The study was conducted between May and October 2020, at the experimental base of the National Engineering and Technology Research Center for Grains in the High-tech Zone, Daqing City, Heilongjiang Province, China (46°39'N, 124°51'E). Soybean (Glycine max L. Merr.) seeds of uniform size were manually selected. The seeds were germinated in a germination box at 22°C for 2 days in the dark. Seedlings with consistent growth were chosen and transplanted at a density of three buds per 33 × 26 cm bucket that was filled previously with fine river sand. The sand was first rinsed with tap water to remove turbidity and then rinsed three times with distilled water. Six 1 cm holes were drilled through the base of each bucket and gauze was placed at the base inside the bucket before filling with sand. Plants were watered with 0.5 L of distilled water per bucket per day until the opposite leaves were fully unfolded.

After the full expansion of opposite leaves, the plants were randomly divided into two groups, which were treated with low nitrogen (LN) and control nitrogen (CK). A nitrogen (N) concentration of 7.5 mM, contained in 1/2 Hoagland nutrient solution, was used as a control (Li et al., 2017; Qiao et al., 2019). CaCl₂ and KCl were used to supplement insufficient elements (Wang et al., 2019) (see Supplementary Table 1 for formula of nutrient solution). Each bucket was treated with 0.5 L of solution. To prevent the accumulation of salt in the sand culture, buckets were rinsed thoroughly with distilled water every 3 days. The soybeans were treated with 100 μM MT when they developed to V3 or R5 stage. Physiological metabolism-related indices, including nitrogen assimilation, related enzyme activity, net photosynthetic rate, chlorophyll content, and soluble sugar content, were measured at 6, 12, 18, and 24 days after MT treatment (Supplementary Figures 1, 2), and the dry matter accumulation of plants and nodules at each investigated growth stage after MT treatment was analyzed (Supplementary Table 2). The time duration with the most obvious difference between MT treatment and non-MT treatment under LN was selected as the reference for sampling time in the 2nd year.

The experiment was conducted with three nitrogen concentration treatments (i.e., 0N, LN, and CK, Supplementary Table 1) in 2021. At three nitrogen levels, MT was sprayed at V3 or R5 stage or not applied, and 9 treatments were obtained. Samples were collected 12 days after the MT application (based on the results of the screening experiment in 2020) for the determination of physiological indices. The functional leaf samples were collected 24 h after MT treatment and stored at −80°C for quantitative reverse transcription PCR (RT-qPCR) analysis. The functional leaf samples were directly determined or wrapped in tin foil and stored at −20°C for physiological and biochemical analysis, and the remaining plants were used to determine the nitrogen and ureides content in each organ of the plant after maturity. Three repeated values were recorded for each physiological and biochemical index and averaged (from three mixed samples) for those in the same container, and 15 repeated values were recorded for the determination of yield components.

The tissue of fresh functional leaves was weighed into 0.5 g portions to determine enzyme activity. Nitrate reductase (NR: EC 1.7.1.3) activity was measured according to the method of Sherrard and Dalling (1979). Glutamine synthetase (GS: EC 6.3.1.2) enzyme activity was determined according to Zhang et al. (2017). The activities of glutamate synthase (GOGAT: EC 1.4.7.1) and glutamate dehydrogenase (GDH: EC 1.4.1.2) were determined according to the method of Lin and Kao (1996).

Fresh leaf samples were weighed to 1 g (repeated three times). The extract used for biochemical analysis was prepared using the method of Oliveira et al. (2013). The content of NO3- was determined spectrophotometrically at 410 nm by the salicylic acid method described by Cataldo et al. (1975). The level of NO2- was determined according to the method described by Hageman and Reed (1980).

The tissue stored at −20°C was divided into 0.5 g samples. The content of soluble protein was determined by spectrophotometry according to the method described by Cao et al. (2019); the content of total free amino acids was determined by using ninhydrin reagent (Moore and Stein, 1954).

The content of total nitrogen in plant tissues was measured by the Kjeldahl method (Zhang et al., 2010). The mature samples dried at 105°C were ground into powder. The samples were weighed to 0.1 g, digested with H₂SO₄-H₂O₂, and then N content was determined by a Kjeldahl nitrogen determination instrument. Each treatment was repeated at least three times.

