Endowing plants with the capacity for autogenic nitrogen xation

Biologically available nitrogen is a common limitation to crop productivity in modern agriculture. The endowment of higher plants with the ability to produce their own nitrogenous fertilizers has been attempted for nearly half a century 1–4 . Here we report that a minimal nitrogen xation system from Paenibacillus polymyxa 5–8 can be used to create an autogenic nitrogen-xing plant through synthetic biology. We found that the genetically modied Arabidopsis containing the cassette of all nine nif genes (nifBHDKENXhesAnifV) showed some activity of nitrogenase and caused higher biomasses and chlorophyll contents than wild-type plants grown in low-nitrogen or nitrogen-free medium. Then we found that the engineered Arabidopsis displayed resistance to KCN and NaN 3 , two substrates of nitrogenase 9 . Furthermore, overexpression of electron transfer component 10 in the engineered nif gene-carrying plants resulted in higher nitrogen xation eciency. Isotopic labeling analysis using liquid chromatography-tandem mass spectrometry showed that the xed nitrogen can ow to amino acids and chlorophyll 11, 12 . This study represents a milestone toward realizing the goal of endowing plants with the capacity for self-fertilization.

selected the compact BNF system from Paenibacillus sp. to design a strategy enabling the constitutive expression of components required for the biosynthesis and activity of nitrogenase in the cytoplasm. We also designed to introduce the ferredoxin, which can act as an electron donor for the Fe protein of nitrogenase and reductant for oxygen protection [44][45][46] , into the transgenic nif plants to improve nitrogen xation e ciency.
Gene synthesis, vector construction, and plant transformation To develop a rst-generation nitrogen-xing plant, the simplest nitrogen xation system in the grampositive bacterium Paenibacillus sp. (GenBank accession number CP017967.3) was chosen for plant transformation. Its minimal compact nif gene cluster contained nine genes organized in the order nifBHDKENXhesAnifV ( Figure 1A). The structures of the Fe protein and MoFe protein in the nitrogenase of Paenibacillus are similar to those of Azotobacter vinelandii 47, 48 ( Figure 1B), but the processing of nitrogenase metalloclusters and the synthesis of Mo cofactor (FeMo-co) is relatively simple 49 ( Figure 1C). To test whether the minimal nif gene cluster can be used in plant nitrogen xation, we reconstructed the expression system of the nitrogenase with the constitutive CaMV 35S Ω promoter after modifying and synthesizing all genes using codons at high frequency in plants ( Figure 1D, Supplementary materials and methods). Then the entire nitrogenase synthetic pathway was introduced into Arabidopsis. More than 100 hygromycin B-resistance plants were further screened at a low level of nitrogen. Three lines showing a normal vegetative phenotype were selected from homozygous T 2 plants containing a stable T-DNA insertion for further analysis. Molecular analysis by PCR ampli cation con rmed that all nif genes were present in these transgenic plants ( Figure S1A) and all lines exhibited single insertions ( Figure 1E). Furthermore we revealed that transgenic lines 37 and 54 exhibited a T-DNA insertion on chromosome 5, while line 58 exhibited a T-DNA insertion on chromosome 3 by resequencing method ( Figure 1F, S1B).

Expression of nitrogenase genes in plants
To determine the transcription e ciency of every nif gene in individual transgenic plants, quantitative real-time PCR and Northern blotting were conducted. The results of Northern blotting analysis indicated that the RNAs were stable and intact (Figure 2A), but the expression level of each exogenous gene was different ( Figure 2B, S1C). The levels of the nif genes were lower in transgenic line 58. For revealing the mechanism of different transcription e ciency of transgenic lines, we assessed the methylation of the CG, CHG and CHH sites in the CaMV 35S promoter following the reaction of genomic DNA with bisulphate and the ampli cation and sequencing of the promoter 50, 51 . The average DNA methylation level in transgenic line 58 reached 93.1%, which is higher than those in line 37 (45.8%) and line 54 (53.6%). The CpG islands in the CaMV 35S promoter showed a higher level of methylation than those of CHG and CHH ( Figure 2C).
