Soybean KTI genes belong to a gene family with 38 members ( Figure S1 ). As displayed in the genetic map, 38 KTI genes are located on 9 chromosomes (Chr) in the soybean genome. Specifically, Chr1 carries 2 KTI genes; Chr3 carries 1 KTI gene; Chr6 carries 1 KTI gene; Chr8 carries 13 KTI genes; Chr8 carries 10 KTI genes; Chr12 carries 1 KTI gene; Chr16 carries 8 KTI genes; Chr18 carries 1 KTI gene; and Chr19 carries 1 KTI gene ( Figure S1 ).

To identify the KTI genes expressed in the seed, we analyzed the expression patterns of all KTI genes in cv. WM82 based on the expression data acquired through the Gene Networks in Seed Development database (http://seedgenenetwork.net/sequence) ( Figure 1A ). According to the expression patterns of KTI genes in various soybean tissues as displayed in Figure 1A , four KTI genes (Gm01g095000 (KTI1), Gm08g341000, Gm08g342300, and Gm08g341500 (KTI3)) were identified as seed-specific KTI genes. Soybean breeding lines V98-9005 (normal TI) and V03-5903 (low TI), presenting significantly different amounts of KTI concentration in seeds, were used to validate the tissue specific expressions of the four KTI genes by real-time PCR. Gm01g095000 (KTI1) and Gm08g341500 (KTI3) were predominately expressed in seeds compared to other tissues ( Figure 1B ). Both Gm08g341000 and Gm08g342300 had a relatively lower expression level in seeds than KTI1 and KTI3 but had higher expression in other tissue types ( Figure 1B ). Interestingly, both KTI1 and KTI3 had a relatively low expression level in the seeds of V03-5903 (low TI line), but higher expression in V98-9005 (normal TI line). Thus, we conclude that KTI1 and KTI3 are two major genes that may directly contribute to the TI contents in soybean seeds.

To knock out the KTI1 and KTI3 genes from cv. WM82 genome and create a new soybean cultivar with low TI content in soybean seeds, we developed a CRISPR/Cas9 construct, pBAR-Cas9-kti13, where the nuclease gene Cas9 is expressed by Arabidopsis ubiquitin 10 (U10) promoter. A bar gene driven by a MAS (mannopine synthase) promoter was used for selection of the putative transformants with bialaphos or phosphinothricin ( Figure 2A ). A tandem array of two sgRNAs targeting KTI1 and one sgRNA targeting KTI3 was expressed by the U6 RNA promoter ( Figure 2B ).

pBAR-Cas9-kti1/kti3 was transformed into WM82 via Agrobacterium-mediated transformation (Plant Transformation Facility at Iowa State University). Seventeen putative transgenic shoots were regenerated. Six shoots elongated and were transferred to rooting mediums. After further selection, they were transplanted into soil. Four lines, No. #2, #5, #11 and #17 were confirmed to be true transformants by positive amplification of the bar gene and a part of the Cas9 gene ( Figures 2C, D ). The gene editing events in T0 plants were identified by amplification and sequencing of DNA fragments covering the sgRNA binding sites of KTI1 and KTI3. The double peaks in the sequencing chromatograms suggest that both KTI1 and KTI3 genes were mutated and resulted in heterozygous alleles in the edited plant cells ( Figure 2E ). T0 seeds were harvested from T0 lines #2, #5, #11 and #17. Four T0 seeds of each line were randomly picked for DNA extraction and genotyping of the KTI1 and KTI3 genes via PCR amplification and DNA sequencing. The KTI1 gene editing was completed and resulted in homozygous mutant alleles in all tested T0 seeds of the four lines. In addition, an identical gene editing pattern in KTI1 was detected in all tested T0 seeds, in which a small DNA fragment (66bp) between two sgRNAs was lost after the gene editing ( Figures 3A–C ). Homozygous KTI3 mutant alleles were only detected in T0 seeds from 2-3, #5-4, #11-2, and #11-4 ( Figures 3D–G ). The gene editing patterns in KTI3 included both small deletions and insertions that all resulted in frameshift mutations in KTI3.

