The experiments were conducted using the flax cultivar ‘Longyan 10’ from the Gansu Academy of Agricultural Sciences (Zhang et al., 2020). Seeds were surface-sterilized by immersion in 75% ethanol for 10 minutes, followed by three rinses with sterile deionized water, and subsequently sown into autoclaved nutrient soil. Growth chamber conditions were maintained at 26°C (day)/18°C (night) with a 16-hour photoperiod. Stress treatments were initiated when seedlings reached 6–7 cm in height. For drought simulation, plants were gently uprooted, washed with distilled water to eliminate soil particles, and transferred to hydroponic systems containing 10% (w/v) polyethylene glycol-6000 (PEG-6000). Salt stress was imposed using an identical protocol with 100 mM sodium chloride (NaCl) solution. Control plants were maintained in distilled water, while low-temperature stress groups were incubated at 4°C. Leaf tissues were harvested at 0, 6, 12, and 24 hours post-treatment using synchronous sampling across all groups to eliminate diurnal rhythm interference, with three biological replicates per time point. All samples were flash-frozen in liquid nitrogen and archived at -80°C for subsequent molecular analyses.

All experimental procedures were conducted with a minimum of three biological replicates. Quantitative data are presented as mean ± standard deviation (SD) derived from triplicate measurements. Statistical analyses and graphical representations were performed using GraphPad Prison 8 software (version 8.4.3; GraphPad Software). Significant differences between groups were determined by two-tailed Student’s t-test, with asterisks denoting the following probability thresholds: *P < 0.05, **P < 0.01, ***P < 0.01.

The complete genome assembly of flax (Longya10) was retrieved from the NCBI database under accession number QMEI02000000, while genome annotation files were sourced from the Figshare repository (https://figshare.com/articles/dataset/Annotation_files_for_Longya-10_genome/13614311). Eight Arabidopsis SOD homologs were analyzed via the TAIR database (Garcia-Hernandez et al., 2002). Candidate flax SOD genes were identified through BLASTP homology searches against the flax proteome (E-value cutoff: 1e-5). Hidden Markov Models (HMMs) for Cu/Zn-SOD (PF00080) and Fe/Mn-SOD (PF02777, PF00081) domains were acquired from the Pfam database (http://pfam.xfam.org/) (Mistry et al., 2021). The HMMER3.0 hmmsearch algorithm (Potter et al., 2018) was employed to predict SOD homologs in flax. Predicted sequences were further validated using SMART (http://smart.embl.de/smart/batch.pl) and the Conserved Domain Database (CDD; https://www.ncbi.nlm.nih.gov/cdd/), yielding a final set of 12 non-redundant LuSOD genes. Biophysical parameters of the LuSOD proteins—including coding sequence length, amino acid count, molecular weight (MW), theoretical isoelectric point (pI), and grand average of hydropathy (GRAVY)—were computationally derived using ExPASy ProtParam (https://web.expasy.org/protparam/) (Fink and Scandalios, 2002). Subcellular localization predictions were performed via the BUSCA web server (http://www.busca.cn).

The full-length genomic sequences of rice and soybean were acquired from the Phytozome platform (https://phytozome-next.jgi.doe.gov/) (Goodstein et al., 2012). SOD gene families in these species were characterized using identical bioinformatics workflows. Multiple sequence alignment of SOD proteins from Arabidopsis, rice, soybean, and flax was performed using ClustalW in MEGA11 with default parameters (Yuan et al., 1999). A maximum likelihood (ML)-based phylogenetic tree was generated in MEGA11 under standard configurations (neighbor-joining algorithm; 1,000 bootstrap replicates) (Tamura et al., 2021). Chromosomal localization of LuSOD genes was mapped by integrating the flax genomic FASTA file with GFF3 annotations. Conserved structural domains were predicted via the NCBI CD-Search tool (https://www.ncbi.nlm.nih.gov/Structure/BWRPSB/BWRPSB.cgi). Conserved protein motifs were identified using MEME Suite (http://alternate.meme-suite.org/tools/meme), and visualization was executed with TBtools v2.069 (Chen et al., 2020).