To determine whether all Nif proteins required for nitrogenase biosynthesis and function could be expressed in plants given the successful transcription of all nif genes in transgenic plants, Western blotting and immunogold labeling were conducted. Western blotting analysis with rabbit polyclonal antibodies against each protein showed that each encoded Nif polypeptide in transgenic lines was approximately the predicted size though the abundance of every polypeptide varied ( Figure 2D).
Immunogold labeling of structural proteins of nitrogenase and their cofactors in TEM sections using their speci c antibody showed associations of Au particles in the cytoplasm, cell membrane, endoplasmic reticulum, and other organelle membranes ( Figure 2E, S2). These nding proved that all Nif proteins were expressed and distributed across the cells of nif transgenic Arabidopsis.

Detection of activity of nitrogenase in transgenic plants
To test this hypothesis that the complete Paenibacillus Nif system is capable of N 2 xation in transgenic plants, we sowed Arabidopsis seeds in low-or no-nitrogen medium and compared the growth potential of the plants. Under low-nitrogen conditions (50 mg/L KNO 3 ), the most evident difference between transgenic and wild-type plants was that the former exhibit greener and larger leaves ( Figures 3A, S3). After growth for 15 days, the fresh weight of the transgenic plants was 1.12-1.28 times of wild-type plants ( Figure 3D); moreover, the total chlorophyll content of the former was 1.32-1.56 times of the latter ( Figure   3F). After growth for 60 days, the dry weight of the transgenic line 54 was increased to 1.37 times of wildtype plants ( Figure S3). An identical phenotype change was detected in the six-leaf stage. When the plants were transferred to medium with low-nitrogen conditions (50 mg/L KNO 3 ) and grown for 30 days, the total chlorophyll content of the transgenic plants was 2-4 times that of the wild-type plants ( Figures  3C, 3G). Surprisingly, when Arabidopsis was grown on nitrogen-free medium, the leaves of the transgenic seedlings presented watery lesions, and their biomass declined signi cantly ( Figure 3B). The symptoms may be related to ammonium toxicity in cells 52 . The fresh weight of the wild-type seedlings in turn was 16%-24% higher than that of transgenic plants grown in nitrogen-free conditions. With the increase in growing time, the symptoms worsened ( Figure S3B). The fresh weight of transgenic line 54 was only 72% that of wild-type seedlings when seedlings were grown in nitrogen-free conditions for 60 days. However, the dry wet ratios of transgenic plants were obviously higher than that of wild-type plants ( Figure S3C). In addition, the content of the proteins in solution and ammonium concentrations in the nif gene-transgenic plants were higher than those of wild-type plants regardless of whether they were grown under lownitrogen conditions or nitrogen-free conditions ( Figure 3E, S4).
To verify that the growth difference between transgenic and wild-type plants come from nitrogenase, we changed the oxygen concentration in the environment where plants grow based on its sensitive to oxygen. We found that the growth difference between transgenic and wild type seedlings became more obvious with the decrease in time of illumination from 16 h to 8 h under low-nitrogen conditions ( Figure   S5). Another experiment showed that the fresh weight and total chlorophyll content of healthy seedlings of transgenic lines increased more obviously than that of wild-type with the decrease in oxygen concentration from 60% to 5% ( Figure S6).
To determine whether the nine Nif proteins could be assembled in transgenic plants and function as in Paenibacillus, we adopted two independent methods for assessing nitrogenase activity: (1) the acetylene reduction method, wherein acetylene is reduced to ethylene, which can be readily quanti ed by gas chromatography 53 , and (2) the 15 N 2 isotope enrichment method, which directly measures the incorporation of this tracer into organic nitrogen 54 . When grown in nitrogen-de cient and high-Moconcentration medium, the crude proteins of all transgenic plants exhibited acetylene reduction. Transgenic line 37 and 54, which expresses the nif genes at signi cantly higher levels, showed approximately 2-3-fold-higher acetylene reduction activity than line 58. The high nitrogenase activity of line 37 and 54 was con rmed by the plant's level of assimilation of 15 N 2 ( Figures 3H, 3I). In addition, the nitrogenase activity in underground parts was higher than that in its aboveground parts ( Figure S7). The poor expression of NifB and NifK in line 58 may interfere with the assembly of the multiprotein complex of nitrogenase ( Figure 2D).
