Soybean protein datasets and complete genome sequences were obtained from the Phytozome v12.1 database (https://phytozome.jgi.doe.gov/pz/portal.html). To identify putative CAD family members, we used characterized CAD protein sequences from Arabidopsis thaliana, retrieved from The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org/), as query sequences for BLASTP searches against the soybean proteome. An initial BLASTP search was performed using an E-value cutoff of 1.0 to maximize sensitivity in candidate retrieval. Putative CAD sequences were then verified using InterProScan and SMART to assess specificity. All candidate sequences were validated for the presence of characteristic CAD domains (PF00107 and PF08240) using InterProScan (version 5.52-86.0) with default parameters and SMART databases (version 9.0). Only sequences with E-values < 0.01 for both domains and covering at least 80% of the domain length were retained (Ma et al., 2018). Duplicate entries were manually eliminated from the dataset. Physicochemical properties including protein length, molecular mass, and theoretical pI values were determined using the ExPASy proteomics server (https://www.expasy.org/) (Gasteiger et al., 2005). Predicted subcellular distribution of GmCAD proteins was determined using WoLF PSORT (https://www.genscript.com/wolf-psort.html).

For phylogenetic reconstruction, CAD amino acid sequences from six plant species Arabidopsis thaliana, Glycine max, Zea mays, Oryza sativa, Phaseolus vulgaris, and Sorghum bicolor were collected (Supplementary Table S1). Sequence alignment was performed using ClustalW, followed by phylogenetic tree construction in MEGA 5.0 with the neighbor-joining method (Tamura et al., 2011). Domain conservation was verified through sequence analysis. The Gene Structure Display Server v2.0 was used to visualize the intron-exon organization of soybean CAD genes. Syntenic relationships between CAD genes from the six species were analyzed using the Plant Genome Duplication Database (PGDD, https://chibba.agtec.uga.edu/duplication/).

Upstream regulatory sequences (2000 bp) of each GmCAD gene were extracted and submitted to PlantCARE (https://bioinformatics.psb.ugent.be/webtools/plantcare/html/) for cis-acting element prediction (Liu et al., 2015).

Expression data for GmCADs in different tissues (leaves, roots, seeds, pods, flowers) were retrieved from RNA-seq datasets available in the Phytozome database. Heatmap visualization was performed using TBtools software (Chen et al., 2020). For pathogen challenge experiments, the susceptible cultivar ‘DN50’ was inoculated using a sorghum-based inoculation system adapted from Trivedi et al (Trivedi et al., 2023). Fusarium oxysporum was propagated on potato dextrose agar (PDA) at 26°C in darkness for 7 days. The culture was then transferred onto autoclaved sorghum grains and incubated under identical conditions for 17 days until complete colonization. Plastic pots (11 cm height) were filled two-thirds full with sterilized vermiculite, and colonized sorghum grains were incorporated at 4% (w/w). After adding a 0.5 cm vermiculite layer, 15 soybean seeds were planted per pot and wholly covered with sterile vermiculite. Three biological replicates were established for each treatment. Plants were grown in controlled-environment chambers at 25°C under a 14 h light/10 h dark cycle. Root tissue was harvested at 8, 11, 14, and 17 days after inoculation, immediately frozen in liquid nitrogen, and stored at -80°C. Total RNA extraction was performed using the Quick Total RNA Isolation Kit (Huayueyang Biotechnology, Beijing). RNA integrity and concentration were assessed using a Nanodrop2000 spectrophotometer. First-strand cDNA synthesis was carried out using PrimeScript^® RT Master Mix (TaKaRa, Dalian) with 500 ng total RNA as template, followed by dilution to 500 ng/μL. Quantitative RT-PCR reactions were conducted on a Roche LightCycler^® 96 platform using SYBR Green chemistry (TaKaRa). Each 20 μL reaction mixture contained 10 μL of 2× SYBR Green Master Mix, 0.5 μL of each primer (10 μM), and appropriate cDNA template. Thermal cycling parameters were: 95°C for 30 s, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. GmACTIN (GenBank: AF049106) served as the internal reference gene. Relative transcript abundance was calculated using the 2^ΔΔCt method. ΔΔCt = (Ct target gene - Ct reference gene) treatment - (Ct target gene - Ct reference gene)control. The control group (CK, non-inoculated plants) at each time point served as the calibrator sample. Primer sequences are provided in Supplementary Table S1 (Sang et al., 2023).