The genome and annotation files of Arabidopsis were retrieved from the TAIR database (https://www.arabidopsis.org/). Collinear relationships were predicted using the Multiple Collinearity Scan (MCScanX) algorithm (Wang et al., 2012). Genome-wide duplication (WGD) events involving LuSOD genes were identified through whole-genome synteny analysis. Tandem duplication events were defined as chromosomal regions harboring two or more homologous genes within a 100-kb span, with no intervening non-homologous genes. Segmental duplication events (BLASTN E-value < 1e-5) were detected by analyzing 100-kb genomic regions (50 kb upstream and downstream) flanking coding sequences (CDS) using BLASTN alignments. Repetitive genes were classified based on sequence alignment length ≥200 base pairs (bp) and nucleotide sequence identity exceeding 85% (Khaja et al., 2006).

Potential miRNA targets of the LuSOD gene family were predicted by aligning miRNA sequences with the 5’ and 3’ untranslated regions (UTRs) and coding sequences (CDS) using the psRNATarget platform (Plant MicroRNA Target Analysis Server; https://www.zhaolab.org/psRNATarget/analysis?function=3) under default parameters (Griffiths-Jones et al., 2006). The 2,000-bp upstream promoter regions of LuSOD genes were isolated from the flax genome using the TBtools software suite. To elucidate the transcriptional regulatory mechanisms of LuSOD genes under environmental stress, cis-acting elements within the 2.0-kb promoter sequences upstream of the translation start site were annotated via the PlantCARE database (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) (Lescot et al., 2002). Identified motifs were visualized using TBtools (version 2.069) for comparative analysis.

To delineate the protein interaction network of the LuSOD gene family, orthologous Arabidopsis SOD genes were employed as reference sequences. Functional protein-protein interaction (PPI) networks were reconstructed using the STRING database (v11.5; https://string-db.org/) under standard configurations (Franceschini et al., 2013). For Gene Ontology (GO) enrichment analysis, the flax proteome was annotated via the eggNOG-mapper platform (http://eggnog-mapper.embl.de/) using the GO-base.ob reference file integrated within TBtools, yielding comprehensive functional annotations (Cantalapiedra et al., 2021). Resultant datasets were visualized using TBtools.

In this study, five flax transcriptional groups were sequenced: (a) pistil, stamen, fruit and stem tip tissues (PRJNA1002756) (https://www.ncbi.nlm.nih.gov/sra/?term=); (b) floral tissues at 30, 20, 10 and 5 days after anthesis (PRJNA833557); (c) different flax embryo tissues, anther and seed tissues (PRJNA663265); (d) root and leaf tissues after salt stress (PRJNA977728) (e) Stem tissue (PRJNA874329) after heat stress. The data were filtered by fastp, then compared to the Longya10 reference genome (Chen et al., 2018). Transcript abundance was quantified as FPKM (Fragments Per Kilobase Million), and log2-transformed FPKM values were visualized as clustered heatmaps using TBtools.

Transgenic Arabidopsis was obtained via the floral dip method (Clough and Bent, 1998). The Agrobacterium harboring the intact 35S::LuCSD3 construct was inoculated into 150 mL of LB liquid medium and cultured at 200 rpm and 28°C for 16 hours. Cells were collected by centrifugation at 5000 rpm for 10 minutes and resuspended in a 5% sucrose solution adjusted to OD600 = 1.0, supplemented with 0.01% Silwet-77 surfactant. Flower buds at the pre-bolting stage were immersed in the transformation solution for 2 minutes, with three rounds of infection per week. T1 generation plants were obtained and screened. T2 lines that produced 100% hygromycin-resistant plants in the T3 generation were identified as homozygous transgenic lines.

Total RNA was isolated from flax leaf tissues using the Trizol reagent-based protocol. The SPARKscript II RT Plus Kit (With gDNA Eraser) (Shandong Sparkjade Biotechnology Co., Ltd.) was utilised to create cDNA. Gene-specific primers ( Supplementary Table S5 ) were designed using Oligo 7 primer design software, and quantitative real-time PCR amplification was conducted with TB Green^® Premix Ex Taq™ II. Relative gene expression levels were calculated via the 2^−ΔΔCT method (Livak and Schmittgen, 2001), with GAPDH serving as the internal reference gene. Three technical replicates were analyzed per sample to determine cycle threshold (Ct) values.

