Multiple amino acid sequence alignments were generated using Clustal Omega (https://www.ebi.ac.uk/jdispatcher/msa/clustalo). Protein domains were identified and analyzed with PROSITE (https://prosite.expasy.org/) and PredictProtein (https://predictprotein.org/).

The soybean cultivar Williams 82 (cv. W82) served as the wild-type control in all experiments. Seeds were germinated in a growth chamber under long-day photoperiod conditions (16 h light/8 h dark) at 25°C, then transferred to a greenhouse and maintained under natural environmental conditions.

Seed samples were harvested from pods at the reproductive stage R6. Total RNA was isolated using the TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions. Contaminating genomic DNA was eliminated using DNase I (Thermo Fisher Scientific), and first-strand cDNA was synthesized from 1 μg of RNA using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). Quantitative real-time RT-PCR (qRT-PCR) was conducted with gene-specific primers ( Supplementary Table 2 ) using the QuantiSpeed SYBR No-Rox Kit (PhileKorea, Seoul, Korea). GmPBB2 (Glyma.14G014800) was used as the internal reference gene. Relative expression levels were automatically calculated from triplicate reactions using the CFX Real-Time PCR Detection System and CFX Manager Software v2.0 (Bio-Rad, Hercules, CA, USA). All experiments were performed in a minimum of three independent biological replicates with three technical replicates per biological replicate. Statistical significance was determined using the Student’s t-test.

Total protein was extracted from soybean cotyledons using an extraction buffer containing 100 mM Tris-HCl (pH 7.5), 1 mM EDTA, 150 mM NaCl, 3 mM DTT, and 1 mM PMSF. One microgram of total protein per sample was resolved on a 15% SDS-PAGE gel and transferred onto Immobilon-P PVDF membranes (Merck Millipore, Co. Wicklow, Ireland). The membranes were incubated with a polyclonal anti-GmP34 antibody and visualized using an HRP-conjugated anti-rabbit IgG secondary antibody (Proteintech, Rosemont, IL, USA) in combination with ECL detection reagent (TransLab, Daejeon, Korea). As a loading control, five micrograms of protein were stained with Coomassie Brilliant Blue.

Genomic sequences of GmP34 (Glyma.08g116300), GmP34h1 (Glyma.08g116400), and GmP34h2 (Glyma.05g158600) were retrieved from the Phytozome v13 database (https://phytozome-next.jgi.doe.gov/). Candidate guide RNAs (gRNAs) were designed using CRISPR-P v2 (http://crispr.hzau.edu.cn/CRISPR2/) and CRISPR RGEN tools (http://www.rgenome.net/). Specifically, guide RNAs (gRNAs) with a GC content between 30% and 70% were initially selected using CRISPR-P v2. These candidates were further refined by identifying gRNAs with zero to three potential off-target mismatches across the genome, as predicted by Cas-OFFinder from RGEN Tools. Finally, we selected gRNAs that did not target the exon regions of any genes, except for our intended target genes. Three gRNAs—gRNA1, gRNA2, and gRNA3 were selected to simultaneously target either GmP34 and GmP34h1 (gRNA1/gRNA2) or all three genes (GmP34, GmP34h1, and GmP34h2) using the gRNA1/gRNA3 combination, with minimal predicted off-target activity ( Supplementary Table 1 ). The constructs were assembled into the pECO201 binary vector using the Golden Gate cloning strategy (Oh et al., 2020). This vector features the Arabidopsis ubiquitin6 (AtU6) promoter for multi-tRNA-gRNA expression, the NOS promoter for Bar gene selection, and the CaMV 35S promoter for expressing Arabidopsis codon-optimized Cas9 (acoCas9). The gRNA sequences are listed in Supplementary Table 1 .

