Introduction

With an annual production of approximately 508 million tons (Mt), rice (Oryza sativa L.) feeds nearly 50% of the global population1. As the global population grows, the demand for rice is expected to increase by 28% between 2020 and 20502. Tillering is an important agronomic trait affecting rice yield3 and rice quality. The tiller number of rice plants is positively associated with nitrogen (N) concentration in soil4,5 and one strategy to increase rice tiller number is to increase the activity of N transporter and the expression of assimilation genes6,7. Another strategy involves modulating phytohormones and related genes, such as reducing gibberellin acid (GA) levels and up-regulating the MONOCULM 1 (MOC1) gene in rice8,9. Traditional agricultural methods are insufficient for simultaneously improving N utilization efficiency, hormone signaling, and gene regulation to enhance tillering and yield.

Nano-enabled agricultural technology has revolutionized modern agriculture10,11. The use of engineered nanomaterials (ENMs) shows promising potential to improve crop photosynthesis, yield, quality, and resistance to abiotic stresses12,13. Multiple studies showed that ENMs can not only increase crop uptake of nutrients by affecting the efficiency of fertilizer nutrients in soil and plant absorption14,15,16, but also regulate plant hormones and key gene expression. For instance, applying multi-walled carbon nanotubes (MWCNTs, 10 mg·kg–1) enhanced N-related metabolic enzyme activity in soil and plants, improving N utilization efficiency (NUE) and the nutritional quality of Brassica campestris L. compared to untreated controls. Similarly, soil application of carbon dots (5 mg·kg−1) enhanced the N fixation ability of soybean nodules17, resulting in a 18.5% and 14.8% increase in NUE of soybean shoots and roots, respectively. Moreover, soil application of SiO2 NMs (50 mg·kg−1) enhanced the N content by optimizing root exudates and rhizosphere soil bacterial community, as well as up-regulated the tillering-related gene expressions (e.g., OsMOC, OsFON, and OsTB) by 50.6 − 282.8%18.

Selenium (Se) is beneficial for human health, and appropriate intake of Se can enhance immune and antioxidant functions11,19,20. Se ENMs-enabled agricultural technology is developing rapidly21,22,23. For example, soil application of Se ENMs (0.5 mg·kg–1, Brassica chinensis L. or 0.1 mg·kg−1, Oryza sativa L.) was shown to promote secretions from beneficial rhizosphere microorganisms, improve the effectiveness of nutrient utilization, and consequently facilitate the growth and Se enrichment. Rice (Oryza sativa L.), being a global staple crop, feeds about 56% of the world’s population and sustains 40% of the world’s poor21,23. A systems-based, sustainable increase in rice quantity and quality is imperative for environmental and biodiversity benefits. However, the mechanisms by which Se ENMs enhance Se enrichment and yield remain unclear.

Most studies on ENM-enabled agriculture have been conducted in greenhouses or potted conditions; field trials are needed to validate their effectiveness in real-world soils. This study focused on rice (Oryza sativa L.) and aimed to verify the practicality of Se ENMs-enabled agriculture technology. We hypothesized Se ENMs could regulate the microbial community to increase the abundance of rhizosphere bacteria and promote nitrogen uptake and the expression of tiller-related genes by adjusting phyto-hormone activity. Meanwhile, field experiments applied Se ENMs could not only increase the Se content but also improve the final yield and quality; furthermore, no negative effects on the soil. Therefore, this study aims to (1) elucidate the mechanisms behind Se ENMs effects on rice yield and Se enrichment, (2) validate their effectiveness in open field environmental conditions, and (3) assess their potential for large-scale application in global rice production. In addition, the economic and health benefits arising from the application of Se ENMs in rice production, its optimization of nitrogen fertilizer use, and greenhouse gas emission prediction are evaluated. It is forward-looking to explore further the applicability and scalability of Se ENMs in rice production around the world.

