Plant material

Eight soybean (G. max L. Merr.) genotypes were used, among which five (SL 958, SL 1074, DS 3105, DS 3106, and SL 955) were resistant, and three (JS 335, AMS 19-01, and JS 25-08) were susceptible to yellow mosaic disease. The resistant and susceptible nature of the soybean cultivars used was established through prior field screening as well as whitefly-mediated and agroinoculation-based artificial challenge (Chavan et al. 2025). Seeds were procured from the Division of Genetics at the Indian Agricultural Research Institute, New Delhi. Plants were grown in an environmentally controlled growth chamber maintained at 24 ± 2°C, with average humidity of 65%, and a light/dark cycle of 16 h/8 h.

Amplification and cloning of complete CDS of GmRDR1 and GmSGS3

Following the standard protocol, total RNA was isolated using Trizol reagent (Invitrogen, Carlsbad, CA, USA) from both resistant (SL 1074) and susceptible (JS 335) soybean cultivars. RNA samples were treated with DNase I (Ambion) to eliminate residual genomic DNA. An aliquot of 1 μg DNA-free RNA was used for cDNA synthesis. The 20 μL reaction mix for cDNA synthesis included 1× buffer, 1 mM dNTPs, 25 ng μL⁻¹ Oligo dT primer, 1 U μL⁻¹ RiboLock RNase Inhibitor and 10 U μL⁻¹ RevertAid reverse transcriptase (Thermo Fisher Scientific, Waltham, MA, USA). The complete CDSs of GmRDR1 and GmSGS3 were amplified from the cDNA using GmRDR1 F/GmRDR1 R and GmSGS3 F/GmSGS3 R, respectively (Table S1). PCR amplification used ~200 ng cDNA with 0.02 U μL⁻¹ Phusion high-fidelity polymerase (Thermo Fisher Scientific) with 1x HF buffer, 0.2 mM dNTPs, and 0.5 μM primers (forward and reverse). The PCR amplicons were purified using Promega Wizard® Genomic DNA Purification Kit (Promega, Madison, WI, USA) and ligated into the pUC18 vector following restriction enzyme-based cloning, where SalI and SacI were used for GmRDR1 and BamHI and EcoRI were used for GmSGS3. The ligated products were then used to transform E. coli DH5α cells. Recombinant colonies were screened through colony PCR and restriction digestion of recombinant plasmids. Confirmed clones for GmRDR1 and GmSGS3 were sequenced through outsourcing (Central Instrumentation Facility (CIF), University of Delhi South Campus, New Delhi, India).

Analysis of GmRDR1 and GmSGS3 sequences

The GmRDR1 and GmSGS3 CDS sequences were subjected to NCBI BLASTn to identify their homologs. SNPs were identified through pairwise sequence alignment. The amino acid sequences of GmRDR1 and GmSGS3 proteins were deduced using the ExPasy Server (https://www.expasy.org/). BioEdit (Hall 1999), and MEGA X (Kumar et al. 2018) software were used for multiple sequence alignment using the amino acid sequences of RDR1 and SGS3 proteins obtained from resistant and susceptible soybean genotypes and related species, selected based on relevance to the study. Phylogenetic analysis was conducted using the neighbor-joining method with 1000 bootstrap replications.

Development and validation of SNP-based CAPS marker

Based on sequence analysis, SNPs were identified in GmSGS3 CDS between resistant and susceptible genotypes. Consistency of SNPs was confirmed through sequencing a fragment of GmSGS3 from four more resistant and two more susceptible genotypes. Using one SNP in GmSGS3 CDS, a CAPS marker was developed, where XhoI was the enzyme that distinguished resistant and susceptible genotypes. To validate the CAPS marker, the complete CDS was initially amplified and digested with XhoI from two resistant (SL 1074, SL 958) and two susceptible (JS 335, AMS 19-01) soybean genotypes. For validation of SNPs from genomic DNA, this was extracted from leaf tissues of resistant (SL 1074, SL 958, DS 3106, DS 3105, and SL 955) and susceptible (JS 335, AMS 19-01, and JS 25-08) soybean genotypes using the CTAB method (Doyle & Doyle 1987). DNA quality and concentration were confirmed using a NanoDrop spectrophotometer (Genetix, India). A fragment of the GmSGS3 gene containing two SNPs was amplified using the primer pair SmSGS3F and SmSGS3R (Table 1). The amplified product was directly digested with the XhoI enzyme (Thermo Scientific) to generate genotype-specific SNP fragment pattern.

