For the lentil × rhizobia strain experiment, the crop variety was Nepalese Simal, and there were four Rhizobium leguminosarum biovar viciae strains (wild type: 3841, VF39, 248; mutant 17-B) (Yost et al., 2006; Vanderlinde et al., 2011). For the cowpea × rhizobia strain experiment, the crop variety was Nepalese Surya and there were two strains of Bradyrhizobium yuanmingense (TTC9, TSC10), one strain of B. japonicum (DTB4) and one strain of B. elkanii (DTC9) (Sarr et al., 2009). For the soybean × rhizobia strain experiment, the crop variety was Nepalese Puja and there were three strains of B. japonicum (USDA 110, USDA 123, USDA 510) (Sadowsky et al., 1987; Bhagwat et al., 1991; Kaneko et al., 2002) and two strains of B. elkanii (USDA 94, USDA 76) (Kuykendall et al., 1992). Different rhizobia strains for each crop were selected to capture a diversity of nodulation and N fixation efficiency.

Seeds of all three crops were surface-sterilized with 70% ethanol for 2 min, 4% sodium hypochlorite for 3 min, and washed with six changes of sterile distilled water. Seeds were pre-germinated on sterilized, wetted filter paper in the dark at 28°C for 2 days, and individual seeds were transferred to germination pouches (17.8 × 16.5 cm, Mega International, Minneapolis, MN, United States) on light shelves containing 50 ml of double distilled water. Each plant was grown in a separate germination pouch. There was only one treatment (one rhizobia strain or non-inoculated control) per pouch, and each germination pouch was considered to be a single replicate. To further reduce cross-contamination, pouches were used that were sealed at the bottom, and nutrient solutions were not shared between treatments. For the lentil and cowpea studies, per treatment, five to six germination pouches were placed in a stand and placed in an open tupperware container. Different rhizobia treatments were not placed within the same container to avoid cross contamination of rhizobia strains. There were three containers per treatment, which were continuously randomized within and across light shelves. For the soybean study, five rhizobia treatments and the non-inoculated control (in individual germination pouches) were randomly allocated within a plastic container. There were eight trays (n = 8), which were continuously randomized within and across light shelves. After 1 week of germination, the water was removed and plants were supplied with 50 (lentil) and 100 ml (cowpea and soybean) of one-fourth strength N-free Hoagland’s nutrient solution (pH = 6.6), containing 0.5 mM K¹⁵NO₃ (98 atom% ¹⁵N; 335134-1G; Sigma Aldrich, Oakville, ON, Canada) as starter N. One week later, this nutrient solution was removed and replaced with 50 ml of the one-fourth strength N-free Hoagland’s solution; the nutrient solution was replaced each week. Lentil plants were grown at 23 ± 2°C with supplemental lighting (range: 180–200 μmol m^-2 s^-1 at the top of the growth pouch, EcoLux SP65, 40W, F40SP65ECO), maintaining a photoperiod of 16 h/8 h light/dark cycles. Cowpea and soybean plants were grown at 23 ± 2°C with supplemental lighting (range: 250–300 μmol m^-2 s^-1 at the top of the growth pouch, EcoLux SP65, 40W, F40SP65ECO), maintaining a photoperiod of 16 h/8 h light/dark cycles. For the rhizobia treatments, strains for lentil (R. leguminosarum biovar viciae strains wild type: 3841, VF39, 248; mutant 17-B), cowpea (B. yuanmingense TTC9, B. yuanmingense TSC10, B. japonicum DTB4, B. elkanii DTC9), and soybean (B. japonicum USDA 110, B. japonicum USDA 123, B. japonicum USDA 510, B. elkanii USDA 94, B. elkanii USDA 76) were grown in tryptone-yeast (TY) extract and modified arabinose-gluconate medium (MAG) (cowpea and soybean) agar media, respectively, in Petri dishes for 3 days at 30°C. To prepare each inoculum, the agar plates were scraped and cells added to sterilized ddH₂O, adjusted to a cell density of OD₅₉₅ = 0.1. Plants were inoculated 1 week after plant growth with 1 ml of each inoculum, placed directly on roots. All the three legume species had a negative control, comprised of plants that were not inoculated with rhizobia.

