Diurnal switches in diazotrophic lifestyle increase nitrogen contribution to cereals_INSIGHT

Characterization of a GS variant conferring altered nitrogen regulation

Guided by the phenotypes of mutations in the glnA gene of S. typhimurium (encoding GS) that exhibit growth rates similar to the wild-type strain, but have lower levels of intracellular glutamine, we selected the glnA424 allele, originally identified in strain SK3130³⁰, for further investigation. Sequencing of the mutant allele revealed the amino acid substitution P95L (Fig. 2a), which is located in a highly conserved region of GS amongst γ-proteobacteria (Fig. 2b). The P95 residue is surface exposed in the structure of the S. typhimurium GS dodecamer³¹, and is not positioned within the active site formed between adjacent subunits. However, since P95 is located in a loop in the vicinity of the active site residue D65, the leucine substitution at position 95 could potentially have an impact on catalysis (Fig. 2c).

a Part of the amino acid sequence of GS showing the location of the P95L substitution (red lettering) encoded by the glnA424 allele, originally identified in the S. typhimurium strain SK3130³⁰. b Conserved sequence around position P95 (indicated in red) among diverse GS proteins (protein IDs provided in Supplementary Data 1). WebLogo 3 was used to draw the sequence logo. c The S. typhimurium GS dodecamer (PDB code 1F52) has 12 active sites formed between adjacent subunits within two eclipsed hexameric rings³¹. The P95 residue (red star) is surface exposed in the structure of the dodecamer and is in the vicinity (~3.0 Å) of the active site residue D65, which increases inter-subunit stability via interaction with R345. ADP in the active site is colored in yellow and the two manganese ions are shown in light brown. Note: the numbers in the expanded figure are the residues resolved in the structure, which are 1 number lower than in the protein sequence.

To further investigate the properties of this mutation, we engineered the glnA gene in E. coli strain NCM3722 to encode the GS-P95L substitution (designated as strain Ec424) (see Methods). In agreement with the previous results in S. typhimurium, the GS-P95L substitution in strain Ec424 exhibited little impact on aerobic growth at 37 °C in the presence of excess fixed nitrogen (10 mM ammonium) (Supplementary Table 1). In contrast to the intracellular level of glutamate, which was unchanged in comparison with the wild-type strain (designated Ec), the intracellular glutamine concentration in the Ec424 strain decreased by more than twofold, which correlated with increased levels of glnA expression, measured with a PglnA::lacZYA reporter (Supplementary Table 1, aerobic conditions). This implies that the lower levels of intracellular glutamine accumulated as a result of the GS-P95L substitution, enable activation of glnA expression in the presence of excess ammonium, suggesting that the Ntr system in strain Ec424 is a least partially blind to the nitrogen status. To investigate if these characteristics of the GS-P95L substitution were retained under conditions appropriate to analyze the regulation of nitrogen fixation, we repeated these experiments under anaerobic conditions in L medium at 30 °C in the presence of ammonia. This analysis was carried out at the lower temperature of 30 °C since nitrogen fixation by K. oxytoca is temperature sensitive. Similar results were obtained, but in this case, we observed a more significant decrease in the internal glutamine level in the Ec424 strain compared with the wild-type (~4.4-fold) and a greater impact on the growth rate, which may be a consequence of the decreased glutamine levels (Supplementary Table 1, anaerobic conditions).