Soybean W82 seeds were transformed using an Agrobacterium-mediated half-seed method, with minor modifications based on a previously published protocol (Kim et al., 2017). Seeds were surface-sterilized inside a desiccator for 5 min using chlorine gas generated from 1% sodium hypochlorite. Sterilized seeds were then germinated on germination medium in the dark for 20 h. Under sterile conditions, germinated seeds were halved longitudinally, and the seed coats were carefully removed. The half-seeds were then immersed in a suspension of Agrobacterium tumefaciens strain EHA105 harboring the CRISPR/Cas9 construct for 30 min at room temperature, followed by co-cultivation at 23°C for 5 h under a 16 h light/8 h dark photoperiod. After co-cultivation, explants were sequentially transferred to shoot induction medium and subsequently to root induction medium. Regeneration was carried out at 23°C under 16 h light/8 h dark conditions until the shoots exceeded 4 cm in height and roots reached lengths >5 mm.

Genomic DNA was isolated from leaves of transformed soybean plants using the Exgene™ Plant SV Kit (GeneAll Biotechnology, Seoul, Korea). Transgenic lines were initially screened by PCR amplification of the Bar and Cas9 genes using gene-specific primers ( Supplementary Table 2 ). The PCR conditions were as follows: an initial denaturation at 95 °C for 10 min; 30 cycles of 95°C for 30 s, 58°C for 10 s, and 72°C for 30 s; followed by a final extension at 72°C for 10 min. Genomic regions targeted for editing were PCR-amplified for subsequent InDel detection and deep sequencing analysis. The PCR conditions for InDel detection were: 94°C for 5 min; 35 cycles of 94°C for 45 s, 60°C for 30 s, and 72°C for 1 min; followed by a final extension at 72°C for 10 min. Amplicons were further processed by incorporating adapter sequences via an additional round of PCR (Oh et al., 2020). The adapter PCR conditions were: 94°C for 3 min; 5 cycles of 94°C for 30 s, 50°C for 30 s, and 72°C for 1 min; followed by 25 cycles of 94°C for 30 s, 65°C for 30 s, and 72 °C for 1 min; with a final extension at 72°C for 10 min. Deep sequencing was carried out using the Illumina MiSeq platform (v2, 300-cycle; San Diego, CA, USA), and the resulting data were analyzed using Cas-Analyzer, part of the CRISPR RGEN Tools suite (http://www.rgenome.net/).

The soybean P34 gene (GmP34, Glyma.08G116300) encodes a papain-like cysteine protease and is recognized as one of the major seed protein allergens (Kalinski et al., 1990, 1992). Analysis of the soybean genome via the Phytozome database (https://phytozome-next.jgi.doe.gov/ identified approximately 100 proteins exhibiting sequence similarity to GmP34. Of these, the ten most closely related proteins were subjected to phylogenetic analysis, which identified two highly similar homologs: GmP34h1 (Glyma.08G116400) and GmP34h2 (Glyma.05G158600), sharing 79.7% and 72.7% amino acid similarity with GmP34, respectively ( Supplementary Figure 1 ). GmP34 is initially produced as a pre-pro-protein and undergoes processing that includes the removal of a 122-amino acid N-terminal signal peptide, ultimately yielding the mature 34 kDa protein localized in the protein storage vacuoles of soybean seeds (Kalinski et al., 1992). Multiple amino acid sequence alignment and domain prediction using PROSITE (https://prosite.expasy.org/) and PredictProtein (https://predictprotein.org/) revealed that GmP34, GmP34h1, and GmP34h2 possess conserved structural motifs. These include endoplasmic reticulum signal peptides, protein kinase C phosphorylation sites, myristoylation sites, a PFTA domain, N-glycosylation sites, and a thiol protease Asn active site ( Figure 1 ). Notably, allergen-associated motifs such as allergen representative peptides (ARPs) and IgE-binding epitopes were also conserved across all three homologs, suggesting a potential shared allergenic function ( Figure 1 ).

Previous studies have shown that GmP34 is specifically expressed during soybean seed maturation (Kalinski et al., 1992; Koo et al., 2011, 2013). To investigate the expression profiles of the three GmP34 homologs, we conducted qRT-PCR using gene-specific primers ( Supplementary Table 1 ) on total RNA extracted from developing soybean seeds of various sizes (2–4 mm, 5–6 mm, 7–8 mm, 9–10 mm, and 11–12 mm). All three homologs exhibited comparable expression patterns, with transcripts first detectable in 7–8 mm seeds and progressively increasing with seed maturation. Notably, GmP34h1 transcript levels declined more rapidly in 11–12 mm seeds compared to GmP34 and GmP34h2 ( Figure 2A ). These transcriptional patterns were further supported by western blot analysis using an anti-GmP34 polyclonal antibody ( Figure 2B ), suggesting functional similarity among the three proteins.