Results and discussion

Se ENMs increase rice yield and Se content in pot experiment

The Se ENMs showed spherical morphology with a size distribution of 50–80 nm (average size, 65 ± 5 nm) (Fig. S1a and S1b). XRD and XPS analyses (Fig. S1c and S1d) showed the Se ENMs are amorphous and their chemical valence state is zero. Se0 ENMs have higher mobility, solubility, prominent bioactivity, and biosafety in plants compared with selenite compounds24. A recent report showed that Se ENMs increased the chlorophyll a and b contents of tomato leaves by 126.3% and 79.5%, respectively, whereas Na2SeO4 only increased them by 40.9% and 2.9% at the same concentration25. Therefore, the as-prepared Se0 ENMs may have potential for developing nano-enabled agriculture technology.

Se ENMs at 0.1, 0.5, and 1 mg‧kg–1 in soil improved rice growth (Fig. S2a and S2b). Upon Se ENMs exposure, the root fresh weight increased by 31.2%, 46.5%, and 31.8%, and the shoot fresh weight increased by 26.7%, 43.5%, and 43.7% (Fig. S2b) compared with control (CK), respectively. The impact on dry weight was even larger: the root dry weight increased by 67.4%, 55.2%, and 46.1%, and shoot dry weight increased by 70.6%, 34.9%, and 72.0% (Fig. S2b), respectively. Non-nano selenite (SeO32‒, Se (IV)) also increased the production of roots and shoots, but the effect was much smaller than that of Se ENMs. Similarly for rice yield, Se ENMs (0.1, 0.5, and 1 mg‧kg–1) led to an increased in grain production by 57.6%, 36.4%, and 40.7% (Fig. S3), whereas Se (IV) only resulted in 11.0% increase. These results thus clearly demonstrated the superiority of the Se ENMs treatment compared to CK and SeO32‒.

To verify that Se ENMs have the same promoting effect in various paddy soils, we selected rice soils from Jilin, Jiangsu, Jiangxi, Chongqing, and Guangdong for further research. Phenotypic images of rice upon different treatments (CK, 0.1 mg·kg‒1 of Se ENMs and 0.5 mg·kg‒1 of Se (IV)) revealed significant differences in the rice growth stage (Fig. S4). As shown in Fig. 1a, rice yield varied significantly across soil type (P < 0.05), but the Se ENMs treatment systematically increased rice yield compared with CK (i.e. by 30.8%, 82.6%, 65.3%, 100.5%, and 114.6% in Guangdong, Jiangsu, Jilin, Chongqing, and Jiangxi, respectively). Conversely, yields for the SeO32‒ treatment was generally similar to CK (Fig. 1a). The images of rice ear and 1000-grain weight also showed that Se ENMs significantly promoted the rice yield (Fig. S5). The results thus demonstrated that the application of Se ENMs significantly boosted rice yield across a range of soils. Both the Se ENMs and SeO32‒ treatment significantly increased the concentration of Se in the grains, but the effects of Se ENMs were superior to those obtained with SeO32‒. Specifically, Se concentration in rice grains treated with Se ENMs in Jiangsu and Jiangxi provinces were 574.9% and 132.5% greater than CK, significantly higher than that the increase observed for SeO32‒ (i.e. 381.7% and 104.1%, (Fig. 1b). In the same two soils, the Se concentration in rice grain reached 63.2 μg·100 g−1 and 72.2 μg·100 g−1 after treatment with Se ENMs, respectively. This level of Se content can well meet the recommended human daily intake of Se25.

Fig. 1: Rice yield and Se content upon different soil treated by Se ENMs and SeO32‒.
figure 1

Rice yield (a) and Se content (b) upon treatment by Se ENMs (0 and 0.1 mg·kg‒1) and SeO32‒ (0.5 mg·kg‒1) of five soils (Guangdong, Jiangsu, Jilin, Chongqing, and Jiangxi). *Represent the significance difference between each treatment group (n = 5, P < 0.05).

Full size image

Overall, the pot experiments demonstrated that Se ENMs can increase rice yield and Se content across a range of different soils. The impact of Se EMNs on yield was much greater than that of SeO32‒.