GmCAD3 haplotype variation was examined across 289 Chinese soybean accessions using unpublished whole-genome resequencing data. The reference GmCAD3 sequence was obtained from Phytozome and aligned against resequencing data to detect single-nucleotide polymorphisms (SNPs). DnaSP 5.0 was used to analyze haplotype structure, with low-frequency haplotypes (<5%) excluded (Ruihong et al., 2006). Statistical comparisons of phenotypic differences between haplotypes were performed using one-way ANOVA followed by Duncan’s post-hoc test in GraphPad Prism 9.5.1.

To evaluate selective pressures on GmCAD3 during soybean domestication, we analyzed publicly available genomic data from 2,790 accessions in the SoyMD database (https://yanglab.hzau.edu.cn/SoyMD/#/). Population genetic parameters, including fixation index (Fst), nucleotide diversity (π), and Tajima’s D, were calculated for GmCAD3 regions. The dataset included 591 wild soybean accessions, 439 Huang-Huai-Hai landraces, 379 Huang-Huai-Hai cultivars, 539 northern landraces, and 842 northern cultivars (Yang et al., 2023).

Complete coding sequences of GmCAD3 haplotypes were amplified from roots of representative soybean varieties (Hap1/3 and Hap2). These sequences were fused to GFP at the C-terminus and cloned into pBI121 vectors through homologous recombination. Arabidopsis thaliana mesophyll protoplasts were isolated and transformed via PEG-calcium-mediated transfection, with a 16-hour incubation period (Yoo et al., 2007). An empty GFP vector served as control. Fluorescence imaging was conducted using a Carl Zeiss LSM710 confocal microscope. AtHSBP-RFP was used as a cytoplasmic reference marker (Hsu et al., 2010).

Promoter regions (2000 bp upstream of start codon) from each GmCAD3 haplotype were amplified and inserted upstream of the luciferase reporter gene in CP461 vectors via homologous recombination, generating proGmCAD3-Hap1::Luc, proGmCAD3-Hap2::Luc, and proGmCAD3-Hap3::Luc constructs. These plasmids were introduced into Agrobacterium tumefaciens EHA105 (Weidi, Shanghai) by chemical transformation. Bacterial suspensions were infiltrated into Nicotiana benthamiana leaves. At 48 hours post-infiltration, leaves were sprayed with Luciferin substrate (Promega, USA) and incubated 5–10 minutes before imaging with a Tanon system (Tanon, China). For quantitative measurements, infiltrated leaf tissue was collected at 48 hpi, ground in liquid nitrogen, and lysed. Luminescence was quantified using a dual-luciferase reporter assay system (Promega, USA) following manufacturer guidelines. Firefly luciferase (LUC) signals were measured first, then Renilla luciferase (REN) signals after LUC quenching (Zhou et al., 2025b).

Coding sequences of GmCAD3 haplotypes (Hap1/3 and Hap2) were cloned into pCAMBIA3301 under CaMV 35S promoter control, creating pCAMBIA3301-GmCAD3-Hap1/3 and pCAMBIA3301-GmCAD3-Hap2. These constructs were transformed into Agrobacterium rhizogenes K599 and used to infect hypocotyls of ‘DN50’ seedlings. Hairy root induction and selection followed established protocols (Zhou et al., 2025c). Non-infected roots served as negative controls. QRT-PCR and fluorescence microscopy confirmed the presence of transgenic roots, and non-transgenic roots were removed. At 17 days post-pathogen inoculation, disease severity, root length, and fresh weight were measured in both control hairy roots (CHR) and overexpression hairy roots (OHR).

