Glycine max [L.] Merr. cv. Pusa-16 procured from Division of Genetics, IARI, New Delhi, India was used for isolation of GmIPK2 gene and as the recipient genotype for subsequent genetic modification. Quantitative characteristics of this cultivar grown primarily in northern plain and hill zones of India have been described in studies made by Karnwal and Singh (2009) and Ramteke et al. (2010; Ramteke and Murlidharan, 2012).

Based on the transcript sequence of soybean IPK2 gene available in the plant comparative genomics portal Phytozome v9.1 (Glyma.12G240900) we designed primers to amplify and clone a 305 bp fragment including 250 bp of its 3′ end coding sequence and a conserved sequence of 55 bp from its 3′ untranslated region. This nucleotide fragment was used to amplify and directionally clone sense (GmIPK2_S; FP 5′-CGCGGATCCGCGTTGCAGAAGCTCAAG-3′ and RP 5′-TCCCCGCGGGG AGCGACACTAATTCAAG-3′) and anti-sense (GmIPK2_As; FP 5′-CCGCTCGAGCGGAACGTC TTCGAGTTC-3′ and RP 5′-CCATCGATGGCGCTGTGATTAAGTTCGTA-3′) strands around a 395 bp spacer of soybean fatty acid oley1Δ12 desaturase gene intron (GmFad2-1; FP 5′-TCCCCG CGGGGAAGGTCTGTCTTATTTTGAATC-3′ and RP 5′-CCATCGATGGTATACCGCACTAGT AAACCAC-3′) cloned in pCR2.1-TOPO cloning vector to form a hairpin (ihp) structure which was confirmed by automated sequencing. The binary vector pCWAK that will carry this silencing cassette was constructed inhouse in a previous work by ligating fragments from the expression vector pCW66 and binary vector pAKVS both kindly obtained from Dr. Craig Atkins Laboratory, University of Western Australia, Australia. The final RNAi construct pCWAK-ipk2 was generated by restriction cloning (BamHI/XhoI) of ihp cassette in pCWAK under the control of soybean seed specific vicilin gene promoter and terminator (Figure 1). The plasmid pCWAK-ipk2 also bears phosphinothricin acetyltransferase (PAT) encoding marker gene bar within the T-borders driven by the cauliflower mosaic virus (CaMV) 35S promoter to confer tolerance to the herbicide glufosinate for transgenic selection. After extensive verification of pCWAK-ipk2 by restriction digestion, we subsequently mobilized the binary construct into disarmed Agrobacterium tumefaciens strain EHA105 by triparental mating using Escherichia coli DH5α (pRK2013) helper strain for further use in preliminary transformation experiments.

Soybean transformation was performed following Agrobacterium-mediated cotyledonary-node method described by Zhang et al., 1999 and Paz et al., 2004 with minor modifications. Briefly, sterilized soybean seeds germinated on half B5 medium were excised to derive two cotyledonary explants per seed and infected with A. tumefaciens harboring the RNAi construct (Figure 2A). Following 3 days of co-cultivation in dark in the presence of thiol compounds (3.3 mM L-cysteine, 1 mM dithiothreitol, 1 mM sodium thiosulfate) to improve cell transformation efficiency (Olhoft and Somers, 2001; Olhoft et al., 2001), the explants were sub-cultured twice for 12–14 days on full B5 medium (shoot induction medium, SIM) supplemented with 4 mg/l glufosinate for selective shoot induction (Figures 2B,C). Four weeks later, healthy explants were transferred to full MS medium (shoot elongation medium, SEM) containing 5 mg/l glufosinate for continued selection and sub-cultured in fresh medium every 2 weeks until shootlets approximately 2–3 cm high were developed (Figures 2D,E). Elongated shootlets were subsequently induced to develop roots by the application of 2mg/l Indole-3-butyric acid (IBA; auxin phytohormone) supplemented in half MS medium (rooting medium, RM) (Figure 2F). Well rooted plantlets were transplanted to soil pots for hardening and grown to maturity under 16 h/8 h light/dark regime at National Phytotron Facility, IARI, Delhi, India (Figures 2G,H).

Southern positive T₂ plants were further analyzed by semiquantitative RT-PCR and quantitative real-time PCR (qRT-PCR) to estimate IPK2 transcript levels. Total RNA was isolated from fresh soybean seed (8–10 mm) tissue of T₂ transgenic and non-transgenic plants using TRI Reagent (Sigma-Aldrich, United States). First-strand cDNA was synthesized from 5 μg of DNaseI-treated total RNA using M-MLV Reverse Transcriptase (Promega, United States) following manufacturer’s instructions. qRT-PCR analysis was subsequently performed following SYBR Green (DyNAmo Flash SYBER Green qPCR Kit, Thermo Scientific, United States) chemistry on PikoReal 96 Real-Time PCR platform (Thermo Scientific, United States). Gene-specific primers were designed to amplify a 124 bp fragment from its conserved region (qIPK1F 5′-GGAGCGCTTGCAGAAGC-3′, qIPK1R5′-GACCAGAGGGTTGGT AGC-3′). To normalize the variance among samples, house-keeping gene phosphoenolpyruvate carboxylase (PEPCo) (qPEPCoF5′-CATGCACCAAAGGGTGTTTT-3′, qPEPCoR 5′-TTTTGCG GCAGCTATCTCTC-3′) was used as endogenous control. Amplification was achieved by a two-step PCR reaction with an initial denaturation at 94°C for 4 min followed by 40 cycles of 94°C for 30 s and annealing/extension at 60°C for 30 s. Triplicate quantitative assays were performed on biological triplate corresponding to each sample. Relative mRNA abundance was calculated using the 2^-ΔΔCT method described by Livak and Schmittgen, 2001. The specificity of each unique amplification product was determined by melting curve analysis.