Mechanisms leading to the promotion of rice yield by Se ENMs

Tillering number is a key factor in determining rice yield26,27. Furthermore, the concentration of N can also increase the tillering number in rice28,29. As shown in Fig. 2a-c, the number of tillers increased by 41.6% upon treatment by Se ENMs (Fig. S6), and the N content of rice root and leaves increased by 4.9% and 48.4% upon exposure to Se ENMs, respectively (Fig. 2b). The results may suggest that the tillering number was associated with increased and the uptake of N by Se ENMs. Microorganisms play a crucial role in regulating the N cycle in soil30, though N absorption and utilization efficiency of rice mainly depends on the content of available N in soil (i.e., ammonia nitrogen (NH4+-N) and nitrate nitrogen (NO3--N))31,32. Alpha diversity (Chao1, Shannon, Simpson, Observed_species) of rice rhizosphere and bulk soil microorganisms were affected by Se ENMs treatment, indicating that Se ENMs were possible to adjust the structure of rhizosphere microorganisms (Fig. S7). Moreover, previous work on Brassica chinensis L. pieces also showed that soil application of Se ENMs affected microorganisms and promoted plant nutrient absorption and utilization efficiency by stimulating the secretion of plant root exudates24. At the genus level, the relative abundance of denitrifying bacteria decreased significantly upon application of Se ENMs, while the abundance of soil N-fixing bacteria (Anaeromyxobacter_sp., Geobacter, Herbaspirillum_sp., and Methylophilus_sp.) increased significantly (Fig. 2c). Anaeromyxobacter_sp. can fix and assimilate N2 in the paddy soil environment33. Geobacter and Methylophilus_sp. can reduce nitrate into ammonium salt34. Herbaspirillum_sp. can colonize rice and contribute to in vivo N fixation35. Therefore, Se ENMs can improve the denitrification process via soil microorganisms to improve the N absorption and utilization in paddy soil.

Fig. 2: Mechanisms leading to the promotion of rice yield by Se ENMs.
figure 2

a Tillering images of rice after treatment by Se ENMs (0.1 mg·kg‒1); (b) Mechanism schematic illustration of increased rice yield N content in rice root, stem, and leaf, relative expression of GA2ox and key tillering genes, and GA1, GA3, and GA4 content in rice (n = 3). c At the genus level, the changes in rhizosphere microbial communities upon exposure to Se ENMs (0.1 mg·kg‒1) (n = 3).

Full size image

Meanwhile, GA 2-oxidases (GA2oxs) are a class of enzymes that inhibit the biosynthesis of bioactive GAs (GA1 and GA4) in plants36. In our study, Se ENMs significantly up-regulated the expression of GA2ox1, GA2ox2, GA2ox3, GA2ox4, and GA2ox6 by 85.4%, 65.2%, 726.8%, 347.0%, and 52.5% compared with CK (Fig. 2b), respectively. Bioactive gibberellic acids (GA1 and GA4) act as plant signaling molecules, and the contents of GA1 and GA4 in rice plants were also reduced by 42.0% and 44.3% in rice treated by Se ENMs (Fig. 2b). GAs can inhibit rice tillering by down- regulating the key tillering genes MOC1, which in turn regulated the expression of genes TB1 and OSH1 to promote tillering27,36. Three key genes (MOC1, TB1, and OSH1) were also significantly up-regulated by 2.5, 8.1, and 10.4 folds in the Se ENMs treatment, respectively (Fig. 2b). These results revealed Se ENMs regulated gibberellic acids hormone synthesis (GA1 and GA4) the expression of tillering genes (MOC1, TB1, and OSH1) in rice. Consequently, Se ENMs largely contributed to the increased tillering number per plant by promoting N uptake, inhibiting the content of active hormones, and up-regulated the key gene expression of tillering.

