Plant transformation, regeneration, and characterization
Maize transformation of the inbred line A188 with the Agrobacterium tumefaciens strain LBA4404 harboring a super-binary plasmid was performed according to the protocol of Ishida et al.31. In particular, the composition of all media cited hereafter is detailed in the reference above. The protocol was slightly modified concerning the selective marker, which was the NPTII gene instead of the bar gene. The super-binary plasmid used for transformation was the result of a recombination within the Agrobacterium tumefaciens strain between plasmid pBIOS 1459 and plasmid pSB1 (harboring the virB and virG genes isolated from pTIBo54232). Plasmid pBIOS1459 contains between the T-DNA borders, a neomycin resistance cassette (NPTII gene33) flanked by an Oryza sativa (Os) actin promoter with its first intron and 3′Nos terminator). This selectable marker cassette was flanked by 2 DS elements (5′Ac and 3′Ac34). Two copies of a synthetic Gln1-3 full length cDNA15 (accession numbers D14577.1 and X65928) including their 3′ non-coding regions, one flanked by the cassava vein mosaic virus promoter (CsVMV promoter35) fused to the rice actin1 first intron36 to enhance the activity of CsVMV promoter and the 3′Nos terminator and the other one flanked by the promoter of the maize Rubisco small subunit (RbcS)37 and the 3′Nos terminator. The nucleotide sequence of Gln1-3 was 98% similar at the amino acid level to that of Gln1-4 the other gene encoding GS113. The resulting recombinant plasmid used for transformation was pBIOS1459 (Supplementary Fig. 1). Transgenic plants were then cultivated in a glasshouse (18–24 °C) from April to September 2008 and either selfed or pollinated with the WT line A188 to produce seeds. For each transgenic line, the number of T-DNA copies was determined by q-PCR using a primer amplifying a T-DNA specific fragment located between the selection cassette and the cassette with the gene of interest (Forward: CCGTCCCGCAAGTTAAATATGA and Reverse: GCTTAGATCTGAGATCGGTAAGGAA), (See Supplementary Fig. 1, for the position of the primers). Twelve independent transgenic hybrid lines (H12, H14, H17, H18, H20, H22, H23, H27, H31, H32, H39, and H40) containing up to two inserted T-DNA copies were selected for the different field trials. After an initial cross of the primary transformant (T0 plant) with pollen of the wild type (WT) A188, 2 rounds of self-pollination were performed in order to obtain plants for which the cob carried only homozygous seeds. Finally, in order to perform the field trials under agronomic conditions, homozygous Gln1-3 over-expressing lines were crossed with a non-transgenic tester line (RBO1) to obtain hybrid seeds. In addition, transgenic hybrids were produced by crossing the line overespressing GS1 with two other different testers AAX3 and AAX7. The RBO1 line belongs to the Stiff Stalk heterotic group, whereas AAX3 and AAX7 belong to the OH43 and Iodent*OH43 heterotic groups respectively.
Plant material for agronomic and physiological studies
Twelve maize (Zea mays, L.) hybrids in which the plasmid pBIOS1459 was introduced (H12, H14, H17, H18, H20, H22, H23, H27, H31, H32, H39, and H40), two bulks of hybrid null segregants (named T01581-BULK-NS and T01594-BULK-NS and a control untransformed wild type hybrid (CH) were grown in the field. The two bulks of hybrid null segregants were obtained from the cross between the tester line RBO1 and various homozygous plants derived from transgenic events for which the transgene was outcrossed during the first self-pollination. Untransformed control hybrids corresponding to the cross between A188 and RBO, were used either as the female or male parental lines. In addition, maize hybrids (H14, H20, H22, H32, and H39) were produced, using two other testers (AAX3 and AAX7). Transgenic and control hybrids were grown at different locations in the USA: Alleman (IA, N 41.834911/W -93.656433). in 2010, Visala (CA, N 36.539310/W -119.242780), Stewardson (IL, N 39.383640/W -88.552070), Finch (MA, N 41.968333/W -93.605097) and Mason City (IA, N 43.331338/W -93.088322) in 2011, Gibson (IL, N 38.659080/W -121.751742), Ashkum (IL), Findlay (OH, N 39.521080/W -88.775610), Finch (MA, N 41.968333/W -93.605097), George (WA, N 38.712022/W -121.763260) in 2013, Findlay (OH, N 39.521080/W -88.775610) in 2014, Sleepy Eye (MN, N 44.369494/W -94.675001) and Findlay in 2015. The locations and their yearly average temperature are shown in Supplementary Data 1b. The list of the hybrids tested over different years for either their biochemical characterization or for their agronomic performance is presented in Supplementary Data 1a. The number of independent transgenic hybrid lines used for the different field trials was varied from one year to the other depending on seed availability. The level of N fertilization was 175 kg/ha supplied before sowing and N provided by the soil was estimated at 60 kg/ha. Both phosphorus (P205) and potassium (K20) were also applied at 100 kg/ha. Water deficit experiments were conducted in order to attain a decrease in kernel yield of approximately 20–30%. The intensity of the drought stress was controlled by monitoring soil moisture using a watermark granular matrix sensor38.