Analysis of amino acid sequences of GmRDR1 and GmSGS3 proteins and molecular docking

Physicochemical properties, including molecular weight, isoelectric point (pI), and amino acid composition, of GmRDR1 and GmSGS3 proteins were analyzed using the ProtParam tool (https://web.expasy.org/protparam/) (Swiss Institute of Bioinformatics, ExPASy). Conserved domains and functional motifs within these proteins were identified using InterPro-Pfam (https://www.ebi.ac.uk/interpro/) (Mitchell et al. 2019) and MOTIF Search (https://www.genome.jp/tools/motif/) (Pagni et al. 2007) servers, providing insights into functional regions. Homology models for both GmRDR1 and GmSGS3 proteins were generated using the SWISS-MODEL server (https://swissmodel.expasy.org/) (Guex et al. 2009). These models were validated through the SAVES server (https://saves.mbi.ucla.edu/), with PROCHECK used for Ramachandran plot analysis (Laskowski et al. 1993), ERRAT for assessing overall model quality (Colovos & Yeates 1993), and ProSA-web to predict Z-scores, ensuring structural reliability (Wiederstein & Sippl 2007). To study the protein–protein interactions between GmRDR1 and GmSGS3, molecular docking was performed using the Gramm docking server (https://gramm.compbio.ku.edu/) (Singh et al. 2024), and further validated using the ClusPro server (https://cluspro.org/) (Kozakov et al. 2017). Active site predictions were obtained using the PrankWeb server (https://prankweb.cz/) (Jendele et al. 2019) to highlight potential interaction regions. Both homology models were submitted for docking analysis, and the resultant complex evaluated based on clustering and energy scores. The optimal docking model was selected for detailed structural and interaction analysis, displaying most favorable interaction and stability. The PDBsum server (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/) (Laskowski et al. 2005) was used to predict interacting amino acids of GmRDR1 and GmSGS3 proteins using the docked file generated by the Gramm server.

Yeast two-hybrid (Y2H) assay

The yeast two-hybrid (Y2H) assay was employed to investigate the interaction between GmSGS3 and GmRDR1 proteins of soybean, following the protocol described by Peng & Neff (2020). The coding sequences of GmRDR1 and GmSGS3 were amplified for cloning into yeast vectors using pGAD-dRDR-F/ pGAD-dRDR-R and pGBK-dSGS3-F/ pGBK-dSGS3-R, respectively (Table S1). The amplified GmRDR1 and GmSGS3 CDS were subsequently cloned into the pGADT7-AD (activation domain) and pGBKT7-BD (binding domain) vectors, respectively, using the NEBuilder^® HiFi DNA Assembly Cloning Kit, which ensures high efficiency and specificity of gene insertion. The pGADT7-GmRDR1 (prey) and pGBKT7-GmSGS3 (bait) constructs were used to transform the yeast strains Y187 and Y2HGold (Takara Bio Inc.), respectively, using the lithium acetate transformation method according to the manufacturer's instructions. Successful transformants were selected on synthetic dropout (SD) media lacking leucine (SD/-Leu) for pGADT7-AD, or lacking tryptophan (SD/-Trp) for pGBKT7-BD constructs. An empty pGBKT7-BD or pGADT7-AD vector was used as a negative control to confirm specificity of the interactions. For interaction analysis, the Y2HGold cells harboring pGBKT7-GmSGS3 were mated with Y187 cells harboring pGADT7-GmRDR1 on SD medium lacking leucine and tryptophan (SD/-Leu/-Trp). The resulting diploids were subsequently selected on triple dropout (SD/-Leu/-Trp/-His) medium then plated on quadruple dropout medium (SD/-Leu/-Trp/-His/-Ade), which lacks leucine, tryptophan, histidine, and adenine, to confirm positive protein–protein interactions. Colony growth was monitored for 5 days at 30°C to assess interaction viability. To further confirm stringency of the interaction, colonies from four dropout media were plated on a quadruple dropout medium supplemented with X-α-gal and Aureobasidin A, following the manufacturer's protocol (Clontech Laboratories). Yeast cells grown on selective SD media were then analyzed for blue color development, indicating a positive interaction. Standard autoactivation analysis were performed using empty vector as one partner.