The manual counting method was used to calculate the number of root nodules per plant of soybean. The fresh nodules were dried at 105°C for 0.5 h, then dried in an oven at 80°C to constant weight, and the dry weight of the nodules was measured. Nodule nitrogenase activity was evaluated by measuring hydrogen production in living soybean nitrogen fixation tissues using a nitrogen fixation analyzer (Q-BOXNFILP) to determine the rate of nitrogen change. According to the method described by Zhang et al. (2018), the content of ureides in the dried organ at the mature stage was determined.

The content of chlorophyll a (Chl a), chlorophyll b (Chl b), and total chlorophyll (total Chl) was determined by the method of Cao et al. (2019). The net photosynthesis rate (Pn), transpiration rate (Tr), stomatal conductance (Gs), and intercellular CO₂ concentration (Ci) of the 2nd leaf from the top of the main stem were measured using a Li-6400 photosynthesis system (Li-6400, LICOR, Lincoln, NE, USA) equipped with a red-blue LED artificial light source leaf chamber using the method described by Wang et al. (2021a). During the measurement, photosynthetically active radiation (PAR) was 1,000 μmol/m² s, atmospheric CO₂ concentration was 400 μmol/mol, leaf temperature was 25°C, and the relative humidity was ~25%. All measurements were made on sunny days between 09:00 and 16:00 (Liu and Li, 2016).

Total RNA of the functional leaf sample (50–100 mg) was extracted using an Axygen reagent kit (AxyPrep, USA) and reverse-transcribed according to the manufacturer's instructions (Prime Script RT Enzyme Mix I for TB Green qPCR, Takara, China). The gene-specific primers were designed with Primer Premier 5.0 (Supplementary Table 3), and RT-qPCR was performed using a 10 μl mixture (5 μl TB Green Premix Ex Taq II; 3 μl distilled water; 0.5 μl PCR Forward Primer; 0.5 μl PCR Reverse Primer; 1 μl cDNA) according to the manufacturer's instructions (CFX96 Real-Time PCR Detection System, Takara, China). A two-step PCR method was used with the following conditions: pre-denaturation at 95°C for 30 s, and 40 cycles at 95°C for 5 s and 60°C for 30 s. The specialist software Bio-Rad iQ5 was used to collect data. Actin (gene ID: Glyma18g52780) was adopted as an internal reference for the normalization of gene expression levels. In the RT-qPCR analysis, samples of two plants from each pot treated with MT for 24 h were mixed as a single replicate, and three pots per treatment were considered as three biological replicates. Each biological replicate had three technical replicates.

Five pots of soybean under each treatment were randomly selected at harvest for the assessment of the number of pods per plant, the number of grains per plant, total grain weight per plant, and hundred-grain weight.

The experiment was conducted in a randomized design. The data obtained were analyzed by Microsoft Excel 2017. LSD and Duncan's multiple range test (P ≤ 0.05) were performed using SPSS 20.0 (SPSS Inc., Chicago, IL, USA). The charts including the PCA Biplot were formed with Origin 2019 software.

As shown in Figures 1, 2, the activities of NR, GS, GOGAT, and GDH treated with LN were significantly lower than those in the CK group. The application of MT effectively promoted the activity of these nitrogen assimilation enzymes under LN level. At 12 days after spraying MT, the enhancement of enzyme activity in LN+V3MT was the highest, and the activity levels of NR, GS, GOGAT, and GDH increased by 18.20, 34.42, 21.16, and 15.22%, respectively, compared with those in LN (Figure 1). Moreover, MT spraying at R5 stage showed a similar result to V3 stage, i.e., compared with LN, MT addition had the largest increase in nitrogen metabolic enzymes at day 12 after MT treatment (Figure 2).

Compared with the CK, MT significantly enhanced the activity of GDH, while the activity of nitrogen metabolism enzymes decreased under 0N and LN treatments (Figure 3). Significant increases in the activity of GS, GOGAT, and GDH were observed under MT application for both nitrogen deficiency treatments, and the promoting effect was more obvious under 0N (Figures 3B–D). However, the alleviation effect of MT on the inhibited NR activity was mainly observed in the low nitrogen supply group, in which the NR activity in LN+V3MT and LN+R5MT increased by 24.08 and 35.59%, respectively, compared with that in LN at 12 days after MT treatment (Figure 3A). Furthermore, 12 days after MT treatment at R5 stage, the GS and GOGAT activities of 0N+V3MT and LN+V3MT were markedly higher than those of the treatment without MT under the corresponding nitrogen level. Specifically, after the MT treatment, GS activity increased by 23.75 and 14.59%, and GOGAT activity was enhanced by 43.86 and 19.58%, respectively, when compared with the corresponding nitrogen level (Figures 3B₂,C₂).