Eight healthy representative plants per treatment were sampled from lentil, cowpea, and soybean 4 weeks after inoculation with rhizobia as indicated. Leaf punches were collected from the fully expanded youngest leaf, across the leaflet main vein from lentil (from one leaflet at the tip), cowpea, and soybean using a 3 mm (lentil) and 6.35 mm (cowpea and soybean) hand puncher (235270975; Fiskars Brands Inc., Middleton, WI, United States) and immediately frozen in liquid N. Leaf punches were stored at -80°C.

The same eight plants per treatment harvested for GlnLux measurements were used. The number of nodules per plant was counted manually. Roots were scanned using an Epson Expression 1640× scanner (Epson Canada Ltd., Markham, ON, Canada), and a detailed root morphological analysis was undertaken using WinRHIZO software (Regent Instruments Inc., Quebec City, QC, Canada) including root volume, total length, surface area, and average diameter. Shoot and root dry weights were measured after drying the plant materials in an oven at 60°C for 3 days.

The same eight plants per treatment harvested for GlnLux measurements were used. The dried shoot samples (see above) were ground using a coffee grinder followed by a Bead Ruptor 12 Homogenizer (OMNI International, Kennesaw, GA, United States) and analyzed for ¹⁵N and total N% using a mass spectrometer (Costech ECS4010 Elemental Analyzer coupled to a Delta V mass spectrometer, Costech, CA, United States) at the Stable Isotope Facility, University of Saskatchewan, and University of British Columbia, Canada, using a standard protocol (Thilakarathna et al., 2012, 2016b). The %Ndfa of the lentil and soybean was calculated using the following formula according to the isotope dilution technique:

where atom % ¹⁵N excess = atom % ¹⁵N_{(lentil/soybean)} - 0.3663.

The amount of shoot N derived from SNF was calculated based on the shoot N content and %Ndfa (shoot N content × %Ndfa/100).

Construction of the GlnLux biosensor strain was previously reported (Tessaro et al., 2012). Briefly, a Gln-auxotrophic E. coli strain (JW3841-1, KanR) was generated by inserting a kanamycin cassette into GlnA [glnA732(del)::kan] (Baba et al., 2006). The strain was subsequently transformed with ampicillin-resistant plasmid pT7-lux (Meighen and Szittner, 1992) which contained a constitutive T7 promoter from Xenorhabdus luminescens driving expression of the luxCDABE operon from Vibrio fischeri to create strain GlnLux.

GlnLux bacteria were cultured in LB medium, which consisted of 5 g/l NaCl (BP358-212, Fisher Scientific), 5 g/l yeast extract (DF0127179, Fisher Scientific), and 10 g/l tryptone (BP1421-500, Fisher Scientific), with or without 12 g/l Bacto-Agar (BD; DF0140010, Fisher Scientific), pH 7.2. M9 minimal medium consisted of 20 ml/l 20% (w/v) D-(+)-glucose (G5767, Sigma), 100 μl/l 1 M CaCl₂ (C-79, Fisher Scientific), 2 ml/l 1 M MgSO₄ (230391, Sigma), and 200 ml/l 5× M9 salts (A-0171, Sigma), pH 7.0. All liquid and solid plate media were supplemented with 50 μg/ml kanamycin monosulfate (K378, PhytoTech, United States) and 100 μg/ml carbenicillin disodium salt (C346, PhytoTech, United States) to select for the disrupted glnA chromosome and reporter plasmid, respectively.

Cells were prepared as previously described (Tessaro et al., 2012). Briefly, GlnLux bacteria were inoculated into 15 ml of LB medium in a 50 ml Falcon tube, and incubated overnight at 37°C with shaking at 250 × g. The culture was spun down at 2500 × g at 21°C for 10 min and the supernatant was decanted. The culture was then washed 3× in sterile M9 minimal medium with centrifugation as above. Finally, the culture was resuspended in 15 ml sterile M9 minimal medium in a 50 ml Falcon tube, and the GlnLux density was adjusted to OD₅₉₅ = 0.025 using sterile M9 minimal medium. The resuspended GlnLux culture was incubated at 37°C with shaking at 250 × g for 14 h to deplete any endogenous Gln.

The procedure was adapted from a previous protocol (Tessaro et al., 2012). Individual frozen leaf punches were ground in a 2 ml conical bottom microcentrifuge tube placed on ice using a micropestle (K7495150000, Kimble Chase, Fisher Scientific) with silica sand in 20 μl of 0.1% final (v/v) protease inhibitor cocktail (PIC) for plant cell extracts (100% stock; #P9599, Sigma). Plant extracts were centrifuged for 20 min at 4°C at 13,000 × g, and the supernatant was transferred to a microcentrifuge tube placed on ice. The plant extracts were diluted 100-fold in ddH₂O. The diluted plant extracts were used for luminometer assays.