Gene expression analysis using real-time PCR

To quantify expression of GmRDR1 and GmSGS3, cDNA was synthesized from total RNA extracted from resistant (SL 1074 and SL 958) and susceptible (JS 335 and AMS 19-01) soybean genotypes at 15, 30, and 45 days post-inoculation (dpi), following the standard protocol described above. Quantitative reverse transcriptase PCR (qRT-PCR) was performed using the Bio-Rad CFX96 Touch Real-Time PCR Detection System, with GmRDR1 primers (GmRDR1 RT F/GmRDR1 RT R) and GmSGS3 primers (GmSGS3 RT F/GmSGS3 RT R), while for the housekeeping gene GmACT, primers used were GmACT F/GmACT R (Table S1). Each analysis was performed using three independent biological replicates, each with three technical replicates. Relative gene expression levels were calculated using the 2^−ΔΔCT method (Livak & Schmittgen 2001), with the soybean actin gene as internal control. Statistical significance between treatment and control conditions was evaluated using a two-tailed Student's t-test, assuming equal variances with significance at P < 0.05, 0.01, and 0.001. The analysis was performed and graphs generated in Microsoft Excel. Data were visually inspected for consistency and approximate normality through residuals and distribution plots. Error bars represent standard errors of the mean (SEM) unless otherwise noted.

Development of hairpin RNAi constructs

siDirect v. 2.1 (http://sidirect2.rnai.jp/) was used to identify siRNA hotspot regions within GmRDR1 and GmSGS3 CDS, leading to design and validation of hairpin RNAi constructs, hpRDR1 and hpSGS3, targeting these genes. The hairpin RNAi constructs were developed by amplifying target regions through PCR using GmRDR1 hairpin primers (RDR1-GG F/RDR1-R) and GmSGS3 hairpin primers (SGS3-GG F/SGS3-GG R) (Table S1). The amplified products were then cloned into the pRNAi-GG vector (obtained from ABRC, Ohio State University) using the Golden Gate assembly method (Yan et al. 2012). Recombinant clones were confirmed through amplification with P21, P22, and gene-specific reverse primers (Table S1). The confirmed hairpin RNAi constructs were introduced into Agrobacterium tumefaciens strain EHA 105 using the freeze–thaw method, and transformed cultures used for plant inoculation. Agrobacterium cultures of the hairpin RNAi constructs were prepared in MES buffer (10 mM MES, pH 5.7, 10 mM MgCl₂, 150 μM acetosyringone) and adjusted to OD₆₀₀ 0.5.

Agroinoculation of hairpin constructs in resistant genotypes followed by challenge inoculation with agroinfectious clones of MYMIV

To assess effects of GmRDR1 and GmSGS3 knockdown on MYMIV resistance, seedlings of the resistant genotypes SL 1074 and SL 958 at the two-leaf stage were infiltrated with agrocultures containing hairpin constructs of these genes. For each genotype, 10 plants were inoculated with each hairpin construct. Five days following infiltration of the hairpin constructs, agroinfectious clones of DNA-A (accession no. OR050594) and DNA-B (accession no. OR050595) of MYMIV, previously developed in our laboratory, were introduced to the hairpin-primed plants. Differential expression of GmRDR1 and GmSGS3 transcripts was analyzed at 15, 30, and 45 dpi using qRT-PCR, as described above. Additionally, MYMIV accumulation in GmRDR1 and GmSGS3 knockdown plants was evaluated after isolating total DNA, followed by qPCR using BV1 F/BV1R primers (Table S1).

GmRDR1 shows no nucleotide sequence variation, while GmSGS3 has two synonymous SNPs that differentiate resistant and susceptible genotypes

Amplicons of ca. 3.4 kb and ca. 1.9 kb from cDNAs of GmRDR1 and GmSGS3, respectively, were successfully obtained from both resistant (SL 1074) and susceptible (JS 335) soybean genotypes (Fig. S1A) using primers GmRDR1 F/GmRDR1 R and GmSGS3 F/GmSGS3 R (Table S1). The recombinant plasmids containing GmRDR1 cDNA were further validated through digestion with SalI and SacI restriction enzymes, which released the expected ca. 3.4 kb insert from the vector backbone (2.7 kb) (Fig. S1B). Similarly, digestion of the recombinant plasmid containing GmSGS3 cDNA with BamHI and EcoRI yielded an anticipated insert of ca. 1.9 kb (Fig. S1C).