The Ntr system controls around 2% of genes in the E. coli genome³², but these genes are differentially expressed in a sequential manner during the transition from high to low levels of glutamine, primarily dependent on the expression level and phosphorylation status of NtrC and the affinity of NtrC-P for its enhancer binding sites in the respective promoters. For example, the glnAp2 promoter, contains a potent enhancer with two high affinity binding sites, whereas the glnK, glnHp2, and nac promoters have weaker enhancers³³. Previous studies have suggested that glnAp2, glnK, glnHp2, nac, and serA promoters belong to different classes of NtrC-activated promoters and thus transcription from all 5 promoters was analyzed in order to assess the hierarchy of Ntr-dependent gene expression³⁴. We sought to examine whether the reduction in intracellular glutamine, resulting from the GS-P95L substitution, was sufficient to promote transcriptional activation of target NtrC-dependent promoters in the presence of ammonia. qRT-PCR experiments with RNA isolated from strains grown anaerobically at 30 °C (see Supplementary Method 1), revealed that the glnA promoters (glnAp1 and glnAp2) in the wild-type strain Ec are relatively insensitive to ammonia within the initial 1–5 mM concentration range as anticipated by the strong NtrC enhancers, and transcript levels increased in the Ec424 mutant strain as expected (Supplementary Fig. 1). We observed a ~3-fold increase of glnA transcript levels in Ec424 in the presence of 1 mM and 2 mM ammonium which decreased slightly to 2.5-fold when the ammonium concentration was raised to 5 mM. Similarly, activation of the glnH promoter in the wild-type strain was insensitive to ammonium within this concentration range, but was more strongly activated (by ~4-fold) in the Ec424 mutant at low ammonium concentrations. In contrast, the nac promoter was silent in the wild-type strain irrespective of the ammonium concentration, but large increases in nac transcription (~60-fold) were observed in the Ec424 mutant at 1 mM ambient ammonium concentration. This pattern was inverted for serA by a 7 to 50-fold decrease as expected, since the nac gene product, Nac, is a repressor of serA transcription. Similar to the nac promoter, activation of glnK transcription was inhibited by ammonium in the Ec wild-type strain, but was strongly activated by ~70-fold in the Ec424 mutant at 1 mM ambient ammonium concentration. This is important in the context of ammonia tolerant nitrogen fixation since, for example, strong activation of the glnK promoter by NtrC is required to express sufficient GlnK to prevent inhibition of NifA transcription by NifL in Klebsiella oxytoca (Fig. 1). Activation of the K. oxytoca nifLA promoter is also important in the context of nitrogen regulation of nif transcription since this promoter contains tandem low-affinity NtrC binding sites that are only activated by high concentrations of NtrC-P^35,36,37. To examine the influence of the GS-P95L substitution on nifLA transcription, we introduced plasmid pKU805 carrying a PnifLA::lacZYA fusion, into the Ec and Ec424 strains. The pattern of regulation in this case was similar to the nac promoter, with activation of nifLA transcription being strongly inhibited by ammonium in the wild-type strain, but activated in the Ec424 strain, particularly within the lower concentration range (1–2 mM ammonium by ~2.5 and ~1.3-fold respectively). Overall, these results demonstrate that the GS-P95L substitution influences the activation of NtrC-dependent promoters in the presence of ammonium to varying extents, presumably reflecting the lower glutamine level in the Ec424 strain, which in turn influences the phosphorylation status of NtrC, and its ability to activate promoters dependent on enhancer binding affinities.

The GS-P95L substitution confers ammonium tolerant nitrogen fixation and ammonia excretion in a temperature-dependent manner

Given that the P95L substitution in GS confers NtrC-dependent activation of the nifLA and glnK promoters in the presence of ammonia in E. coli (Supplementary Fig. 2), we sought to determine if nitrogen fixation is also ammonium tolerant in the Ec424 strain carrying the reconstituted nif gene cluster on plasmid pKU7824 (Fig. 3a). Nitrogenase assays (measured by acetylene reduction) were initially carried out under anaerobic conditions at 30 °C, which is conducive for nitrogen fixation, along with wild-type controls also carrying plasmid pKU7824. In contrast to the wild-type control in which nitrogenase expression and activity were prevented even at low levels of ammonia (1 mM), the Ec424 (pKU7824) strain exhibited nitrogenase activity in the presence of ammonia, although the activity declined with increasing ammonia concentrations (Fig. 3b). Interestingly, this pattern of nitrogen regulation parallels that of nifLA expression (Supplementary Fig. 1D), suggesting that ammonia tolerance might be limited by activation of the nifL promoter by NtrC-P. We also measured the external ammonia during the time course of the experiment and observed that less ammonia was consumed by the Ec424 (pKU7824) strain (Supplementary Fig. 2A). This suggests that this strain can fix nitrogen under nitrogen rich conditions, as anticipated from the acetylene reduction assays, although we cannot entirely rule out the possibility that the rate of ammonia consumption reflects differences in the growth rates of the strains. To our surprise, when these experiments were carried out at 23 °C, rather than 30 °C, we observed stronger tolerance to elevated ammonia concentrations in the Ec424 (pKU7824) strain, representing at 5 mM ammonium, about 70% of the nitrogenase activity observed in the wild-type strain (Ec (pKU7824)) in the absence of added ammonium (Fig. 3c). This was reflected by even lower consumption of external ammonium during the time course of the experiment (Supplementary Fig. 2B).