Role of soil properties on the bioavailability of Se ENMs

The above results confirmed that Se ENMs increased the Se concentration in rice grains in different soils. There was a considerable variation in the Se concentration in rice grains across the different soils (Fig. 1b), suggesting soil properties affect the bioavailability of Se ENMs. While other studies have also shown that soil properties can significantly impact the effectiveness of ENMs37, determining the specific soil properties and processes involved is challenging due to the complexity of soil. A correlation analysis was conducted between paddy soil properties and Se contents (Fig. 3a), and a multiple linear regression (MLR) algorithm was developed with different paddy soil properties as input and rice grain Se content as output (Fig. S8). Soil total organic carbon (TOC) was the key factor affecting Se content in rice negatively (p < 0.001) (Fig. 3b), much likely via a decrease in Se bioavailability. For instance, Guangdong soil exhibited the highest TOC content (74.9 g·kg‒1) (Fig. S10a) and was associated with significantly lower grain Se contents than the other three soils upon treatment with Se ENMs (Fig. 1b).

Fig. 3: Factors affecting the bioavailability of Se ENMs.
figure 3

a Correlation matrix of soil properties; the direction of the circle represents positive or negative, and the color represents the size of correlation coefficients, with purple representing −1 and green representing 1. b A scatter plot in three dimensions (top) and the accuracy of the MLR model (bottom) comparing observation versus prediction of LnTF. The LnTF referred to the transport effectiveness of Se in soil-plant system. The TF was defined as TF = (concentration of Se in grain – concentration of Se in soil) / concentration of Se in soil. In 3D-plot, the X-axis represents total organic carbon (TOC), the Y-axis represents available potassium (AK), the Z-axis represents LnTF, and the MLR algorithm was LnTF = −0.01*AK − 0.21*TOC + 43.62 (R2 = 0.89, p < 0.001). c Se content in iron plaque upon different TOC treatments (n = 5). d Iron plaque content (n = 5). e EDS elemental mapping of Fe and Se in rice root upon exposure to Se ENMs and CK. f XRD patterns of iron plaque in rice root upon different treatments, where ST stands for Se ENMs+TOC.

Full size image

Soil organic matter possesses strong adsorption capacity38, including for Se (Fig. S9 and Fig. 3c). Organic matter can also promote the formation of iron plaque on the root surface39 (Fig. 3d), which could further prevent the uptake of Se ENMs into rice roots (Fig. S10b and S10c). EDA maps show the overlap of Se and Fe on the root surface of rice (Fig. 3e), and XRD analyses indicate that FeSe and FeSe2 were formed on the root surface (Fig. 3f). The overlapping of Fe and Se on the root surface was less pronounced when Se ENMs and TOC were treated separately (Fig. S10d). Collectively, these results suggest that Se ENMs were retained at the root surface by iron plaque and/or Fe-Se complexes, which reduced Se bioavailability in the presence of higher organic matter content.

Field experiment and soil risk

The growth of rice plants and rice panicles was significantly improved upon the application of Se ENMs in the field (Fig. S11). Compared to the control (CK), the number, weight, and height of rice spikes increased by 18.8%, 14.0%, and 14.5%, respectively (Fig. S12). Additionally, root length, average diameter, surface area, and the number of root tips increased by 7.2%, 20.6%, 6.9%, and 34.4%, respectively (Fig. S13). The rice yield ( + 10.7%) and Se concentration in rice grains ( + 309.8%) were increased compared with CK. The quality of rice grains was also improved upon Se ENMs application, as indicated by the increased starch content ( + 5.4%), protein content ( + 2.4%), polysaccharide content ( + 17.7%), and total amino acid content ( + 8.0%) in rice grains compared with CK (Fig. S14). All these indicators portended that Se nano-enabled agriculture technology can be applied to produce high-quality rice with essential trace elements and nutrients for human health.

