rice (O. sativa) calli were transformed via Agrobacterium-mediated transformation 47 . Three independent homozygous T2 or T3 lines (OSA1#1, OSA1#2 and OSA1#3) were used for all phenotypic analyses.
Quantitative reverse-transcription PCR. Total RNA was isolated from the roots of WT and transgenic plants using TRIzol reagent according to the manufacturer's instructions 48 (Invitrogen Life Technologies, Carlsbad, CA, USA). Quantitative PCR was performed using an SYBR Premix Ex Taq II (Perfect Real Time) Kit (TaKaRa Biotechnology, Dalian, China) on a Step One Plus Real-Time PCR System (Applied Biosystems, Bio-Rad, CA, USA), and the data were analysed using the 2 −ΔΔCT method. The OsActin and OsGAPDH genes were used as internal references to normalise the test gene expression data. All analyses were repeated at least three times. PCR primer sets for gene amplification are listed in Supplementary Table 7.
Immunodetection. Leaf and root samples were harvested separately. The samples were immediately homogenised in liquid N and then in ice-cold homogenisation buffer with a mortar and pestle 49 . The membrane proteins were collected by centrifugation and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and immunoblot analysis; PM H + -ATPase was detected using anti-H + -ATPase antibody 50 . Actin was used as an internal control protein and was detected using anti-actin antibodies (1:3000 dilution, Sigma-Aldrich, St. Louis, MO, USA, Cat#057M4548). Relative PM H + -ATPase levels (Figs. 2d and 3d and Supplementary Fig. 4c) were estimated from the ratio of the signal intensity of PM H + -ATPase to that of actin from the same sample. WT and OSA1-ox seedlings were grown in IRRI nutrient solution containing 2 mM NH 4 + for 3 weeks in a growth chamber. Immunohistochemical detection of PM H + -ATPase was performed 9 (Supplementary Fig. 3). Roots of 3-week-old seedlings were harvested separately and placed in fixation buffer (4% paraformaldehyde, 60 mM sucrose and 50 mM cacodylic acid; pH 7.4) for 2 h at room temperature. Fixed samples were washed five times with phosphate-buffered saline (PBS) and embedded in 5% agar dissolved in PBS. Sections of 100-mm thickness were prepared using a vibratome (ZERO1; Dosaka EM, Kyoto, Japan) and placed on a glass slide. Samples were treated with enzyme solution (0.1% pectolyase and 0.3% Triton X-100 in PBS) for 2 h, washed five times with PBS, washed once with blocking solution (5% bovine serum albumin) for 10 min and incubated with primary antibody (anti-H + -ATPase) diluted 1000-fold in PBS overnight. On the second day, samples were washed five times with PBS, washed in blocking solution for 10 min and incubated with secondary antibody (Alexa 546, diluted 1000-fold in PBS) for 2 h. Finally, the samples were observed under confocal laser scanning microscopy (FV-10i; Olympus, Tokyo, Japan).