At 14 days after Fusarium oxysporum inoculation, CAD enzyme activity and total lignin content were determined in wild-type and GmCAD3-overexpressing hairy roots using commercial kits (CAD Activity Assay Kit, AKSU053U; Lignin Content Assay Kit, AKSU010U; BOXBIO, China). For each assay, 0.1 g of root tissue was processed according to kit protocols.

Statistical analyses were conducted using SPSS 25 and GraphPad Prism 9.5.1. Data normality was assessed using Shapiro-Wilk tests, and homogeneity of variance was evaluated using Levene’s test. For pairwise comparisons, Student’s t-tests were performed. For multiple group comparisons, one-way ANOVA followed by Duncan’s post-hoc test was used. For haplotype disease severity index (DSI) analysis, one-way ANOVA with Tukey’s HSD post-hoc test was applied. Significance threshold was set at P < 0.05. For multiple comparisons in expression profiling across tissues and timepoints, false discovery rate (FDR) correction was applied using the Benjamini-Hochberg method.

To identify members of the GmCAD gene family, the known CAD protein sequences from Arabidopsis thaliana were compared against the Phytozome v12.1 database. As summarized in Table 1 and Supplementary Table S2, a total of seven cytosolic GmCAD genes (GmCAD1-7) were identified in the soybean genome, with ORF lengths ranging from 870 bp (GmCAD5) to 1095 bp (GmCAD3). These genes encode proteins of 289–364 amino acids, with predicted molecular weights of 31.43–39.37 kDa and theoretical isoelectric points (pI) between 5.51 and 6.86 (Table 1).

To elucidate the evolutionary relationships of CAD genes, a phylogenetic tree was constructed using 45 CAD proteins from six species: rice (10), soybean (7), Arabidopsis thaliana (6), maize (5), common bean (8), and sorghum (9). The CAD proteins were classified into four distinct clusters based on their phylogenetic relationships (Figure 1A). Cluster 1 primarily contained four sorghum, four maize, and eight rice CAD members. Cluster 2 was mainly composed of five soybean members (GmCAD1, GmCAD2, GmCAD3, GmCAD5, GmCAD6), six common bean members, and AtCAD6 from Arabidopsis thaliana. Cluster 3 largely comprised three sorghum members, three Arabidopsis thaliana members, the common bean member PvCAD1, and the rice member OsCAD9. Cluster 4 contained two soybean members (GmCAD4 and GmCAD7), two Arabidopsis thaliana members, and one member each from rice, maize, sorghum, and common bean. Domain analysis revealed that the CAD proteins contain two conserved domains (PF08240, alcohol dehydrogenase N-terminal; PF00107, alcohol dehydrogenase C-terminal), consistent with previously reported domains and responsible for CAD catalytic activity (Figure 1B).

Comparative analysis of CAD genes from Arabidopsis thaliana, Glycine max, Zea mays, Oryza sativa, Phaseolus vulgaris, and Sorghum bicolor provided valuable insights into the evolutionary history of GmCADs. As shown in Figure 2A, the GmCADs were distributed across 6 of the 20 soybean chromosomes, with 1 or 2 genes located on each of these chromosomes. A total of 12 orthologous CAD gene pairs were identified among the six species (Supplementary Table S4). Notably, except for the syntenic relationship between GmCAD4 and GmCAD7, no synteny was detected between soybean and the other species, suggesting that the diversification of soybean CAD genes primarily occurred after species divergence. Tissue-specific expression analysis revealed distinct tissue specificity among GmCADs (Figure 2B; Supplementary Table S5). GmCAD2, 3, 5, and 6 showed negligible transcript abundance across roots, stems, leaves, flowers, and seeds, whereas GmCAD1, 4, and 7 were highly expressed in all examined tissues.