T0 seeds were also used for quantification of the KTI content by using a HPLC-based approach (Rosso et al., 2018). The tested seeds of #2-3, #5-4, #11-2, and #11-4, which carried mutations on both KTI1 and KTI3 genes had the lowest KTI content ( Figure 4 ). The tested seeds of #2-1, #5-1, #11-1, and #17-1, with only the KTI1 mutation, also had lower KTI content than the wild-type WM82 seeds ( Figure 4 ). The KTI content in other genotyped seeds with editing only on KTI1 was also lower than that in WM82 seeds (data not shown). We further tested the trypsin inhibition activity (TIA) using crude protein extracts from the T0 seeds. As shown in Figure 5 , the crude proteins of seeds with mutant kti1 and kti3 had the lowest TIA. The seeds with mutant kti1 only also had reduced TIA ( Figure 5 ) in comparison with WM82 and Glenn (a commercial soybean cultivar as a control). The KTI content and TIA were ranked in order as: kti1/3 double mutant < kti1 single mutant ≤ PI 547656 (low TI accession) < WM82 < Glenn. Taken together, we conclude that KTI1 and KTI3 are two major genes responsible for the KTI content and TIA in soybean seeds. Therefore, knockout of KTI1 and KTI3 reduced the KTI content and impaired the TIA in soybean seeds.

The edited KTI1 gene lost 66 bp that may result in mutant proteins with deletion of 22 amino acids. Truncated KTI1 may still possess some TIA. To rule out this possibility, we also tested the TIA of truncated KTI1_Δ22aa protein in vitro. To this end, we cloned the open reading frames of KTI1 _Δ66bp and wild-type KTI1 and KTI3 into a protein expression vector, in which a 6xHis tag is fused to C-terminus of the expressed proteins. The purified proteins were subjected to a TIA assay which showed that while KTI1 and KTI3 both could inhibit trypsin activity, the truncated KTI1_Δ22aa failed to suppress trypsin activity ( Figures S2A, B ). Therefore, the new kti1 allele (KTI1 _Δ66bp) encodes a truncated protein that loses its TI function.

To examine whether the mutant kti1/3 could significantly affect plant growth and maturity period days of soybean, we planted T0 seeds of line #2 and #5, and WM82 in a greenhouse. By bialaphos-mediated screening, we classified the T1 plants from line #2 and #5 as transgene-free plants or transgenic plants. We measured the agronomic traits of the transgenic plants including plant height, the number of main branches per plant, number of pods bearing branches, number of pods, leaf length, leaf width and petiole length. There was no significant difference in terms of all measured agronomic traits among the plants of WM82, Line 2 and Line 5 ( Table 1 ). We also measured the maturity period days of the soybean plants by recording the dates from planting to beginning bloom (R1), to beginning pod (R3), to beginning seed (R5), to full seed (R6), to maturity (R8), and the total lifespan (from planting to maturity). There were no remarkable differences in terms of R1, R3, R5, R6, R8 and total life span among all tested plants ( Table 1 ). Therefore, we conclude that knockout of KTI1 and KTI3 did not alter plant growth or the maturity period of soybean lines tested.

A double homozygous kti1 and kti3 mutant plant #5-26 that did not carry the Cas9 transgene was selected from T1 generation plants ( Figures 3A, C, D ). The ‘transgene-free’ soybean plants can be used to breed the low TI trait into other elite soybean cultivars. In order to co-select the kti1 and kti3 mutant alleles in the derived progenies, we attempted to develop co-dominate molecular markers that can distinguish between wild-type KTI1/KTI3 and mutant kti1/kti3 alleles.

In the genome of #5-26, the mutant allele of kti1 had a 66 bp deletion. We designed three PCR primers, ZW1, ZW2 and ZW3 ( Figure 6A and Table S2 ). ZW1 was a common reverse primer that can bind to the same region in both KTI1 and kti1 alleles, while ZW2 and ZW3 were both forward primers binding to unique sequences of KTI1 and kti1 alleles, respectively ( Figure 6A ). Soybean cultivars that carry the wild-type KTI1 gene (WM82) amplified a 180 bp DNA fragment when hybridized with ZW1 and ZW2, but failed to amplify any fragments when hybridized with ZW1 and ZW3. On the contrary, the soybean lines carrying a homozygous kti1 mutant allele (#5-26) amplified a 134 bp DNA fragment with ZW1 and ZW3, but failed to amplify any fragments with ZW1 and ZW2 ( Figure 6B ).