For luminometer assays, white opaque 96-well reader plates (#07-200-589, Fisher Scientific) were loaded with 160 μl/well M9 minimal medium followed by 20 μl/well of each plant extract. Finally, 20 μl of 14 h-Gln-depleted GlnLux culture (pre-depletion OD₅₉₅ = 0.025) was added to each well. Plates were sealed with non-breathable sterile film (#361006008, Fisher Scientific) to prevent media evaporation, and centrifuged for 5 s at 2000 × g to mix the GlnLux bacteria with plant extracts and M9 minimal medium. The plates were incubated at 37°C without shaking. For lux quantification, plates were read in a MicoLumatPlus LB96V luminometer with WinGlow Software (Berthold Technologies, Germany). Samples were read after 2 h of incubation for 1 s in a luminometer chamber temperature of 37°C in an endpoint assay using the integrate software function. Readings were taken every hour after incubation until lux values saturated. The plates were transferred back and forth from a 37°C incubator, and non-breathable films were replaced following each read. The 0 μg/ml Gln standard reading was subtracted from all lux values.

To test the biosensor response to exogenous Gln, six different concentrations of Gln standards (L-Gln G229, PhytoTech, 0, 625 × 10^-9, 125 × 10^-8, 25 × 10^-7, 5 × 10^-6, 1 × 10^-5 M) were also tested in the same plates along with plant samples, following the same method (Supplementary Figure S1).

The effect of different rhizobia strains was analyzed using analysis of variance (ANOVA) set at p < 0.05. Means were compared using Tukey. To test for the linearity of the biosensor response, linear regression analysis was performed using a Goodness of Fit (R²) test. Total shoot N%, amount of shoot N fixed, and %Ndfa versus GlnLux correlation analysis were performed using the Pearson correlation test. All statistical analyses were performed using GraphPad Prism Software (v5, GraphPad Software, United States).

To test the GlnLux leaf punch assay with lentil, four strains of R. leguminosarum bv. viciae (VF39, 248, 3841, and 17-B), which differ in their N-fixation capacities, were inoculated onto a variety of lentil grown by smallholder Nepalese hillside farmers (Figure 2A). The different rhizobia strains all produced a similar number of root nodules (Figure 2B and Supplementary Figure S2A). A single punch per plant from a young leaf (from low-N grown lentils) was ground in PIC, co-incubated for 2 h with GlnLux cells in 96-well plates, and then the lux signal quantified after a 2-min read (Figure 1B). The GlnLux output (Figure 2E) showed a similar trend as the shoot N% (Figure 2F), and the percentage N derived from the atmosphere (%Ndfa) (Figure 2G). GlnLux output strongly correlated with SNF measured using the ¹⁵N stable isotope assay (%Ndfa) (Pearson correlation coefficient R² = 0.9972) (Figure 2H), plant shoot N% (R² = 0.8910) (Figure 2I), and amount of shoot N fixed (R² = 0.890) (Figure 2J). The GlnLux result was not an artifact of altered Gln demand, as significant differences were not observed among the different rhizobia treatments for normalized shoot dry weight (Figure 2C), normalized root dry weight (Figure 2D), normalized total dry weight, shoot/root biomass ratio, root length, root surface area, or volume (Supplementary Figures S2B–F).

To test the GlnLux leaf punch assay with cowpea, a ureide exporter, two strains of B. yuanmingense (TTC9 and TSC10), one strain of B. japonicum (DTB4) and one strain of B. elkanii (DTC9), which differ in their nodulation capacity, were inoculated onto a variety of cowpea grown by smallholder Nepalese hillside farmers (Figure 3A). The different rhizobia strains produced a significantly different number of root nodules (P < 0.0001) (Figure 3B and Supplementary Figure S3A). The GlnLux output showed a similar trend (Figure 3E) as the shoot N% (Figure 3F) and the %Ndfa (Figure 3G). GlnLux output correlated with SNF measured using the ¹⁵N stable isotope assay (%Ndfa) (Pearson correlation coefficient R² = 0.7038) (Figure 3H), plant shoot N% (R² = 0.9873) (Figure 3I), and amount of shoot N fixed (R² = 0.8641) (Figure 3J). Significant differences were not observed among the different rhizobia treatments for normalized root dry weight (Figure 3D), root length, root surface area, or volume (Supplementary Figures S3B–D). Plants inoculated with strain TTC9 had the highest shoot (Figure 3C) and total biomass (Supplementary Figure S3E), whereas the lowest biomass and shoot/root ratio were associated with strain DTC9 (Supplementary Figures S3E,F).