Total phosphorous in the seeds was determined colorimetrically by a method described by Chen et al. (1956). Samples (transgenic and non-transgenic control) for the assay were prepared by wet ashing dried ground seeds (250 mg) in 2 ml of concentrated sulfuric acid (Larson et al., 2000; Raboy et al., 2000).

For the analysis of inorganic phosphorus (Pi) levels, we followed a protocol described by Al-Amery et al. (2015) which uses single seed chips (1–2 mg) derived from cotyledonary segment opposite to the embryonic axis. Briefly, seed samples were extracted using 50 μl buffer (25 mM magnesium chloride and 12.5% trichloroacetic acid) for 14–16 h at 37°C with gentle shaking. 10 μl subsample of extracts were loaded in triplicates onto a 96 well plate and diluted with 90 μl water. Each sample was incubated with 100 μl of Chen’s reagent at 37°C for 1 h and the absorbance was read at 882 nm on GloMax^®-Multi Detection System (Promega, United States).

Phytic acid estimation in transgenic seeds was performed by reversed-phase high-performance liquid chromatography (RP-HPLC) as described by Lehrfeld (1989) with some modifications (Pandey et al., 2016). 500 mg dried seed tissue was homogenized in a pestle-mortar, hexane-defatted and subsequently extracted with 0.78 M HCl by sonication (3 min) and mechanical agitation for 1 h at 250 rpm. The extract was centrifuged and one part of clear supernatant was mixed with four parts of HPLC grade water. The diluted sample was passed through conditioned SAX column (Hypersep, Thermo Scientific, United States) connected to a vacuum manifold (Millipore, United States) and eluted with 2 M HCl. The filtrate was evaporated in a vacuum rotary evaporator (Hei-VAP Value Digital G3, Heidolph, Germany), re-dissolved in 1 ml of mobile phase (acetonitrile, formic acid and tetrabutylammonium hydroxide, 4.8:5.1:0.1, v/v/v) and filtered using a 0.2 micron PVDF syringe filter (Millipore, United States). For HPLC analysis, sample was injected onto the C18 RP-column (250 × 4.6 mm, 5 μ; Shimadzu, Japan) equilibrated with isocratic mobile phase at a flow rate of 1.0 ml min^-1. PA signals were monitored with a UV-VIS photodiode array detector (SPDM-20Avp, Shimadzu, Japan) at a wavelength of 197 nm. The PA concentration was calculated using the calibration curve prepared with a PA dodecasodium salt standard (Sigma-Aldrich). The software used for analysis was LC-Solutions (version 1.25).

Mature transgenic and non-transgenic seeds were milled (Wiley, Thomas Scientific, United States), weighed (1 g) and digested in triplicates under simulated gastrointestinal conditions following method described by Kiers et al., 2000 to determine the bioavailability of iron (Fe), zinc (Zn), and calcium (Ca). The digest was centrifuged at 9000 rpm for 15 min and supernatant obtained was passed through a 0.45 mm filter. The filtrate was analyzed for Fe, Zn, and Ca using an Atomic Absorption Spectrophotometer (AAnalyst 200, Perkin Elmer, United States). A blank consisting of distilled water was processed in a similar manner and used for sample correction.

Every experiment was carried out in biological and experimental replications (three to six) for each non-transgenic control and transgenic sample and represented as mean ± standard error. All statistical evaluations were performed using SAS software (version 9.2). Segregation patterns analyzed with chi-square (χ²) goodness of fit were tested at 5% level of significance. We also conducted ANOVA for agronomic traits of transgenics and non-transgenic control and compared the group means by Tukey’s test at 5% level of significance.

A preliminary screening was conducted on the T₀ plants to identify putative lpa transgenic lines by PCR amplification of ∼700 bp fragment of ihp construct. The assay characterized eight independent transformation events each showing the expected amplicon size which was absent in the case of untransformed control plants (Figure 3).

Seeds from each independent event identified were grown successfully under containment conditions to obtain T₁ progeny plants by self pollination. All the derived progenies appeared phenotypically normal and fertile. We performed segregation analysis on them by studying bar gene expression and found that five of the events (P2, P4, P5, P6, and P8) held a good fit to the stable Mendelian inheritance of a single locus (3:1) (Table 1) whilst three (P1, P3, and P7) of them showed a segregation ratio of two transgene loci (15:1).