The promoter regions of each GmCAD gene were examined to identify potential cis-acting elements involved in transcriptional regulation. As illustrated in Figure 3, and Supplementary Table S5, most GmCAD genes contained hormone-responsive elements, including the auxin-responsive element (AuxRE), abscisic acid-responsive elements (ABRE and GARE-motif), gibberellin-responsive element (P-box), salicylic acid-responsive element (TCA-element), and methyl jasmonate-responsive elements (CGTCA-motif and TGACG-motif). Interestingly, while the CAT-box element (associated with meristem expression) was identified in GmCAD1, 2, and 4, GmCAD2 showed lower expression across most tissues examined, suggesting that additional regulatory mechanisms beyond the presence of this cis-element may control its tissue-specific expression pattern. Furthermore, the majority of GmCAD gene promoters contained stress-responsive cis-elements. Among them, the low-temperature responsive element (LTR) was present in GmCAD3 and 4; the anaerobic response element (ARE) was found in GmCAD1, 4, 5, and 7; and the TC-rich repeat element, involved in defense and stress responses, was identified in GmCAD3. These bioinformatic analyses of cis-regulatory elements suggest that GmCAD genes play significant roles in modulating plant stress responses and growth development.

To further investigate the potential role of GmCADs in response to Fusarium oxysporum stress, their expression patterns were examined at 8, 11, 14, and 17 days post-inoculation (dpi). The expression levels of all detected GmCADs were upregulated following pathogen challenge, except for GmCAD5, which showed no detectable expression. Among them, GmCAD3 exhibited the most significant differential expression compared to the CK group. Its expression showed a time-dependent increase upon infection, peaking at 14 and 17 dpi, with levels 8.64-fold and 6.68-fold higher than those in the CK, respectively (Figure 4).

To investigate the association between natural variation in GmCAD3 and disease resistance phenotypes, we performed haplotype analysis using genome resequencing data from 289 soybean accessions. Sequence analysis identified seven single-nucleotide polymorphisms (SNPs) within a 5,433 bp genomic region encompassing the GmCAD3 gene, including five located in the 2,000 bp promoter region and two in the 1,095 bp CDS (Figure 5A). Based on these SNPs, the accessions were classified into three distinct haplotypes. Hap 1 was the most common (n = 110, 38.1%), while Hap 2 (n = 106, 36.7%) and Hap 3 (n = 73, 25.3%) were less frequent. Phenotypic evaluation revealed significant differences in disease resistance among haplotypes (Figure 5B). Accessions carrying Hap2 and Hap3 exhibited significantly lower disease indices than those carrying Hap1. Although Hap3 showed a numerically lower disease index than Hap2, this difference was not statistically significant.

To investigate whether GmCAD3 experienced selective pressure during soybean domestication and improvement, we analyzed population genetic parameters using genomic data from 2,790 soybean accessions retrieved from the SoyMD database (Figure 5C). The dataset comprised 591 wild soybeans (G1), 439 Huang-Huai-Hai landraces (G2), 379 Huang-Huai-Hai improved cultivars (G3), 539 Northeast China landraces (G7), and 842 Northeast China improved cultivars (G8). Pairwise fixation index (Fst) analysis between wild and domesticated populations revealed weak to moderate genetic differentiation (Fst = 0.05–0.21) across the GmCAD3 genomic region, indicating the absence of strong directional selection at this locus during domestication. Notably, nucleotide diversity (π) was consistently higher in cultivated populations compared to wild accessions, a pattern contrary to expectations under strong positive selection, which typically reduces genetic variation surrounding selected alleles. Tajima’s D statistics provided further evidence against a selective sweep at the GmCAD3 locus. Cultivated populations displayed significantly positive Tajima’s D values (ranging from 2.36 to 2.97), suggesting either balancing selection or recent population expansion rather than directional selection for a single favorable allele. These positive values indicate an excess of intermediate-frequency alleles relative to neutral expectations, consistent with the maintenance of multiple functional haplotypes.