Sequencing of GmRDR1 and GmSGS3 coding sequence (CDS) revealed that GmRDR1 CDS is 3378 bp, while GmSGS3 is 1917 bp. The GmRDR1 CDS from SL 1074 and JS 335 were deposited in GenBank (accession nos PP663024, and PP601412, respectively). Similarly, GmSGS3 CDS from SL 1074 and JS 335 were deposited in GenBank (accession nos PP601410 and PP601411, respectively). Sequence analysis revealed no variation in the GmRDR1 CDS between SL 1074 and JS 335. However, in GmSGS3 CDS, two single nucleotide polymorphisms (SNPs) at CDS coordinates 441 and 543 were identified between resistant (SL 1074) and susceptible (JS 335) genotypes. At position 441, cytosine (C) was found in JS 335, whereas thymine (T) was present in SL 1074. Similarly, at position 543, JS 335 had thymine (T), while SL 1074 contained cytosine (C) (Fig. 1A). Notably, these SNPs do not alter amino acid sequence, indicating they are synonymous SNPs. To determine whether these SNPs were the result of sequencing errors, two additional clones were sequenced, and the same SNPs were obtained. Furthermore, to explore the association of these SNPs with the resistant trait, a fragment of GmSGS3 CDS containing both SNPs was amplified using the primers smSGS3 F and smSGS3 R (Table 1), and amplicons sequenced from four more resistant and two more susceptible genotypes. Interestingly, both SNPs (Fig. 1B) consistently and clearly distinguished GmSGS3 CDS between resistant and susceptible genotypes.

A CAPS marker involving SNP 441 of GmSGS3 distinguishes resistant and susceptible soybean genotypes

The SNP position 441 in the GmSGS3 coding sequence creates an XhoI restriction endonuclease site in susceptible soybean genotypes, making it a promising candidate for Cleaved Amplified Polymorphic Sequences (CAPS) markers to differentiate between resistant and susceptible genotypes (Fig. 2A). To verify this, initially, the complete GmSGS3 CDS from two resistant genotypes (SL 1074 and SL 958) and two susceptible genotypes (JS 335 and AMS 19-01) were amplified. The resulting PCR amplicons were then digested with XhoI. As anticipated, the amplicon from the susceptible genotypes yielded two fragments of 1474 bp and 439 bp, while the amplicon from the resistant genotypes remained intact, producing a single fragment of 1917 bp (Fig. 2B).

To further validate efficacy of the CAPS marker in distinguishing resistant and susceptible genotypes at DNA level, a fragment of the GmSGS3 gene was amplified using primers smSGS3 F and smSGS3 R (Table 1) from five resistant genotypes (SL 1074, SL 958, DS 3106, DS 3105, and SL 955) and three susceptible genotypes (JS 335, AMS 19-01, and JS 25-08). All genotypes produced an amplicon of 433 bp, which was then subjected to XhoI digestion. As expected, the three susceptible genotypes yielded two digestion products of 299 bp and 134 bp, while the five resistant genotypes retained an undigested 433 bp band. These results confirm efficacy of the SNP-based CAPS marker in differentiating between resistant and susceptible genotypes (Fig. 2C).

Amino acid sequence of GmRDR1 and GmSGS3 contains important catalytic motifs and domains with a close phylogenetic relationship within the Leguminosae

The GmRDR1 protein consists of 1125 amino acids with a molecular weight of 129,182.74 Da. The ProtParam tool revealed GmRDR1 protein had a theoretical isoelectric point (pI) of 8.14, indicating a slightly basic nature. It contains 139 negatively charged residues (Asp + Glu) and 144 positively charged residues (Arg + Lys), resulting in a net positive charge at physiological pH. Phylogenetic analysis using amino acid sequences of GmRDR1 from resistant genotype SL 1074 and susceptible genotype JS 335 placed them in the RDR1 clade, clearly distinct from other RDR clades, such as RDR2 and RDR6 (Fig. 3A). GmRDR1 shared high amino acid sequence identity with Cajanus cajan (CcRDR1, 85%) and Vigna unguiculata (VunRDR1, 80%), reflecting the close evolutionary relationships among legume crops (Table S2). In contrast, GmRDR1 had lower amino acid sequence identity with other RDR1 proteins reported from Arabidopsis thaliana (AtRDR1, 63%), Nicotiana glutinosa (NgRDR1, 62%), and Solanum lycopersicum (SlRDR1, 62%), showing moderate divergence in non-legume species (Table S2). As expected, GmRDR1 had significantly lower amino acid sequence identity with RDR2 (38.2%, 38.1%, and 36% with G. soja, A. thaliana, and G. max, respectively) and RDR6 (32%, 32%, 31%, and 30.5% with G. max, V. unguiculata, N. glutinosa, and A. thaliana, respectively) proteins (Table S2).