To assess the salt stress tolerance phenotypes of T3 transgenic Arabidopsis, two-week-old wild-type (Col-0) and overexpression lines (OE) were divided into treatment and control groups. The treatment cohort was subjected to 200 mM NaCl irrigation every three days for 15 days, while controls received equivalent volumes of distilled water. Following the stress regimen, the third fully expanded rosette leaves from both groups were harvested for nitroblue tetrazolium (NBT) staining. The NBT working solution was formulated by dissolving 0.05 g NBT powder in 0.5 mL phosphate-buffered solution (PBS, pH 7.8), followed by dilution to 50 mL with deionized water. Leaf samples were incubated in the NBT solution for 30–60 minutes under dark conditions. Subsequently, stained leaves were destained in 95% ethanol until complete chlorophyll removal for histological observation. All experimental procedures were performed with three biological replicates to ensure statistical robustness.

The physiological and biochemical indicators of the control group and OE lines (OE-1 and OE-5) were measured after 15 days of 200 mM NaCl stress. The parameters included malondialdehyde (MDA) content, proline (Pro) content, superoxide dismutase (SOD) activity, peroxidase (POD) activity, and hydrogen peroxide (H₂O₂) content. All reagent kits were provided by Beijing Boxbio Science & Technology Co.,Ltd. The measurements were conducted following the methods described in previous studies (Lu et al., 2025). Each sample was analyzed in triplicate.

Utilizing Hidden Markov Model (HMM) profiles of Cu/ZnSOD (PF00080) and Fe/MnSOD (PF02777, PF00081), 12 SOD genes were identified in the flax (Longya10) genome. Phylogenetic orthology with Arabidopsis SOD homologs led to their nomenclature as LuCSD1–LuCSD6 (Cu/Zn-SOD), LuFSD1–LuFSD3 (Fe-SOD), and LuMSD1–LuMSD3 (Mn-SOD) ( Table 1 ; Figure 1 ). Biochemical characterization of the encoded proteins revealed substantial variation: LuMSD1 encoded the longest polypeptide (384 amino acids), while LuCSD5 represented the shortest (152 amino acids). Molecular weights of LuSOD proteins spanned 15.35–42.36 kDa. Isoelectric point (pI) analysis indicated that only four proteins (LuCSD3, LuMSD1, LuMSD2, and LuMSD3) exhibited pI values >7, suggesting a predominance of acidic amino acids in the LuSOD family. Instability indices ranged from 15.74 (stable) to 51.0 (unstable), with all LuSOD proteins except LuFSD3 classified as stable. Aliphatic indices varied between 72.71 and 91.04, reflecting differences in thermostability. Hydropathicity analysis predicted LuCSD2 as the sole hydrophobic protein, whereas others were hydrophilic. Subcellular localization predictions assigned five genes (LuFSD1, LuFSD2, LuMSD1–LuMSD3) to mitochondria, with remaining members localized to chloroplasts.

To resolve evolutionary relationships, a phylogenetic tree was reconstructed from 40 SOD protein sequences across Arabidopsis (8), Glycine max (13), Oryza sativa (7), and Linum usitatissimum (12) ( Figure 2 ). According to Arabidopsis SOD protein subfamily classification ( Supplementary Table S1 ), 12 SOD genes in flax were divided into three subfamilies, named Mn-SOD, Fe-SOD and Cu/Zn-SOD respectively. Twelve LuSOD genes were distributed in each subfamily, and the largest number of genes was in the Cu/Zn-SOD subfamily, including 6 members (LuCSD1-LuCSD6). Further studies showed that there was a close relationship between flax and SOD family of Arabidopsis among the three middle species.

To characterize the protein structural features of the flax SOD gene family, the amino acid sequences of 12 LuSOD homologs were subjected to conserved motif analysis using the MEME online platform. Ten evolutionarily conserved motifs (annotated as Motif 1–Motif 10) were systematically identified across the LuSOD protein sequences ( Figures 2A, B ; Supplementary Table S2 ). LuCSD gene contains only three domains (motif1, motif3 and motif6), among which motif3 and motif6 are unique domains of LuCSD gene. We also found that both LuMSD and LuFSD genes contain motif2 and motif4, in which motif7 and motif8 are unique domains of LuMSD gene. Among LuFSD1 and LuFSD2 genes, motif9 and motif10 are their unique domains. There are great differences in protein domains among different subfamilies of LuSOD genes, but the same subfamilies have the same domain, which also proves that different subfamilies have different biological functions.

