Genetic remodeling of soil diazotrophs enables partial replacement of synthetic nitrogen fertilizer with biological nitrogen fixation in maize

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Scientific Reports volume 14, Article number: 27754 (2024 ) Cite this article apple trees fertilizer

Increasing biological nitrogen (N) fixation (BNF) in maize production could reduce the environmental impacts of N fertilizer use, but reactive N in the rhizosphere of maize limits the BNF process. Using non-transgenic methods, we developed gene-edited strains of Klebsiella variicola (Kv137-2253) and Kosakonia sacchari (Ks6-5687) bacteria optimized for root-associated BNF and ammonium excretion in N-rich conditions. The aim of this research was to elucidate the mechanism of action of these strains. We present evidence from in vitro, in planta and field experiments that confirms that our genetic remodeling strategy derepresses BNF activity in N-rich systems and increases ammonium excretion by orders of magnitude above the respective wildtype strains. BNF is demonstrated in controlled environments by the transfer of labeled 15N2 gas from the rhizosphere to the chlorophyll of inoculated maize plants. This was corroborated in several 15N isotope tracer field experiments where inoculation with the formulated, commercial-grade product derived from the gene-edited strains (PIVOT BIO PROVEN® 40) provided on average 21 kg N ha-1 to the plant by the VT-R1 growth stages. Data from small-plot and on-farm trials suggest that this technology can improve crop N status pre-flowering and has potential to mitigate the risk of yield loss associated with a reduction in synthetic N fertilizer inputs.

Synthetic N fertilizers derived from the Haber–Bosch fixation of dinitrogen gas (N2) are credited to have enabled the sustenance for about half of the world’s population1, contributing circa 48% of the N harvested in major cereal crops worldwide2,3,4,5. However, this benefit has come with steep environmental costs. This is because only a portion of applied synthetic N fertilizer is typically recovered by cereal crops6,7, while the excess reactive N is prone to loss from agricultural soils. Nitrate (NO3) leachate pollutes freshwater sources and marine ecosystems7,8, and soil nitrous oxide (N2O) emissions are an important contributor to stratospheric ozone depletion9and a potent greenhouse gas (GHG)10. Soil N2O emissions associated with N fertilizer applications, along with the embedded GHG emissions from their manufacturing, transport and application, collectively account for nearly 2.5% of global GHG emissions11.

Reducing synthetic N fertilizer use is considered key to effectively reducing NO3 leaching and N2O emissions11,12. However, doing so without negatively impacting cereal yields is challenging, as it requires tight synchronization between N supply and crop N demand to achieve sustainable N fertilizer use efficiency13,14,15. Various strategies have been proposed to accomplish this, including fine tuning crop N supply by improving fertilizer timing, rate, source, and placement16; optimizing the timing of crop N demand via higher planting density, earlier planting dates, and crop genetic improvement17; and enhancing soil mineral N retention for subsequent crop uptake with crop rotations, cover crops, reduced tillage, and drainage management18. While several of these strategies have been shown to improve N fertilizer use efficiency13, there remain critical socio-economic barriers to grower adoption13,19. Consequently, the United States (U.S.) and other industrialized countries have made little progress towards reducing N fertilizer consumption2,13,20. Therefore, there is a need for innovative solutions that enable sustainable reductions in N fertilizer use to abate its environmental impacts without compromising crop productivity.

Biological N fixation (BNF) is one of the dominant pathways for acquiring newly available N in most terrestrial ecosystems and is a major supplier of N in legume crops21,22. Although cereals lack the specialized structures needed to form symbiotic relationships, BNF performed by free-living, associative, and endophytic diazotrophs can still make significant contributions to cereal N uptake in some conditions23; nearly a quarter of the N harvested globally in maize, wheat and rice is estimated to have originated from BNF from 1961–20105. However, nitrogenase, the main enzyme responsible for the reduction of N2to ammonia, is downregulated by the presence of reactive N21, such that N fertilizer application greatly reduces the contribution of BNF to cereal N uptake24. Thus, BNF has not been compatible with modern cereal production that is heavily reliant on N fertilizers. Augmenting BNF in these conditions could diminish our reliance on energy- and GHG-intensive synthetic N sources. Additionally, because BNF performed by root-associated or endophytic diazotrophs has the potential to better match crop N uptake demand spatially and temporally compared to other sources of N22,25, the risk of environmental loss of reactive N could be mitigated.

Re-engineering diazotrophs’ genetic constraints can enhance BNF under high N fertility conditions25,26,27. Previous research27 demonstrated the successful genetic remodeling of the root-associated diazotroph Klebsiella variicola (Kv137-1036) isolated from field-grown maize. Removing the repression of nitrogenase in N-rich environments (i.e., derepression) in its parent strain, Kv137, resulted in > 100-fold increase in nitrogenase activity in vitro in N-rich media27. Furthermore, Kv137-1036 exhibited colonization levels similar to Kv137 on the roots of three-week-old maize plants while displaying higher BNF activity in the rhizosphere. The commercial utility of the formulated product (PIVOT BIO PROVEN®) was validated for biosafety, product stability, and an average grain yield boost of 0.3 \(\pm\) 0.1 Mg ha-1. Additionally, there was an 8 to 25% reduction in within-field yield variability relative to an uninoculated control in multiyear and multisite commercial-scale on-farm field trials27. Although Kv137-1036 was confirmed to excrete ammonium (NH4+) due to passive excretion driven by higher rates of BNF27, further increases in NH4+ excretion may be possible with targeted edits to downregulate N assimilation, and thereby increase N efflux into the maize rhizosphere. Further in-vitro research with a lineage of Kosakonia sacchari (Ks6) showed robust colonization of the maize rhizosphere across diverse field conditions and elevated levels of BNF in the presence of exogenous N with remodeled strains26, providing additional avenues for increasing BNF in fertilized cereal crops.

Although the N-fixation potential of these genetically remodeled strains under N-rich conditions has been confirmed in vitro26,27, and maize grain yield advantages have been observed in the case of Kv137-103627, direct validation of plant-associated BNF (i.e., crop assimilation of fixed N2) is still needed. Furthermore, a deeper understanding of how inoculation affects maize plant N status and productivity can provide us with a roadmap for decoupling synthetic N fertilizer use from cereal grain production (Fig. 1a). The aim of this research was to elucidate BNF as the mechanism of action of two genetically remodeled strains (Fig. 1b-c), Kv137-2253 and Ks6-5687 (previously referred to as PBC6.99 in ref.26) for improving N supply and productivity in maize. (Details on the genetic remodeling strategy pursued to engineer these strains is provided on Fig. 1b-c and supplemental information text S1.) To do this, we used data from in vitro, in planta, and field experiments, and evaluated the response of maize plants to BNF, as well as the potential of the formulated, commercially available inoculant (PIVOT BIO PROVEN® 40, hereafter referred to as PROVEN 40) to replace a portion of synthetic N fertilizer inputs. Establishing evidence for the latter has tangible and immediate implications for mitigating reactive N pollution and GHG emissions associated with the use of synthetic N fertilizers.

Redesigning maize nitrogen (N) nutrition with genetically remodeled root-associated soil diazotrophs for biological N fixation (BNF) under N-rich conditions. (a) Conceptual diagram of expected changes to N cycling in maize cropping systems resulting from better synchronization of crop N demand and N supply, and reduction in synthetic N requirements, and associated environmental N losses. (b) BNF and its regulatory pathways. NifA is the transcriptional activator of the nitrogenase complex and its accessory genes, collectively named nif regulon. Under abundant N sources, NifL protein inhibits NifA’s transcriptional activity. This inhibition is alleviated when low glutamine levels activate the Urydylyl-transferase activity of GlnD, the glutamine-sensing bifunctional protein (dashed green lines). GlnD, in turn, urydylylates PII signaling proteins, which represses the inhibitory effect of NifL and increases the expression of nifA (green arrows). Active NifA leads to the expression of nif regulon and BNF. The uridylated PII proteins also activate the adenylyl removal function of the GlnE protein, which in turn activates the glutamine synthetase (GlnA) activity for the assimilation of the fixed N via glutamine. Increased glutamine levels trigger the uridylyl removal activity of GlnD, which removes the uridylyl groups from the PII signaling proteins. In this study, we have edited four gene targets (numbered circles) to bypass the regulations and enable constitutive BNF (yellow circles) and ammonium excretion (green circles) in the diazotrophic strains. (c) Description of the derepression and excretion edits. To constitutively express nifA, we have deleted the negative regulator, nifL (1), and introduced a strong constitutive promoter upstream of the nifA gene (2). To decrease assimilation and increase ammonium excretion, we deleted the whole glnD gene in Ks6-5687 (3a) or removed its uridylyl transferase (UT) domain in Kv137-2253 (3b). Lastly, to further downregulate the assimilation of ammonium via glutamine synthetase (GlnA), we inactivated this protein by removing the adenylyl removal domain of the glnE gene, which kept GlnA in the inactive, adenylated state (4).