The InterPro-Pfam and Motif Search server predicted domain architecture of GmRDR1, which consisted of an RNA recognition motif 3–60 amino acids (aa) essential for interacting with single-stranded RNA, contributing to RNA processing, RNA-dependent RNA polymerase domain (375–945 aa), which is essential for RNA replication, and RNA binding domain (565–596 aa), likely having a role in regulating RNA transcription and translation (Table 2). Amino acid sequence-based multiple alignment of different RDR sequences revealed the presence of a DLDGD catalytic motif, a hallmark of RDRα clade proteins and essential for RNA-dependent RNA polymerase (RdRP) activity, is conserved across all RDR1, RDR2, and RDR6 aa sequences (Fig. S2). However, there was a notable variation in RDR5 from Solanum lycopersicum, where the motif was DFDGD, reflecting phylogenetic divergence. Besides this conserved stretch of aa, all RDR1 sequences had six other conserved aa sequence stretches (NVSNRV, NRQLITLLSTLGV, FRASKLL, MMGCLDE, NYIVN, IANAH) within the RNA-dependent RNA polymerase domain (Fig. S2). Additionally, two other conserved aa sequence stretches (KLGNLM, YGIKTEAEIL) were present at the C-terminus of all RDR1 proteins (Fig. S2).

The GmSGS3, with 638 amino acids and MW 73,529.77 Da, has a pI of 6.00, indicating slight acidity. It contains 109 negatively charged (Asp + Glu) and 96 positively charged (Arg + Lys) residues, giving it a net negative charge at physiological pH, favoring interactions with positively charged molecules under neutral to mildly acidic conditions. Phylogenetic analysis of GmSGS3 aa sequences revealed GmSGS3 clustered within the legume SGS3 clade, indicating minimal genetic divergence among legumes (Fig. 3B). GmSGS3 shared 100% identity with GsSGS3 (Glycine soja), highlighting the close evolutionary relationship within Glycine species (Table S3). Besides Glycine species, GmSGS3 shared high amino acid sequence identity with other legume species, displaying 88.7% identity with VunSGS3 (V. unguiculata) and 87.8% identity with PvSGS3 (Phaseolus vulgaris) (Table S3). In contrast, GmSGS3 had less amino acid sequence identity with SGS3 homologs from non-legume species such as N. tabacum (55%), N. benthamiana (53.7%), A. thaliana (54.1%), and S. lycopersicum (52.2%) (Table S3).

Like GmRDR1, InterPro-Pfam, and Motif Search server predicted GmSGS3 had XS domain (316–430 aa) and the zinc finger XS (zf-XS) domain (245–283 aa), both of which are crucial for RNA binding and stabilization in gene silencing processes (Table 2).

GmRDR1 protein interacts with GmSGS3 protein

The predicted 3-D structure of GmRDR1 and GmSGS3 proteins was developed through homology modeling using the SWISS-MODEL server (Fig. S3Ai,Bi). The structural integrity of these models was evaluated using SAVES server by analyzing ERRAT scores and Ramachandran plot, confirming quality and accuracy. Ramachandran plot showed that 94.8% and 81.3% of amino acid residues of GmRDR1, and GmSGS3, respectively, were in the favoured regions, indicating good structural integrity (Fig. S3Aii,Bii). Additionally, ERRAT scores provided further validation, with GmRDR1 and GmSGS3 achieving 94.38% and 98.54%, respectively, affirming high reliability of both models (Fig S3Aiii,Biii). The ProSA-web Z-scores aligned with expectations for well-constructed protein structures, with scores of −14.31 and −6.41 for GmRDR1 and GmSGS3, respectively, further confirming their structural quality and suitability for subsequent analyses (Fig. S3Aiv,Biv).