Differences in inorganic nitrogen contents were not evident between CK and V3MT/R5MT. Regulation by MT of NO3- and NH4+ in functional leaves was mainly reflected in 0N and LN levels (Figure 4). As a result of V3 application, MT primarily promoted the accumulation of NO3- at the low nitrogen level. At the two sampling time points, the NO3- content of LN+V3MT was 25.61 and 43.68% higher than that of LN, respectively, but lower than that of the CK treatment (Figure 4A). In contrast, the contribution of exogenous MT to NH4+ accumulation was more significant at 0N level (Figure 4B).

Soluble protein and free amino acid contents were affected by insufficient nitrogen supply, with the levels lower than those in the CK group (Figure 5). Compared with LN, the content of soluble protein was enhanced by 16.64% when MT was applied at V3 stage (Figure 5A₁). The application of MT at R5 stage had an insignificant effect on soluble protein content (Figure 5B₁). The content of free amino acids was raised in both MT application stages compared with their corresponding nitrogen levels. Moreover, the contents of free amino acids in functional leaves treated with 0N+V3MT and LN+V3MT were still significantly higher than those treated with 0N and LN, respectively, at 12 days after R5 spraying (Figure 5B₂).

Melatonin increased the nodule number and nodule dry weight by 23.08 and 35.03%, respectively, 12 days after V3 stage treatment (Figures 6A₁,B₁) when compared with the CK. Notably, the activity of nitrogenase in root nodules was only promoted by MT application at V3 stage under 0N level (Figure 6C). In the condition of nitrogen deficiency (0N and LN), the number, dry weight, and nitrogen fixation ability of root nodules were higher than those of CK, and the optimal parameters were reflected in the LN level (Figure 6). The MT treatment at V3 stage further increased the number and dry weight of nodules under nitrogen deficiency. The number and dry weight of nodules treated by LN+V3MT were highest, even 12 days after MT was applied in R5 stage, and were 56.59 and 28.07% more than that of CK, respectively (Figures 6A₂,B₂). However, the application of MT at R5 stage only affected the dry weight of nodules at low nitrogen levels, increasing it by 14.11% compared with LN (Figure 6B₂).

Exogenous MT application at R5 stage (R5MT) resulted in more ureides content in roots at harvest when compared with the CK treatment (Figure 7D). Nitrogen deficiency caused the transfer of ureide to seed and reduced the accumulation of ureides in stem and pod skin at harvest (Figures 7A–C). Regardless, when seedlings were treated with MT at V3 stage, the ureide accumulation in the stem, pod skin, and root increased under 0N level (Figures 7A,B,D). In contrast, the MT treatment at R5 stage mainly promoted the ureides content in pod skin, grain, and root both under 0N and LN levels (Figures 7B–D). The ureides content in seeds under 0N+R5MT and LN+R5MT conditions was 0.59 and 0.63 μmol/g higher than that of the control, respectively (Figure 7C).

The nitrogen content of each organ at 0N level during harvest was decreased when compared with CK, while the difference in its accumulation between LN and CK was mainly reflected in the stem and root (Figures 8A,D). When MT was applied alone at V3 or R5 stage of soybean, the change in nitrogen content was insignificant in comparison with CK. However, the increases in nitrogen content in the stem of 0N+V3MT and 0N+R5MT were 32.88 and 27.40%, respectively, when compared with 0N (Figure 8A). Likewise, increases in the seed for those same treatment groups were 2.53 and 3.76%, respectively (Figure 8C). Furthermore, the application of MT at V3 and R5 under LN treatment increased the nitrogen content of stems by 25.51 and 14.29%, respectively (Figures 8A,B).

Compared with CK, insufficient nitrogen supply significantly suppressed the levels of Chl a and Chl b, and the inhibitory affect was more evident in 0N (Figures 9A–C). The degree of decrease in Chl a content was lower than that of Chl b both in 0N and LN, so the ratio of Chl a/Chl b increased when compared with CK at the seedling stage (Figure 9D). MT alone at V3 stage was beneficial to increase the content of Chl b that plays an auxiliary role. However, the application of MT markedly alleviated the inhibition effect of N deficiency both at V3 and R5 stages, particularly on the content of Chl a (Figure 9A). Notably, MT treatment at R5 stage further increased the ratio of Chl a/Chl b, which showed that the increase of Chl a content was greater than Chl b in 0N+R5MT and LN+R5MT treatments compared with corresponding nitrogen treatments (Figure 9D₂).