a Schematic diagram of the GS-P95L substitution in E. coli containing the K. oxytoca engineered nitrogen fixation gene cluster on plasmid pKU7824. The derivative of E. coli encoding the GS-P95L variant is designated as Ec424. b, c Acetylene reduction activities determined by the K. oxytoca nif gene cluster on plasmid pKU7824 in either wild-type E. coli (Ec) (black bars) or the GS-P95L variant of E. coli (Ec424) (orange bars), in response to the ammonia concentrations indicated on the x axes. Cultures were grown anaerobically in L medium either at 30 °C (b) or 23 °C (c) for 10 h prior to assay. Data are normalized to 100% of the nitrogenase activity of Ec (pKU7824) in the absence of ammonium (0 mM NH₄⁺) at each temperature. Means and SDs were calculated. n = 3 biologically independent samples. Source data are provided as a Source Data file.

We postulated that at the lower temperature, the GS activity of the P95L substitution could be further decreased, resulting in even further activation of the Ntr system as a consequence of lower glutamine levels signaling nitrogen deprivation. We also considered the possibility that further decreases in GS activity might uncouple ammonia assimilation from nitrogen fixation, resulting in ammonia excretion. When comparing the temperature dependency of growth, dependent upon nitrogen fixation with gaseous nitrogen as the sole nitrogen source, we observed that the Ec424 (pKU7824) strain exhibited significantly higher growth penalties compared with the wild-type strain Ec (pKU7824), as the temperature decreased (Supplementary Table 2). These temperature-dependent growth penalties were far less substantial when strains were grown aerobically in the presence of excess ammonia, suggesting that the growth defects might be associated with nitrogen fixation (Supplementary Table 3). Furthermore, at 27 °C and lower temperatures, the Ec424 (pKU7824) strain excreted ammonia. We observed a negative correlation between the amount of ammonia excreted and the growth temperature, with maximal ammonia excretion detected at 20 °C, which also correlated with the temperature-dependent decrease in growth rate (Supplementary Table 2). Thus Ec424 (pKU7824) excretes ammonia in a temperature-dependent manner.

To translate this potential ammonia delivery system from a model organism to a plant-associated diazotroph, we introduced the mutation encoding GS-P95L into the glnA gene of K. oxytoca strain M5al, thus generating a strain designated Ko424. In order to further stabilize the mutation, double nucleotide mismatches (CCA to CTG), instead of a single mismatch (CCA to CTA) were introduced at the mutation site of the glnA gene in K. oxytoca strain M5al (Fig. 4a). This strain also exhibited a temperature-dependent phenotype for ammonium tolerant nitrogen fixation, similar to that of the engineered E. coli strain Ec424 (pKU7824) (Fig. 4b, c). However, nitrogen fixation in this strain was notably more ammonium tolerant than in the E. coli strain at 23 °C, resulting in higher nitrogenase activities in the presence of ammonia and minimal consumption of exogenously added ammonium during the time course of the experiment. We also noted that the growth penalty of the Ko424 strain was less severe than that of the Ec424 (pKU7824) strain at temperatures above 25 °C (compare Table 1 and Supplementary Table 2). The Ko424 mutant strain also exhibited temperature-dependent ammonia excretion, with a positive correlation between decreasing growth rate and the amount of ammonia released, resulting in maximum excretion of ~3.3 mM NH₄⁺ at 20 °C (Table 1), compared with ~1.1 mM NH₄⁺ in E. coli (Supplementary Table 2). Time courses of ammonia excretion in both E. coli and the Ko424 strain cultured at 23 °C revealed that peak rates of ammonium accumulation occurred after maximum nitrogenase activities had been reached (Supplementary Fig. 3). The level of ammonia excretion also correlated with temperature-dependent decreases in the intracellular glutamine concentration in the Ko424 strain (from ~ 1.6 mM at 33 °C to ~0.5 mM at 20 °C) (Table 1 and Supplementary Fig. 4). This is consistent with increased ammonia tolerant nitrogen fixation at the lower temperatures resulting from signals of nitrogen deficiency, as a consequence of low glutamine levels. These results also imply that the catalytic activity of GS-P95L decreases significantly at the lower temperatures, thus decreasing the rate of ammonia assimilation by GS, resulting in ammonia excretion. In contrast to the differences in glutamine levels between the wild-type and mutant strains (~5.9-fold at 30 °C and ~6.3 fold 23 °C), glutamate concentrations were less than twofold lower in strain Ko424 compared with the wild-type strain Ko, consistent with maintenance of the glutamate pool irrespective of nitrogen deprivation³⁰. Overall, these results demonstrate that critical uncoupling between nitrogen fixation and nitrogen assimilation occurs in the Ko424 strain, resulting in ammonia excretion at temperatures lower than 30 °C, most likely as a consequence of decreased activity of the GS-P95L variant at the lower temperatures.