In the genome of line #5-26, the mutant kti3 allele had a 38 bp deletion, which allowed us to design PCR primers ZW4, ZW5 and ZW6 ( Figure 6A and Table S2 ). ZW4 was a common reverse primer for both KTI3 and kti3, while ZW5 and ZW6 were forward primers matched with unique sequences of KTI3 and kti3, respectively ( Figure 6A ). A soybean cultivar carrying the wild-type KTI3 gene (MW82) amplified a 264 bp DNA fragment with primers ZW4 and ZW5, but not with primers ZW4 and ZW6. In contrast, the soybean line carrying homozygous kti3 allele (#5-26) can amplify a 233 bp DNA fragment with primers ZW4 and ZW6, but not ZW4 and ZW5 ( Figure 6B ).

We further tested these PCR primers by amplifying DNA fragments from two T1 plants that were genotyped by DNA sequencing. Line #5-9 had homozygous kti1 alleles and heterozygous KTI3/kti3 alleles, where the kti3 allele was identical to the one in #5-26 ( Figure S3 ). Line #2-30 had homozygous kti3 alleles and heterozygous KTI1/kti1 alleles, where the kti1 allele was identical to the one in #5-26 ( Figure S3 ). As shown in Figure 6B , PCR amplification with the different combinations of ZW1, ZW2, ZW3, ZW4, ZW5 and ZW6 can accurately identify the KTI1/kti1 and KTI3/kti3 genotypes of #5-9 and #2-30 ( Figure 6B ). Therefore, we successfully developed molecular markers to select the kti1 and kti3 mutant alleles generated by CRISPR/Cas9 mediated mutagenesis. These molecular markers can assist in the breeding selection of low TI soybean plants harboring kti1/3.

To simplify the procedure of marker-aided selection, we tested a gel-electrophoresis-free protocol that can be implemented for high throughput screening of progenies derived from a cross between a soybean cultivar carrying wild-type KTI1/3 and one carrying the kti1/3 mutant. In brief, all PCR products as described above were mixed with 1X SYBR Green and heated at 75°C for 10 mins and then visualized under UV light. As shown in Figure 6B , the fluorescent signals were the indications of positive amplifications in WM82 with primers ZW1/ZW2 and ZW4/ZW5, while in #5-26, the fluorescent signals can only be observed with primers ZW1/ZW3 and ZW4/ZW6 (Hirotsu et al., 2010). Therefore, we identified the homozygous kti1 and kti3 alleles by directly staining the PCR products without the need of gel electrophoresis, which can significantly reduce the cost of labor and time.

Soybean plants were grown in 2.5-gallon pots using Miracle-Gro all-purpose potting soil mix in Keck Greenhouse at Virginia Tech (14h/10h light/dark cycle at 25°C/20°C) for the experiments described herein. The plants were watered by an automatic irrigation system. Soybean transformation was performed at the plant transformation facility at Iowa State University as previously described (Paz et al., 2006; Luth et al., 2015; Ge et al., 2016). The plant growth indicators and maturity period days of WM82 and progeny plants of the T1 generation derived from lines #2 and #5 were measured in the green house. The 4-week-old T1 soybean plants were used for genotyping. The seeds of T1 plants were used for seed weight analysis.

The gene map showing locations of KTI genes on soybean chromosomes was made using MapInspect. Locations of all KTI genes were obtained from the Phytozome database (https://phytozome-next.jgi.doe.gov/) and plotted on their respective chromosomes.

E. coli strains DH5α and C41 (DE3) (Lucigen, Middleton, WI) were grown on Luria agar medium at 37°C. Agrobacterium tumefaciens (A. tumefaciens) EHA105 was grown on Luria agar medium at 28°C (Zhao et al., 2011; Traore et al., 2019). E. coli antibiotic selections used in this study were as follows: 50 μg/ml kanamycin, 100 μg/ml carbenicillin, 100 μg/ml spectinomycin. A. tumefaciens antibiotic selection were 100 μg/ml rifampicin, and/or 100 μg/ml spectinomycin.