The leaf gas exchange parameters were severely inhibited when nitrogen supply was reduced (0N and LN), particularly in 0N conditions (Figure 10). The exogenous MT at V3 and R5 stages enhanced Pn under all nitrogen levels including the control nitrogen level (Figure 10A). When MT was applied in V3 stage under LN condition, Pn returned to the same level as the control (Figure 10A₁). However, spraying MT at V3 stage significantly promoted Cond, Ci, and Tr under 0N level only at 12 days after application (Figures 10B₁–D₁). On the 12th day after R5 treatment, the Pn of 0N+V3MT was significantly higher than that of 0N by 13.67%, while there was no distinction in Cond, Ci, and Tr between 0N+V3MT and 0N (Figures 10B₂–D₂). In contrast, for 0N+R5MT, the photosynthetic correlation indices Pn, Cond, Ci, and Tr increased by 16.67, 25.43, 9.97, and 22.90%, respectively, in comparison with 0N (Figures 10A₂–D₂).

Gene expression related to MT-induced nitrogen metabolism and photosynthesis is presented in Figure 11. When MT was applied alone, the expression levels of PsaF and PsbE related to photosynthetic system I (PSI) and photosynthetic system II (PSII) were 1.57 and 1.22 times higher, respectively, than that of CK. Under nitrogen deficiency, the genes effectively upregulated by MT included not only those encoding PSI (PsaA, PsaF) and PSII (PsbE, PsbO) but also those encoding nitrogen assimilation enzymes (NR2, NiR, GS1β, and GOGAT) and amino acid permease (GmAAP6a), which could induce amino acid transport from source to sink. Compared with CK, the inhibition of the above genes' expression by nitrogen deficiency (0N and LN) was more significant at 0N level than at LN level. Among them, under 0N treatment, NR2, NiR, GS1β, GOGAT, GmAAP6a, PsaA, PsaF, PsbE, and PsbO were downregulated by 5.75, 7.43, 4.67, 2.18, 5.67, 5.25, 4.68, 2.33, and 4.60 times, respectively, compared with the control. In contrast, when the seedlings under zero nitrogen stress were treated with MT, the expression of the above genes was only downregulated by 2.95, 6.20, 1.11, 0.79, 0.79, 3.36, 3.54, 1.83, and 3.67 times, respectively, in comparison with the CK.

Compared with the CK, pod number, seed number, and seed weight individually increased by, respectively, 8.15, 9.75, and 8.74% in 2020 and by 10.86, 8.44, and 10.96% in 2021, when MT was applied alone in V3 stage (Figure 12 and Table 1). Insufficient nitrogen supply inhibited the yield of soybean, and the inhibitory effect of pod number, seeds number, grain weight, and 100-seed weight per plant of 0N were much more inhibited than LN when compared with CK. The negative effect of insufficient nitrogen supply on plant yield was alleviated with the use of MT in nitrogen-deficient plants. Among them, pod number, seed number, and grain weight of 0N+V3MT increased by 13.33, 16.55, and 16.69%, respectively, compared with 0N. Combined with the two-year production data to calculate the average efficiency, the above yield indices of LN+V3MT were 17.80, 15.21, and 12.41% higher than those of LN, respectively, and the number of pods and seeds returned to the level of CK treatment. The results showed that the application of MT at R5 stage increased the grain weight under 0N and LN levels by 3.07 and 3.22% (2-year average), respectively. Although the grain weight of LN+R5MT was 7.72% (2-year average) higher than that of LN, it was still lower than that of CK treatment.

The PCA biplot of exogenous MT 12 days after V3 and R5 treatments explained the total variability by 88.4 and 88.7%, respectively (Figure 13). The effect of MT was analyzed based on the principal component 1 (PC1) that accounted for a larger proportion in the cause of differences among treatments. The score plot reveals that the distance between the points representing 0N+V3/R5MT and CK projected to the PC1 axis was less than that between 0N and CK. Similarly, the distance between the points representing LN+V3/R5MT and CK projected to the PC1 axis was smaller than that of LN and CK. MT treatment effectively reduced most of the differences in 0N or LN with CK due to nitrogen deficiency. In addition, the correlation between physiological characters and yield was analyzed by loading plot. All indices had a positive correlation with yield after MT treatment at V3 stage, except for the dry weight and the nodule number. Similarly, after MT treatment at R5 stage, while the correlation between yield and dry weight of nodule or nitrogenase activity decreased to almost negligible, the other indices were all significantly positively correlated with yield parameters.