In-vitro acetylene reduction assay confirmed that Ks6-5687 and Kv137-2253 exhibit BNF activity under both N-free and N-rich conditions, whereas the wild-type Kv137 and Ks6 strains displayed BNF only in the absence of N (Fig. 2a). However, the edited Kv137-1036 and Kv137-2253 strains still showed some repression of nitrogenase in the presence of NH4Cl. To amplify NH4+ excretion of the previously engineered Kv137-1036 strain, we deleted the uridylyl transferase domain of GlnD in Kv137-2253, preventing this protein from activating the assimilation pathways (Fig. 1b-c and supplemental text S1). The edited strain Kv137-2253 excreted about 15 times more ammonium than the Kv137-1036 strain, underscoring the effectiveness of the ΔglnD edit over the previous ΔnifL::PinfC modification in Kv137-1036 (Fig. 2b).

Evaluation of remodeled strains in vitro. (a) Biological N fixation activity was assayed using the Acetylene Reduction Assay (ARA) of individual strain growth cultures with wild-type (WT), nif gene knockout (nif-KO), and remodeled strains (supplemental Table S1) under N-free and N-rich conditions (10 mM NH4Cl). Bars represent sample means, and error bars show the standard error of four replicates. Letters denote the mean separation for the remodeled strain at the α = 0.05 level, using the Tukey adjustment. Wild-type strains (WT) with N-rich conditions and nif-KO strains were excluded from the test of hypothesis due to measurements often falling below the detection limit. (b) Ammonium excretion was assayed using the Megazyme Ammonia Assay (MAA) Kit (P/N K-AMIAR) in individual strain growth cultures. Bars represent the mean, and error bars show the standard error of eight replicates. Text annotation indicates the magnitude changes between K. variicola generation (Gen) 1 (Kv137-1036) and Gen 2 (Kv137-2253) edits as estimated by the test of hypothesis.

We have shown previously that both Ks6 and Kv137 strains have robust colonization and nitrogenase activity in the maize rhizosphere26,27. Inoculating agricultural fields with an assemblage of these two strains holds the potential to create a highly effective product capable of delivering more consistent performance across a broad spectrum of field conditions. Therefore, we measured the N-fixation ability of the remodeled strain assemblage (Ks6-5687 + Kv137-2253) when inoculated in maize seedlings grown in N-free hydroponic nutrient solution and gas-tight vials incubated with acetylene. Headspace gas analysis revealed increased acetylene reduction with the remodeled strains compared to plant tubes inoculated with the nif-KO strains (p < 0.001; Fig. 3a). This demonstrated the ability of these strains to fix N2 in a controlled plant environment. The small background ethylene signal observed in the nif-KO inoculated treatment could have originated from endogenous diazotroph nitrogenase activity given that this in-planta system was not completely sterile, or plant ethylene biosynthesis as a response to stress.

Evaluation of genetically remodeled strains in planta. (a) Measured biological N fixation activity of combined enhanced strains (Ks6-5687 + Kv137-2253) compared to combined nif gene knockout strains (nif-KO; Ks6-7023 + Kv137-7036) using the Acetylene Reduction Assay (ARA). Each point represents an individual plant replicate (n = 48 or 47). The p-value displayed corresponds to a two-tailed t-test and Welch’s correction for unequal variance. (b) Measured δ15N in chlorophyll extracted from 10-day-old plant shoots inoculated with enhanced strains and nif-KO combinations after incubating in sealed, 10 L bags supplemented with nonenriched (0.36% 15N2) or enriched gas (10% 15N2) for five days. Each box plot represents data from three experiments, comprising six or seven plants per treatment.

Labeling of N2 gas with 15N isotopes allows us to trace the movement and transformation of N, providing direct evidence of whether fixed N2 is incorporated into plant tissues. Results from 15N2 gas enrichment experiments conducted in a sterilized, N-free plant-growth pouch system (Fig. 3b) showed elevated atom% 15N concentration in the chlorophyll extract of plants inoculated with remodeled strains (p < 0.001), indicating the incorporation of atmospheric N2 into aboveground plant tissues. Conversely, no significant elevation in atom% with nif-KO strains was observed. Similarly, as expected, no significant increase in atom 15N for the enhanced or nif-KO treatments was observed when plants were incubated with non-enriched (0.367% 15N2) gas (Fig. 3b).

In small-plot field experiments conducted at the University of Wisconsin-Madison Hancock and Arlington Agricultural Research Stations, enriched 15N fertilizer was applied to soil and treated plots were inoculated with commercial-grade product containing both Ks6-5687 and Kv137-2253 strains (PROVEN 40; see supplemental Table S2 for experimental setup). Here, significant dilution δ15N compared to the uninoculated control was interpreted as a signal of crop N uptake of BNF. Samples collected from the 15 cm tip of the uppermost collared leaf (henceforth leaf tips) during across various growth stages showed marginally significant δ15N dilution in one out of three experiments but otherwise inconclusive results (Fig. 4). In the 2021 Hancock experiment, leaf tip dilution across all stages translated to 8.2% N derived from the atmosphere (Ndfa; p = 0.08). No significant dilution was detected in 2022 at either Hancock or Arlington. Because no significant interaction with the growth stage was detected in any of the experiments (p > 0.1), %Ndfa estimates across stages are reported for each experiment (Fig. 4).

Estimated changes in leaf tip 15N abundance of field-grown maize using the 15N enrichment-dilution method. Leaf tips were collected from three experimental field trials in Wisconsin at five growth stages, from early vegetative to flowering. Symbols represent the estimated marginal means derived from fitted linear mixed effect models, back-transformed from the log scale. Error bars show 95% confidence limits around the means. Post-hoc pairwise comparisons of estimated marginal means between uninoculated control and PROVEN 40 were performed across growth stages, since their interactions by inoculation treatment were not significant (p > 0.1).

We now turn to results from a two-year N fertilizer rate experiment conducted at Purdue University Agronomy Center for Research and Education (ACRE) farm (supplemental Table S3 for experimental setup). Here we used the δ15N natural abundance approach, which relies on natural variations in 15N fractionation in atmospheric and soil pools (see Methods). We interpret the dilution of δ15N compared to the reference plots (i.e., uninoculated control with reduced N fertilizer rate) as a signal of BNF. Results revealed lower levels of δ15N in the leaf tips of V8 plants treated with PROVEN 40, compared to the uninoculated control in one year (14.5%; Ndfa p < 0.001) but not in the second year (Fig. 5a). No interaction with N fertilizer rate was observed. Additional measurements during the V8, flowering (R1), and physiological maturity (R6) growth stages showed substantial δ15N dilution in the above ground biomass in the first year (2021, Fig. 5b). The response was less pronounced in the second year (2022), with dilution in the aboveground biomass marginally detected only during the R1 stage (Fig. 5c). Note that 2022 at the ACRE farm was abnormally dry, particularly during the first 8 weeks post planting (supplemental Fig. S1). No interaction with the N fertilizer rate was detected in any of the years or growth stages.

Estimated changes in δ15N natural abundance from a two-year N fertilizer rate trial at the Purdue University ACRE farm. (a) δ15N concentrations in leaf tips and above ground biomass in (b) 2021 and (c) 2022. Symbols represent the estimated marginal means derived from fitted linear mixed-effect models, back-transformed from the log scale. Error bars show 95% confidence limits. Percent N derived from the atmosphere (Ndfa) was calculated as the change from control δ15N abundance, divided by the marginal mean δ15N abundance of the control, assuming the “B” value of the equation is zero. Note that Ndfa values are only shown if a significant dilution in δ15N was observed in the inoculated control as assessed by a post-hoc pairwise comparison t-test with p < 0.10. At the ACRE farm, contrasts between treatments represent the mean effects across the six fertilizer rates, since fertilizer rate by inoculation treatment interactions were not significant (p > 0.1).

Similarly, a positive response in aboveground biomass N was only detected at V8 in 2021 (18.5% increase p < 0.001) but not at the R1 and R6 stages (Fig. 6). Based on the %Ndfa estimates for the R1 stage (Fig. 5b-c) and biomass N measurements, we estimate that PROVEN 40 provided 19.4 (95% CI: 3.25, 35.6) and 14.0 (95% CI: -0.95, 28.8) kg N ha-1 to the plant in 2021 and 2022, respectively (supplemental Table S4).

Estimated changes in aboveground biomass N from a two-year N fertilizer rate trial at the Purdue University ACRE farm. Symbols represent the estimated marginal means derived from fitted linear mixed-effect models, back-transformed from the log scale. Error bars show 95% confidence limits. Contrasts between treatments at each year and growth stage represent the mean effects across the six fertilizer rates, since fertilizer rate by inoculation treatment interactions were not significant (p > 0.1).