To test the GlnLux leaf punch assay with soybean, another ureide exporter, a Nepalese subsistence farmer variety of soybean was inoculated with three strains of B. japonicum (USDA 110, USDA 123, 510) and two strains of B. elkanii (USDA 94 and USDA 76), which differ in their N-fixation capacities (Figure 4A). The different rhizobia strains produced a significantly different number of root nodules (P < 0.0001) (Figure 4B and Supplementary Figure S4A). The GlnLux output showed a similar trend (Figure 4E) as the shoot N% (Figure 4F), and the %Ndfa (Figure 4G). GlnLux output strongly correlated with SNF measured using the ¹⁵N stable isotope assay (%Ndfa) (Pearson correlation coefficient R² = 0.9772) (Figure 4H), plant shoot N% (R² = 0.9805) (Figure 4I), and amount of shoot N fixed (R² = 0.9747) (Figure 4J). Significant differences were not observed among the different rhizobia treatments for root length, root surface area, or volume (Supplementary Figures S4B–D). However, slight differences were found among the rhizobia strains for normalized shoot (Figure 4C), root (Figure 4D), and total dry weight (Supplementary Figure S4E). Plants inoculated with USDA 110 (high fixer) had the highest shoot biomass, total biomass, and shoot/root ratio (Supplementary Figure S4F).

Nitrogen fixed by rhizobia inside legume nodules is secreted as ammonium into the host plant cell, where it rapidly assimilated into Gln (Atkins, 1987). Gln is further metabolized into asparagine (Asn) in amide exporting legumes or as the ureide compounds, allantoin and allantoic acid, in ureide exporting legumes. Amide exporting legumes then transfer the majority of their fixed N through xylem as Asn (80%) and Gln (10%), whereas in ureide exporting legumes, the major export compounds are allantoin and allantoic acid (90%) (Schubert, 1986; Unkovich et al., 2008; Aranjuelo et al., 2011; Howieson and Dilworth, 2016). In ureide exporting legumes, aside from the ureides, Gln and Asn are the major amino acids found in xylem sap, as well as minor proportions of aspartic acid, lysine, valine, leucine, and glutamic acid (Pate et al., 1980). Indeed, xylem Gln as an immediate product of N fixation was shown to be a useful indicator of SNF activity across various legumes (Amarante et al., 2006; Amarante and Sodek, 2006; Justino and Sodek, 2013). Xylem sap can be obtained either from the stump of intact roots following decapitation of the shoot (root bleeding sap), or from freshly harvested shoots by applying a mild vacuum (vacuum-extracted sap) (Unkovich et al., 2008). In the former method, root xylem is sampled, whereas in the latter, shoot xylem is sampled. Different amide exporting legumes were shown to transfer around 10% of fixed N as Gln in the absence of external N (i.e., when SNF dependent), and low genotypic variability was shown to exist for the concentration of Gln in xylem sap across legumes (e.g., comparing Pisum sativum, Lupinus albus, and Crotalaria juncea) (Amarante et al., 2006). Similarly, low genotypic variability was found for the Asn:Gln ratio among different ureide exporting legumes (Vigna angularis, G. max, V. unguiculata, Cyamopsis tetragonoloba, Macrotyloma uniflorum, Psophocarpus tetragonolobus, Vigna umbellata, Vigna mungo) when plants were grown without external N (Pate et al., 1980). Nevertheless, there is variation in the type and amount of organic N exported from nodules and transferred through xylem in different legumes (Udvardi and Poole, 2013). Furthermore, complex amino acid cycling takes place in legumes in order to maintain SNF while allowing the host plant to regulate the symbiotic process (Lodwig et al., 2003). Interestingly, SNF output measured using the GlnLux biosensor method was highly correlated to relative Gln output in both the amide and ureide exporting legumes tested. Therefore, the GlnLux results suggest that Gln can be a useful indicator of symbiotic N fixation across various legumes under controlled conditions.