The expressed nitrogenase in transgenic plants shows wide substrate speci city In addition to reducing N 2 and protons, nitrogenase can reduce azide and a wide array of carboncontaining compounds such as alkynes and carbon-nitrogen substrates 9 . In these substrates, KCN and NaN 3 can decrease plant respiration rates by inhibiting the electron transfer pathway through cytochrome 55 . To verify the activity of nitrogenase to catalyze the decomposition of the two respiratory inhibitors, we tested the tolerance of transgenic plants to KCN and NaN 3 . We found that the death rates of wild-type Arabidopsis seedlings reached approximately 63.5% and 41.6%, while the death rates of nif gene-transgenic Arabidopsis seedlings were only 24.2%-28.5% and 3.5%-14.7% when treated with 20 mg/L KCN and 1.35 mg/L NaN 3, respectively ( Figure 4AB). In addition, the respiratory inhibitors KCN and NaN 3 exerted more signi cant effects on the growth of wild-type plants than on transgenic plants. The fresh weights of the wild-type seedlings were reduced by 12% at low nitrogen concentrations (30 mg/L KNO 3 ) and 16.7% at higher nitrogen concentrations (200 mg/L KNO 3 ) when 10 mg/L KCN was added to the medium. By comparison, the fresh weight of the transgenic plants remained unchanged or even increased when KCN was added to the medium. As the reduction of HCN was observed with the production of ammonia (NH 3 ) for Mo-nitrogenases, low KCN concentrations can stimulate the growth of nif gene-transgenic plants at low nitrogen concentrations. The fresh weight of wild-type seedlings was 4.5%-11.5% lower than that of engineered nif gene-carrying plants when 0.65 mg/L NaN 3 was added to the medium containing a higher nitrogen concentration (200 mg/L KNO 3 ), but the reduction was increased to 24.4%-29.7% at a low nitrogen concentration (30 mg/L KNO 3 ) ( Figure 4C, 4D).
NifXVhaesA assists in nitrogenase assembly and ferredoxin is a useful electron donor for nitrogenase in plants NifH, NifD, NifK, NifE, NifN, and NifB are conserved in diazotrophs 6, 56, 57 . In theory, the expression of these six proteins could allow the breeding of a nitrogen-xing plant 3 . To determine whether plants can complete the assembly of nitrogenase with the six proteins, we constructed a constitutive expression vector with four nifBHDK genes and six nifBHDKEN ( Figure 5A). No difference in plant growth in nutrient medium was found among the wild-type, nifBHDK, nifBHDKEN, and nifBHDKENXVhesA transgenic plants. However, when the growth environment exhibited low-nitrogen conditions, wild-type and transgenic seedlings containing incomplete Paenibacillus nitrogen xation gene clusters showed reduced growth and leaf chlorosis compared with plants expressing all genes in the cluster ( Figure 5B). The fresh weight of the plants containing only nifBHDKEN and nifBHDK was 24%-33% lower, and their total chlorophyll content was 30%-37% lower than those of plants containing the complete nitrogen xation gene cluster nifBHDKENXVhesA when grown with 50 mg/L KNO 3 for 15 days ( Figures 5D, 5E). In addition, transgenic seedlings containing nifBHDKEN showed no resistance to KCN and NaN 3 , as observed in the wild-type ( Figure 5C). When grown on low-nitrogen medium containing 10 mg/L KCN or 0.65 mg/L NaN 3 , more than 20% of nifBHDKEN transgenic plants died, and their fresh weights were only 52% and 64% of that of nifBHDKENXVhesA plants. Similar results were obtained in studies of the nifBHDK and nifXVhaesA transgenic lines ( Figure 5F). No nitrogenase activity was detected in the nifBHDK and nifBHDKEN transgenic lines by either the acetylene reduction method or the 15 N 2 enrichment method. NifXVhaesA might participate in the assembly of nitrogenase 49, 58 and act as activating factor for its activity in plants 59-62 .