Employing a mixed-effect quadratic linear plateau modeling approach, we also examined differences in the maize yield response to the N fertilizer rate among inoculation treatments and site years as characterized by three curve parameters: yield with no N fertilizer, the slope of the response, and maximum attainable yield (i.e., when yield is no longer restricted by N; see supplemental information text S2 for more details on curve fitting). Specifically, we found that PROVEN 40 mainly affected the yield with no N fertilizer curve parameter, but the magnitude of the effect differed between the two site years (p = 0.064). The yield benefit of PROVEN 40 at zero N fertilizer was 0.7 Mg ha-1 less in 2022 than in 2021 (95% CI: -1.43, 0.024). Conversely, there was no evidence that PROVEN 40 affected the slope of the response (p = 0.995) or the maximum attainable yield (p = 0.67) across the two years (supplemental Tables S5 and S6). Utilizing the estimated parameters, we computed response curves across the synthetic N fertilizer rate continuum and determined the agronomic optimum N fertilizer rate (AONR; i.e., the minimum N rate to achieve the maximum attainable yield) at each combination of site-year and management treatment (Fig. 7a). This analysis revealed that PROVEN 40 reduced AONR by 27.4 kg N ha-1 (95% CI: -40.3, -15.9) in 2021 and by 10.5 kg N ha-1 (95% CI: -25.3, 4.37) in 2022 (Fig. 7b), compared to the uninoculated control. The AONR for the control was estimated at 270 kg N ha-1 (95% CI: 252, 294) and 287 (95% CI: 260, 308) in 2021 and 2022, respectively (Fig. 7a).

Effect of PROVEN 40 on the yield response to synthetic N fertilizer application rate and the agronomic optimum N fertilizer rate (AONR). (a) Estimated yield responses derived from experiments conducted in 2021 and 2022 at the ACRE farm. Symbols represent sample means, with their associated error bars denoting standard errors. Solid lines represent the best-fit curves using a mixed-effect nonlinear quadratic plateau model with replicate within a year as a random effect, with the surrounding shaded region representing the 95% confidence interval. Vertical dash lines intersecting the x-axis represent each respective AONR (i.e., the minimum N rate to achieve maximum yield), and error bars represent their estimated 95% confidence intervals. (b) Changes in the AONR relative to the uninoculated control. Error bars represent the 95% confidence intervals. All confidence intervals were estimated using residual-resampling bootstrapping (n = 300).

We examine data from 29 field trials (supplemental Table S7) conducted in 2022 in which maize seeds were inoculated prior to planting with PROVEN 40 with a ladder of inoculation rates (see Methods and supplemental Figure S2). At each site, inoculated plots received fertilizer at a reduced rate of 45 kg N ha-1 below the local business-as-usual (BAU) N fertilizer rate to elicit a small but measurable crop response to N (see supplemental Table S7 for details on N fertilizer management). Uninoculated plots with the reduced and BAU N fertilizer rates were also included to serve as controls.

Analysis of genomic DNA (gDNA) extracted from maize root tissue samples collected at a subset of sites (n = 9) and amplified via qPCR (see Methods) show significantly elevated levels of abundance of both Ks6-5687 and Kv137-2253 strains in the roots of V4-V5 maize plants compared to those in the uninoculated control (Fig. 8a), confirming that the remodeled strains can successfully colonize maize roots in fields. Minimal background amplification in the uninoculated control indicates that our primer sets were specific for our target strains in these field sites. Increasing inoculation rates show greater abundance in both remodeled strains (p < 0.001). We detected 130 and 37 times more gDNA copies in the high inoculation rate compared to the uninoculated control in Ks6-5687 and Kv137-2253, respectively. We also found evidence of differential abundance between the two strains, with populations averaging 4.8 times greater in Kv137-2253 compared to Ks6-5687 (p = 0.012).

Estimates of root colonization, δ15N natural abundance and grain yield effects of treatments co-inoculated with Ks6-5687 + Kv137-2253 in small plot experiments. (a) Genomic DNA (gDNA) extracted from maize root samples collected during the V4-V5 growth stages at nine field sites. The inoculation treatments (Low, Medium and High) and control received 45 kg N ha-1 less than the business-as-usual (BAU) N fertilizer rate. Bars represent the estimated marginal means with the combined conditional and zero-inflated models, back-transformed from the log(y + 1) scale. Error bars show 95% confidence interval of the means. P-values shown are for the contrast of each inoculated treatment versus their respective uninoculated control, adjusted using a Dunnett’s test approximation. (b) Estimated population-level effects of inoculation treatments and the BAU on δ15N abundance at VT-R1 (16 sites) and yields (28 sites). The control (uninoculated with reduced N fertilizer) was used as the baseline for all contrasts and had an estimated mean δ15N of 3.32 ‰ (95% CI: 2.73–3.99 ‰) and yield of 12.7 Mg ha-1 (95% CI: 11.8, 13.9). Percent N derived from the atmosphere (Ndfa) was calculated as the change from the negative control δ15N abundance, divided by the marginal mean δ15N abundance of the control, assuming the “B” value of the equation is zero. Error bars show 95% confidence intervals for the estimated effects.

Above ground biomass samples collected around flowering (VT-R1 growth stages) at a subset of trials (n = 16) and analyzed for δ15N natural abundance suggest significant δ15N dilution for the medium (mean = -0.36‰, 95% CI: -0.71, 0.0077) and high (mean = -0.37‰, 95% CI: -0.72, -0.0026) inoculation rate treatments (Fig. 8b). These translated to 11% Ndfa across both treatments (95% CI: 0, 22%). Note that, as expected, the dilution observed under the BAU (i.e., 45 kg of N ha-1 of additional N fertilizer over the reduced N) was also significant (mean = 0.46‰, 95% CI: -0.80, -0.098; Fig. 8b). Analysis of grain yields measured in 28 trials (Fig. 8b) show a significant yield increase compared to the negative control at the highest inoculation rate (0.21 Mg ha-1; 95% CI: 0.017, 0.41), but not for the medium and low inoculation rates. The BAU also exhibited a similar yield effect (0.21 Mg ha-1; 95% CI: 0.016, 0.41). No significant effects on aboveground biomass N at flowering (VT-R1) were detected for any of the treatments (p > 0.1; Fig. 8b). Based on these measurements, we estimate that PROVEN 40 provided on average 21.2 (95% CI: -0.16, 42.6) kg N ha-1 to the plant by VT-R1 (supplemental Table S8).

We analyzed maize aboveground fresh biomass and leaf chlorophyll concentration in samples collected from the sixth leaf stage (V6) up to tasseling (VT) to explore maize plant response to inoculation with PROVEN 40 on real-world conditions. Our study encompassed 135 commercial-scale grower fields during 2022 and 2023 (supplemental Fig. S3), comparing uninoculated field zones with BAU N fertilizer rate against inoculated zones that received 39 to 45 kg N ha-1 less than the BAU. Analysis indicates that inoculation with PROVEN 40 led to positive changes in plant fresh weight (39 g plant-1; p < 0.0001), as well as in leaf chlorophyll concentrations (12.6 µmol m-2; p < 0.0001; Fig. 9a). We estimate a median increase in aboveground biomass N of 14% across all sites (Fig. 9b), based on a regression model of data from a subset of nine sites where aboveground biomass N was also measured (cross-validated R2 = 0.89; see methods and supplemental Fig. S4).

Early-season plant response to N supplied by remodeled strains. (a) Measured changes in aboveground fresh weight and leaf chlorophyll concentrations between plant samples that received the local business-as-usual (BAU) N fertilizer rate and samples inoculated with PROVEN 40 with 39 to 45 kg N ha-1 less than the BAU. Samples were collected during vegetative growth (V6 to VT) at 135 on-farm side-by-side grower trials (supplemental Fig. S3). Points indicate the average treatment difference at each site. The hypothesis was tested using a non-parametric Wilcoxon paired t-test with continuity correction. (b) Estimated changes in aboveground biomass N at the sites between PROVEN 40 and BAU treatments. Sites are shown ranked from low to high. Points show the predicted change at each site, and error bars show 95% confidence intervals. Estimates are extrapolated using the multiple linear regression model of aboveground fresh weight and leaf chlorophyll concentration measurements to aboveground biomass N at a subset of nine sites (supplemental Fig. S4).

Finally, we examine yield impacts of practice change using a dataset of 58 commercial-scale fields in 2021. Here, each field also included zones inoculated with PROVEN 40 and received 39 to 45 kg N ha-1 less N fertilizer than the BAU, and BAU control zones (supplemental Fig. S5). Field-level mean yield comparisons between BAU and PROVEN 40 with reduced N fertilizer plots showed no significant reduction in yield across all sites (p = 0.87, Table 1). The minimum detectable effect of the test is 0.166 Mg ha-1 or 1.2% of the population mean yield estimate, underscoring the robustness of this finding. Notably, field areas inoculated with PROVEN 40 at the reduced N fertilizer exhibited an 8% decrease in the coefficient of variation (CV) of yield compared to uninoculated BAU counterparts (p = 0.013, Table 1).