Molecular docking analysis using ClusPro and Gramm servers predicted stable protein–protein interactions between GmRDR1 and GmSGS3 proteins (Fig. 4A). ClusPro server-generated top-ranked docking complex from Cluster 0 yielded a center-weighted score of −1309.2 and lowest energy score of −1493.2 Kcal mol⁻¹ (Table S4), indicating strong binding affinity and stable interaction. The interacting residues between GmSGS3 and GmRDR1 proteins were analyzed using PDBsum. This analysis revealed that they interact through a substantial interface area, with GmSGS3 covering 2527 Ų and GmRDR1 covering 2509 Ų. Within this interface, 36 residues from GmSGS3 and 43 residues from GmRDR1 were involved (Fig. 4C). The interaction includes formation of two salt bridges and nine hydrogen bonds, together with 428 non-bonded contacts (Fig. 4B), highlighting key interactions that likely contribute to the stability and specificity of the GmSGS3–GmRDR1 complex.

The yeast two-hybrid (Y2H) assay was used to validate the predicted docking result between the GmSGS3 and GmRDR1 proteins (Fig. 5). The pGADT7-RDR1 (prey) and pGBKT7-SGS3 (bait) constructs were successfully transformed into yeast strains Y187 and Y2HGold, respectively, as confirmed by growth of transformants on SD selective media lacking leucine (SD/-Leu) for pGADT7 and lacking tryptophan (SD/-Trp) for pGBKT7 constructs. There was no growth for negative controls containing empty vectors, demonstrating specificity of transformation and vector insertion. Following mating, diploid yeast cells containing both bait and prey constructs were selected on SD/-Leu/-Trp media, confirming successful mating. Diploids were subsequently plated on SD/-Leu/-Trp/-His medium to assess the potential interaction between GmSGS3 and GmRDR1 proteins. Colony growth on the SD/-Leu/-Trp/-His medium indicated a positive interaction between GmSGS3 and GmRDR1, as evidenced by colony formation, compared to negative controls. To confirm and further evaluate the strength of the interaction, diploid colonies were plated on SD/-Leu/-Trp/-His/-Ade medium. Colony growth on the quadruple dropout medium lacking leucine, tryptophan, histidine, and adenine provided further evidence of the specific interaction between GmRDR1 and GmSGS3 proteins. The negative controls had no colony growth under these stringent selection conditions, while significant colony growth was observed in experimental setups containing both bait and prey constructs. For additional confirmation, diploid yeast cells were plated on SD/-Leu/-Trp/-His/-Ade medium supplemented with X-α-gal and Aureobasidin A. Yeast diploid colonies harboring both GmRDR1 and GmSGS3 constructs turned blue in the presence of X-α-gal, demonstrating β-galactosidase activity indicative of protein–protein interaction. In contrast, no blue color development was observed in the negative controls. Moreover, colony growth in the presence of Aureobasidin A further supported the specificity and robustness of the GmRDR1-GmSGS3 interaction.

GmRDR1 and GmSGS3 expression upregulated in resistant soybean genotypes after MYMIV inoculation

The transcript dynamics of GmRDR1 and GmSGS3 were analyzed using qRT-PCR in two resistant soybean genotypes, SL 1074 and SL 958, and two susceptible soybean genotypes, JS 335 and AMS 19-01, at 15, 30, and 45 days post-inoculation (dpi) with MYMIV. Corresponding non-inoculated (healthy) plants for these time points were used as controls. For GmRDR1, SL 1074 markedly increased expression at 15 dpi, with a 4.76-fold change relative to its healthy control (Fig. 6Ai). SL 958 also exhibited a substantial response, with a 3.55-fold change at 15 dpi (Fig. 6Aii), while JS 335 (Fig. 6Aiii) and AMS 19-01 (Fig. 6Aiv) showed lower fold changes of 1.43 and 1.19, respectively, indicating weaker upregulation of GmRDR1. At 30 dpi, GmRDR1 expression was 8.83- and 6.80-fold for SL 1074 (Fig. 6Av) and SL 958 (Fig. 6Avi), respectively. Conversely, JS 335 (Fig. 6Avii) and AMS 19-01 (Fig. 6Aviii) showed only modest increases with fold changes of 2.66 and 1.70, respectively, after 30 dpi. Results at 45 dpi were also consistent for both groups. At 45 dpi, relative fold changes of GmRDR1 expression were 8.40 and 5.44 for SL 1074 (Fig. 6Aix) and SL 958 (Fig. 6Ax), respectively, whereas for JS 335 (Fig. 6Axi) and AMS 19-01 (Fig. 6Axii) values were 2.35 and 1.89, respectively.