Taken together, these population genetic signatures demonstrate that GmCAD3 has retained substantial haplotypic diversity throughout soybean domestication and modern breeding. The persistence of multiple GmCAD3 alleles in contemporary cultivars suggests that genetic variation at this locus provides adaptive benefits for disease resistance under diverse environmental conditions, rather than a single “optimal” allele being fixed through artificial selection.

To investigate the functional consequences of SNPs among different GmCAD3 haplotypes, we first analyzed their promoter sequences. Comparative analysis revealed distinct SNP profiles: GmCAD3-HAP1 and GmCAD3-HAP2 differed at four positions (Chr9_42543767, Chr9_42543441, Chr9_42543020, and Chr9_42542588), while GmCAD3-HAP1 and GmCAD3-HAP3 differed at Chr9_42543356. GmCAD3-HAP2 and GmCAD3-HAP3 exhibited variations at all five promoter SNP positions (Supplementary Figure 1; Figure 5A). Promoter activity was assessed using Luc assays (Figures 6A, B). The luminescence intensity of proGmCAD3-HAP2::Luc and proGmCAD3-HAP3::Luc was significantly stronger than that of proGmCAD3-HAP1::Luc. Quantitative analysis revealed that the LUC/REN ratio for GmCAD3-HAP2 (3.25 ± 0.18) was 1.18-fold and 1.44-fold higher than those of GmCAD3-HAP3 (2.76 ± 0.15) and GmCAD3-HAP1 (2.26 ± 0.12), respectively.

Analysis of translated amino acid sequences (Supplementary Figure 2) indicated that the SNP at Chr9_42539607 in GmCAD3-HAP1 and GmCAD3-HAP3 is synonymous, resulting in identical amino acid sequences. In contrast, GmCAD3-HAP2 carries a nonsynonymous mutation at Chr9_42541316, leading to an amino acid substitution at position 46 from isoleucine (I) to valine (V). Subcellular localization analysis demonstrated that this I46V substitution did not alter protein localization, with both GmCAD3-HAP1/3 and GmCAD3-HAP2 variants localized to the cytoplasm (Figure 6C).

To isolate the functional effects of coding sequence polymorphisms from regulatory differences between haplotypes, the CaMV35S constitutive promoter was used to drive expression of both Hap1/3 and Hap2 coding sequences at comparable levels. This design complements our promoter activity analysis (Figures 6A, B), which demonstrated that the five promoter SNPs result in differential transcriptional activity among haplotypes. By standardizing expression levels through a common promoter, we could directly assess whether the nonsynonymous I46V substitution in Hap2 (Supplementary Figure 2) affects CAD enzyme activity or disease resistance function. To investigate the biological functions of different haplotypes of GmCAD3 in disease resistance, transgenic soybean hairy roots overexpressing this gene were generated via Agrobacterium rhizogenes-mediated transformation (Figure 7C). qRT-PCR analysis confirmed that the transcriptional levels of GmCAD3 in four independent overexpression lines (OHR1/3-1, OHR1/3-2, OHR2-1, OHR2-2) were 2.3 to 4.15-fold higher than those in the control wild-type (WT) hairy roots. All four independent overexpression lines were used in subsequent phenotypic, enzymatic, and biochemical analyses to ensure reproducibility of observed effects across independent transformation events. Following inoculation with Fusarium oxysporum, all GmCAD3-overexpressing hairy roots, regardless of haplotype, exhibited significantly enhanced disease resistance compared to the WT control (Figure 7A). Disease severity index (DSI) analysis revealed that the average DSI values of GmCAD3-Hap1/3 and GmCAD3-Hap2 were 36.46 and 36.11, respectively, both significantly lower than that of the WT (62.41). However, no significant difference in DSI was observed between GmCAD3-Hap1/3 and GmCAD3-Hap2. Morphological trait measurements showed that GmCAD3-Hap1/3 and GmCAD3-Hap2 maintained longer root lengths (27.61 cm and 28.57 cm, respectively) and higher root fresh weights (1.59 g and 1.28 g, respectively), which were significantly greater than those of the WT (15.08 cm, 0.68 g; Figure 7B). No significant morphological differences were detected between the two haplotypes.