The control and transgenic hybrids were grown in a random block design, completely balanced lattice or split plot depending on the year of experimentation and on the location, with four to five replicates for each hybrid line (Supplementary Data 4). An outside border area of at least six rows (commercial variety 356M70 from Blue River Organic Seed, IA, USA) was planted surrounding the different plots of the field trial. Each plot was composed of two rows of maize of 6 m length. The number of plants in each plot was between 60 and 75, depending on the location.
The seeds were sown in May in the five years of experimentation. Kernel yield components at plant maturity were measured using a plot combine harvester (Juniper Systems, Inc., Logan, UT, USA). Kernel yield (KY), thousand kernel weight (TKW), and kernel moisture (KM) were measured using the on-board equipment of the plot combine harvester. Kernel weight was then normalized to moisture at 15% using the following formula: Normalized kernel weight = measured KW × 100-measured moisture (expressed in %)/85 (100-normalized soil moisture at 15%). In all the field experiments kernel yield was expressed in quintal per hectare (Ql. ha−1).
For all biochemical analyses performed at the vegetative stage (V) and at 15DAS leaf and kernel samples were harvested in the year 2011 in Finch (MA, USA) using a pool of the two BULKS-NS (NS) and the wild type hybrid (CH) as controls, and the transgenic lines H12, H14, H18, H20, H22, H23, H27, and H32 (Supplementary Data 1a). As 25% of the plants were removed from the different plots or used for leaf sampling, this 2011 field trial which was partly destructive was not included in the evaluation of yield over the five years of experimentation. At the vegetative (V) stage, half of the 6th fully emerged leaf without the main central midrib was harvested at the 7–8 leaf stage between 9 a.m. and noon. For the grain filling stage, half of the leaf below the ear was harvested 15 days after silking (15DAS), each plant being harvested at the same developmental stage. The plant developmental stage at 15DAS has been shown to provide a good indication of the transition occurring when both C and N metabolites start to be actively translocated to the developing kernels6. Moreover, the leaf below the ear was selected, since it provides a good indication of the sink to source transition during grain filling39. No major delays in silking dates were observed between the different hybrids. The leaf samples were immediately placed in liquid N2 and stored at −80 °C until further analysis. In the same experiment, in order to evaluate the agronomic performance of the Gln1-3 overexpressors in comparison to the controls, shoot DW 15DAS, shoot DW at maturity, and kernel mass per plant (g/plant) were measured. The leaf C and N contents were also measured on plants harvested 15DAS and at maturity (kernel harvesting) and in the kernels.