The evidence presented here broadly supports the hypothesized mechanism of enhanced BNF by the remodeled strains, representing a source of N typically not available to maize grown with high levels of synthetic N fertilizer. Our research confirms that the genetic remodeling strategy for Ks6-5687 and Kv137-2253 (Fig. 1) effectively derepresses BNF activity in N-rich systems and enhances ammonium excretion by orders of magnitude above the respective wildtype strains, displaying a 15-fold increase relative to our previously edited Kv137-1036 strain27 (Fig. 2b). In-planta BNF is demonstrated by the transfer of 15N-enriched N2 gas from the rhizosphere to the chlorophyll of maize in controlled plant environments (Fig. 3b). This was partially corroborated by the 15N dilution detected in a portion of the field experiments (Figs. 5 and 7b). We also show that maize N status during mid vegetative growth was positively affected by inoculations in small-plot experiments (Fig. 6) and grower fields (Fig. 9), although differences in aboveground biomass N could not be detected during the VT-R1 and R6 growth stages (Figs. 6 and 8b). Nonetheless, we observed similar yield responses of the BAU fertilizer rate and PROVEN 40 under a fertilizer reduction of up to 45 kg N ha-1 in both small-plot experiments (Fig. 8b) and on-farm trials (Table 1). The latter suggests that the gene-edited strains have the potential to mitigate the risk of yield penalties when synthetic N fertilizer inputs are reduced, although the environmental conditions that both crop and microbes experience likely impact the fertilizer replacement potential.

Evidence from the field studies suggest that the Ks6-5687 and Kv137-2253 strains can successfully colonize maize roots, persist to at least the V4-V5 stage (Fig. 8a) and retain the potential to provide BNF to maize in N-rich field environments. These remodeled strains are estimated to contribute 11% Ndfa in the above ground biomass by the VT-R1 crop growth stage, which translates to 21.2 kg N ha-1 (supplemental Table S8). This %Ndfa estimate is comparable to the R1 %Ndfa (14–19 kg N ha-1; supplemental Table S4) and the estimated reduction in AONR (10.5–27.4 kg N ha-1; Fig. 7b) observed at the ACRE farm. These are also within the range of rates previously achieved with wildtype strain inoculation in cereal crops23, which can contribute 13–26 kg N ha-1 in maize, 3–40 kg N ha-1 in wheat, and 18–51 kg N ha-1in rice31. However, these previous estimates originate from experiments in soils receiving little or no N fertilizer, except for one study in rice24 which reported a > 80% reduction in BNF activity in soils fertilized with 125 kg N ha-1 or more. Therefore, our results represent one of the few datasets suggesting heightened BNF activity in N-rich soils, which we accomplished through our nitrogenase derepression and ammonium excretion gene edits.

The large uncertainty in Ndfa estimates, however, remains a major barrier in our analysis. Note that population-level estimate of Ndfa across 16 sites in the inoculation rate trials was found to be within -0.16 and 42.6 kg N ha-1with a 95% confidence (supplemental Table S8). Another study28 reported an even wider range of Ndfa in indigenous maize landraces (29–82%; 4–122 kg N ha-1). Rates of free-living and associative BNF are known to be highly variable, due to spatial–temporal gradients in its underlying drivers such as temperature, moisture, pH, oxygen concentration and carbon substrate availability29. Indeed, the N fertilizer rate study at the ACRE farm found limited evidence of BNF in 2022 (Figs. 5 and 6), a year in which the site received only 77 mm of rainfall within the first 50 days post-inoculation (~ 36% of the climatic mean; supplemental Fig. S1). Dry soil conditions are known to have adverse effects on the soil microbiome, both directly and indirectly due to changes in the composition and flux of carbohydrates from roots30,31,32. We expect the effects to be more pronounced in diazotrophs given the potential for oxygen damage to the nitrogenase enzyme, as dry soil conditions allow for increased oxygen diffusion into the rhizosphere33. However, we lack data to determine the extent to which survival and BNF by the remodeled strains is modulated by drought.

Leaf tip measurements from the 15N enrichment-dilution experiments, although directionally consistent with our other results, were much less conclusive (Fig. 4). Yet, the generalizability of %Ndfa estimates from leaf tips to the entire plant remains unclear. While a previous study found that %Ndfa in the third-youngest leaf was representative of the whole plant (including the roots) during the second to sixth month post-planting28, these findings were on indigenous Mexican landraces of maize grown in N-depleted soils. In contrast, our trials evaluated commercial hybrids (supplemental Tables S2, S3, and S7) in well fertilized conditions, and tissue collected was from the youngest collared leaf. A key assumption is that these leaf tip samples reflect a recent and transient flux from BNF rather than a whole-plant estimate. To the best of our knowledge, however, this hypothesis has not been validated, and remains a focal point for future research.

Free-living and associative diazotrophs, including K. variicola, are also known to offer plant health benefits beyond BNF, including mineral solubilization, phytohormone production and stress tolerance34,35. It is then possible that additional mechanisms could also be influencing our Ndfa estimates. In all our field 15N tracer studies, we assume that the inoculated crop and its respective reference crop (i.e., the uninoculated control) have similar rooting characteristics and N assimilation dynamics from fertilizer and soil (apart from BNF from the inoculum). Thus, we attribute the relative decrease in 15N in the inoculated crop to the proportion of plant N uptake from root-associated BNF carried out by Ks6-5687 and Kv137-2253. However, since synthetic N fertilizers typically exhibit lower δ15N abundance than soil36,37, the observed decrease in δ15N could partly stem from greater fertilizer recovery efficiency if larger root biomass or changes in root architecture are occurring in response to other growth promoting mechanisms. Although this scenario could mean that Ndfa is being overestimated, greater fertilizer recovery would nonetheless have positive implications for mitigating NO3 leaching and N2O emissions. Further research therefore is needed to disentangle these complex processes. Given that Ndfa methods (e.g., 15N enrichment-dilution, 15N natural abundance, N balance) are all susceptible to large uncertainty, especially in cereal crops, here we use multiple methods as previously suggested38. We refer to ref.23 for additional review of the methodological challenges of measuring BNF in cereal crops.

Results also demonstrate a potential positive influence of PROVEN 40 inoculation on the early-season N status of maize plants in small-plot trials (Fig. 6) and on-farm trials (Fig. 9). In the latter, we construe that the enhancements in fresh biomass and leaf chlorophyll content in inoculated plants suggest heightened rates of N assimilation during the initial growth stages, given that these parameters are known to be correlated to crop N status39,40,41,42. This correlation was corroborated by ground truth, aboveground biomass N measurements in a subset of sites (supplemental Fig. S4). That these improvements are driven by inoculation is particularly convincing given the concurrent 39–45 kg N ha-1fertilizer rates reduction in inoculated plants. This early-season advantage could occur because BNF is more spatially accessible to plant uptake than soil N during the early season. Not only are maize root systems small at this point, but microbially driven soil N mineralization is typically limited by cool temperatures, particularly in poorly drained soils with no-till or thick residue cover because these conditions maintain lower soil temperatures longer in the spring43.

This N uptake advantage appears to largely disappear by VT-R1 (Figs. 6 and 8b), perhaps because N mineralization or other sources of soil N can typically fill the gap. Nevertheless, the increase in early-season N uptake resulting from BNF could still have positive impacts on yield potential. Pre-V8 N deficiency can cause irreversible reductions in ear size and kernel number44, primarily due to its detrimental impact on floral primordia differentiation around V5-V645. Although compensatory mechanisms can increase individual kernel weights44,46, reduced kernel count generally leads to lower yields. In this context, BNF could help maximize kernel number potential by decreasing early-season N stress. However, further research is needed to test this hypothesis. Moreover, the final effect on grain yields may depend on the inoculum’s ability to sustain a continued supply of fixed N beyond the early-season and on sufficient soil N supply required to meet the augmented N supply driven by increased yield potential. There is little evidence supporting BNF activity after flowering, given the lack of difference in aboveground biomass N at VT-R1 (Figs. 6 and 8b), and Ndfa between R1 and R6 at the ACRE farm (Fig. 5b-c). Although there is evidence of strains persisting into the V4-V5 stages (Fig. 8a), fitness costs associated with the nitrogenase derepression and ammonium excretion gene edits26could hinder their long-term competitiveness in the soil microbiome, such that declining populations later in the growing season would be expected. If the above is true, then inoculation might be most effective in environments where N uptake occurs mainly during vegetative growth, similar to what has been shown for early in-season N fertilizer application47.

Results from Table 1 could suggest that, on the aggregate, on-farm reductions of N fertilizer inputs of 39–45 kg N ha-1 with PROVEN 40 inoculation exhibit no deleterious yield effects compared to the BAU. Note that the power of the test on Table 1should allow us to detect any difference greater than 0.17 Mg ha (or 1.2% of the population mean). This evidence, however, should be interpreted with caution given that it is likely that some of the BAU fertilizer rates exceeded their site-year AONR, in which case a lack of response is to be expected. Indeed, grower-determined fertilizer rates are based on a combination of site-specific knowledge and risk tolerance19 and on average exceed optimum rates by 17 to 28 kg N ha-148 . However, other data presented here may also support this conclusion, including: 1) the comparable yield results from the reduced N fertilizer high inoculation rate treatment and the BAU (Fig. 8b); 2) decreases in AONR in 2021 at the ACRE farm (Fig. 7); and 3) the reduction in within-field yield variability with inoculation (Table 1).