a Schematic diagram of the construction of the glutamine synthetase P95L substitution in K. oxytoca. The derivative of K. oxytoca encoding the GS-P95L variant is designated as Ko424. b, c Acetylene reduction activities of K. oxytoca (Ko) (black bars) or the GS-P95L variant Ko424 (green bars), in response to the ammonia concentrations indicated on the x axes. Cultures were grown anaerobically in L medium either at 30 °C (b) or 23 °C (c) for 10 hours prior to assay. Data are normalised to 100% of the nitrogenase activity of Ko in the absence of ammonium (0 mM NH₄⁺) at each temperature. Means and SDs were calculated. n = 3 biologically independent samples. d The ammonia excreted by Ko424 is derived from nitrogen fixation. Cultures were grown in nitrogen-free medium, in which air in the anaerobic flask was evacuated with a vacuum pump and then filled with either ¹⁵N₂ or ¹⁴N₂ as the sole nitrogen source as indicated on each spectrum. A further culture was incubated with the inert gas argon (Ar) in the headspace to provide a negative control. After 72 h of anaerobic cultivation of Ko424 at 23 °C, the supernatant was collected and deuterated dimethyl sulfoxide (DMSO) (10% final concentration) and concentrated hydrochloric acid (5% final concentration) were added. To obtain the peak map of the ¹H-NMR spectrum, samples were analyzed with a selective pulse sequence for ammonia using a high-resolution nuclear magnetic resonance spectrometer (600 MHz Bruker Avance NMR instrument)^27,38 as described in the Methods. Source data are provided as a Source Data file.

To demonstrate whether the ammonia detected in the culture supernatants originates from nitrogen fixation during anaerobic growth or for example, from turnover of other intracellular nitrogen sources, we incubated Ko424 in N-free L medium under an atmosphere of ¹⁵N₂ gas and demonstrated with ¹H-NMR³⁸ that the supernatant exhibited the same predominant doublet as the ¹⁵NH₄Cl standard (Fig. 4e). Thus, we conclude that the ammonia in the medium is derived de novo from nitrogen fixation by Ko424.

As both the expression and activity of GS are regulated in response to the nitrogen status, we considered the possibility that feedback regulation of the activity of the P95L variant via post-translational modification by the adenylyltransferase GlnE^17,18, might be responsible for the ammonia excretion phenotype. In order to investigate this, we constructed a derivative of strain Ko424 in which the glnE gene was deleted (designated Ko424ΔglnE). However, no detectable changes in ammonia excretion were observed in the Ko424ΔglnE strain when compared with the parental strain Ko424 at 23 °C (Supplementary Fig. 5). Thus, either the P95L variant of GS is not adenylylated by GlnE, or the intracellular glutamine concentration in this variant is not sufficient to activate the feedback mechanism, which seems highly likely given the low internal glutamine levels in Ko424.