Metabolite extraction and biochemical analyses
Leaf GS activity was measured using the biosynthetic reaction using hydroxylamine instead of ammonium as a substrate leading to the formation of γ-glutamyl hydroxamate as previously described13. The total C and N content of 25 mg of frozen leaf material and dry kernels was determined in an elemental analyzer using the combustion method of Dumas (Flash 2000, Thermo Scientific, Cergy-Pontoise, France). For nitrate and ammonium measurements, 100 mg of frozen leaf powder were extracted in 1 mL of 80% ethanol at room temperature for an hour. During extraction, the samples were continuously agitated and centrifuged at 12,000 × g for 5 min. The supernatants were removed and the pellets were subjected to a further extraction in 60% ethanol and finally in the water. All the supernatants were combined to form the water/ethanol extract. Nitrate and ammonium were determined by the method of Cataldo et al.40 and by the phenol hypochlorite assay41.
15N-labeling experiment
The 15NH4+ labeling experiment was conducted using detached young developing leaves collected at the vegetative (V) stage of plant growth and development. The atom %15N of each amino acid was then determined by GC-MS analysis according to the protocol described by Cukier et al.42. In this protocol, the methods are described for tracing the pathway by which plants are able to take up 15N-labeled ammonium and convert them into amino acids. Following amino acid extraction, purification, and separation by GC/MS, a calculation of the 15N enrichment of each amino acid is carried out on a relative basis to identify any differences in the dynamics of amino acids accumulation. Plants were grown under hydroponic conditions on a complete nutrient solution30 and the 6th emerged leaves from four individual maize plants for the control and transgenic hybrid lines were used to perform a 6 h labeling experiment with 4 mM NH4Cl, enriched with 50% 15NH4Cl.
Protein gel blot analysis
Soluble proteins were extracted from frozen leaf powder that had been previously stored at −80 °C. Extraction was performed at 4 °C in a buffer containing 50 mM Tris-HCl pH 7.5, 1 mM EDTA, 1 mM MgCl2, 0.5% (w/v) polyvinylpyrrolidone, 0.1% (v/v) 2-mercaptoethanol and 4 mM leupeptin, and separated by SDS-PAGE43. The percentage of polyacrylamide in the gels was 10% and equal amounts of protein (10 μg) were loaded onto each track. Proteins were electrophoretically transferred to nitrocellulose membranes for protein gel blot analysis. GS1 and GS2 polypeptides were detected using polyclonal antisera raised against GS2 of tobacco (Nicotiana tabaccum L.) diluted 10 times using the protocol described by Hirel et al.44. Soluble protein was determined using a commercially available kit (Coomassie Protein assay reagent, Biorad, München, Germany) using bovine serum albumin as a standard.
Immunolocalization studies
For light microscope immunological studies, leaf pieces (2–3 mm2) were fixed in freshly prepared 1.5% (w/v) paraformaldehyde in phosphate buffer 0.1 M, pH 7.4 for 4 h at 4 °C. For immunolocalization, material was dehydrated in an ethanol series (final concentration 90% v/v ethanol) and embedded in London Resin white resin (Polysciences, Warrington, PA). Polymerization was carried out in gelatin capsules for 10 h at 54 °C. For light microscope immunological studies, thin sections of 1 µm were floated on drops of sterile water on slides and treated at room temperature in 1% periodic acid for 1 h, followed by 0.5 M NH4Cl for 1 h. Sections were then washed three times with distilled water and incubated for 1 h at room temperature in T1 buffer (0.05 M Tris-HCl buffer containing 2.5% (w/v) NaCl, 0.1% (w/v) BSA, and 0.05% (v/v) Tween 20, pH 7.4) containing 10% BSA. Sections were first incubated with 5% normal goat serum in T1 buffer for 1 h at room temperature and then with anti-GS rabbit serum diluted 1:70 in T1 buffer for 6 h at room temperature. Sections were washed five times with T1 buffer, twice with T2 buffer (0.02 M Tris-HCI buffer containing 2% NaCl, 0.1% BSA, and 0.05% Tween 20, pH 8) and incubated with 10 μl colloidal