Measurement of PM H + -ATPase activity. We determined ATP hydrolytic activity of PM H + -ATPase 51 . Root and leaf tissues were ground in ice-cold homogenisation buffer to isolate the PM in a two-phase partitioning method 51 . PM H + -ATPase hydrolysis activity was determined as the difference between assay results with and without the addition of 0.1 mM Na 3 VO 4 to the reaction solution (Figs. 2e and 3e). The assay was performed in 0.5 mL of reaction solution containing 30 52 . Colour development was completed after 30 min and measured spectrophotometrically at 720 nm. In each test, H + -ATPase activity was calculated as the amount of phosphate liberated within 30 min mg −1 membrane protein in excess of the boiled-membrane protein control. 15 N absorption rates in roots of WT, OSA1-overexpressing and osa1 plants. WT, OSA1-ox and osa1 mutant seedlings were grown in IRRI nutrient solution containing 2 mM NH 4 + for 4 weeks in a growth chamber (NC-410HC, Nippon Medical & Chemical Instruments Co., Ltd.) under~150 μmol m −2 s −1 fluorescent light at 30°C/24°C (12 h/12 h) and 60-80% relative humidity. To determine 15 NH 4 + absorption rates within 30 min, seedlings were rinsed in 0.1 mM CaSO 4 for 1 min, transferred to modified IRRI nutrient solution containing 2 mM ( 15 NH 4 ) 2 SO 4 (atom% 15 N: 98%) incubated with mock 5 μM FC (Fig. 1a) or 350 μM vanadate ( Supplementary Fig. 5) for 30 min 53 and rinsed again with 0.1 mM CaSO 4 for 1 min 48 . To determine 15 NH 4 + absorption rates within 5 min in roots of WT, OSA1-ox and osa1 mutant plants under different NH 4 + concentrations, seedlings were incubated with 0.5, 1, 2, 4 and 8 mM 15 NH 4 + for 5 min (Figs. 2f and 3f).
Roots and shoots were separated for weighing, and then immediately frozen in liquid N 2 . After grinding, an aliquot of the powder was dried to a constant weight at 70°C, and 10 mg of each sample was analysed using the MAT253-Flash EA1112-MS system (Thermo Fisher Scientific, Inc., USA). Each experiment was performed with three independent biological replicates, and statistical analyses were performed using two-tailed Student's t tests.
Nutrient element analysis of plant samples. Roots and leaves were harvested from 6-week-old plants, washed three times with tap water and rinsed twice (5 min each) with deionised water to remove any adhering nutrients. The leaves and roots were dried in a forced-air oven at 70°C for~48 h to a constant weight for dry weight measurements (Figs. 2b and 3b and Supplementary Fig. 4a). The dried samples were ground and passed through a 1.0-mm screen. Total N/C contents (Figs. 2g, h and 3g, h) were determined via the dry combustion method using an Element Analyser (vario EL, Elementar, Langenselbold, Germany). For the analysis of mineral elements, the dry biomass was digested in H 2 SO 4 or HClO 4 . P concentrations were determined using the molybdate yellow method and K concentrations were determined by flame emission photometry ( Supplementary  Fig. 6) 54,55 . The other nutrient elements were measured by ICP (Agilent 710 ICP-OES). At least three plants per treatment were harvested, and three independent biological replicates were analysed for each treatment. Figs. 7a and 8c) was quantified as the number of open stomata per total stomata observed. At least 100 stomata were observed per treatment; three biological replicates were analysed for each treatment, and statistical analysis was conducted using two-tailed Student's t tests.
Gas-exchange measurements. Gas-exchange measurements were performed using the LI-6400 System (Li-Cor) with a standard chamber. Light and CO 2 response curves were constructed based on data obtained using measurement processes and light sources 12,56 . The flow rate, leaf temperature and relative humidity were kept constant at 500 μmol s −1 , 24°C and 60-75% (Pa/Pa), respectively. Under each light/CO 2 condition, photosynthetic rate and stomatal conductance data were collected after these values reached a steady state (15-30 min). Fully expanded leaves from 6-week-old plants were used in these experiments (Figs. 1b and 4 and Supplementary Figs. 7b, c and 4d, e). WL was provided by a fibre optic illuminator with a halogen projector lamp (15 V/150 W, Moritex, San Jose, CA, USA) as a light source powered by an MHAB-150W (Moritex) power supply. For CO 2 response curves, leaves were measured at saturating WL conditions (~1500 μmol m −2 s −1 ) (Fig. 4g and Supplementary Fig. 7b, c). To obtain the stomatal conductance and CO 2 assimilation rate data shown in Fig. 4c, d and Supplementary Table 2, leaves were measured under saturating WL conditions (~1000 μmol m −2 s −1 ).