Biochemical properties of the P95L variant of glutamine synthetase

To characterize the K. oxytoca GS-P95L enzyme in vitro, we expressed N-terminal his tagged derivatives of the wild-type and variant proteins in a glnA glnE double mutant of E. coli BL21 and purified both proteins by nickel affinity chromatography (see Supplementary Method 2). Measurements of GS biosynthetic activity using a phosphate release assay (see Supplementary Method 3) revealed that the K_m for glutamate for the wild-type GS enzyme varied with temperature (between ~11 mM at 20 °C to ~3.5 mM at 37 °C) in agreement with the published data at 25 °C for the E. coli GS enzyme (~4 mM¹⁷) (Supplementary Fig. 6A and Supplementary Table 4). (We note that the amino acid sequences of GS from both K. oxytoca and E. coli are very similar to each other, for details see Supplementary Fig. 7). The kinetic properties of the GS-P95L enzyme with respect to glutamate are clearly different, although it was not possible to accurately determine K_m values at the different temperatures, possibly because the enzyme activities were low (Supplementary Fig. 6B). At saturating glutamate concentrations (250 mM L-glutamate for details, see Methods), the apparent K_m for NH₄⁺ of the wild-type enzyme did not appear to differ substantially from 20 °C to 37 °C, but differential decreases in enzyme activity were observed between the wild-type and variant enzymes at 30 °C compared with 23 °C (Supplementary Fig. 6C and D). Overall, these data suggest that the GS-P95L enzyme is more temperature sensitive than wild-type GS in vitro, although the mechanistic basis for temperature sensitivity is not resolved.

Influence of temperature on ammonium donation by Ko424 to the eukaryotic alga Chlorella sorokiniana

We anticipated that ammonia excretion by Ko424 might support the growth of recipient organisms under nitrogen-limiting conditions in a temperature-sensitive manner. Ammonia excretion by Ko424 could be visualized by anaerobic co-culture with the photosynthetic eukaryotic alga C. sorokiniana on nitrogen-free solid L media, when grown in an illuminated cabinet with glucose as carbon source. Only trace growth of Chlorella adjacent to the Ko424 strain of K. oxytoca was observed at a constant temperature of 30 °C (Supplementary Fig. 8A) in contrast to incubation at 23 °C, where Chlorella adjacent to Ko424 turned green, as an indication of obvious growth. However, wild-type K. oxytoca enabled only trace growth of Chlorella at 23 °C, with no evidence of photosynthesis (Supplementary Fig. 8B). In agreement with the plate phenotype, the total amount of chlorophyll a synthesized by Chlorella adjacent to Ko424 was significantly higher at 23 °C than at 30 °C (Supplementary Fig. 8C). This implies that Ko424 expressing the P95L substitution in GS can support the growth of a eukaryotic alga in a temperature-dependent manner, presumably by providing fixed nitrogen at 23 °C.

We considered the possibility that oscillating temperature shifts might enable more nitrogen donation by the bacteria to Chlorella if the donor is given time to recover and proliferate at 30 °C, rather than being incubated continuously at 23 °C, since constitutive ammonia excretion at this temperature is likely to lead to severe nitrogen starvation in the bacteria. When we subjected the co-culture to 12 hr oscillating temperature shifts between 30 °C and 23 °C, the growth of Chlorella was apparently enhanced, resulting in an increase in the chlorophyll a content of Chlorella adjacent to Ko424 compared with incubation at constant 23 °C (Fig. 5a–d). In contrast Chlorella grew normally in the presence of ammonia with high chlorophyll levels, irrespective of the temperature (Fig. 5e–h). The importance of fluctuating temperature shifts in supporting nitrogen donation by Ko424 is further explored below.