gold-goat anti-rabbit immunoglobulin complex (Sigma, St. Louis, MO, USA) diluted 1:50 in T2 buffer for 2 h at room temperature. Immuno-gold labeling was revealed by silver enhancement as described by the supplier (British Biocell International) and sections were back-stained with 1% fuchsine before microscopical examination under bright field and/or epipolarised light on a Nikon Eclips 800 epifluorescent photomicroscope. Controls were run either by omitting the primary antibody or by its substitution with preimmune serum. For immunolocalization examination by transmission electron microscopy (TEM) immunological studies, leaf pieces (2 to 3 mm2) were fixed in freshly prepared 1.5% (w/v) paraformaldehyde in phosphate buffer 0.1 M, pH 7.4 for 4 h at 4 °C. Leaf material was dehydrated in an ethanol series (final concentration 90% v/v ethanol) and embedded in London Resin white resin (Polysciences, Warrington, PA). Polymerization was carried out in gelatin capsules for 10 h at 54 °C. For immunotransmission electron microscopy studies, ultra-thin sections were mounted on 400 μm mesh nickel grids and allowed to dry at 37 °C. Sections were first incubated with 5% (v/v) normal goat serum in T1 buffer for 1 h at room temperature and then for an additional 6 h at room temperature with the GS antiserum, also used to perform protein gel blots, diluted 100 times in T1 buffer. Sections were then washed three times with T1 buffer and incubated for 2 h at room temperature with 10 nm colloidal gold goat anti-rabbit immunoglobulin complex (Sigma, St Louis, Mi) diluted 70 times in T1 buffer. After several washes, grids were treated with 5% (w/v) uranyl acetate in water and examined with a Philips CM12 electron microscope (Philips, Eindhoven, The Netherlands) at 100 kV. Negative controls were conducted by substituting the serum-containing GS antibodies with preimmune rabbit serum.
Production of the MAGIC panel and genotyping
The maize MAGIC panel used to perform association studies, was created from a funnel cross of 16 founder lines (B96, EP1, DK105, FV2, CO255, F492, A654, FV252, DK63, C103, OH43, A632, B73, W117, ND245, and VA85) representing the most significant heterotic groups used for maize hybrid production in temperate regions (similar to the panel produced by Dell’Acqua et al.45). The original population called BALANCE was used to extract 375 DHIL at the 3rd generation of mixing (G8). The crossing scheme used to create the MAGIC panel is illustrated in Supplementary Fig. 5. The 375 DHILs along with the 16 founder lines and the tester were genotyped with the Axiom maize 600 K array46. The parental origin of each allele in the DHILs was inferred using the R/qtl package47 and each DHIL was represented as a mosaic of the founder genomes. Knowing the parental mosaics of each DH lines, the SNP detected from the sequencing of the founders were projected on each DHIL (allowing in silico genotyping). The MAGIC panel allowed to detect more than 12 million single nucleotide polymorphisms (SNPs) following next generation sequencing (NGS) data of the 16 founders and of the tester line. Among these 12 million SNPs, approximately 8 million were selected using different criteria of quality control recommended by the US HapMap48 including identification by descent (IBD)49 and linkage disequilibrium (LD). These selected markers were positioned on the maize reference genome B73 RefGen_v450. The SNPs located in the coding sequences of the genes Gln1.3 (65 SNPs) and Gln1.4 (56 SNPs), in the promoter (5′) and 3′ terminal region, were used to perform the association genetics studies. Thus, a given Filtered Gene Set (FGS), (gene coordinate RefGen_v4)50 was extended with 5000 bp into the putative promoter region51 and extended with 500 bp into the 3′ terminal region, for the final selection of SNPs.
Association study
The association genetics study was performed using SNP markers identified by sequencing the founder lines of the MAGIC panel, targeting two candidate genes Gln1-3 and Gln1-4, including their promoter and 3′ terminal regions as described above. The SNP markers were anchored on the maize genome RefGen_v450.
