The two host plants used were the soybean variety HN89 and the maize variety Tiannuo. The AM fungal partner was Rhizophagus irregularis (DAOM 197198; accession no. AUPC00000000; Tisserant et al., 2013; Wang et al., 2016), and the rhizobial isolate was belonging to Bradyrhizobium elkanii strain BXYD3.

One greenhouse experiment and one growth chamber labeling experiment were conducted. The greenhouse experiment was carried out to check soybean growth performance in response to the varying N and P levels, and rhizobium and AM inoculation treatments, which included four N and P treatments (LN and P, LNLP, 265 μM N and 25 μM P; LN and medium P, LNMP, 265 μM N and 250 μM P; MN and LP, MNLP, 2650 μM N and 25 μM P; MN and P MNMP, 2650 μM N and 250 μM P), two culture modes (monoculture and intercropping) and three combinations of AM fungal and rhizobium inoculation treatments, including an uninoculated control (-A-R), AM fungal inoculation (+A-R), and inoculation of both the AMF and rhizobium (+A+R). There were four replicates of each treatment. The growth chamber labeling experiment had one N and P treatment (530 μM N and 50 μM P) with two cropping systems (monoculture and intercropping) and three combinations of AM fungal and rhizobium inoculation (-A-R, +A-R and +A+R) in order to explore N transfer and C allocation in CMNs. N and P concentrations were increased compared to LNLP levels in greenhouse experiment, but had not been increased to those at MNMP levels since the plants in MNMP were too large to handle for ¹³C labeling. There were six replicates of each treatment. Three were used for ¹⁵N and ¹³C labeling, and the other three were used as controls to examine natural ¹⁵N and ¹³C abundances, and adjust for the ¹⁵N and ¹³C abundance of plant samples from the background contribution.

A two-compartment growth system was employed for both the greenhouse experiment and the growth chamber labeling experiment (Supplementary Figure S1). Soybean and maize plants were grown side by side in the two-compartment system constructed out of perspex boxes (13 cm × 12 cm × 12 cm). The two compartments were evenly separated by a perforated perspex divider (0.5 cm wide) to form a 0.5 cm air gap in order to prevent ion diffusion between compartments. Every side of the divider was enclosed by a 30 μm nylon mesh that allowed only fungal hyphal connection while preventing root contacts. Every compartment was filled with 1 kg of growth substrate consisting of a 4:1 mix of sand and soil that was autoclaved in two 1 h 121°C treatments in order to inhibit microbial contamination. The growth substrate had a pH of 5.0 when mixed with water, along with 1.2 mg kg^-1 of available P as determined by the Bray II method, and 47.2 mg kg^-1 of available N as determined by the alkali-hydrolyzed diffusion method.

In the greenhouse experiment, plants were grown under natural light in the intensity range of 500–1800 μmol m^-2 s^-1. Diurnal air temperatures ranged consistently between 20 and 28°C. Prior to planting, 10% dry inoculum was incorporated into the growth substrate. The inoculum consisted of colonized millet root fragments and a soil:sand mix containing the spores and extraradical hyphae of R. irregularis. The uninoculated control also incorporated the 10% soil:sand mix from millet grown without mycorrhizal inoculation. Soybean seedlings were inoculated with 2 mL rhizobium solution 1 week after germination (Wang et al., 2011). Plants were watered with soybean nutrient solution (Zhang et al., 2015) containing different N and P concentrations every week throughout the growth period.

In the growth chamber labeling experiment, plants were grown in a walk-in growth chamber with a 16 h photoperiod, a 24°C/22°C day/night temperature cycle, and a light intensity of 455 μmol m^-2 s^-1. After germination, seedlings were inoculated by adding 500 spores of R. irregularis into the soil close to the roots. After 1 week, soybean seedlings were inoculated with 2 mL of rhizobium solution (Wang et al., 2011). Plants were watered every week with soybean nutrient solution containing 50 μM P and 530 μM N.