Although various strategies have been developed to reduce GmP34 protein accumulation in soybean seeds—including transgene-induced gene silencing, extensive screening of natural accessions, and CRISPR/Cas9-mediated genome editing (GE) (Herman et al., 2003; Bilyeu et al., 2009; Sugano et al., 2020)—these methods have exclusively targeted the GmP34 gene. To concurrently eliminate GmP34 along with its closely related homologs, GmP34h1 and GmP34h2, we employed the CRISPR/Cas9 system and designed three guide RNAs (gRNAs) using CRISPR-P v2.0 and RGEN tools, referred to as gRNA1, gRNA2, and gRNA3. Among them, gRNA1 and gRNA3 were designed to target all three GmP34 homologs, whereas gRNA2 specifically targeted GmP34 and GmP34h1 ( Figures 3A, B ). Using these gRNAs, we constructed two multiplex genome-editing vectors in the pECO201 backbone: one expressing gRNA1 and gRNA2 to target GmP34 and GmP34h1, and another expressing gRNA1 and gRNA3 to simultaneously target all three homologs. These constructs were designated GmP34 GE Common Target 1 (GmP34 GE_CT1) and Common Target 2 (GmP34 GE_CT2), respectively ( Figure 3C ).

The embryonic axis of soybean W82 was inoculated with Agrobacterium strains carrying either the GmP34 GE_CT1 or GmP34 GE_CT2 construct, and transgenic T₀ plants were selected using phosphinothricin (PPT) ( Figure 3D ). Fifteen T₀ plants were obtained for each construct, and stable integration of the T-DNA into the genome was confirmed via Bar gene PCR ( Figures 3E, F ). To characterize mutation patterns in each T₀ line, the targeted genomic regions of the GmP34 homologs were analyzed using InDel PCR and targeted deep sequencing. InDel PCR analysis revealed that T₀ lines harboring the GmP34 GE_CT1 construct exhibited a range of deletions in the GmP34 gene. Notably, lines #8 and #10 showed simultaneous deletions in both GmP34 and GmP34h1 ( Figure 3E ). Deep sequencing confirmed these findings, revealing multiple mutation types—including deletions and substitutions—with one or two predominant mutation patterns in both GmP34 and GmP34h1 in lines #8 and #10 ( Supplementary Figure 2A ). In the case of GmP34 GE_CT2 T₀ lines, InDel PCR identified clear deletions in GmP34 and GmP34h1 in lines #1 and #14, respectively, while no distinct deletion patterns were observed in GmP34h2 ( Figure 3F ). Sequencing analysis of CT2 line #1 showed a predominant 61-nucleotide deletion in GmP34, while GmP34h1 and GmP34h2 contained only small deletions (1–4 nucleotides) that were indistinguishable from the W82 control by InDel PCR. Similarly, CT2 line #14 exhibited a 76-nucleotide deletion in GmP34h1, along with minor deletions in GmP34 and GmP34h2 as the primary mutation patterns ( Supplementary Figure 2B ). Based on these InDel PCR and sequencing results, we selected CT1 lines #8 and #10, which carried mutations in GmP34 and GmP34h1, and CT2 lines #1 and #14, which harbored mutations in all three GmP34 homologs. These T₀ lines were subsequently advanced to the next generation.

Using the GmP34 GE_CT1 T1 plants (#8 and #10), we performed Bar PCR, InDel PCR, and targeted deep sequencing. Based on this screening, line #10—characterized by a single T-DNA insertion and higher editing efficiency (data not shown)—was selected and advanced to the T₂ generation. We then screened T₂ progeny (#10–1 to #10-28) for GmP34 protein expression via western blotting. Three days after sowing, when GmP34 protein is still detectable, a small portion of the cotyledons was sampled for protein analysis ( Figure 4A ). Western blot analysis revealed that none of the #10–6 progeny expressed detectable levels of GmP34 protein in their cotyledons ( Figure 4B ). In contrast, the #10–22 progeny showed variable GmP34 protein expression ( Figure 4C ), whereas all #10–14 progeny exhibited clear GmP34 protein signals ( Figure 4D ). Subsequent Bar PCR analysis of these lines indicated that #10–6 was T-DNA homozygous, #10–22 was heterozygous, and #10–14 lacked the T-DNA insertion (null) (data not shown).