Electron transfer component (ETC) is important for the supply of reducing power for catalysis of nitrogen xation 63 . However, the nifF gene encoding avodoxin, a direct electron donor to nitrogenase, is deleted in the minimal Paenibacillus nif gene cluster genome. In addition, avodoxin does not exist alone in plants 64 . To provide e cient electron donors to Paenibacillus nitrogenase expressed in the plant cytoplasm, we introduced exogenous ferredoxin to the engineered nif gene-carrying plants based on the fact that ferredoxins were found to be competent electron donors to nitrogenase in Rhodospirillum rubrum 44 . At rst, we transferred genes of ETC into nif gene-carrying plants by hybridization. Transgenic plants containing ferredoxin and the ferredoxin oxidoreductase of the naphthalene dioxygenase from Pseudomonas putida 10 were used as the male parents in crosses with engineered nif gene-carrying plants. The hybrids grew better than the engineered nif gene-carrying plants under low-nitrogen conditions ( Figure 5G). In addition, the engineered nif gene-carrying plants showed greater resistance to KCN and NaN 3 upon the introduction of ETC ( Figure 5H). The level of 15 N 2 assimilation in the hybrids was higher than that in engineered nif gene-carrying plants ( Figure 5I). Then we introduced genes of ETC into nif gene-carrying plants by Agrobacterium-mediated transformation. The activity of nitrogenase also increased in plants with high expression of genes of ETC. In addition, we found no signi cant difference in effects on the active level of nitrogenase between ferredoxin and avodoxin ( Figure 5J). We speculated that ferredoxin can be act electron donor as avodoxin to nitrogenase in vivo.
Mass spectrometry reveals that 15 N 2 can be transformed to the source of nitrogen for amino acids and chlorophyll biosynthesis in plants Inorganic nitrogen produced by nitrogen xation can be involved in the synthesis of amino acids and other important nitrogen-containing compounds, such as amino acids, chlorophyll, and many products of secondary metabolism in nature 65 . To test whether some intracellular pool of nitrogen is indeed synthesized from N 2 xation, we investigated the metabolism of organisms using tracer techniques with stable isotopes 66 . We found that the nine tested amino acids and chlorophyll of nif transgenic plants have higher level of 15 N than that of wild type plants after cultured in nitrogen free condition supplemented with isotopically labeled 15 N 2 (20% air). Moreover, the hybrids of line 37 and ferredoxin plants (404-37+FD) have a higher level of 15 N labeled amino acids and chlorophyll than nif transgenic line 37 ( Figure 6 and Supplementary Table S6). For example, glutamate, the most important amino acid for nitrogen metabolism, showed a 15 N isotopic abundance of 4.9% in the hybrids 404-37+FD, while at natural abundance levels (about 0.37%) in wild-type plants, asparagine, the principal nitrogen transport amino acid in higher plants 67 , also showed a higher 15 N isotopic abundance in the hybrids 404-37+FD, alanine, which played a key role in storing carbon and nitrogen 68 , showed a 15 N isotopic abundance more than 5 times higher in the hybrids 404-37+FD than in wild-type plants. Other amino acids produced by intermediates in the EMP pathway, such as isoleucine and valine formed from pyruvate, serine and glycine from 3-phosphoglyceric acid, and phenylalanine from phosphoenolpyruvate (PEP), have more or  Table S6).

Discussion
Despite the results of pioneering work, the ability to reliably engineer micro-BNF systems in plants remains limited 2, 4, 69 . Here, we constructed a plant multigene expression vector with Paenibacillus nitrogen xation genes to constitutively and simultaneously express nitrogenase catalytic proteins and their assembly cofactors in cells. When we grew plants in a low-nitrogen environment, the complete-nif gene-transgenic plants overcame the general nitrogen de ciency and achieved a greater biomass and higher chlorophyll content. We have detected the activity of nitrogenase via protein extraction from the transgenic lines in vitro through the acetylene reduction method and N 2 assimilation in vivo by the 15 N enrichment method. Furthermore, we found that the transgenic lines show resistance to the respiratory inhibitors KCN and NaN 3 , two substrates of nitrogenase. Through further research using tracer techniques, we also found that nitrogen can ow to the anabolism of amino acids and chlorophyll in engineered plants. Based on our analysis, we are con dent that the complete set of Paenibacillus nitrogen xation genes enables plants to synthesize their own nitrogen fertilizer.