For field gas-exchange measurements ( Supplementary Fig. 10), the flow rate of the Li-6400 system was kept constant at 500 μmol s −1 at a leaf temperature and relative humidity of 28°C and 40-50% (Pa/Pa), respectively. All measurements were performed before the heading stage. At least three plants were selected for measurement, and three biological replicates were analysed for each treatment.
Stomatal density and size. Three to four fully expanded leaves of 6-week-old rice plants were selected. At least five microphotographs were randomly taken of the adaxial or abaxial surface of the leaf lamina. The average stomatal density and size (long axis of each stoma) were calculated 57 (Supplementary Fig. 8a, b).
High-throughput RNA-seq analysis. For RNA-seq analysis, total RNA was extracted from leaves and roots collected from 4-week-old WT, OSA1-ox (OSA1#2) and osa1 (osa1-2) mutant rice lines using a TRIzol Plus RNA Purification Kit (Thermo Fisher Scientific, Waltham, MA, USA). Complementary DNA libraries were constructed using a TruSeq RNA Sample Prep Kit v. 2 (Illumina, San Diego, CA, USA) and sequenced using a NextSeq 500 system (Illumina). Base-calling of sequence reads was performed using the NextSeq 500 pipeline software. Only highquality sequence reads (50 continuous nucleotides with quality values >25) were used for mapping (Fig. 5a, b and Supplementary Fig. 9). Reads were mapped to O. sativa (IRGSP v. 1.0 2019.8.29) transcripts using the Bowtie software 58 . Experiments were repeated three times separately. We obtained 10.1-13.6 million sequence reads per experiment. Gene expression values were reported in RPM (reads per million mapped reads) units. Normalisation of read counts and statistical analyses were performed using the EdgeR software package 59,60 and the Degust Ver. 3.1.0 web tool (http://degust.erc.monash.edu). Obtained RPM values were further analysed using MS Excel software. Only genes with log 2 fold change ≥1 or ≤−1, and an FDR < 0.05 were considered to be significant DEGs. GO term enrichment was conducted using GO Term Enrichment tool in the Plant Transcriptional Regulatory Map (Plan-tRegMap) website 61 (http://plantregmap.gao-lab.org/go.php). GO category (http:// geneontology.org/) FDR ≤ 0.05 was regarded as significantly enriched.
Detection of rhizosphere acidification in roots. Rhizosphere acidification in WT and OSA1-oxs roots was determined 51 . The roots of 7-day-old plants were thoroughly washed with deionised water and spread on an agar sheet containing 0.7% (w/v) agar, 0.02% (w/v) bromocresol purple, 2 mM NH 4 Cl and 1 mM CaSO 4 at pH 5.6. The roots were carefully pressed into the agar to avoid damage. For visualisation of rhizosphere acidification, incubation was conducted in a growth chamber in the dark for 12 h. The relative area of rhizosphere acidification ( Supplementary  Fig. 2b) was estimated as a ratio to the WT area (yellow area on agar sheet; Supplementary Fig. 2a). At least three plants per treatment were harvested, and three independent biological replicates were analysed for each treatment.
Quantification of H + extrusion rate. The H + efflux rates from the WT and OSA1-ox rice roots were measured using the scanning ion-selective electrode technique (SIET System BIO-003A, Younger USA Science and Technology Corp., Applicable Electronics Inc., Science Wares Inc., Falmouth, MA, USA) (Supplementary Fig. 2f) 15,21 . Briefly, seedlings were placed in 50 mL of growth solution with 2 mM NH 4 + for 12 h. Then, the rice roots of 7-day-old plants were washed with deionised water and equilibrated in the measuring solution for 10 min. The equilibrated roots were then transferred to a measuring chamber, which contained 3 mL of a solution comprising 0.2 mM CaCl 2 , 0.1 mM KCl, 0.1 mM NaNO 3 and 0.5 g L −1 MES (2-morpholinoethanesulfonic acid sodium salt) (pH 5.7). At least three plants per treatment were analysed, and three independent biological replicates were performed.