For GmSGS3, a similar trend was observed, with resistant genotypes displaying notably higher expression post-inoculation of MYMIV. At 15 dpi, SL 1074 (Fig. 6Bi) and SL 958 (Fig. 6Bii) exhibited fold changes of 4.25 and 4.907, respectively, relative to their healthy controls. In contrast, JS 335 (Fig. 6Biii) and AMS 19-01 (Fig. 6Biv) had minimal increases, with fold changes of 1.66 and 1.584, respectively. At 30 dpi, SL 1074 (Fig. 6Bv) and SL 958 (Fig. 6Bvi) had further elevated GmSGS3 expression levels, reaching fold changes of 7.36 and 7.49, respectively, while JS 335 (Fig. 6Bvii) and AMS 19-01 (Fig. 6Bviii) remained low at 1.606 and 1.53, respectively. By 45 dpi, high expression levels were sustained in SL 1074 (Fig. 6Bix) and SL 958 (Fig. 6Bx), with fold changes of 6.08 and 7.21, respectively, while JS 335 (Fig. 6Bxi) and AMS 19-01 (Fig. 6Bxii) showed minimal increases of 1.77 and 1.97, respectively.

Transient knockdown of GmRDR1 and GmSGS3 transcripts in resistant cultivars of soybean result in increased MYMIV accumulation

Bioinformatic analysis using siDirect software was employed to identify siRNA hotspot regions within the GmSGS3 and GmRDR1 of resistant varieties SL 958 and SL 1074. A sequence of 632 bp in GmSGS3 and 601 bp in GmRDR1 exhibited a high density of potential siRNA targets with minimal off-target effects, making them optimal candidates for subsequent siRNA-mediated gene silencing studies. Targeted 632 bp of GmSGS3 and 601 bp of GmRDR1 were amplified using primers RDR1-GG F/RDR1-GG R and SGS3-GG F/SGS3-GG R, respectively (Fig. S4A). These amplicons were cloned in the pRNAi-GG vector to generate hairpin RNAi constructs hpSGS3 and hpRDR1. The recombinant colonies were confirmed through colony PCR using primers P21 and gene-specific reverse primers (Table S1). Colony PCR yielded expected amplicons of 899 bp for hpSGS3 and 868 bp for hpRDR1 (excess size due to addition of vector backbone sequence using p21 primer) (Fig. S4B). Amplification of the recombinant plasmids using P21, P22, and gene-specific reverse primers showed 899 and 632 bp amplicons for hpSGS3 and 868 bp and 601 bp amplicons for hpRDR1, respectively, validating cloning of the amplicons in sense and antisense orientations (Fig. S4C).

The hpRDR1 and hpSGS3 constructs were mobilized to Agrobacterium tumefaciens strain EHA 105, and the Agrobacterium cultures containing both constructs were infiltrated into seedlings of the resistant cultivars SL 1074 and SL 958; followed by agroinfiltration of a partial tandem repeat construct of MYMIV. The effectiveness of gene silencing and virus accumulation was assessed by real-time PCR. In SL 1074, hpRDR1 led to a 2.44-fold reduction in GmRDR1 gene expression compared to the control (Fig. 7Ai). This downregulation corresponded with a 4.74-fold increase in MYMIV accumulation (Fig. 7Bi). Similarly, in genotype, SL 958, GmRDR1 expression decreased 1.91-fold (Fig. 7Aii), which was associated with a 3.72-fold increase in MYMIV viral load (Fig. 7Bii). For GmSGS3, silencing in SL 1074 resulted in a 1.89-fold decrease in its expression (Fig. 7Aiii), leading to a 5.67-fold increase in MYMIV accumulation (Fig. 7Biii). In SL 958, GmSGS3 expression was reduced by 1.75-fold (Fig. 7Aiv) due to hpSGS3 treatment, resulting in a 4.91-fold increase in viral load (Fig. 7Biv).