Meanwhile, Se ENMs also have potential effects on soil, which will provide support for the integration of agriculture and nanotechnology. Assessing the potential impact of Se ENMs on soil properties is crucial for guiding sustainable soil management practices, promoting the health of soils and crops, and supporting various ecosystem services40,41. Additionally, Se-metabolising bacteria are ubiquitously present within the environment, where they can utilise Se oxyanions and different compounds as terminal electron acceptors. Through this metabolism they also enhance degradation of various organic compounds or the dominant role of Se in ecosystem during the nitrogen and carbon cycle42. Soil health is considered “the frontier of soil science”43. In this study, linear scoring function methods were applied to evaluate soil health, and the descriptive statistics for each soil health indicator and total score are presented in Table S3 and Table S4. At the start of cultivation, the relative soil health (RSH) was 0.63. Following cultivation, the RSH values for the different treatments varied. The CK treatment showed a decline in soil health, with an RSH value of 0.53. In contrast, the application of SeO32‒ and Se ENMs resulted in a decrease in soil health during the growing stage, but the soil health recovered at the maturing stage, with RSH values of 0.66 and 0.63, respectively (Table S4 and Fig. 4a). These results showed that soil quality significantly declined in the CK treatment, whereas SeO32‒ and Se ENMs applications had a similar positive impact on soil health during the later stages of cultivation.

Fig. 4: Impacts of Se ENM application on soil properties.
figure 4

a Soil properties over time. b The relative importance of different soil indexes by using the RF and ANN. c Effects of Se ENMs on soil function at D120 compared with CK.

Full size image

However, the inadvertent accumulation of selenite in crops and the environment poses a significant concern for food safety and environmental sustainability. Compared to SeO32‒, Se ENMs have great advantages due to nanoscale, highly stable, low toxicity, and high bioavailability. Random forests and artificial neural network models were used to identify the most informative soil indicators with Se ENMs application (Fig. 4b). Following the application of Se ENMs, the soil exhibited improved biodiversity, enhanced nutrient utilization, and better maintenance of soil physicochemical properties compared with CK (Fig. 4c). Specifically, the indexes of Faith-pd and Pielou_e showed significant increases compared with CK (Fig. S15). Although the P content decreased significantly, no significant differences were observed in other nutrient elements such as N and Mg (Fig. S16). Moreover, indicators related to soil physicochemical properties, such as pH, showed significant improvements within an acceptable range (Fig. S17). Therefore, compared with CK, the application of Se ENMs enhanced biodiversity and improved nutrient utilization while maintaining soil physicochemical properties, thereby contributing to the preservation of soil health.

The prospects of Se ENMs-enabled agriculture

Se is an essential trace element in human diet due to its role in antioxidation, prevention of diabetes, and detoxification processes44. However, approximately 1 billion people worldwide suffer from Se deficiency44. The Se ENMs treatment (0.1 mg g–1, field experiment) significantly increased the content of Se in rice from 4.9 μg·100 g−1 for the CK group up to 20.2 μg 100 g−1 (Fig. S18). According to China’s Dietary guidelines43, the daily consumption of 300–500 g of Se-rich rice would provide individuals with 61−101 μg·day−1 of Se, meeting the daily recommended Se intake (55 μg·day−1). Therefore, the developed Se ENMs could provide a sustainable solution to address Se deficiencies.

In addition to Se fortification, the soil application of Se ENMs (0.1 mg‧kg–1) increased rice yield by 10% (about 885 kg·ha–1) (Fig. 5). The global rice production in 2022 is approximately 508 Mt (https://www.fao.org/). Therefore, the application of Se ENMs could potentially increase rice production by 50.8 Mt, which could provide enough food for nearly 623.3 million population annually. With a projected global population of around 9.7 billion by 205045, there is a pressing need to increase food production by more than 70% to meet the growing demand. Se ENMs-enabled agriculture could help alleviate this food crisis and bridge the gap between demand and supply. Also, we estimate that Se ENMs would not increase the overall cost of rice production. The current production cost of Se ENMs is approximately US$ 26.1 ha–1 (0.1 mg‧kg–1), which could be well compensated by the increase production generating a net profit of approximately US$ 231.5 ha–1 (Fig. 5, Text S5). Therefore, the developed

Fig. 5: Potential application of Se ENMs-enabled agriculture.
figure 5

a Effects of Se nano-enabled agriculture technology (economic, human health, and environmental). b Global prediction of Se ENMs-enabled enhancement of Se concentration in rice production.