In the growth chamber labeling experiment, ¹⁵N labeling was conducted according to Xiao et al. (2004) with some modifications. Six weeks after planting, soybean plants were labeled through petiole injection with (¹⁵NH₄)₂SO₄ enriched with 99% ¹⁵N. Ten microliter of 88 mM (¹⁵NH₄)₂SO₄ solution was injected through a 25 μL syringe each day for 9 days in succession, and thereby supplying every soybean plant with a total of 0.22 mg ¹⁵N.

¹³C labeling was conducted according to Lu et al. (2002a) with some modifications. Briefly, at the time of labeling, the surface of the perspex boxes was sealed with plastic film with a narrow slot for the shoot to ensure the isolation of the shoots from the roots, and to create a shoot-labeling box. The Perspex boxes were transferred to transparent labeling chambers (50 cm × 52 cm × 50 cm), which was also sealed to prevent ¹³CO₂ leakage. ¹³CO₂ gas inside the chamber was generated by adding lactic acid into vial of Ba¹³CO₃ (98% ¹³C) with a syringe. The mean ¹³CO₂ concentrations were 400–450 μL L^-1 in the chamber. Four fans were used to recirculate the air within the closed labeling chambers. Plants were labeled at midmorning for 6 h, and then taken out from the chamber for 3 days before harvesting.

After 50 days, plants in the greenhouse experiment were harvested. Shoots, roots, and nodules were separated for determining plant dry weight, plant N and P concentrations, nodule number, and nodule fresh weight. One fresh root subsample was weighed and then cleared with 10% KOH solution and stained with 5% ink-vinegar solution for assessing AM colonization (Wang et al., 2016). Nitrogen and P concentrations were measured using a San++ SKALAR continuous flow analyzer (Skalar, The Netherlands) after acid digestion.

In the growth chamber labeling experiment, plants were harvested 3 days after ¹³C labeling, and the shoots, roots and nodules were separated. The roots were washed, weighed, and divided into two portions; one subsample was used for determination of AM colonization (Wang et al., 2016); and the rest of the samples were dried to determine ¹⁵N and ¹³C concentrations, along with N, C, and P concentrations. The concentrations of ¹⁵N and ¹³C in the shoots, roots, and nodules were determined using Delta V Advantage isotope ratio mass spectrometry (Thermo Finnigan, German). Total carbon was determined using a Euro EA3000 single elemental analyzer (Euro Vector, Italy).

N transfer was calculated as follows (He et al., 2009; Meng et al., 2015):

Percentage N transfer (% N_transfer):

Where ¹⁵N content = atom% ¹⁵N excess × total N; The atom% ¹⁵N excess was calculated by subtracting the natural abundance of ¹⁵N in unlabeled soybean or maize from the measured atom% ¹⁵N of labeled soybean or maize.

Amount of N transferred from soybean to maize (N_transferred, mg plant^-1):

Percentage N in maize derived from transfer (% NDFT):

The percent N derived from the atmosphere fixed by soybean (% Ndfa, Xiao et al., 2004):

where mono maize is the maize grown in monoculture. Mixed soybean is the soybean grown in soybean/maize intercropping.

The content of ¹³C incorporated into different tissues (shoots, roots, or nodules) of soybean or maize was calculated based on the difference in atom% ¹³C excess of the labeled and non-labeled tissues as follows (Lu et al., 2002b):

where TC indicates the total tissue C content.

All of the data were analyzed to calculate mean and SE using Microsoft Excel 2007 (Microsoft Company, USA). SAS (SAS Institute Inc., Cary, NC, USA) was used for two-way (inoculation treatment and cropping system) or three-way (nutrient treatment, inoculation treatment, and cropping system) analysis of variance (ANOVA). Mean separation was conducted using the Duncan Multiple Range Test (DMRT) following two- or three-way ANOVA, and the least significant difference (LSD) for the two cropping systems or for two inoculation treatments.

The results showed that soybean and maize growth was significantly affected by N and P availability (P < 0.0001; Table 1, Supplementary Table S1). Biomasses of monocropped soybean and maize in the moderate N and P (MNMP) treatment increased by 68–173 and 311–502%, respectively, and those IS and maize biomasses increased by 53–158 and 247–766%, respectively, compared with their biomasses in the corresponding LN and P (LNLP) cultures (Table 1).