The transgenic lines described in this work may serve as a unique starting point for the introduction of the nif gene cluster from bacteria to enable nitrogen xation activity in plants. Although numerous challenges and many barriers remain before plants can e ciently x atmospheric nitrogen, for example protecting nitrogenase from toxicity of oxygen and ne-tuning the expression of target genes 70 , the work here let us see a glimmer of light to produce nitrogen autotrophic plants in future.
14. Dos Santos, P.C. et al. Distribution of nitrogen xation and nitrogenase-like sequences amongst microbial genomes. BMC Genom. 13, 162 (2012  Each read was 150 bp in length, and at least 12 GB of data were obtained. The Q30 value of each sample was up to 80 (the general perception of a high error rate [> 0.1%] is less than 20%).
The T-DNA sequence of the vector was compared with the sequence obtained by using Bowtie2 software RNA (15 μg) was separated on 1% denaturing agarose-formaldehyde gels. Equal loading was con rmed by staining the gels with ethidium bromide. After the RNA was transferred to nylon membranes, it was probed with digoxigenin (DIG)-labeled cDNA probes obtained by PCR (PCR DIG Probe Synthesis Kit, Roche, Mannheim, Germany). To amplify the respective probes, sequence-speci c primers were used.
Colorimetric detection was performed using nitro blue tetrazolium (NBT) and 5-bromo-4-chloro-3-indolylphosphate (BCIP) as substrates for alkaline phosphatase. Quantitative analysis of the northern blot results was performed by using Gel Analyser 2010.

Western blot analysis
To demonstrate the production of Nif proteins, the total protein fraction generated from the Arabidopsis seedlings was extracted with SDS-Tris buffer (0.1 M Tris-HCl, 5% SDS, 2% β-mercaptoethanol and pH 6.8) and adjusted for consistency of the protein concentration. For western blot analysis, the protein samples were separated on a 3% gel and a 10%-15% separation gel using Tris/glycine/SDS buffer. The protein was transferred to 0.2 µm nitrocellulose (NC) membranes using a Trans-Blot Turbo Transfer System (Bio-Rad). Three percent skim milk was used as the blocking agent, and TBST (10 mM Tris, 150 mM NaCl, 0.05% Tween-20, pH 7.6) was used as the washing buffer. Speci c antibodies (Abclonal Biotechnology Co, China) raised in rabbit against nine different nif proteins were used for analysis at a 1:1,000 dilution.
The NC membranes were xed, closed, treated with the anti-nif protein antibody and then incubated overnight at 4 °C. Finally, the NC membranes were reacted with a horseradish peroxidase-conjugated goat anti-rabbit IgG antibody and colored via the ECL method.
Immunogold labeling 75  To assay nitrogenase activity, transgenic plants were transferred to nitrogen-de cient or low-nitrogen medium at the germination and seedling stages. To test the inhibition of oxygen on nitrogenase, we cultured plants in a sealed bag (Mitsubishi Gas Chemical Company, Japan), in which the mixed gas with different concentrations of oxygen were lled. After 2 weeks, the transgenic Arabidopsis plants were used for biomass, chlorophyll, proteins in solution and free amino acid analyses.
Chlorophyll content was determined spectrophotometrically at 649 and 665 nm, according to the Lichtenthaler and Welburn method 77 . The protein content was determined from the OD 280 78 ; bovine serum albumin was used as the standard. Ammonium concentrations in Arabidopsis seedlings were determined by uorescence spectroscopy at a neutral pH based on o-phthaldialdehyde (OPA) 79 . Arabidopsis seedlings (50 mg) were ground and homogenized with 1 ml of cold 10 mM formic acid solution. The homogenate was centrifuged at 13,000×g (4 °C) for 15 min, and the supernatant was transferred to a new eppendorf tube and stored on ice. OPA (16 mg) was dissolved in ethanol, and 40 ml of 0.1 M phosphate buffer containing 10 mM β-mercaptoethanol (pH 6.8). A 10 µl volume of the extract was mixed with 400 µl of OPA solution, reacted at 80 °C for 15 min and immediately cooled on ice. Fluorescence was measured at 470 nm with excitation at 410 nm using an F-2700 uorescence spectrophotometer (Hitachi, Japan). Ultrapure (NH 4 ) 2 SO 4 was used as the standard.