Full size image

The risk of accumulation of Se in soil is considered minimal, at least in the short term. The Se content in the original soil was 0.21 mg kg‒1; it raised to 0.36 mg kg‒1 with Se ENMs and 0.44 mg·kg‒1 with SeO32‒ (Table S5), which is far below the maximum value for Se-contaminated soil ( >30 mg·kg‒1)46. Repetitive long-term application would require continuous monitoring of the change of Se content in soil.

Our results also showed that the application of Se ENMs significantly increased the N content in rice roots and leaves by 4.9% and 48.4%, respectively, in the greenhouse study (Fig. 2b). Despite a 30% reduction in N fertilizer application, both the grain protein content and total amino acid content increased in the Se ENMs field treatment (by 2.5% and 8.0%, respectively, Fig. S14). The global annual use of N fertilizers for rice is approximately 18.9 Mt47 and a reduction by 30% is equivalent to 5.67 Mt per year. Considering that the production of 1 ton of N fertilizers is associated with the emissions of 8.3 tons of CO248, a 30% reduction in N fertilizers would avoid the emission of 45.36 Mt of CO2 (Fig. 5). Durability of carbon dioxide removal is critical for Paris climate goals49. These considerations are in addition to other important environmental damages associated with N fertilizer usage that could be mitigated by Se ENMs (e.g. N2O emission, nitrate leaching and runoff, Text S6). Therefore, the implementation of Se ENMs in rice production offers significant economic, environmental, and human health potential benefits.

Machine learning techniques were utilized to assess the global applicability of Se ENMs in rice production. The global soil conditions were utilized to predict the impact of Se concentration in rice on the improvement of Se ENMs amendment. The random forest (RF) model demonstrated good accuracy and robustness (Rtraining2 = 0.86, qtraining2 = 0.81, Rtest2 = 0.57). Additional metrics relative to model performance are shown in Table S6-7. The SHAP analysis (Fig. S18) revealed that application Se level (A_Se), TOC, and total nitrogen (TN) were the three most influential factors, consistent with experimental results. For global prediction, the application method was set as a foliar application at 0.1 mg L‒1 (4 times, 25 L·time‒1). The detailed information about the data extraction, calculation of effect size, and global prediction are described in Text S4. The results showed that Se concentration of rice grains was increased by 2.43 to 4.57 times (corresponding to the effect size 0.89−1.52) compared with CK, with an average increase of 3.92 times (corresponding to the effect size 1.36) (Fig. 5b). Among the major rice-growing countries, China showed a 3.44 times increase (effect size, 1.23), India showed a 3.66 times increase (effect size, 1.29), the United States showed a 3.99-fold increase (effect size, 1.38), and Australia showed a 3.50 times increase (effect size, 1.25) (Fig. S19-20), indicating remarkable effects of Se ENMs application in different regions of the world. Nevertheless, it should be noted that these effect values represent the multiples relative to the CK, rather than absolute values. Consequently, caution should be exercised to prevent Se levels in rice from exceeding recommended levels. For specific locations, adjusting the A_Se can help maintain Se levels within safe ranges (0.1–1 mg·kg‒1) in plants. Overall, the newly developed Se ENMs have substantial benefits to society, the economy, and the environment and look promising for sustainable rice production. Meanwhile, addressing socio-economic and regional challenges, such as infrastructure, cost, and regulatory frameworks, and tailored recommendations for specific agricultural systems, considering diverse farming practices. One challenge may be around the perception of nano-enabled agricultural technology by the public and practitioners, and more efforts should be devoted to increasing education and communication to reduce the risk of rejection50.

Conclusion

In summary, pot and field experiments demonstrated that Se ENMs can improve rice production by increasing both rice yield and grain Se content. The processes involved in yield were elucidated and shown to be related to increased rice tillering through the alteration of the rhizosphere microbial community, improvement of N availability, and regulation of plant endogenous hormones (GAs) and gene expressions (MOC1, TB1, and OSH1). Soil TOC reduced the bioavailability of Se ENMs via the formation of iron plaque. Se ENMs did not negatively impact physical, chemical and biological soil properties, whereas soil health indicators deteriorated when Se ENMs were not applied. Importantly, Se ENMs-enabled agriculture offers great economic (profits, +$231.5 ha−1), environmental (CO2, −1.12 Tons ha−1), and human health (Se-rich rice, ~20 μg·100 g−1) by machine learning. These results revealed that the application of the newly developed Se ENMs in rice production could have significant economic and health benefits while reducing greenhouse gas emissions and nitrogen losses. In the future, further work should be dedicated to explore further the applicability and scalability of Se-ENMs in rice production around the world.