The analysis of the structure of LuSOD gene shows that the number of exons is between 5 and 11, and the number of introns is between 4 and 10 ( Figure 2C ). Both LuMSD1 and LuMSD2 contain 11 exons and 10 introns, four genes (LuCSD1, LuCSD2, LuCSD3 and LuFSD3) contain 8 exons and 7 introns, two genes (LuCSD6 and LuFSD2) contain 7 exons and 6 introns, LuCSD5 and LuCSD4 both contain 6 exons and 5 introns. The least number of exons and introns in LuMSD3 are 5 and 4 respectively.

Chromosomal localization of the LuSOD gene family was mapped using the flax reference genome, revealing an uneven distribution of 12 LuSOD genes across six chromosomes ( Figure 2D ). Chromosome 7 harbored the highest number of LuSOD genes (3 genes, 25% of the total), followed by chromosomes 1 and 5 (2 genes each, 16.67%), while chromosomes 3, 9, and 11 each contained a single LuSOD locus (8.33% per chromosome). To investigate duplication events, a collinearity analysis of the LuSOD family was performed via Circos visualization ( Figure 3A ), identifying eight segmental duplication pairs, indicative of substantial gene family expansion. To elucidate evolutionary conservation, syntenic relationships between flax and Arabidopsis SOD homologs were analyzed ( Figure 3B ). Nine collinear ortholog pairs were identified, with flax chromosomes 1, 3, 5, 7, 9, and 11 exhibiting synteny to Arabidopsis chromosomes 2, 3, and 5. Notably, no collinear SOD gene pairs were detected on Arabidopsis chromosomes 1 and 4. These findings collectively demonstrate strong chromosomal conservation of SOD genes between flax and Arabidopsis, with lineage-specific divergence in genomic organization.

To elucidate the regulatory mechanisms of the LuSOD gene family under abiotic stress, promoter regions spanning 2000 bp upstream of the LuSOD loci were analyzed and visualized ( Figure 4 ; Supplementary Table S3 ). Due to genomic proximity, the upstream sequences of LuCSD1 and LuCSD6 could not be isolated, likely resulting from overlapping or truncated intergenic regions. Consequently, promoter analyses focused on the remaining 10 LuSOD genes, excluding ubiquitous elements such as CAAT-box and TATA-box. A total of 259 cis-regulatory elements were identified and categorized into four functional groups: developmental regulation, environmental stress adaptation, phytohormone signaling, and light responsiveness ( Figure 4 ). Light-responsive elements constituted the predominant category (91 elements, 35.14%), with conserved motifs including G-box, I-box, and Box 4. The second largest group comprised hormone-related elements (89 elements, 34.36%), dominated by methyl jasmonate (MeJA)-responsive motifs (TGACG and CGTCA), alongside abscisic acid (ABRE), auxin (TGA-element, AuxRR-core), gibberellin (GARE-motif), and salicylic acid (TCA-element) response elements. Notably, MeJA-associated motifs were the most abundant hormonal regulators in LuSOD promoters. Environmental stress-responsive elements (57 elements, 22.01%) included anaerobic induction (ARE), drought-inducible MYB binding sites (MBS), low-temperature response (LTR), MYBHv1 recognition sites (CCAAT-box), and defense/stress-related TC-rich repeats. Developmental elements (22 elements, 8.5%) encompassed zein metabolism regulators (O2-site), meristem-specific motifs (CAT-box), and endosperm activity markers (GCN4_motif).

MiRNA prediction results showed that among the 12 LuSOD gene families, only 7 family members (58.33%) predicted 17 miRNA targets ( Table 2 ). Among them, LuCSD1 gene predicted the most miRNA targets, including lus-miR159b/c and lus-miR398b/c/d/e. LuFSD2 gene has the least target and contains only one miRNA target (lus-miR828a). We found that the same LuSOD gene can be targeted by different miRNA. For example, LuCSD1 can be targeted by both lus-miR159 and lus-miR398, and LuMSD2 gene can be targeted by lus-miR156 and lus-miR530 at the same time. We also found that different LuSOD genes can be targeted by the same miRNA. For example, lus-miR319 can target both LuCSD3 and LuCSD4 genes, lus-miR156 can target both LuMSD1 and LuMSD2 genes, and lus-miR159 can target LuCSD1, LuCSD3 and LuCSD4 at the same time. The results show that lus-miR159 is the main target miRNA of LuSOD gene family.light response elements.