To further examine whether GmCAD3 enhances disease resistance in soybean through CAD enzyme activity and lignin accumulation, we measured both CAD activity and lignin content in the four overexpression lines. As shown in Figure 7B, CAD activities in GmCAD3-Hap1/3 and GmCAD3-Hap2 were 695.90 U/g and 535.00 U/g, respectively, significantly higher than that in the WT (263.18 U/g). Similarly, after inoculation with F. oxysporum, lignin content in the overexpression hairy roots of GmCAD3-Hap1/3 and GmCAD3-Hap2 increased significantly to 369.51 mg/g and 340.62 mg/g, respectively, compared to 196.68 mg/g in the WT (Figure 7B). These results indicate that GmCAD3 enhances resistance to pathogen infection by elevating CAD activity and promoting lignin accumulation, thereby alleviating root damage. Furthermore, the SNP variations in the coding sequence among different GmCAD3 haplotypes did not lead to significant functional differences.

The CAD family has undergone species-specific expansion throughout evolution, as demonstrated by comparing CAD families from various plant species (Zhao et al., 2020; Zhang et al., 2021; Li et al., 2022). Recently, new functions of CAD have been reported in wheat (Rong et al., 2016), cotton (Li et al., 2022), and Arabidopsis (Li et al., 2022), including roles in lignin biosynthesis, cell wall formation, and disease resistance. Tissue-specific expression analysis revealed distinct patterns among GmCAD members. GmCAD1, 4, and 7 were found to have high expression levels across all examined tissues, whereas GmCAD2, 3, 5, and 6 showed negligible transcript abundance in roots, stems, leaves, flowers, and seeds. The numerous potential physiological roles of GmCADs in soybean development are supported by their expression patterns in various tissues. The transcriptional responses of GmCADs to Fusarium oxysporum infection were examined, showing significant upregulation of GmCADs under pathogen attack, especially GmCAD3. Likewise, TaCAD12 (Rong et al., 2016), and GhCAD7 (Tuerxun et al., 2025), transcripts are significantly upregulated in response to pathogen infection and play crucial roles in disease resistance. Cis-regulatory elements in the promoter regions may mediate transcriptional activation of CAD genes in response to biotic stress (Ijaz et al., 2020; Xu et al., 2024). Most cis-acting elements in GmCAD promoters likely regulate stress and hormone responses, including ABRE, GARE-motif, TCA-element, CGTCA-motif, and TGACG-motif, as well as stress-response elements such as LTR, TC-rich repeats, and ARE. Phylogenetic analysis shows that GmCADs form four groups, mainly in Cluster 2 with common bean and AtCAD6, suggesting conserved functions in legumes. Synteny analysis reveals limited cross-species relationships, except for GmCAD4-GmCAD7, suggesting soybean CAD gene diversification occurred mainly after species divergence. This pattern differs from many other gene families that show widespread synteny across plant lineages, implying that lineage-specific selection pressures may shape CAD genes (Zhao et al., 2017; Wang et al., 2025).

GmCAD3 responded more rapidly and strongly to Fusarium oxysporum than other genes, with maximal induction (~8.64-fold) at 14 days post-inoculation, indicating a key role in pathogen stress response. Using the hairy root system, it was confirmed that GmCAD3 encodes a cinnamyl alcohol dehydrogenase. Overexpressing GmCAD3 in soybean hairy roots increased root length and weight, suggesting improved disease tolerance. These results align with earlier studies, where overexpressing CAD genes markedly enhanced disease resistance in wheat and other plant species (Rong et al., 2016; Zhang et al., 2022; Zhou et al., 2025a, b).