Although a 39–45 kg N ha-1 reduction in N fertilizer rate represents a modest fraction of the total amount of N fertilizer applied to maize, widespread implementation could significantly reduce fertilizer consumption. The Ks6-5687 + Kv137-2253 strain assemblage has been commercialized as PIVOT BIO PROVEN® 40 in the U.S49 with an estimated 2 million ha of adoption area in 2023. In many cases, inoculation has been accompanied by reductions in N fertilizer rates, which avoids GHG emissions from fertilizer manufacturing (⪝ 2.6 kg CO2equivalents per kg of N applied)50, and has potential to avoid soil N2O emissions and NO3 leaching. However, further research is necessary to confirm and quantify this impact. Nevertheless, the prospect of reducing N fertilizer requirements in maize through BNF represents an incremental step towards decoupling U.S. cereal crop production from synthetic N fertilizer use. However, as discussed above, environmental constraints could be important in modulating product efficacy. Therefore, a comprehensive assessment of the role of management and environmental conditions in its effectiveness and impacts on N2O emissions and NO3 leaching are the crucial next steps for developing a more mechanistic understanding of how remodeled strains affect the cropping system. This new knowledge will enable us to optimize efficacy and evaluate the complete life cycle impacts of the technology, getting us closer to the goal of replacing synthetic N fertilizers.

Klebsiella variicola (Kv137) and Kosakonia sacchari (Ks6) parent strains, previously identified as robust maize root colonizers26,27, were engineered through non-transgenic methods to enhance their BNF activity under N-rich conditions, resulting in modified strains Kv137-2253 and Ks6-5687. The engineering involved editing the nifLA operon to boost nitrogenase protein activity and editing the glnD and glnE genes to facilitate the extracellular release of NH4 (Fig. 1b-c). The rationale and details behind the gene edits are provided in the supplemental information (supplemental text S1).

A modified version of the acetylene reduction assay (ARA) described by ref.27 was used to measure nitrogenase activity in pure culture conditions. Strains were cultured from single colonies into 5 mL of super-optimized broth (SOB) for 24 h (30 °C, aerobic). The growth culture (1 mL) was then added to 25 mL of minimal media supplemented with 10 mM glutamine and grown for 24 h (30 °C, aerobic). The growth culture (0.6 mL) was then added to 2.4 mL of minimal media or 2.4 mL of minimal media supplemented with 10 mM NH4Cl in airtight culture tubes prepared in an anaerobic chamber and grown for 4 h (30 °C, anaerobic). A headspace of 10% was replaced by an equal volume of acetylene, and incubation continued for an additional hour. A gas-tight syringe was used to remove 2 mL of headspace in preparation for ethylene production quantification using an Agilent 7890B gas chromatograph equipped with a flame ionization detector. The initial culture biomass was compared to the end biomass by measuring optical density at 590 nm (OD590). The change in biomass and ethylene concentration was used to calculate mmol ethylene OD-1, a proxy for BNF. To establish a negative control, we knocked out the central nitrogenase subunit nifH, deleting the entirety of the nifH gene in the background of both remodeled strains (Ks6-7023 and Kv137-7036). The absence of BNF in the nifH gene knockout (nif-KO) strains was confirmed via ARA prior to use as a negative control. BNF in strains of the same wildtype lineage was compared using a one-way ANOVA and post-hoc Tukey test across four replicate experiments. Treatments with a substantial number of measurements below the detection limit of the assay were excluded from the analysis.

The ammonium excretion assay measures the excretion of fixed N in the form of ammonium using batch cultures in DeepWell plates. Strains were propagated from a single colony in 1 mL well-1 SOB in a 96-well DeepWell plate. The plate was incubated for 24 h (30 °C, 200 rpm) and then diluted 1:24 into a fresh plate containing 1.15 mL well-1 of minimal media containing 10 mM glutamine. Cells were incubated for 24 h (30 °C, 200 rpm) and then diluted 1:10 into a fresh plate containing minimal medium without glutamine. The plate was transferred to an anaerobic chamber (Coy Laboratory Products, Inc. Grass Lake, MI, U.S.) with a gas mixture of > 98.5% N, 1.2–1.5% hydrogen, and < 30 ppm oxygen and incubated at 1320 rpm at room temperature for 72 h. The initial culture biomass was compared to the end biomass by measuring OD590. Cells were then separated by centrifugation. The supernatant from the reactor broth was assayed for free ammonium using the Megazyme Ammonia Assay Kit (NEOGEN Corp., Lansing, MI, U.S.) and normalized to biomass at each time point to calculate mmol NH4+ OD-1. The experiment was replicated eight times. After removing measurements below the assay’s detection limit, the ammonium excretion of remodeled strains of the same wildtype lineage was compared using a two-tailed t-test with log-transformed measurements to correct for heteroskedasticity between treatments.

Maize seeds used in the experiments correspond to commercially available maize hybrids sourced from local seed dealers, unless stated otherwise. Seeds were sterilized before use in experiments to reduce the effects of environmental microorganisms. Maize seeds were surface sterilized with 70% isopropanol for two minutes and 2.5% bleach for 15 min. Seeds were rinsed thrice in sterile water and then immersed in a fungicide solution (Captan at 60 mg mL-1) overnight with gentle shaking. After three rinses with water, seeds were germinated on Petri dishes in the dark for three to four days on wet paper towels.

This experiment aimed to investigate the N-fixing capabilities of the engineered bacterial strains and their impact on ethylene production in maize plants. For this, maize seeds were sterilized and germinated before planting in 70 mL individual tubes with Turface® (calcined, non-swelling illite clay) as the growth medium and N-free Hoagland’s hydroponic nutrient solution for plant nutrition. The plants were grown uncapped for four days in a growth chamber (16 h, 28 °C days and 8 h, 25 °C nights). Bacterial cell suspensions were grown from frozen glycerol stocks and normalized to OD600 = 0.8. The cell suspensions were combined in equal volume into two groups: 1) enhanced strains and 2) nif-KO strains, both in enhanced strain backgrounds. The combined bacterial suspensions were mixed 1:2 with semisolid Fahräeus medium and used to inoculate the four-day seedlings (8 mL plant-1). The tubes were sealed, and acetylene gas was introduced into the headspace. After six days, 2 mL of headspace samples were collected to measure ethylene production. After testing for normality with a Shapiro-Wilks test, data were analyzed with a two-tailed t-test with Welch’s correction for unequal variance. We excluded one tube in the nif-KO treatment from the analysis because of unexpectedly low amounts of acetylene gas, indicating a possible leak in the tube.

This experimental setup aimed to directly measure the extent to which growing maize plants incorporate N fixed by the engineered strains. Enhanced strains (Ks6-5687 and Kv137-2253) and their nif-KO counterparts (Ks6-7023 and Kv137-7036) were cultured on TY solid medium at 30 °C for 12 h. Maize hybrid PHJ89 x PH20751,52 seeds were provided by the University of Wisconsin Corn Silage and Biofeedstock Breeding Program. Seeds were sterilized and germinated before plantlets were transferred into CYG seed germination pouches, each containing three plantlets and sterile water. These pouches were placed in a plant growth chamber with controlled lighting and humidity. The five-day-old plantlets were inoculated with 10 mL of each bacterial strain and grown individually in tall plastic containers for seven days before being transferred into gas sampling bags.

For isotopic labeling, 15N gas (15N2) or industrial-grade N gas (N2) was injected into the gas bags to create control and 15N enriched conditions at the same oxygen level. To ensure accurate ratios of 15N:14N, the GB6000 gas mixer (MCQ Instruments, Rome, Italy) was utilized, resulting in 10.33% 15N in the 15N-enriched bags and 0.3663% in the control. Shoots and roots of the inoculated plants were harvested separately after five days in the growth chamber and prepared for analysis. Chlorophyll was extracted from two to three shoots following the methodology in ref.53 and sent to the isotope ratio mass spectrometry facility for 15N analysis. 15N incorporation between plants inoculated with nif-KO and remodeled strains was compared using two-tailed student t-tests. Prior to analysis of variance, Shapiro–Wilk and Levene tests were used to establish normality and equal variance, respectively.