a–c Growth promotion of the eukaryotic algae Chlorella sorokiniana by Ko424 and Ko in N-free L agar at constant 30 °C, constant 23 °C or with 12 h temperature shifts between 30 °C and 23 °C (30 °C−23 °C), respectively. Strains were streaked in a fan-like arrangement on the plates with C. sorokiniana alternating between Ko424 and Ko as indicated. Plates were incubated in transparent anaerobic tanks and grown under light illumination for 7 days. d Chlorophyll a content of C. sorokiniana adjacent to Ko424 at constant 30 °C, constant 23 °C or with 12 hr temperature shifts between 30 °C and 23 °C (30 °C−23 °C), respectively. The means and SDs were calculated. n = 4 biologically independent samples. Statistical significance is indicated as **P ≤ 0.01 analyzed using one-way ANOVA with Student’s t-test. P values from left to right: 0.000000094 (30 °C vs 23 °C), 0.00000000029 (30 °C vs 30–23 °C), and 0.00107 (23 °C vs 30–23 °C). e–g Growth of the eukaryotic algae Chlorella sorokiniana, Ko424, and Ko in L agar supplemented with 3 mM ammonium at constant 30 °C, constant 23 °C or with 12 hr temperature shifts between 30 °C and 23 °C (30 °C−23 °C), respectively. Plates were incubated in transparent anaerobic tanks and grown under light illumination for 7 days. h Chlorophyll a content of C. sorokiniana adjacent to Ko424 at constant 30 °C, constant 23 °C or with 12 hr temperature shifts between 30 °C and 23 °C (30 °C−23 °C), respectively. The means and SDs were calculated. n = 4 biologically independent samples. Statistical significance is indicated as **P ≤ 0.01 analyzed using one-way ANOVA with Student’s t-test. NS, non-significant. P values from left to right 0.0087 (30 °C vs 23 °C) and 0.011 (30 °C vs 30–23 °C). Source data are provided as a Source Data file.

Fixed nitrogen delivery to plants by K. oxytoca Ko424

When considering the potential benefit of strain Ko424 in agriculture as a plant associative diazotroph that exhibits thermosensitive ammonia excretion, we examined the detailed seasonal temperature profile for cereal plantations in central China. Interestingly, the average daily temperature shift between day and night is from 30 °C to 23 °C during the summer months (Supplementary Table 5), which fits well with the temperature profile of ammonia excretion by Ko424. As a prelude to performing plant experiments, we first examined the stability of the GS-P95L phenotype in both the Ec424 (pKU7824) and Ko424 strains after a more prolonged period of incubation at 23 °C. (Nucleotide substitutions encoding this GS variant in the different strains is shown in Supplementary Fig 9A). We observed in both cases that after reaching stationary phase, the ammonia excreted into the culture medium remained constant for up to 144 hr (Supplementary Fig. 9B). This implies that although the bacteria are subjected to severe nitrogen stress under these conditions, escape mutants that acquire the ability to reassimilate the excreted ammonia do not arise frequently in the population. In order to further consider the potential use of the bacteria as inoculants in the environment, we carried out stability studies of Ko424 after growth for up to 40 generations at either 23 °C or 30 °C (see Supplementary Method 4). In each case we screened 400 individual colonies of the mutant strain and observed that all of these remained competent to support the growth of C. sorokiniana (Supplementary Fig. 9C). Thus, reversion of the mutant phenotype was not detectable for up to 40 generations of bacterial cell proliferation.

To investigate the potential of Ko424 for plant growth promotion under conditions that simulate the summer temperature profile in central China, we carried out coculture experiments with japonica rice grown under hydroponic conditions with 30 °C day and 23 °C night temperature shifts (see Supplementary Method 5 and 6), representing the temperature profile in the air as well as in the hydroponic system (Supplementary Fig. 10). After an initial growth period of 4 days, the root system of the rice was incubated for 1 hour with a suspension of inoculant bacteria at a bacterial cell density of 10⁹ cells per mL. Plants were then transferred back to the hydroponic system and grown for a further 12 days prior to harvesting. In the absence of any added carbon or nitrogen, Ko424 increased the apparent plant biomass by 12% when compared with wild-type K. oxytoca and also significant increases in comparison with the uninoculated control (CK), or a derivative of the wild-type strain carrying a complete deletion of the nif gene cluster (KoΔnif) (Supplementary Fig. 11A and B) This plant growth promotion by Ko424 was also reflected in an increase in the total nitrogen content (18%) of rice plants in comparison with the wild-type strain (Supplementary Fig. 11C).

Similar plant growth experiments as described above were carried out with maize, but in this case as more maize cultivars were available to us, we performed preliminary screening with eight different maize inbred lines, in order to determine if Ko424 has a preference for specific plant genotypes. Maize plants were grown under hydroponic conditions with the same day-night temperature shifts used for the rice experiments, with no added carbon source (Supplementary Fig. 12). In general, wild-type K. oxytoca or the KoΔnif mutant provided little impact on maize growth compared with the uninoculated control, but three maize inbred lines responded positively to Ko424, exhibiting increases in dry weight of 14% (with B73), 20% (with Fu8701) and 29% (with 93-63) in comparison with the wild-type strain (Supplementary Fig. 12).