Cell wall reinforcement through lignin deposition is a crucial defense mechanism under pathogen attack (Gao et al., 2024; Uddin et al., 2024). In pathogen-challenged situations, CAD is essential for the production of lignin monomers (Zhang et al., 2022). During Fusarium oxysporum infection, the CAD enzyme activity in GmCAD3-OHR increased considerably, and a significant accumulation of lignin was observed. Consistent with earlier studies, our observations confirmed the role of GmCAD3 in enhancing disease resistance by catalyzing the final step in monolignol biosynthesis and promoting lignin accumulation (Roychowdhury et al., 2025; Wang et al., 2025). Recent evidence indicates that lignin functions as a physical barrier that strengthens cell walls and limits pathogen invasion (Zhou et al., 2025b). The temporal expression pattern of GmCAD3, which shows progressive induction over time and peaks at 14–17 days post-inoculation, aligns with the establishment of defense lignification, which requires sustained gene expression and metabolite accumulation. In contrast, constitutively expressed GmCAD members (GmCAD1, 4, and 7) are likely to contribute to developmental lignification rather than to stress-induced responses.

The widespread application of whole-genome sequencing-based haplotype analysis in crops has been used in recent years to mine beneficial alleles from natural variability, narrow the target range, and improve the accuracy of target gene identification (Yang et al., 2023). This approach has been increasingly employed to assess candidate genes and study the role of natural variation in disease resistance traits (Sang et al., 2023). The GmCAD3 gene harbored 7 significant SNPs, 5 in the promoter region and 2 in the coding sequence. Three haplotypes were identified in 289 soybean accessions: Hap2 and Hap3 showed lower disease severity than Hap1, indicating that variants confer resistance to Fusarium oxysporum. Promoter analysis showed that these SNPs differentially affected transcriptional activity, with proGmCAD3-HAP2:Luc exhibiting the strongest luminescence, followed by hap3 and hap1. The LUC/REN ratio was 1.18 and 1.44 times higher for HAP2 than for HAP3 and HAP1, respectively, indicating that promoter polymorphisms affect gene expression, likely by altering transcription factor binding sites, thereby influencing GmCAD3 transcription, lignin levels, and disease resistance.

The two SNPs in the coding sequence differentiated GmCAD3-HAP2 from the other haplotypes. Analysis revealed that one SNP represents a synonymous mutation in GmCAD3-HAP1 and GmCAD3-HAP3, while the other causes a nonsynonymous mutation in GmCAD3-HAP2, resulting in an amino acid substitution from isoleucine to valine at position 46. Subcellular localization analysis demonstrated that this substitution did not alter protein localization, with both variants remaining in the cytoplasm.

To determine whether this I46V substitution affects protein function, we conducted functional validation through hairy root overexpression using a constitutive CaMV35S promoter. This experimental design was specifically chosen to isolate coding sequence effects from the regulatory differences already characterized through promoter activity analysis (Figures 6A, B). The results showed that both GmCAD3-Hap1/3 and GmCAD3-Hap2 exhibited comparable disease resistance (DSI: 36.46 vs 36.11), CAD enzyme activity (695.90 vs 535.00 U/g FW), and lignin content (369.51 vs 340.62 mg/g DW) when expressed at similar levels (Figure 7B). These results indicate that the I46V substitution, involving two hydrophobic amino acids with similar chemical properties, does not significantly affect protein catalytic function or disease resistance capacity.

Integrating our findings across multiple experimental approaches reveals a clear mechanistic picture: the differential disease resistance observed among GmCAD3 haplotypes in natural germplasm accessions (Figure 5B) is primarily driven by promoter variation affecting expression levels rather than by coding sequence changes affecting protein function. The promoter activity analysis showed that HAP2 exhibits 1.18-fold and 1.44-fold higher transcriptional activity than HAP3 and HAP1, respectively (Figure 6B), while the protein functional validation demonstrated equivalent enzymatic capacity across haplotypes. This finding has important implications for molecular breeding strategies: marker-assisted selection should prioritize favorable promoter haplotypes that enhance GmCAD3 expression, as the protein variants themselves are functionally equivalent. The higher promoter activity of HAP2, combined with its significantly lower disease severity index in germplasm evaluation, supports this variant as the optimal target for improving Fusarium resistance in soybean breeding programs.