In general, the N transport form in xylem sap changes based on soil and environmental factors (rainfall/water availability, soil nutrient availability, salinity) (Neo and Layzell, 1997; Cramer et al., 2002; Amarante et al., 2006; Bai et al., 2007; Ladrera et al., 2007; Gil-Quintana et al., 2012, 2013). The experiments in this study were undertaken under optimal, controlled conditions with minimal external N. It is important to note that external N (e.g., soil N) is also assimilated into Gln, which has the potential to confound interpretations of GlnLux data with respect to SNF. For example, in soybean, a ureide exporter, exposure to NH₃ was shown to result in increased phloem Gln (Neo and Layzell, 1997) whereas another study showed similar xylem Gln levels when plants were supplied with NO₃ (20 mM KNO₃) compared to non-fertilized plants (McClure et al., 1979). A higher Asn:Gln ratio was observed in eight ureide exporters exposed to NO₃, shown to be as a result of decreased Gln relative to Asn in cowpea (Pate et al., 1980). Xylem sap Gln was shown to change less significantly in amide exporting legumes in response to the source of N (SNF-dependent or external N) compared to ureide exporting legumes (Atkins et al., 1979; Amarante et al., 2006), likely because there is a larger relative background pool of Gln in the former.

In addition to external N, drought stress reduces SNF in legumes (Zahran, 1999; Giller, 2001; Kunert et al., 2016). Generally ureide exporting legumes (soybean, common bean) were shown to be more sensitive to drought stress compared to amide exporters (lentil, lupin) (Arrese-Igor et al., 2011). Therefore, the GlnLux assay may work more reliably with amide exporters than ureide exporters under drought stress conditions. Drought stress was shown to change Gln concentrations in amide exporters (Gil-Quintana et al., 2012) and ureide exporters (Silvente et al., 2012). The alteration in Gln concentration under drought stress varied by plant tissue (leaf/stem/root/nodule) (Ladrera et al., 2007; Gil-Quintana et al., 2012, 2013). It was found that under drought stress, leaf Gln content either decreased or remained close to the control (well-watered plants) in Medicago truncatula (amide exporter) (Gil-Quintana et al., 2012) and soybean (Silvente et al., 2012). Reallocation of leaf N compounds in legumes was shown to occur under drought conditions due to up-regulation of Gln synthetase (Aranjuelo et al., 2011). Further accumulation of total amino acids (Gil-Quintana et al., 2012) and proline during drought stress is a common scenario, where proline has been shown to act as an osmoregulant (Aranjuelo et al., 2011).

In addition, stress conditions that impair SNF can change the Gln concentration in xylem sap. For example, xylem Gln concentrations were shown to be reduced under water logging conditions in both amide and ureide exporting legumes (Amarante et al., 2006; Amarante and Sodek, 2006; Justino and Sodek, 2013). Interestingly, a similar pattern of reduction in xylem Gln was observed in both amide exporting (P. sativum, L. albus, C. juncea) and ureide exporting legumes (G. max, Phaseolus vulgaris, V. unguiculata) under water logging conditions similar to the reduction in SNF activity (Amarante et al., 2006). Therefore, in this study, Gln measurements were shown to be a good indicator of reductions in SNF under waterlogging stress.

Finally, Gln concentration can vary during the growth stage of legume plants. For example, Pate et al. (1980) found that the Asn:Gln ratio increased as cowpea and mung bean plants matured. However, the study did not show whether changes in the Asn:Gln ratio were due to decreased Gln or increased Asn.

The GlnLux leaf punch assay does not provide a direct measurement of SNF (%Ndfa) compared to ¹⁵N-based methods (Table 1). Another disadvantage is that GlnLux reports only the relative amount of Gln, not absolute amount. However, a concentration gradient of pure Gln standards can be used along with the leaf punch assay (Supplementary Figure S1), which was highly replicable. Although we found very strong correlation between GlnLux and ¹⁵N based %Ndfa, the GlnLux method is a temporal snapshot and does not provide an integrated measurement of accumulated SNF output over time (Table 1). When collecting samples, leaf punches have to be flash-frozen in liquid N and kept frozen until processing, which limits its applicability to field conditions. Finally, as noted above, stress conditions such as drought can alter the leaf Gln concentration in legumes, which challenges the use of GlnLux to infer SNF under unpredictable conditions. Given this caveat, and since the leaf punch method cannot distinguish between N derived from SNF versus soil (Table 1), it may be most suited for pre-screening rhizobia inoculants under controlled N supply conditions prior to field experiments. We also expect that the GlnLux leaf punch method will assist with basic research especially to understand the unsolved mechanisms involved in the maintenance of SNF by facilitating high-throughput bacteria/plant mutant screens.