Three experimental field trials were conducted across two growing seasons in 2021 and 2022 at the University of Wisconsin-Madison's Hancock and Arlington Agricultural Research Stations to assess the transfer of biologically fixed N from modified bacterial strains to maize plants using the 15N enrichment-dilution method. The experiments featured three-row plots with maize hybrid PHJ89 x PH207, following a randomized complete block (RCB) design with 12 replications in 2021 and 30 in 2022. To enrich the field with 15N, a 1.0 atom % 15N solution was applied using ammonium sulfate one day before planting. Aiming for uniform distribution of the isotope across the field, 15N fertilizer was mixed with starter fertilizer, evenly distributed during planting. Throughout the growing season, side dress fertilizer applications were made per the farm's recommendations, accompanied by a 1.0 atom % 15N enrichment in either powder or liquid form, ensuring uniform incorporation into each row of the experiment. The process of inoculation with Kv137-2253 + Ks6-5687 involved labeling maize seeds with specific treatments, soaking the seeds before planting, and subsequently applying a second inoculation when plants reached the second and third leaf stages (V2-V3). Special care was taken to avoid contamination using dedicated equipment for each treatment. Leaf tip samples were systematically collected, dried, ground, and sent to the Cornell Stable Isotope Laboratory (Ithaca, NY, U.S.) for stable isotope analysis. Reported values are expressed as δ15N and atom % 15N excess, assuming atmospheric N2 gas is 0.37% 15N.

Field experiments were conducted at Purdue University's Agronomy Center for Research and Education (ACRE) during the 2021 and 2022 growing seasons. These experiments featured six-row plots planted with maize hybrid P1359AM using a RCB split-plot design with eight replications. The main plots comprised six N fertilizer treatments ranging from 0 to 286 kg N ha-1. At the same time, the subplots included the inoculation treatment with commercial-grade inoculant with Kv137-2253 and Ks6-5687 strains (PROVEN 40) and a uninoculated control (NTC). In both years, the experimental fields were planted with soybeans in the previous year, and were tilled in the previous fall before planting. Biomass samples were collected at V8, flowering (R1), and maturity (R6) from 10 consecutive plants per plot. At V8, tips (15 cm) from the uppermost collared leaf were collected from 20 representative plants per plot. Samples were dried at 60 °C for five days, weighed ground to 1 mm, and analyzed for N concentration and δ 15N isotopic concentration (Isotope Tracer Technologies Inc., Ontario, Canada). Grain yield measurements served as a basis for assessing the impact of BNF on the crop's response to synthetic N fertilizer. End-of-season grain yields (adjusted to 15.5% moisture content) were obtained by harvesting the central two rows of each plot using a small combine. We employed a quadratic-plateau nonlinear regression approach to derive three critical parameters of the N response curve: the yield without N fertilizer (a), the initial slope of the yield response (b), and the maximum attainable yield (Ymax). We then tested whether the inoculation treatment (NTC vs. PROVEN 40), the trial year (2021 and 2022), or their interaction influenced these curve parameters. The computed model parameters were used to determine the agronomic optimum N fertilizer rate (AONR) for each combination of year and treatment with bootstrapped 95% confidence intervals for the response curves and AONR54. More details are shown in the supplemental information (supplemental text S2).

Twenty-nine small plot trials were conducted in 2022 by Pivot Bio, Inc. through contract research organizations. Experiments were set up as strip trials following a RCB design with 6 replicates. Treatments included: 1) uninoculated control with the local business-as-usual (BAU) N fertilizer rate; 2) uninoculated control where N fertilizer rate was reduced by 44.8 kg N ha-1; 3) and three inoculation rate treatments (Low, Medium and High) with reduced N fertilizer rates. Further details on the management at each trial are shown in the supplemental Table S7.

In the inoculation rate treatments, maize seeds were inoculated prior to planting with commercial-grade inoculant with Kv137-2253 and Ks6-5687 strains (PROVEN 40) by soaking seeds in the resuspended liquid formulation containing both live strains at various concentrations. Inoculation rates of the combined strains were roughly 104, 105 and 106 CFU per seed. We expected cell viability to decline by about a factor of 10 by planting time (up to 45 days post inoculation). We assessed inoculum cell viability at planting of a subset of trials and determined it adequately reflected the desired inoculation rate ladder (supplemental Fig. S2).

Plot plant population was assessed at around the V2 stage by counting emerged plants in a section of the center 2 rows within a standard length (typically ~ 5 m). At around the V4-V5 growth stages, root ball samples were collected from two non-adjacent maize plants by uprooting using a shovel to retrieve the root to a depth of 30–45 cm. Excess soil was removed by shaking gently, and aerial tissues were removed by cutting the stalk just above the crown. To minimize cross contamination, gloves and shovels were cleaned or changed between plots. Unwashed root samples were shipped to the lab in cold (4 ⁰C) for colonization analysis. At the VT-R1 growth stage, representative plants were collected from each plot by cutting at the soil level, chopped to 10 cm sections and sent to laboratory to determine dry mass and N concentration (Waypoint Analytica LLC, Memphis, TN, U.S.). Aboveground biomass N at VT-R1 was estimated as the product of plant dry mass and N concentration, scaled by the plot plant population (in kg N ha-1). A subsample of the ground (< 1 mm), homogenized plant biomass tissue was sent aside for 15N isotope concentration analysis by a third-party laboratory (Isotope Tracer Technologies Inc., Ontario, Canada). Note that V4-V5 root samples and VT-R1 aboveground biomass samples were collected only in a subset of sites (supplemental Table S7). Grain yield was determined after grain moisture had dropped below 20% by harvesting the center four rows of the plot using a mechanical combine harvester. Yields are reported adjusted to 15.5% moisture. Aboveground biomass N, grain yields and δ15N concentration metrics were tested for the null hypothesis of no treatment differences using a linear mixed-effect model with treatment as the fixed effect and replicate within site as the random effect. If appropriate, measurements were log-transformed to correct for heteroscedasticity and back-transformed to report results.

To quantify the abundance of the target microbial strains in the V4-V5 root samples, we employed a quantitative PCR (qPCR) approach. Genomic DNA (gDNA) was extracted using the following methods. Roots were lightly rinsed to remove bulk soil. Then tissues from a "standard sample" zone (seminal to node 3, ~ 2.5 cm below the seed) were cut into 1 cm pieces and weighed directly into a 24-well plate containing 12.7 mm steel homogenization beads. Root samples (0.5–0.6 g) were then frozen at -80 ⁰C, homogenized using a Spex genogrinder (Cole-Parmer, Metuchen, NJ, U.S.). gDNA was extracted using a custom mix of AT/ATL lysis buffers (QIAGEN GmbH, Hilden, Germany; PN: 19,075, 939,016) and elution buffers (Zymo Research Corp., Irvine, CA, U.S.; PN: D3004-2–400, D3004-4–50). The qPCR reaction was performed using the MDX020 Probe Low Rox kit (Meridian Biosciences Inc., Cincinnati, OH, U.S.) in an Applied Biosystems Quant 7 Thermocycler (Thermo Fisher Scientific Inc., Waltham, MA, U.S.). Amplification primers were designed in-house to target the edits or gene regions that we found to be reasonably unique to Ks6-5687 and Kv137-2253. The amplified products were quantified via dye fluorescence, which then was used to estimate abundance of the target sequences. qPCR readings were expressed as copies per g of root fresh tissue. Following ref.55 we analyzed these data using a zero-inflated gaussian mixed-effect model approach where the dependent variable is transformed with log(y + 1) to control for heteroskedasticity while retaining zero-inflation. Inoculation rate and target strain were set as the fixed effects, while replicate within trial as random intercepts.

Plant samples from the 15N enrichment-dilution trials and 15N natural abundance trials were used to compare the 15N abundance in the biomass of plants inoculated with PROVEN 40 to the uninoculated control (NTC) and estimate the percent of aboveground biomass N derived from the atmosphere (Ndfa). When depending on natural 15N abundance across many sites, sites were weighed according to the mean δ15N of the reference plots as we expect low levels of background δ15N to limit detection of δ15N dilution. All field 15N isotope experiments were analyzed by fitting linear mixed-effects models with fixed and random effects per the experimental design. If models failed assumptions of normality or homoscedasticity according to the Shapiro–Wilk or Levene tests, respectively, or by examination of residual-by-predicted scatter plots, models were refit with log-transformed data and reevaluated for improved fit. When higher-order interactions were not significant, they were removed, models were re-fit, and the improved fit was evaluated using the Akaike information criterion (AIC). Additional details of linear mixed-effects model fitting are provided in supplemental Table S9. All tests of hypotheses of δ15N data were conducted using α = 0.1.

Percent Ndfa was calculated as the change from control δ15N abundance, divided by the marginal mean δ15N abundance of the control, assuming the \(\updelta {\text{N}}_{air}\) is zero using the following equation:

Large-scale, nonreplicated, side-by-side demonstration trials were established in commercial maize fields throughout the United States during 2021, 2022, and 2023. Cooperator growers were voluntarily enrolled through a commercial incentive program. Each field consisted of two treatments: i) business-as-usual (BAU) treatment that followed each grower’s standard practices, and ii) a modified practice where growers reduced synthetic N fertilizer inputs (by 39 to 45 kg N ha-1) and applied PROVEN 40. The growers determined the trial layouts. Yield maps were collected in 2021, and early-season plant traits were measured in 2022 and 2023 from a subset of trials.