We further confirmed integration of the transgene cassette by southern blot analysis using bar gene probe (Figure 4). The genomic DNA was digested with PstI endonuclease which is not present within the T-DNA of the construct. Upon detection, all the transgenic lines analyzed showed a distinct pattern of seperation which indicate that each plant originated from an independent transformation event. The blot showed that all the fragments were above 3 kb in size which suggest that the transgenic lines carry intact copies of T-DNA because the shortest fragments in each line were longer than the T-DNA region (∼2.8 kb). The blot also confirmed single copy integration of target gene into the genome of transgenic events P2, P4, P5, P6, and P8. Since single copy insertions are always desirable, T₂ generation plants were cultivated from seeds of these five characterized events and PCR screened to select homozygous lines. A positive amplification in all the tested T₂ plants derived from the T₁ parents, P2-45 and P6-39, suggest their homozygous nature and thus seeds collected from only these parents were selected to characterize lpa trait by further analysis.

To assess the level of reduction in GmIPK2 expression in transgenic seeds with respect to non-transgenic control, we carried out RT-PCR and qRT-PCR analysis. From RT-PCR results, we observed a variable reduction in the expression of GmIPK2 gene with no variation in PEPCo transcript expression between the non-transgenic and transgenic plants thus supporting its use as a stable endogenous reference (Figure 5A). We further quantified the variations in GmIPK2 transcript levels by qRT-PCR and found a maximum reduction of 2.79-fold and 2.56-fold in T₃ developing seeds from the transgenic lines P2-45-8 and P6-39-10, respectively (Figure 5B) at P ≥ 0.05. We also studied GmIPK2 expression in other plant tissues of these two lines and found no variation in its transcript levels with respect to the control plants (Supplementary Figures S3, S4) which suggest its successful silencing only in the seeds.

To further examine the effect of GmIPK2 silencing on PA synthesis, we quantified PA content in mature grain extracts of above two lines and non-transgenic samples by HPLC analysis. The chromatogram obtained by scanning the samples at 197 nm displayed PA retention peaks around 7.006 ± 0.09 min, the area under which was used to compute the level of PA. A decrease in the PA content was observed in T₃ seeds compared to control which displayed larger peak area indicating its higher concentration in the seeds (Figures 6–8). The mean PA content was 3.48 g 100g^-1 for non-transgenic seeds against 1.91 g 100 g^-1 and 2.02 g 100 g^-1 for the seeds from transgenic lines P2-45-8 and P6-39-10, showing a maximal reduction of 45 and 42%, respectively (Figure 9).

We estimated the total and free phosphorus (Pi) levels in transgenic as well as non-transgenic seeds to correlate it with the reduction in PA content. No significant difference in total seed phosphorus (P) was observed for the RNAi lines as compared to the control. The average total phosphorus content of T₃ seeds was 5.82 mg g^-1 (P2-45-8) and 5.75 mg g^-1 (P6-39-10), which was observed closely similar to that of non-transgenic seeds, 5.89 mg g^-1. However, a significant increase of 39 and 37% in the concentration of Pi was recorded in the T₃ seeds obtained from transgenic lines P2-45-8 and P6-39-10, respectively, which gave a dark blue reaction with chen’s reagent, confirming a reduction in the content of its principal storage form PA (Figure 9).

In order to determine whether mitigation of PA improved the mineral bioavailability in soy grains, we made a quantitative estimation of micronutrients in transgenic seeds compared to non-transgenic control by following in vitro digestion method. From the observations made by atomic absorption spectrophotometer, we noticed an increase in the in vitro Fe, Ca, and Zn availability from 54.8, 44.15 and 58.50% for non-transgenic control seeds to 71.1, 59.2, 66.9% for P2-45-8 and 70.7, 58.42, 64.50% for P6-39-10 seeds (Table 2). As mentioned before, PA binding with minerals result in formation of insoluble salts with poor bioavailability, our result suggests that the mineral bioavailability can significantly be improved by reduction of PA in lpa transgenic lines.

Decrease in seed PA concentration can be accompanied by adverse agronomic consequences. Therefore, we conducted germination tests on the lpa seeds collected from our transgenic lines. The CGT test so conducted confirmed that 98% of the seeds possessed good germination potential whereas AAT test established their high vigor (Supplementary Figure S5A). Also, the morphological analysis of seed germination in each of the tests revealed similar phenotypes in both transgenic and control seeds (Supplementary Figure S5B). Besides, since PA also regulates the growth and development of seedlings, we carried out phenotyping on 75 days old plants. The growth of transgenic plants was compared with the growth of control soybean plants under field growth conditions. Interestingly, P2 and P6 plants exhibited an increase in leaf size which could possibly be explained due to increase in available seed P content (P < 0.05) (Tubana and George, 1997). No significant differences were observed between wild type and transgenic plants in other agronomic traits investigated (P > 0.05) (Table 3). This data confirms that GmIPK2 silenced, lpa soybean seeds are capable of generating plants showing normal morphologies and agronomic traits.