Subsequent experiments focusing on the maize 93-63 line, suggested that all the K. oxytoca strains reduced the oxygen concentration in the hydroponic system to microaerobic levels (~1–2%), which are appropriate for nitrogen fixation by this diazotroph. Moreover, a GFP-expressing version of wild-type K. oxytoca colonized the plant root surface, including the lateral roots and root hairs, suggestive of an efficient plant-microbe interaction (Supplementary Fig. 13). Guided by these findings, detailed experiments were carried out with the maize 93-63 inbred line in the absence of added nitrogen source, again applying temperature shifts between day (30 °C) and night (23 °C). Examples of the plant growth experiments are shown in Fig. 6. An independent iteration of these experiments was also carried out as a measure of reproducibility (Supplementary Fig. 14). The values reported below represent the range between the two independent iterations. After 9 days of bacterial inoculation, a small but statistically significant biomass increase was observed with K. oxytoca wild type (Ko), when compared with non-inoculated controls. However, as a similar increase was observed with the KoΔnif mutant control, this cannot be a consequence of BNF. A more significant increase in dry weight was observed with Ko424, which increased plant biomass by ~30–42% when compared with Ko (Fig. 6a, b and Supplementary Fig. 14A). When the nitrogen content of the plants was compared under the same conditions, only the Ko424 strain resulted in a significant increase in N content (Fig. 6c and Supplementary Fig. 14B). Compared with wild-type Ko, inoculation with Ko424 increased the nitrogen content by ~34–38%. When similar experiments were carried out at a constant temperature of 23 °C, smaller increases in dry weight and nitrogen content were observed with Ko424, when compared with every other control (Fig. 6d, e, Supplementary Fig. 14C and D). Notably, the N content of the maize line decreased to ~22–23% when co-cultured with Ko424 at 23 °C (11–16% less than observed with the diurnal 30 °C−23 °C temperature shift). Moreover, no significant increases in dry weight or nitrogen content were observed in comparison with the other controls when Ko424 was inoculated on maize grown at a constant 30 °C (Fig. 6f, g, Supplementary Fig. 14E and F). Overall, these results indicate that Ko424 can promote better growth and higher N content of maize line 93-63 when plants are provided with a diurnal 30 °C−23 °C temperature shift, compared with plants grown constantly at either 23 °C or 30 °C.

a Influence of inoculation with wild-type or mutant strains of K. oxytoca on the growth phenotype of the maize inbred line 93-63 after 9 days in the absence of added nitrogen source. CK indicates no bacteria were added; KoΔnif, inoculation with the nif gene cluster deletion mutant of K. oxytoca; Ko, inoculation with the wild-type strain of K. oxytoca; Ko424, inoculation with the GS-P95L variant strain of K. oxytoca. b, c Dry weight and nitrogen content, respectively, of maize after 9 days inoculation with or without the wild-type and mutant strains of K. oxytoca with 12 h temperature shifts between day (30 °C) and night (23 °C) (30 °C−23 °C). DW, dry weight. N content, nitrogen content. d, e Dry weight and nitrogen content, respectively, of maize after 9 days inoculation with or without the wild type and mutant strains of K. oxytoca at constant 23 °C. f, g Dry weight and nitrogen content, respectively, of maize after 9 days inoculation with or without the wild type and mutant strains of K. oxytoca at constant 30 °C. The means and SEMs were calculated. n = 5–6 biologically independent samples. Statistical significance is indicated as *P ≤ 0.05, **P ≤ 0.01 analyzed using one-way ANOVA with Student’s t-test. NS, non-significant. P values from left to rignt in b: 0.00031 (CK vs KoΔnif), 0.0099 (CK vs Ko), 0.00022 (CK vs Ko424), 0.021 (KoΔnif vs Ko424) and 0.018 (Ko vs Ko424). P values from left to rignt in c: 0.0061 (CK vs Ko424), 0.0047 (KoΔnif vs Ko424) and 0.012 (Ko vs Ko424). P values from left to right in d: 0.008 (CK vs Ko424) and 0.044 (KoΔnif vs Ko424). P values from left to rignt in e: 0.0041 (CK vs Ko424), 0.0043 (KoΔnif vs Ko424) and 0.034 (Ko vs Ko424). Source data are provided as a Source Data file.