The null hypothesis of no treatment differences in aboveground fresh weight and leaf chlorophyll concentration was evaluated using a nonparametric two-sided paired t-test or a Wilcoxon signed-rank test with continuity correction if the distribution of the differences failed a Shapiro–Wilk normality test with a p < 0.05.

To validate the relationship between total aboveground N (kg ha-1) and measurements of aboveground fresh weight and chlorophyll concentration, we conducted additional plant sampling in a subset of these fields (n = 9) at stages ranging from V7 to VT. Individual plants (n = 18 plants per treatment) were collected, fresh weight and leaf chlorophyll concentration were measured as described above, and subsamples of two immediately adjacent plants were sent to the laboratory for drying, weighing, and tissue N concentration analysis (Midwest Labs, Atlantic, Iowa, U.S.). A predictive model was developed by regressing these field measurements against the log-transformed aboveground biomass N. The model was then cross-validated by iteratively leaving one site out each time and refitting the model, and the full model was used to predict percent changes in aboveground biomass N across all sites. More details about model development and testing are available in the supplemental information (supplemental Fig. S4).

In 2021, yield monitor datasets were acquired from 58 grower fields through a third-party farm data software manager (INTENT, https://intent.ag/). Growers submitted yield and ancillary datasets through INTENT’s data ingestion portal, which subsequently were cleaned and moisture-corrected (15.5% moisture) by the software and spatially joined with plot-treatment information. Growers applied the treatment (39–45 kg ha-1 less than BAU with PROVEN 40) across one or more strips in their fields. The size and number the strips varied among locations, with an average area of 11.3 ha of treated area by location. For the BAU treatment area, we selected an equally sized area adjacent to each strip. This aimed to represent a side-by-side comparison at each location. (An example of the setup is shown in the supplemental Fig. S5). Yield was summarized at the site level using the mean and coefficient of variation. Both site-level metrics were tested for the null hypothesis of no treatment differences using a linear mixed-effect model with treatment as the fixed effect and site as the random effect. The coefficient of variation was log-transformed to correct for heteroscedasticity and back-transformed to report results. Minimum detectable effects were calculated using the standard error of the estimated treatment contrast and the expected t-ratio (two-tailed with α = 0.05).

All statistical analyses were performed and documented with Jupyter Notebooks57using the R statistical software (version 4.1.3)58 kernel augmented with the following packages and their dependencies:

Data munging and visualization: tidyverse, readxl, broom, foreach, ggpubr, ggthemes, scales, png

Model fitting and evaluation: lme4, lmerTest, nlme, glmmtmb, car, DHARMa, caret

Post-hoc mean comparisons: emmeans, multcomp

A complete list of all attached packages and their citations is provided in the supplemental information.

The datasets and computer analysis code used to support the findings of this study are available in the following repository: https://doi.org/https://doi.org/10.5281/zenodo.10658648.

Smil, V. Nitrogen and Food Production: Proteins for Human Diets. Ambio 31, 126–131 (2002).

Article ADS PubMed Google Scholar

Zhang, X. et al. Managing nitrogen for sustainable development. Nature 528, 51–59 (2015).

Article ADS CAS PubMed Google Scholar

Erisman, J. W., Sutton, M. A., Galloway, J., Klimont, Z. & Winiwarter, W. How a century of ammonia synthesis changed the world. Nat Geosci 1, 636–639 (2008).

Article ADS CAS Google Scholar

Fowler, D. et al. The global nitrogen cycle in the Twentyfirst century. Philosophical Transactions of the Royal Society B: Biological Sciences 368, 20130164 (2013).

Ladha, J. K. et al. Global nitrogen budgets in cereals: A 50-year assessment for maize, rice, and wheat production systems. Sci Rep 6, 19355 (2016).

Article ADS CAS PubMed PubMed Central Google Scholar

Ladha, J. K., Pathak, H., Krupnik, T. J., Six, J. & van Kessel, C. Efficiency of fertilizer nitrogen in cereal production: Retrospects and prospects. Advances in Agronomy 87, 85–156 (2005).

Robertson, G. P. & Vitousek, P. M. Nitrogen in agriculture: Balancing the cost of an essential resource. Annu Rev Environ Resour 34, 97–125 (2009).

Melillo, J. M. Disruption of the global nitrogen cycle: A grand challenge for the twenty-first century. Ambio 50, 759 (2021).

Article ADS PubMed PubMed Central Google Scholar

Ravishankara, A. R., Daniel, J. S. & Portmann, R. W. Nitrous oxide (N2O): The dominant ozone-depleting substance emitted in the 21st century. Science 326, 123–125 (2009).

Article ADS CAS PubMed Google Scholar

IPCC. AR6 Synthesis Report: Climate Change 2023. IPCC Fifth Assessment Synthesis Report-Climate Change 2014 Synthesis Report (2023).

Gao, Y. & Cabrera Serrenho, A. Greenhouse gas emissions from nitrogen fertilizers could be reduced by up to one-fifth of current levels by 2050 with combined interventions. Nat Food 4, 170–178 (2023).

Venterea, R. T. et al. Challenges and opportunities for mitigating nitrous oxide emissions from fertilized cropping systems. Front Ecol Environ 10, 562–570 (2012).

Cassman, K. G. & Dobermann, A. Nitrogen and the future of agriculture: 20 years on. Ambio 51, 17–24 (2022).

Article ADS PubMed Google Scholar

Cassman, K. G., Dobermann, A. & Walters, D. T. Agroecosystems, nitrogen-use efficiency and nitrogen management. Ambio 31, 132–140 (2002).

Article ADS PubMed Google Scholar

Dobermann, A. et al. Responsible plant nutrition: A new paradigm to support food system transformation. Glob Food Sec 33, 100636 (2022).

Christianson, L. E. & Harmel, R. D. 4R Water quality impacts: An assessment and synthesis of forty years of drainage nitrogen losses. Journal of Environment Quality 44, 1852 (2015).

Anas, M. et al. Fate of nitrogen in agriculture and environment: agronomic, eco-physiological and molecular approaches to improve nitrogen use efficiency. Biol Res 53, 47 (2020).

Article CAS PubMed PubMed Central Google Scholar

Castellano, M. J. & David, M. B. Long-term fate of nitrate fertilizer in agricultural soils is not necessarily related to nitrate leaching from agricultural soils. Proc Natl Acad Sci U S A 111, E766 (2014).

Article ADS CAS PubMed PubMed Central Google Scholar

Stuart, D., Schewe, R. L. & McDermott, M. Reducing nitrogen fertilizer application as a climate change mitigation strategy: Understanding farmer decision-making and potential barriers to change in the US. Land use policy 36, 210–218 (2014).

van Grinsven, H. J. M. et al. Losses of ammonia and nitrate from agriculture and their effect on nitrogen recovery in the European Union and the United States between 1900 and 2050. J Environ Qual 44, 356–367 (2015).

Reed, S. C., Cleveland, C. C. & Townsend, A. R. Functional ecology of free-living nitrogen fixation: A contemporary perspective. Annu Rev Ecol Evol Syst 42, 489–512 (2011).

Olivares, J., Bedmar, E. J. & Sanjuán, J. Biological nitrogen fixation in the context of global change. Molecular Plant-Microbe Interactions 26, 486–494 (2013).

Article CAS PubMed Google Scholar

Ladha, J. K. et al. Biological nitrogen fixation and prospects for ecological intensification in cereal-based cropping systems. Field Crops Res 283, 108541 (2022).

Article ADS PubMed PubMed Central Google Scholar

Zhang, Y. et al. How do different nitrogen application levels and irrigation practices impact biological nitrogen fixation and its distribution in paddy system?. Plant Soil 467, 329–344 (2021).

Bloch, S. E., Ryu, M. H., Ozaydin, B. & Broglie, R. Harnessing atmospheric nitrogen for cereal crop production. Curr Opin Biotechnol 62, 181–188 (2020).

Article CAS PubMed Google Scholar

Bloch, S. E. et al. Biological nitrogen fixation in maize: Optimizing nitrogenase expression in a root-associated diazotroph. J Exp Bot 71, 4591–4603 (2020).

Article CAS PubMed PubMed Central Google Scholar

Wen, A. et al. Enabling biological nitrogen fixation for cereal crops in fertilized fields. ACS Synth Biol 10, 3264–3277 (2021).

Article CAS PubMed Google Scholar

Van Deynze, A. et al. Nitrogen fixation in a landrace of maize is supported by a mucilage-associated diazotrophic microbiota. PLoS Biol 16, e2006352 (2018).

Article PubMed PubMed Central Google Scholar

Soper, F. M. et al. A roadmap for sampling and scaling biological nitrogen fixation in terrestrial ecosystems. Methods Ecol Evol 12, 1122–1137 (2021).

Steiner, F. A. et al. Rhizosheath drought responsiveness is variety-specific and a key component of belowground plant adaptation. New Phytologist 242, 479–492 (2024).

Article CAS PubMed Google Scholar

Tiziani, R. et al. Drought, heat, and their combination impact the root exudation patterns and rhizosphere microbiome in maize roots. Environ Exp Bot 203, 105071 (2022).