To investigate functional linkages between flax SOD proteins and their regulatory roles, a protein-protein interaction (PPI) network was constructed using Arabidopsis homologs as a reference framework ( Figure 5A ). The analysis revealed pairwise interactions among six LuSOD genes (LuCSD2, LuCSD3, LuCSD6, LuFSD2, LuFSD3, and LuMSD2), whereas the remaining six genes showed no connectivity. The strong evolutionary conservation between LuSOD and AtSOD proteins suggested functional parallels, particularly given the established role of Arabidopsis SODs in mitigating biotic and abiotic stressors. These findings implied that the six interacting LuSOD genes may mediate analogous stress-responsive mechanisms in flax.

Gene Ontology (GO) enrichment analysis further delineated the functional spectrum of LuSOD genes across three domains: molecular functions, cellular components, and biological processes ( Figure 5B ; Supplementary Table S4 ). Molecular functions were primarily associated with antioxidant activity and superoxide dismutase activity. Cellular component annotations highlighted localization to the extracellular space and archived extracellular regions (obsolete classification). Biological processes predominantly involved cellular responses to superoxide radicals, oxygen radical detoxification, and oxidative stress adaptation, underscoring the pivotal role of LuSOD genes in stress resilience.

Transcriptomic analysis of the LuSOD gene family under abiotic stress revealed distinct expression dynamics: under heat stress, only LuFSD1 exhibited elevated expression in stems, while salt stress significantly suppressed most LuSOD genes in leaves and stems ( Figure 6A ). However, four genes (LuCSD3, LuCSD4, LuCSD5, and LuCSD6) were upregulated in roots, and three (LuMSD3, LuCSD5, and LuCSD6) showed marked induction in leaves under salt stress, with the LuFSD subfamily displaying pronounced sensitivity to salt-induced repression. Tissue-specific expression profiling further uncovered spatiotemporal regulation: in floral tissues ( Figures 6B, C ), all LuSOD genes except LuCSD4 and LuFSD2 were highly expressed at 5 days post-anthesis (DPA), followed by seven genes (LuCSD4, LuFSD2, LuCSD2, LuMSD1, LuCSD3, LuFSD1, and LuMSD2) at 10 DPA, three (LuCSD1, LuCSD5, and LuMSD3) at 20 DPA, and universal downregulation by 30 DPA. Vegetative tissues showed dominant LuSOD expression in leaves (notably LuCSD3), contrasting with minimal activity in roots, stems, anthers, stamens, and seeds. Embryo development stages revealed specialized roles: LuMSD3 peaked in mature and cotyledon-stage embryos; LuFSD3 and LuFSD2 in heart-stage embryos; LuFSD2 and LuMSD1 in globular embryos; and the LuFSD subfamily in torpedo-stage embryos. Reproductive organs exhibited coordinated upregulation of LuCSD5 and LuCSD6 in ovaries, pistils, and fruits. Collectively, the robust expression of LuSOD genes, particularly LuCSD3, in leaf tissues underscored their pivotal role in maintaining ROS homeostasis, positioning leaves as central hubs for antioxidant defense in flax.

To evaluate the involvement of LuSOD genes in abiotic stress responses (salt, cold, drought), qRT-PCR was employed to quantify their relative expression levels in flax leaf tissues under salt stress. Transcriptional dynamics were assessed at 0, 6, 12, and 24 hours post-stress induction, with expression levels normalized to the 0-hour control ( Figures 7 – 9 ). Salt stress triggered time-dependent transcriptional reprogramming within the LuSOD family ( Figure 7 ). Six genes (LuCSD4, LuCSD5, LuCSD6, LuFSD2, LuMSD1, and LuMSD3) displayed pronounced upregulation under 12-hour salt stress, peaking at 2.3-, 1.3-, 1.25-, 2.4-, and 3.1-fold increases relative to controls, respectively. Two genes (LuCSD3 and LuFSD1) exhibited maximal induction (2.1- and 2.3-fold increases) after 24 hours of salt exposure. In contrast, LuCSD1 showed no transcriptional response to salt stress in leaves. Conversely, three genes (LuCSD2, LuFSD3, and LuMSD2) demonstrated substantial downregulation, reaching minimal expression levels at 24 hours.