AM fungal and rhizobium inoculation produced great effects on soybean and maize growth. Inoculation with rhizobia and/or AMF dramatically increased biomass of soybean and maize (P < 0.0001; Table 1, Supplementary Table S1). Compared to uninoculated controls, in LNLP, soybean biomass in monoculture and intercropping systems increased by 38 and 52%, respectively, with only AMF inoculation, and by 55 and 66%, respectively, with co-inoculation, while maize biomass increases were 81 and 37%, respectively, with only AMF inoculation, and 81 and 33%, respectively, with co-inoculation. Moderate application of N and P amplified the effect of AM fungal and rhizobium inoculation, with monoculture and intercropped biomass increases compared to uninoculated controls of 124 and 156%, respectively, for soybean inoculated with AMF alone, 141 and 164%, respectively, for co-inoculated soybean, 165 and 152%, respectively, for maize only inoculated with AMF, and 165 and 233%, respectively, for co-inoculated maize.

Cropping system had a significant effect on soybean growth (P < 0.0001; Table 1; Supplementary Table S1). Biomasses of all soybean plants in intercropping treatments were lower than in monoculture across nutrient levels and inoculation treatments. Under moderate N and low P (MNLP) conditions, IS biomass was 27% less than in monoculture. Conversely, all maize plants in monoculture and intercropping systems responded positively to increased nutrient and inoculation treatments (Table 1; Supplementary Table S1). Yet, the only effect of intercropping observed for maize was when the biomass of IM was 17% higher than the biomass of monocultured maize in the MNMP and co-inoculation treatment.

In addition, total biomass of IM and soybean was also significantly affected by nutrient availability and inoculation treatments. Co-inoculation had positive effects on total biomass in both monoculture and intercropping systems, and moderate application of N and P amplified the effect of co-inoculation. Total biomass was obviously higher than those of sole maize and soybean in the MNMP and co-inoculation treatment (P < 0.0001; Figure 1; Supplementary Table S1).

Nitrogen and P contents of all soybean and maize plants were significantly affected by N and P availability, along with AM fungal and rhizobium inoculation. This is consistent with the plant biomass results (Table 1). Moderate N and P application significantly enhanced N and P contents compared to low application rates regardless of inoculation treatment or cropping system (P < 0.0001; Tables 2 and 3; Supplementary Table S1). Both soybean and maize accumulated more N and P with MN and MP availability than those with LN and LP availability. Inoculations of soybean and maize with AMF and rhizobia significantly enhanced the contents of N and P in all cropping and nutrient treatments, except for the N and P contents of IM in the co-inoculation treatment reared in the LN treatment (P < 0.0001; Tables 2 and 3; Supplementary Table S1). With inoculation, the respective N contents of soybean and maize increased by 19–273 and 22–160%, and the respective P contents increased by 40–559 and 74–544% over non-inoculated controls. Moderate N and P application once again amplified the differences among inoculation treatments.

The cropping system also significantly affected both N and P contents in soybean and maize (P < 0.0001 for soybean; P < 0.05 for maize; Tables 2 and 3; Supplementary Table S1). N and P contents of soybean plants had a decreasing trend, whereas those of maize plants showed an increasing trend from monoculture to intercropping. Furthermore, this difference in N and P contents between monocultured and IM/soybean tended to increase with increasing N and P availability, as well as, with AM fungal and rhizobium inoculation. Compared with the corresponding monocultured plants, the respective increases in the N and P contents of IM plants were 19 and 32%, and the decreases in the respective N and P contents of IS were 17 and 19% in the co-inoculation treatment with MNMP availability (Tables 2 and 3). In addition, relative to monocultured maize and soybean, with co-inoculation, there was a 22% increase in the P content of IM in LN and moderate P (LNMP), along with 35 and 42% decreases in N and P contents, respectively, of IS reared in MNLP.

In the greenhouse experiment, all uninoculated plants were non-mycorrhizal (results not shown). All inoculated plants were well colonized by AMF, with AM colonization ranging between 74 and 87% in soybean, and between 76 and 86% in maize (Supplementary Figure S2). Even so, an effect of nutrient levels on AM colonization was observed (P < 0.05; Supplementary Table S1). In soybean, AM colonization of monoculture plants in MNMP was 14.8% lower than in MNLP given only AM fungal inoculation.