Experimental procedures

Synthesis and Characterization of Se ENMs

Se ENMs were synthesized according to our previous method25. The shape and size of Se ENMs were observed by transmission electron microscope (TEM, JEM–2100, Nippon electronics co, JPN) operating at an acceleration voltage of 200 kV. Surface and subsurface chemical status was measured by X–ray photoelectron spectroscopy (XPS, Thermo ESCALAB 250XI; Axis Ultra DLD Kratos AXIS SUPRA; PHI-5000versaprobe).

Pot experiment

Rice seeds (Chuangliangyou 669, Oryza sativa L.) were obtained from the Anhui Royee seeds Co., Ltd. The seeds were sterilized with 5% NaClO for 10 min and then rinsed with deionized (DI) water three times. After immersion in DI water for 8 h, the seeds were germinated on wet filter paper in a petri dish for three days. The rice seedlings were then transplanted into a container filled with 5 kg of paddy soil. Three days before transplanting, different concentrations of Se ENMs (0, 0.1, 0.5, and 1 mg kg−1) and a Na2SeO3 suspensions (0.5 mg·kg−1) were applied to the paddy soil. Each container had two rice seedlings, and there were five replicates (container) for each treatment. The paddy soil was obtained from Jiangsu province, and the soil properties are presented in Table S1. One week before transplanting, the basal fertilizers, including N (CO(NH2)2, 0.2 g kg−1), P (CaH2PO4·H2O, 0.15 g kg−1), and K (KCl, 0.2 g·kg−1) were added to all the culture containers in suspension form. Plant height and tiller numbers were counted every 30 days until harvest. The rice plants were harvested 120 days after transplanting, and the biomass and yield of rice plants were weighed before and after drying.

Five different paddy fields (i.e., Jilin, Jiangxi, Jiangsu, Chongqing, and Guangdong) were selected in order to study the performances of Se ENMs across a range of soil properties, in greenhouse experiments. The details of soil properties are shown in Table S1. The cultivation of rice in greenhouse followed the same methods as described above. The rice images were captured at different growth stages throughout their life cycle. After sampling, the rice yield, grain Se content, and thousands-of-grain weight were analyzed.

Analysis of Se and N contents in rice plant

The Se content in rice grain was analyzed by inductively coupled plasma mass spectrometry (ICP–MS, iCAP–TQ, Thermo Fisher, Germany) after digestion. Briefly, the dried rice plant tissues (rice grain, 25 mg) were digested (200 °C, 1600 W, and 30 min) in a microwave accelerated reaction system (CEM crop, NC) with a mixture of guaranteed reagent HNO3 (14.5 mol L–1) and ultrapure water (V: V = 4:3). The extract was filtered with microporous membrane (0.22 µm) and diluted to 50 mL for the ICP–MS analysis. The internal standard element was Ge (GSB 04-2826-2011, National Center for analysis and testing of nonferrous metals and electronic materials, China). The standard recovery rate of each element ranged from 85.6% to 103.4%.

The N contents of rice root, stem, and leaf were measured by an Element analyzer (Elementar Unicube, Germany). Specifically, 5 mg of dried rice plant tissues were packed in a tin boat and fully combusted at 1850 °C. Benzoic acid was used as the standard to verify the stability of the instrument.