To generate T-DNA-free mutants for the GmP34 homologs, we employed two approaches: (1) advancing T-DNA heterozygous lines to identify segregating T-DNA-free progeny and (2) backcrossing T-DNA homozygous lines with W82 plants and screening the F₂ (BC₁F₂) generation. For the first strategy, we selected the T-DNA heterozygous GmP34 GE_CT1 T₂ line #10-22–2 based on Bar PCR and TaqMan PCR analyses and obtained T₃ seeds (data not shown). From 3-day-old germinating T₃ seedlings, a small portion of the cotyledons was sampled to assess GmP34 protein levels by western blotting. To identify T-DNA-free lines, we applied PPT solution to the leaves of further-grown T₃ seedlings and selected individuals exhibiting a PPT-sensitive necrosis phenotype ( Figure 4E ). Among these, we screened for GmP34-null individuals using western blotting and ultimately identified two T-DNA-free GmP34-null lines: #10-22-2–77 and #10-22-2-124 ( Figure 4F ). Targeted deep sequencing revealed that both lines carried an identical 3-nucleotide deletion in GmP34, resulting in a single amino acid deletion (Val) and a substitution (Lys to Glu) at positions 150 and 151, respectively. No mutations were detected in the GmP34h1 gene ( Figure 4G ). These results demonstrate the successful isolation of T-DNA-free GmP34 single mutant lines.

As the second strategy, we analyzed T₃ lines derived from the T-DNA homozygous GmP34 GE_CT1 #10-6–1 line. All T₃ plants were confirmed to carry T-DNA based on Bar and Cas9 PCR. InDel PCR analysis revealed that six lines (#12, 16, 18, 20, 29, and 30) exhibited deletions in both GmP34 and GmP34h1 genes ( Figure 5A ). Among these, line #10-6-1–18 was selected for backcrossing with W82, and a BC₁F₂ population was generated. T-DNA-free individuals were initially identified through PPT leaf painting and subsequently confirmed by Bar and Cas9 PCR ( Figure 5B ). InDel PCR further revealed that three BC₁F₂ lines (#9-21, #9-25, and #9-43) harbored homozygous deletion patterns in both GmP34 and GmP34h1 identical to those of the parental #10-6-1–18 line ( Figure 5B ), confirming that they were T-DNA-free GmP34/GmP34h1 double mutants. Western blot analysis supported these findings, showing the complete absence of GmP34 protein in cotyledons of all three lines ( Figure 5C ). Targeted deep sequencing revealed that each line carried a 415-nucleotide deletion in the first exon of GmP34, resulting in a frameshift and a premature stop codon at the 12th amino acid position ( Figures 5D, E ). Editing in the GmP34h1 gene was more complex: lines #9–21 and #9–43 harbored a 591-nucleotide deletion spanning the 3’ region of the gRNA2 target site, encompassing the first intron, second exon, and second intron, along with six nucleotide substitutions ( Figures 5D, E ). Line #9–25 exhibited the same pattern with an additional 3-nucleotide deletion in the first intron ( Figure 5D ). Despite these differences, both editing types led to frameshift mutations and a premature stop codon at amino acid position 144 of the GmP34h1 protein ( Figure 5E ).

Additional T-DNA-free GmP34 and GmP34h1 double mutants were identified through the analysis of GmP34 GE_CT2 T₁ lines. Western blot analysis of T₁ seedlings from CT2 #1 and #14 lines revealed 19 and 28 GmP34 protein-null individuals, respectively ( Supplementary Figure 3 ). Among these, one T-DNA-free line, designated #1-65, was identified in the CT2 #1 progeny using Bar and Cas9 PCR ( Figure 6A ). InDel PCR and targeted deep sequencing showed that this line harbored a 61-nucleotide deletion along with a single-nucleotide substitution in the GmP34 gene and a one-nucleotide deletion in the GmP34h1 gene ( Figures 6A, C ). These mutations introduced premature stop codons at the 52nd and 12th amino acid positions of the GmP34 and GmP34h1 proteins, respectively ( Figure 6E ). No sequence alterations were observed in the GmP34h2 gene ( Figures 6C, E ).