To estimate the contribution of nitrogen fixation to plant growth we carried out ¹⁵N dilution experiments with the maize 93-63 inbred line grown in the presence of 0.5 mM ¹⁵N labeled nitrate, again applying temperature shifts between day (30 °C) and night (23 °C). An increase in dry weight was observed in the presence of Ko424 under these conditions, when compared with the other controls (Fig. 7a). Data from ¹⁵N isotope dilution experiments, calculated as the percentage of N derived from the atmosphere (%NF), showed that after 12 days of bacterial inoculation, Ko424 provided ~14% of plant N through BNF in comparison with reference plants inoculated with the non-nitrogen fixing strain KoDnif (Table 2). When the ¹⁴N in the seed is considered and subtracted, Ko424 provided ~26% of plant N through BNF in comparison with reference plants inoculated with the non-nitrogen fixing strain KoDnif (Table 2). Notably, although the wild-type Ko strain is a diazotroph, it did not result in significant ¹⁵N dilution, confirming that this strain does not donate nitrogen to maize via BNF.

a ¹⁵N-dilution experiments with maize inbred line 93-63 and different bacterial inoculants. 0.5 mM ¹⁵NO₃⁻ was added at the time of inoculation. Dry weight (DW) of maize after 12 days inoculation of maize with either wild type or mutant strains of K. oxytoca. The means and SEMs were calculated. n = 5–6 biologically independent samples. Statistical significance is indicated as **P ≤ 0.01 analyzed using one-way ANOVA with Student’s t-test. NS, non-significant. P values from left to right: 0.00088 (KoΔnif vs Ko424) and 0.00071 (Ko vs Ko424). b, c Direct ¹⁵N incorporation into maize from ¹⁵N₂. b Schematic diagram of the maize hydroponics system for ¹⁵N isotopic analysis. c Determination of fractional enrichment of the ¹⁵N isotope into pheophytin from maize inoculated with KoΔnif, Ko and Ko424. 50% of the co-culture gas was displaced with ¹⁵N₂ gas and the final oxygen concentration in the gas mixture was 20%. 1% CO₂ was added daily. The results were subtracted from the fractional ¹⁵N isotope enrichment of the uninoculated control. Asterisks indicate a statistically significant difference relative to Ko. The means and SEMs were calculated. n = 6 biologically independent samples. Statistical significance is indicated as **P ≤ 0.01 analyzed using one-way ANOVA with Student’s t-test. NS, non-significant. Two-sided P value: 0.0000036 (Ko vs Ko424). Source data are provided as a Source Data file.

To demonstrate that nitrogen incorporated into plant biomass is directly derived from nitrogen fixation by Ko424, we incubated either maize or rice plants in coculture with the bacteria in gas-tight bags, in which 50% of the atmosphere was displaced with pure ¹⁵N₂, supplemented with 1% CO₂. After 8 days incubation, we determined the ¹⁵N content of the plant-specific chlorophyll derivative pheophytin and observed 1.4% incorporation of ¹⁵N atoms in the total N of pheophytin for Ko424-inoculated maize (Fig. 7b, c), and 0.4% incorporation in the total N of pheophytin for Ko424-inoculated rice (Supplementary Fig. 15), both of which are statistically significant compared with the KoDnif inoculant and the Ko-inoculant. Although the level of incorporation of ¹⁵N into plant pheophytin was an order of magnitude lower than that demonstrated in the ¹⁵N dilution experiments, we postulate that this is a consequence of the artificial physiological environment imposed by incubating plants in a sealed environment³⁹ in the absence of added carbon source, where in our conditions plant biomass, chlorophyll content, nitrate assimilation and the ability of plant exudates to support the growth of K. oxytoca are significantly compromised (Supplementary Fig. 16). These results therefore confirm that fixed nitrogen derived from dinitrogen gas is directly transferred to plants cultivated in the presence of Ko424. However, given the limitations of both the ¹⁵N dilution and the ¹⁵N incorporation techniques, it is difficult to precisely quantify the amount of nitrogen directly transferred to the plants via BNF.