The open reading frames (ORFs) of KTI1 and KTI3, were amplified from the genomic DNA of WM82. The KTI1 _Δ66bp, truncated ORF of KTI1, was amplified from the genomic DNA of mutant soybean plant #2-1. All PCR primers with annotations are listed in Table S2 . The genes/fragments were then cloned into a pDonr207 plasmid (Thermo Fisher Scientific) for future use.

T1 plant genomic DNA (gDNA) was used as the templates to amplify KTI1 and/or its mutant allele, and KTI3 and/or its mutant allele by PCR. The purified PCR fragments were used for genotyping by Sanger sequencing at Virginia Tech Genomic Sequencing Center and cloned to the PCR8/GW/TOPO vector by TA cloning (Invitrogen) for molecular marker tests.

In order to apply the CRISPR/Cas9 system to gene editing in soybean, we modified our current CRISPR/Cas9 construct (Liu et al., 2016). The cassette consists of a MAS promoter, the bialaphos resistant gene, and a MAS terminator that was amplified using plasmid DNA of pEarleyGate101 as the template. All PCR primers with annotations are listed in Table S2 . The cassette was assembled to the backbone of CRISPR/Cas9 construct using Gibson Assembly^® Cloning Kit (New England Biolabs Inc). Since 38 KTI genes in soybean share conserved sequences, an alignment was conducted for these genes to design gRNAs that can exclusively target on KTI1 and KTI3. The gRNAs were synthesized in one cassette at GenScript Biotech Corp. The backbone of the new CRISPR/Cas9 construct and the fragment of gRNAs were assembled together using Gibson Assembly^® Cloning Kit.

RNA sequencing data, in FPKM (fragments per kilobase of transcript per million fragments mapped), of 38 KTI genes in 30 different tissue types from Williams 82 were acquired through the Gene Networks in Seed Development database (http://seedgenenetwork.net/sequence). Construction of the heatmap to visualize expression data was done using the heatmap.2 function from the ggplot2 package in R. A green/blue color gradient was chosen to show expression with blue representing little to no expression and green representing high expression. The code for the heatmap is as follows:

heatmap.2(x=KTI Expression, main = “KTI Expression In Different Soybean Tissue”, notecol=“black”, density.info=“none”, trace=“none”, margins =c(12,9), col=my_palette, breaks=col_breaks, dendrogram=“row”,Colv=“NA”, ylab= “Genes”, xlab= “Tissue Type”, cexCol=.9,cexRow = .8)

All RNA was extracted from V98-9005 and V03-5903 seeds using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. Any DNA residue was eliminated by treating with UltraPure DNase I (Thermo Fisher Scientific). The integrity and quantity of total RNA were determined by electrophoresis in 1% agarose gel and a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). cDNA synthesis was performed using the SuperScript III First-Strand RT-PCR Kit (Thermo Fisher Scientific) with an oligo-dT primer based on the manufacturer’s instructions. Real-time PCR was conducted with cDNA as the template using the Quantitect SYBR Green PCR kit (Qiagen) according to the manufacturer’s protocol. Oligo primers are listed in Table S2 . The soybean ELF1B gene (Gm02g276600) was used as reference gene, and data is presented as ΔCT (Jian et al., 2008).

The KTI1, KTI1 _Δ66bp, and KTI3 genes in pDonr207 were subcloned into a Gateway compatible pET28a destination vector via a LR^® Gateway cloning kit (Thermo Fisher Scientific) (Earley et al., 2006; Liu et al., 2016). The plasmids were transformed into E. coli C41 cells (Lucigen). KTI1, KTI1_Δ22aa, and KTI3 proteins were expressed and purified following a procedure as previously described (Liu et al., 2020). Protein purity was evaluated by SDS-PAGE. The protein concentration was determined by a protein assay kit (Bio-Rad) using bovine serum albumin as standard (Han et al., 2015).