In contrast, Bar and Cas9 PCR analyses of GmP34 protein-null CT2 #14 T₁ seedlings indicated the absence of T-DNA-free individuals in this population ( Figure 6B ). Subsequent InDel PCR and targeted deep sequencing analyses revealed that the CT2 T₁ #14–1 line harbored 4- and 75-nucleotide deletions in the GmP34 and GmP34h1 genes, respectively. These mutations introduced premature stop codons at the 30th and 12th amino acid positions of the GmP34 and GmP34h1 proteins, respectively ( Figures 6D, F ). To obtain T-DNA-free mutants, this line was backcrossed with W82 plants. Among 192 BC₁F₂ progeny derived from the CT2 T₁ #14–1 line, seven T-DNA-free individuals were identified by PPT leaf painting and further confirmed by Bar and Cas9 PCR ( Supplementary Figure 4A ). Western blot analysis of these lines revealed two GmP34 protein-null lines, #7–109 and #7-138 ( Supplementary Figure 4B ). InDel PCR analysis confirmed that both lines carried homozygous deletions in the GmP34 and GmP34h1 genes ( Supplementary Figure 4C ), and targeted deep sequencing validated the presence of the same 4- and 75-nucleotide deletions identified in the original #14–1 line ( Supplementary Figure 4D ), resulting in premature stop codons at the same amino acid positions ( Supplementary Figure 4E ). No mutations were detected in the GmP34h2 gene. Collectively, these findings led to the identification of two additional T-DNA-free GmP34/GmP34h1 double mutants: CT2 T₁ #14–1 BC₁F₂ #7–109 and #7-138. In summary, we successfully generated six independent T-DNA-free GmP34/GmP34h1 double-mutant lines: three from the BC₁F₂ population of the GmP34 GE_CT1 #10-6-1–18 line (#9-21, #9-43, and #9-25), one from the GmP34 GE_CT2 T₁ line (#1-65), and two from the BC₁F₂ population of the GmP34 GE_CT2 #14–1 line (#7–109 and #7-138).

Targeted deep sequencing revealed that two CT2 T₁ lines, #14–114 and #14–117 carried mutations in all three GmP34 homologs—GmP34, GmP34h1, and GmP34h2. Notably, the two lines exhibited identical mutation patterns, consisting of 4-, 75-, and 8-nucleotide deletions in the GmP34, GmP34h1, and GmP34h2 genes, respectively ( Figure 6D ). These deletions caused frameshift mutations, which introduced premature stop codons at the 30th, 12th, and 15th amino acid positions of the GmP34, GmP34h1, and GmP34h2 proteins, respectively ( Figure 6F ). To generate T-DNA-free triple mutants, both lines are currently being backcrossed with W82 plants.

Taking all results together, we summarized the GE status of the three GmP34 homologous genes in Table 1 . Through CRISPR/Cas9-mediated GE, we successfully generated GmP34 single mutants, GmP34/GmP34h1 double mutants, and GmP34/GmP34h1/GmP34h2 triple mutants. From the screening of the GmP34 GE_CT1 T₃ population, we identified two T-DNA-free GmP34 single-mutant lines, #10-22-2–77 and #10-22-2-124. A total of six T-DNA-free GmP34/GmP34h1 double-mutant lines were obtained—one directly from the GmP34 GE_CT2 T₁ population (#1-65), and five from BC₁F₂ populations derived from CT1 #10-6-1-18 (#9-21, #9-25, and #9-43) and CT2 #14-1 (#7–109 and #7-138) lines. In addition, we identified two GmP34/GmP34h1/GmP34h2 triple-mutant lines (#14–114 and #14-117) from the GmP34 GE_CT2 T₁ population, which are currently being backcrossed with W82 to eliminate the T-DNA.