Cyanide (CN − ) is an inhibitor of electron ow and is toxic to plant cells. To assay nitrogenase activities in vivo, transgenic Arabidopsis plants were transferred to low-nitrogen medium containing 10 mg/L KCN, and the KCN resistance of seedlings was investigated. Similar to cyanide, sodium azide is a highly toxic cellular respiratory inhibitor that can hinder seed germination and growth. Azide can also act as a nitrogenase substrate; it is catalytically reduced to produce N 2 H 2 and NH 3 . Thus, transgenic Arabidopsis plants were transferred to low-nitrogen medium containing 0.65 mg/L NaN 3 to investigate its effects on seed germination and plant growth.

N 2 incorporation assay
Seeds of A. thaliana were surface sterilized and then sown in 50 ml asks containing 20 ml of solid nitrogen-de cient MS medium with 9 µM Na 2 MoO 4 and 50 mM Fe(III)C 6 H 5 O 7 . The asks were sealed with a rubber stopper. Fifty percent of the air in the asks was replaced with 15 N 2 (99%+, Shanghai Engineering Research Centre for Stable Isotope). After 30 days of incubation at 25 °C, the cultivated samples were dried at 80 °C and gradually ground into ne powder.
Five milligrams of sample and 50 mg of copper oxide particles were put into a glass reformer tube, and then the air in the tube was extracted with a vacuum system. The sample reformer tube was sealed when its vacuum degree reached 0.01Pa. The sealed tube was heated in a mu e furnace at 530 °C for 4 h. After the reaction, the tube was cooled for mass spectrometry analysis. and dithiothreitol (0.5 mM) and then centrifuged at 12,000 rpm for 15 min. To maintain anaerobic conditions during protein extraction, the samples and buffers were maintained inside either a bag or sealed centrifuge tubes lled with argon (Ar). All samples and buffers were washed out with Ar to remove O 2 from the solutions and then stored at 4 °C throughout the assay.
In vitro nitrogenase activities were tested by using an ATP-regeneration system with dithionite as the arti cial electron donor 52 . Exactly 0.2 ml of crude protein and 0.8 ml of the enzyme reaction solution containing ATP, MgCl 2 , creatine phosphate (Sigma), creatine phosphokinase (Sigma, 324 u/mg) and 40 mM MOPS-KOH (pH 7.4) were placed in a serum bottle (10 ml), which was sealed with a rubber plug and deoxygenated several times with high-purity Ar. C 2 H 2 (10% of the headspace volume) was injected into the test tubes. After incubating the cultures at 30 °C with shaking at 250 rpm for 1 h, the reaction was stopped with 30% TCA. Thereafter, 1 ml of the culture headspace was withdrawn through the rubber stopper with a gas-tight syringe and manually injected into an Agilent 7890B gas chromatograph to quantify ethylene production. All treatments were performed with three replicates, and all experiments were repeated at least three times.

Sample preparation for GC-MS analysis of isotope-labeled amino acids and chlorophyll
Four-week-old Arabidopsis seedlings were transferred to 50 ml asks containing nitrogen-de cient MS medium with 9 µM Na 2 MoO 4 and 50 mM Fe(III)C 6 H 5 O 7 and grown for 30 days. Then 20% air in the asks was replaced with isotopic-labeled N 2 (99%+, Shanghai Engineering Research Centre for Stable Isotope).
Seedlings were grown under such condition for one week. For isotopic analysis of amino acids, samples must be pretreated with silylation reactions 80 . After freeze-drying treatment, the sample from 3 different seedlings was homogenized with distilled deionized water at a ratio of 1/25 (W/V) and sonicated for 30 min in an ice bath. The homogenized samples were then centrifuged for 10 min at 12000 r/min.  (9) where A mix is the peak intensity of the measured mass spectrum for the sample.
The 15 N isotopic abundance of chlorophyll b can be determined in the same way.     Error bars indicate the standard deviation observed from three independent experiments.