A TI activity bioassay was performed following American Association of Cereal Chemists Official Method 22-40 (AACC, 1999) with some modifications previously reported (Rosso et al., 2018). Briefly, 30 mg of finely ground soybean seed powder was mixed with 3 mL of 9 mM HCl (pH 2.0). The mixture was shaken for 1 h at room temperature. 2 mL of the extracts was centrifuged at 10,350 rpm for 20 min at room temperature, and the supernatant was diluted by 10 times with 9 mM HCl for measuring TI activity. A TI activity assay was performed in a 96-well plate format following the same steps described previously (Rosso et al., 2018). Each sample row was repeated three times. Portions of diluted HCl extracts (0, 20, 30, 40, and 60 µL) or 50 µg recombinant proteins of KTI1, KTI1_Δ22aa, and KTI3 were pipetted into the microplate wells, and the volume was adjusted to 60 µL with 9 mM HCl. 60 µL of extractant was used as a sample blank and 60 µL of water were used as a substrate blank. 60 µL of trypsin (from bovine pancreas, Sigma-Aldrich T8003) solution was added to each sample well, and the microplates were placed in an oven at 37°C for 15 min. After the incubation, 150 µL of BAPNA substrate pre-warmed at 37°C was added to all wells, and the plates were incubated for exactly 10 min at 37°C. The reaction was stopped by adding 30 µL of acetic acid solution to all wells. The absorbance of each well was read on a plate reader (FLOUstar Omega, BMG Labtech) at 410 nm for 30 s after shaking at 700 rpm.

The HPLC method to quantify KTI was performed following the method developed previously (Rosso et al., 2018). Briefly, 10 mg of finely ground soybean seed powder was mixed with 1.5 mL of 0.1 M sodium acetate buffer (pH 4.5). Samples were vortexed and shaken for 1 h at room temperature. The sample was centrifuged at 12,000 rpm for 15 min. 1mL of the supernatant was filtered through a syringe with an IC Millex-LG 13-mm mounted 0.2-mm low protein binding hydrophilic millipore (polytetrafluoroethylene [PTFE]) membrane filter (Millipore Ireland). The KTI in solution was separated on an Agilent 1260 Infinity series (Agilent Technologies) equipped with a guard column (4.6 x 5 mm) packed with POROS R2 10-mm Self Pack Media and a Poros R2/H perfusion analytical column (2.1 x 100 mm, 10 um). The mobile Phase A consisted of 0.01% (v/v) trifluoroacetic acid in Milli-Q water, and the mobile Phase B was 0.085% (v/v) trifluoroacetic acid in acetonitrile. The injection volume was 10 uL and the detection wavelength was 220 nm.

The transgene free and double homozygous mutant line, #5-26, was selected for the development of molecular selection markers. Based on the genotyping data of #5-26, two pairs of markers were designed: ZW1 with ZW2 or ZW3. ZW1 is the common reverse primer for both KTI1 and kti1, while ZW2 and ZW3 are two reverse primers matched with unique sequences in KTI1 and kti1, respectively. Similarly, two pairs of molecular markers, ZW4 with ZW5 or ZW6, were designed. ZW4 is the common reverse primer for both KTI3 and kti3, while ZW5 and ZW6 are two reverse primers matched with unique sequences in KTI3 and kti3, respectively. The gDNA of WM82, #5-26 (homozygous mutants of both kti1 and kti3), #5-9 (homozygous mutant of kti1 while heterozygous mutant of kti3) and #2-30 (homozygous mutant of kti3 while heterozygous mutant of kti1) were used as templates to test the efficiency and reliability of these markers in PCR.

PCR amplifications were performed in a total volume of 20 μl containing 50 ng of gDNA, 0.5 μM each of forward and reverse primers ( Table S2 ), 10 μl 2X BioMix Red (Bioline) and ddH₂O. The PCR program was set to be 95°C for 5 min for pre-denature, followed by 35 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s, extension at 72°C for 30 s, followed by final extension at 72°C for 5 min.

In order to screen the large-scale progenies derived from crosses between soybean elite cultivars carrying wild type KTI1/3 and the newly developed mutant plant carrying kti1/3, a simple gel electrophoresis-free method was designed. 1X sybrgreen dye (Thermo Fisher) was added to complete PCR reactions, and the solution was incubated for 10 mins at 75°C before placing in the gel doc (Biorad) for imaging the fluorescent signals. Only the positive PCR products with the dye will display fluorescent signals while the failed PCR will not show signals.

Analytical experiments were performed with at least three technical replicates. Statistical significance was based on one-way ANOVA test for multiple comparisons. Data was analyzed using JMP Pro14. Values of P<0.05 were considered significant.