Field trials and phenotyping of the MAGIC panel
The DHILs of the panel were crossed to the tester line MBS847. The test-cross progenies were evaluated in the field for yield, in two different locations and over two years of experimentation. The experiments were designed to limit confounding effects due to differences in precocity, notably by reducing the range of flowering time. Kernel yield was adjusted to 15% moisture. Two field trials were conducted in Blois (France, 47.74683°N 1.23296°E, altitude = 123 m) and in Saint-Paul-lès-Romans, France, 45.041260°N 5.070226°E, altitude = 185 m) in 2015 using 346 hybrids. A third field trial was conducted in Saint-Paul (Saint-Paul-lès-Romans, 45.07464°N 5.11341°E, altitude = 185 m) in 2016 using 380 hybrids. Two additional field trials were conducted in Blois (47.444498°N 1.135907°E, altitude = 120 m) and in Nerac (France, 44.101156°N 0.18223°E, altitude 55 m) in 2017. The total level of N fertilization was 175 kg/ha (two successive applications of 150 kg/ha and 200 kg/ha in the form of urea at 1 leaf and 8 Leaves stages respectively and N provided by the soil was estimated at 60 kg/ha. Both phosphorus (P205) and potassium (K20) were also applied at 100 kg/ha. The experiments were carried out following an alpha-lattice design with two replicates for each trial. The plot length was comprised between 5.35 and 5.2 m with two rows spaced by a 0.8 m interval in the different locations. Temperature, rainfall, solar radiation, and water potential at 30, 60 et 90 cm depths were recorded each day to monitor the trials and especially the irrigation status to avoid a water stress to occur. The two first trials were sown on April 23rd 2015 in Blois at a density of 95000 seeds/ha in a clay-sand soil and 86250 seeds/ha in Saint Paul in a sandy silty clay soil. The third trial was sown on May 6th 2016 in Saint Paul at a density of 90000 seeds/ha. In 2017 the two trials were conducted in Blois and in Nerac at a density of 90,000 and 85,000 seeds/ha respectively in a silty clay soil for the later. In Saint Paul in 2015 a thermal stress defined as hot temperature scenario52 occurred at flowering and during the grain filling period whereas in Nerac in 2017 it was only during the later. Two applications of N were scattered over the plots at 1 Leaf and 8 Leaves stages. Weeds, diseases, and pests were controlled using conventional agronomic practices for both trials. The adjusted mean values for kernel yield are presented in Supplementary Data 7.
Statistics and reproducibility
For the combined agronomic and physiological studies performed on plants grown in the field in Finch in 2011, the control and transgenic hybrids were grown in a random block design with four blocks. For all biochemical analyses, shoot, and kernel mass production in each of the four different blocks, six leaf samples or plants were harvested each coming from a different plant. The six leaf samples or plants were pooled, making four replicates in total. All data shown in the graphs are presented as the mean with S.D. One-way ANOVA statistical analyses were performed using a Student−Newman−Keuls test (t-test) to identify differences between the controls and the transgenic hybrids (p ≤ 0.05). For the protein gel blot analysis, the four soluble protein extracts used to measure total leaf GS activity were pooled, each containing 2.5 μg of proteins. For the 15NH4+ labeling experiment, four individual maize plants for the control and transgenic hybrid lines were used to perform the labeling experiment. Statistical analysis was performed by a t-test, which was used to identify amino acids exhibiting a different pattern of 15N-labeling (p ≤ 0.05). For light microscope immunological studies, gold particles were counted on 10–15 different leaf sections. Significant differences at p ≤ 0.05 were identified using the t-test. For all the other field experiments performed to measure differences in yield between the controls and the transgenic, a two-way statistical analysis was performed using a mixed model restricted maximum likelihood. The model is the following: random factors = replicate (complete), sub-block (ibloc) tested in the replicate (for lattice only). Fixed factor (construct). A test of the differences in terms of the least square means was made for the fixed effects at the construct level between the control hybrids (CH and NS) and the transgenic hybrids. The association between the SNPs, and kernel yield measured in different years and locations, was performed using a mixed linear model (MLM) implemented through ASReml statistical package that fits linear mixed models using residual maximum likehood (REML)53, (VSN International Ltd) on R environment54. The model used was the K model55. The association study was conducted using the leave one chromosome out (LOCO) method in which the kinship is calculated from all the SNPs but those located on the chromosome where the tested marker is located56,57. For the field trials and phenotyping of the MAGIC panel, the experiments were carried out following an alpha-lattice design with two replicates for each trial. The adjusted mean values for kernel yield were calculated.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