qRT–PCR analysis for key regulator gene involved in rice tillers

The relative expression of key genes (MONC1, TB1, OSH1, GA2ox1, GA2ox2, GA2ox3, GA2ox4, GA2ox5, and GA2ox6) involved in tillering process was investigated by Quantitative real-time PCR (qRT–PCR). Briefly, total RNA of rice rhizome junction was extracted by using plant RNA extraction kit (TaKaRa minibest) according to the protocol of manufacturer. The concentration of extracted RNA was measured by a ultramicro-spectrophotometer (UltraM-QB200, Gallop tech Co., Ltd. Shanghai, China). The extracted RNA was converted into cDNA using a cDNA synthesis kit (CW Biotech Co., Ltd., Jiangsu, China) following the manufacturer’s protocol. The qRT–PCR reaction system contained cDNA template (2 µL), forward primer (1 µL, 10 µM), reverse primer (1 µL, 10 µM), and 2×Ultra SYBR Mixture (25 µL). The details of primers are listed in Table S2. qRT–PCR reaction system was performed by CFX96TM real-time system (BIO-RAD, USA). The reaction program was 10 min at 95 °C, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Each biological replicate was evaluated with three technical replicates. The relative gene expression was calculated using the 2−ΔΔCT method.

Rhizosphere microbial analysis

Rice rhizosphere soil was collected for analysis of the diversity and composition of microbial communities at harvest47. High-throughput sequencing technique was applied using an Illumina MiSeq platform by Personalbio Technology Co., Ltd. (Shanghai, China). The V3–V4 region of the rhizosphere microbial 16S rRNA gene was amplified with primers 338 F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806 R (5′-GGACTACHVGGGTWTCTAAT-3′) using a quantitative PCR system. Microbial sequencing results were analyzed by the genes cloud tools (www.genescloud.cn), a free online platform for data analysis.

Hormone extracts and quantitative analysis

The equivalent of 1.5 g of fresh plant material was ground into a dry powder in liquid nitrogen, and placed in a glass test tube, to which a mixture of isopropyl alcohol, water, and hydrochloric acid was added. After shaking for 30 min, dichloromethane was added and samples were oscillated for 30 mins. Subsequently, samples were centrifuged at 15,871 g for 5 min, and the organic phase (nitrogen gas) was removed with nitrogen (avoiding light) and re-dissolved with methanol (0.1% formic acid), and centrifuged for 10 min (13,000 g). Then the supernatant was filtered through 0.22 µm membrane, and analyzed by HPLC-MS/MS (Agilent, USA). All testing procedures were performed at Nottingbio Ltd., LTD.

Field experiment

The field study was conducted in Kunshan, Jiangsu province (31.4°N, 120.1°E), with rice seeds (Nanjing 46), which are commonly used in this area. The paddy field (14 × 100 m2) was divided into nine plots, each plot approximately 150 m2. They were randomly distributed to the CK group, Se ENMs group, and SeO32‒ treatment group, with three plots in each group. Each plot had its own irrigation and drainage system. Water and fertilizer management in paddy fields was based on the experience of local agronomists and farmers. A total of 190 kg ha‒1 of N fertilizer was applied throughout the growth period. The field had a 30% reduction in N fertilizer (133 kg ha‒1) application after the application of Se ENMs (0.1 mg/kg, 4 times, 25 L/time) during the rice life cycle. Immediately after harvest, root parameters (root length, root average diameter, root surface area, and root tips) were analyzed using a root scanner (WinRHIZO Pro 2017b, Canada). The starch, protein, polysaccharide, and amino acid content in rice grains were determined using previously described methods48. The protein content in rice was determined by a K9840 automatic nitrogen analyzer (K9840, Hanon, China). The Se content in rice was measured by ICP–MS after digestion. The amino acid content in rice was determined and analyzed by LC–MS/MS (Vanquish Flex, Thermofish Scientific, Germany).

Statistical analysis

All treatments were conducted at least in triplicate. Data are presented as mean ± standard deviation. One-way ANOVA with the least significant difference (LSD) test was used to analyze the significant difference among different treatments using Origin Statistics 2019b at p < 0.05. Heat map data is processed on platform https://www.genescloud.cn/login. The detailed methodologies to characterize rice grains, the specific setting, analysis of LC–MS/MS, soil health assessment, and global prediction are shown in Text S1, Text S2, Text S3, and Text S4.