Zhang, X. et al. The spatial distribution of rhizosphere microbial activities under drought: water availability is more important than root-hair-controlled exudation. New Phytologist 237, 780–792 (2023).

Article CAS PubMed Google Scholar

Fairbairn, L. et al. Relationship between soil CO2 fluxes and soil moisture: Anaerobic sources explain fluxes at high water content. Geoderma 434, 116493 (2023).

Article ADS CAS Google Scholar

Mendes, R., Garbeva, P. & Raaijmakers, J. M. The rhizosphere microbiome: Significance of plant beneficial, plant pathogenic, and human pathogenic microorganisms. FEMS Microbiol Rev 37, 634–663 (2013).

Article CAS PubMed Google Scholar

Duran-Bedolla, J., Garza-Ramos, U., Rodríguez-Medina, N., Aguilar Vera, A. & Barrios-Camacho, H. Exploring the environmental traits and applications of Klebsiella variicola. Brazilian Journal of Microbiology 52, 2233 (2021).

Article CAS PubMed PubMed Central Google Scholar

Vitòria, L., Otero, N., Soler, A. & Canals, A. Fertilizer characterization: Isotopic data (N, S, O, C, and Sr). Environ Sci Technol 38, 3254–3262 (2004).

Article ADS PubMed Google Scholar

Bateman, A. S., Kelly, S. D. & Jickells, T. D. Nitrogen isotope relationships between crops and fertilizer: Implications for using nitrogen isotope analysis as an indicator of agricultural regime. J Agric Food Chem 53, 5760–5765 (2005).

Article CAS PubMed Google Scholar

Unkovich, M., Herridge, D., James, E. K., Giller, K. & Peoples, M. B. Reliable quantification of N2 fixation by non-legumes remains problematic. Nutr Cycl Agroecosyst 118, 223–225 (2020).

Yang, H. et al. Maize canopy and leaf chlorophyll content assessment from leaf spectral reflectance: Estimation and uncertainty analysis across growth stages and vertical distribution. Remote Sensing 14, 2115 (2022).

Wood, C. W., Reeves, D. W., Duffield, R. R. & Edmisten, K. L. Field chlorophyll measurements for evaluation of corn nitrogen status. J Plant Nutr 15, 487–500 (1992).

Nasar, J. et al. The effect of maize–alfalfa intercropping on the physiological characteristics, nitrogen uptake and yield of maize. Plant Biol 22, 1140–1149 (2020).

Article CAS PubMed Google Scholar

Ziadi, N. et al. Chlorophyll measurements and nitrogen nutrition index for the evaluation of corn nitrogen status. Agron J 100, 1264–1273 (2008).

Quinn, D. J., Lee, C. D. & Poffenbarger, H. J. Corn yield response to sub-surface banded starter fertilizer in the U.S.: A meta-analysis. Field Crops Res 254, 107834 (2020).

Subedi, K. D. & Ma, B. L. Nitrogen uptake and partitioning in stay-green and leafy maize hybrids. Crop Sci 45, 740–747 (2005).

Abendroth, L. J., Elmore, R. W., Boyer, M. J. & Marlay, S. K. Corn Growth and Development (PMR 1009) (Iowa State University Extension, 2011).

Borrás, L., Slafer, G. a & Otegui, M. E. Seed dry weight response to source–sink manipulations in wheat, maize and soybean: a quantitative reappraisal. Field Crops Res 86, 131–146 (2004).

Fernandez, J. A., DeBruin, J., Messina, C. D. & Ciampitti, I. A. Late-season nitrogen fertilization on maize yield: A meta-analysis. Field Crops Res 247, 107586 (2020).

Ransom, C. J. et al. Corn nitrogen rate recommendation tools’ performance across eight US midwest corn belt states. Agron J 112, 470–492 (2020).

Temme, K., Shah, N., Eskiyenenturk, B., Bloch, S. & Tamsir, A. Modified bacterial strains for improved fixation of nitrogen (Patent No. WO2021221690A1). (2021).

Wang, M. et al. Summary of Expansions and Updates in GREET 2020. Energy Systems Division (2020).

Hirsch, C. N. et al. Draft assembly of elite inbred line PH207 provides insights into genomic and transcriptome diversity in maize. Plant Cell 28, 2700–2714 (2016).

Article CAS PubMed PubMed Central Google Scholar

Gage, J. L. et al. Multiple Maize Reference Genomes Impact the Identification of Variants by Genome-Wide Association Study in a Diverse Inbred Panel. Plant Genome 12, 180069 (2019).

Kahn, M. L. et al. A mass spectrometry method for measuring 15N incorporation into pheophytin. Anal Biochem 307, 219–225 (2002).

Article CAS PubMed Google Scholar

Francis, H. R., Ma, T. F. & Ruark, M. D. Toward a standardized statistical methodology comparing optimum nitrogen rates among management practices: A bootstrapping approach. Agricultural and Environmental Letters 6, e20045 (2021).

Zhang, X., Guo, B. & Yi, N. Zero-Inflated gaussian mixed models for analyzing longitudinal microbiome data. PLoS One 15, e0242073 (2020).

Article CAS PubMed PubMed Central Google Scholar

Parry, C., Blonquist, J. M. & Bugbee, B. In situ measurement of leaf chlorophyll concentration: analysis of the optical/absolute relationship. Plant Cell Environ 37, 2508–2520 (2014).

Article CAS PubMed Google Scholar

Kluyver, T. et al. Jupyter Notebooks – a publishing format for reproducible computational workflows. in Positioning and Power in Academic Publishing: Players, Agents and Agendas (eds. Loizides, F. & Schmidt, B.) 87–90 (2016).

R Core Team. R: A Language and Environment for Statistical Computing. Preprint at (2022).

We want to thank Dr. Natalia de Leon and Marina Borsecnik from University of Wisconsin-Madison for kindly providing the PHJ89 x PH207 hybrid germplasm used in the 15N gas enrichment assay and field 15N enrichment-dilution experiments.

Present address: Regrow Agriculture, Inc. , Durham , NH, 03824, USA

Rafael Martinez-Feria and Maegen B. Simmonds have lead authorship.

Pivot Bio, Inc., 2910 Seventh St, Berkeley, CA, 94710, USA

Rafael Martinez-Feria, Maegen B. Simmonds, Bilge Ozaydin, Stacey Lewis, Allison Schwartz, Alex Pluchino, Megan McKellar, Shayin S. Gottlieb, Tasha Kayatsky, Richelle Vital, Sharon E. Mehlman, Zoe Caron, Nicholas R. Colaianni, Garrett Verhagen, Clayton Nevins, Lori Reese, Paul Otyama, Alice Robinson, Timothy Learmonth, Christine M. F. Miller, Keira Havens, Alvin Tamsir & Karsten Temme

Department of Plant and Agroecosystem Sciences, University of Wisconsin-Madison, 1575 Linden Drive, Madison, WI, 53706, USA

Jean-Michel Ané, Junko Maeda, Valentina Infante, Bjorn H. Karlsson & Caitlin McLimans

Department of Agronomy, Purdue University, 915 Mitch Daniels Blvd, West Lafayette, IN, 479074, USA

Tony Vyn, Brendan Hanson & Garrett Verhagen

You can also search for this author in PubMed Google Scholar

RM-F and MBS are joint lead authors. Conceptualization: KT, AT, MBS, RM-F, J-MA, TV. Methodology, experimentation and data curation: BO, SL (Gene editing, in-vitro assays); AS, AP, MM, (in-planta ARA); SSG, SM, TK, RV (root samples and gDNA extractions); ZC (qPCR amplification); J-MA, JM, VI, BHK, CM (in-planta 15N incorporation and field 15N enrichment-dilution); TV, BH, CV (field 15N natural abundance, N fertilizer rate trials); LR, CN, PO (grower trial data). Formal Analysis: RM-F, MBS, AR, NRC. Writing – Original Draft: RM-F, MBS, BO, AS, SG, NRC, J-MA, TV, CN, TL, CMFM, KH, AT, KT. Writing – Review and Editing: all authors.

Authors affiliated with Pivot Bio, Inc. declare a competing financial interest given that Pivot Bio, Inc. has obtained patents (World Intellectual Property Organization Patent No. WO2021221690A1) and is developing and commercializing products based on the microbial strains included in this study. All the remaining authors declare no conflict of interest.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Martinez-Feria, R., Simmonds, M.B., Ozaydin, B. et al. Genetic remodeling of soil diazotrophs enables partial replacement of synthetic nitrogen fertilizer with biological nitrogen fixation in maize. Sci Rep 14, 27754 (2024). https://doi.org/10.1038/s41598-024-78243-3

DOI: https://doi.org/10.1038/s41598-024-78243-3

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Scientific Reports (Sci Rep) ISSN 2045-2322 (online)

fertilizer for peach trees Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Genetic remodeling of soil diazotrophs enables partial replacement of synthetic nitrogen fertilizer with biological nitrogen fixation in maize | Scientific Reports