N and P application also affected soybean nodulation (P < 0.0001; Figure 2A; Supplementary Table S2). In comparison with the LNLP treatment, nodule fresh weights increased by 201 and 158% in LNMP, and by 70 and 48% in MNMP, whereas they decreased by 83 and 88% in MNLP for monocultured and IS, respectively. Nodule numbers also varied significantly among N and P treatments (P < 0.0001; Figure 2B; Supplementary Table S2), with decreases of 88 and 246% observed in MNLP relative to the number of nodules observed on soybean growing in LNLP under monocropping and intercropping conditions, respectively. Between the two cropping systems, there were no significant differences observed in nodule fresh weight or nodule number, except that the number of nodules on monocropped soybean was 88% higher than on IS growing with MNLP availability (Figure 2B).

To explore whether the growth differences observed between intercropped and the corresponding monocultured plants in co-inoculation treatments were due to N transfer through CMN after inoculation, the N transfer from soybean to maize was followed by ¹⁵N labeling in two compartment systems. As in the greenhouse experiment, inoculation with AMF and rhizobia significantly increased biomass and N content of soybean and maize irrespective of cropping system (P < 0.0001; Supplementary Figure S3; Supplementary Table S3). No significant difference in AM colonization was observed among treatments, except that colonization was lower in monocultured soybean inoculated only with AMF than it was in co-inoculated and IS and maize (Supplementary Figure S4; Supplementary Table S3). Biomass and N contents of maize plants were obviously higher when grown together with a neighboring soybean in co-inoculation treatments (P < 0.01; Supplementary Figure S3; Supplementary Table S3). In contrast, biomass and N content of IS showed the decreased trend compared with monocultured soybean (Supplementary Figure S3; Supplementary Table S3). As a consequence, soybean was, again, a weak competitor compared with maize in soybean/maize intercropping systems.

Biological nitrogen fixation was measured in the co-inoculation treatment. Soybean displayed a high N-fixing capacity, with more than 58% of its N being derived from the atmosphere (Table 4). This amounted to 13.4 mg plant^-1 in the co-inoculation treatment. Petiole injection of ¹⁵N in the soybean donor was correlated with the appearance of the label in the maize receiver plants in the intercropping system. The ¹⁵N Petiole injection of soybean resulted in a higher N transfer from soybean to maize in both the AM fungal inoculation alone and the co-inoculation treatments. Although there was no significant difference in percentage of N transfer (% N_transfer) between the AM fungal inoculation alone and co-inoculation treatments, the amounts of N transferred from soybean to maize were significantly different, with 1.3 and 2.0 mg plant^-1 observed, respectively (Table 4). Co-inoculation with AMF and rhizobia enhanced N transfer from soybean to maize, and the percentage N in maize derived from transfer (% NDFT) increased from 11.1% with only AM fungal inoculation to 15.2% with co-inoculation (Table 4).

To evaluate the impacts of carbon allocation on the competitive abilities of maize and soybean in the intercropping system, pulse labeling of ¹³C was performed with soybean and maize plants in a walk-in growth chamber. The results showed that ¹³C content in monocultured and IS plants increased by 200 and 194% in shoots, and 130 and 113% in roots, respectively, with co-inoculation compared with uninoculated control plants (P < 0.0001; Figure 3; Supplementary Table S3). However, nodule ¹³C cost was similar in monocultured and IS, and accounted 7% of soybean root ¹³C content, respectively, in co-inoculation treatment. Interestingly, the ¹³C content of IS roots in the AM fungal inoculation treatment exhibited a 34% decrease compared with uninoculated controls, as well as, 75 and 69% decreases in shoots and roots, respectively, compared to the co-inoculation treatment (Figure 3). In contrast, the ¹³C content in IM shoots increased by 83% with AM fungal inoculation compared with uninoculated maize plants (Figure 3). These results indicate that photosynthate assimilation is stimulated by AM symbiosis in maize, and by rhizobial symbiosis in soybean, especially in the intercropping system.