Under prolonged abiotic stress conditions, the LuSOD gene family exhibited dynamic transcriptional regulation in flax leaf tissues. During cold stress ( Figure 8 ), eight genes (LuCSD3, LuCSD4, LuCSD5, LuCSD6, LuFSD1, LuFSD2, LuMSD1, and LuMSD2) showed pronounced upregulation at 12 hours, peaking at 2.8-, 0.8-, 1.2-, 1.6-, 2.5-, 1.4-, 1.9-, and 2.8-fold induction, respectively, followed by gradual attenuation, while LuCSD2 remained unresponsive. Three genes (LuCSD1, LuFSD3, and LuMSD3) were significantly downregulated, reaching minimal expression levels at 24 hours, suggesting their potential role in cold adaptation through negative regulatory mechanisms. Similarly, under drought stress ( Figure 9 ), the same eight genes displayed marked upregulation at 12 hours, with peak expression levels of 3.3-, 0.7-, 1.4-, 1.5-, 2.4-, 1.5-, 2.1-, and 2.8-fold increases, respectively, whereas LuCSD1, LuFSD3, and LuMSD3 exhibited sustained downregulation, implicating their involvement in drought-responsive suppression pathways. Notably, LuCSD3 consistently demonstrated robust upregulation under both stress conditions, highlighting its central role in stress adaptation. These results collectively underscore the functional divergence of LuSOD genes in mediating stress-specific transcriptional reprogramming, with select members acting as positive regulators of antioxidant defense and others contributing to stress tolerance through negative feedback mechanisms.

Our experimental findings revealed that the LuCSD3 gene conferred robust stress tolerance in flax. To functionally characterize LuCSD3, we engineered a binary expression vector (pCAMBIA3301-LuCSD3; Figure 10A ) and stably transformed it into wild-type Arabidopsis (Col-0) via Agrobacterium-mediated floral dip. Primary transformants screening identified eight independent T1 lines showing constitutive LuCSD3 overexpression. Molecular characterization of T3-generation progeny confirmed two homozygous lines (OE-1 and OE-5) with maximal transgene expression levels ( Figures 10B, C ). Salt tolerance tests were conducted on Col and OE lines (OE-1 and OE-5) by exposing plants to 200 mM NaCl stress for 15 days. The results showed that the OE plants exhibited superior growth phenotypes under salt stress, with significantly less wilting compared to the control group (Col) ( Figure 10D ).

To assess the physiological and biochemical impacts of LuCSD3 overexpression, malondialdehyde (MDA) and proline (Pro) levels were quantified in wild-type (Col) and transgenic lines (OE-1, OE-5) under salt stress ( Figures 10F, K ). Transgenic plants exhibited markedly reduced MDA concentrations and significantly elevated Pro accumulation compared to Col controls. To evaluate LuCSD3-mediated regulation of reactive oxygen species (ROS), hydrogen peroxide (H₂O₂) content, superoxide dismutase (SOD), and peroxidase (POD) activities were examined. Under non-stress conditions, no statistically significant variations in SOD, POD, or H₂O₂ levels were observed between Col and OE lines ( Figures 10G, H, J ). However, salt-stressed OE lines demonstrated a pronounced increase in SOD (1.8-fold) and POD (2.3-fold) enzymatic activities, coupled with a 40% reduction in H₂O₂ content relative to controls. These findings were further supported by NBT staining assays, which revealed reduced ROS accumulation in OE leaf tissues under stress ( Figure 10E ). Collectively, these results indicate that LuCSD3 overexpression enhances salt stress resilience in plants by augmenting antioxidant enzyme activities (SOD, POD) to neutralize ROS and modulating osmoprotectant (Pro) synthesis while minimizing oxidative damage (reduced MDA). This dual mechanism underscores LuCSD3 as a key regulator of ROS homeostasis under abiotic stress.

When plants encounter salt stress, they employ various mechanisms to resist adverse conditions. The expression levels of five selected salt stress-related genes (NHX1, HKT1, SOS1, SOS2, and SOS3) were analyzed in OE-LuCSD3 lines ( Figure 11 ). The results indicated that the expression levels of SOS2 and SOS3 were not correlated with the expression of LuCSD3. However, three key genes (NHX1, HKT1, and SOS1) showed significantly upregulated expression under salt stress. These findings suggest that the LuCSD3 gene enhances salt tolerance by mitigating ROS accumulation in plants.