The Zinc-Finger Protein MsCCCH20 Is Predicted to Regulate Salt-Stress Response in Alfalfa (Medicago sativa L.) by Binding to Conserved 3′UTR Motifs

Abstract

Soil salinization is a source of major abiotic stress that severely limits the production of alfalfa (Medicago sativa L.), a globally critical forage legume for sustainable livestock production. Its complex autotetraploid genome and self-incompatibility greatly hinder salt tolerance genetic improvement, while the post-transcriptional regulatory mechanism of alfalfa salt stress response remains largely uncharacterized. This study aimed to narrow the gap between genome-wide genetic signals and causal regulatory mechanisms and identify core post-transcriptional regulators of alfalfa salt tolerance via a multi-layered integrative analysis pipeline. We performed a genome-wide association study (GWAS) using 220 globally collected alfalfa accessions, combined with public transcriptome integration, co-expression network analysis, 3′ untranslated region (3′UTR) motif discovery, and AlphaFold2-based protein-RNA docking simulation. We identified 20 significant salt tolerance-associated loci and prioritized the CCCH-type zinc-finger RNA-binding protein (RBP) MsCCCH20 as the core candidate regulator. We further screened 35 high-confidence target genes of MsCCCH20, detected conserved AU/AG-rich 3′UTR motifs, and provided structural predictions consistent with potential sequence-specific interactions (ipTM 0.70–0.79). Our findings establish a robust framework linking genetic association signals to post-transcriptional regulatory networks and provide high-confidence candidate genes and functional markers for the molecular breeding of salt-tolerant alfalfa.

1. Introduction

Alfalfa (Medicago sativa L.), widely regarded as the “Queen of Forages,” is the most widely cultivated perennial forage legume worldwide [1]. It is characterized by high biomass yield, with an annual dry matter yield of 7.4 t/ha under optimal water management in the agro-pastoral ecotone of Northern China, the core alfalfa planting region in China [2], well-defined nutritional value, and robust symbiotic nitrogen fixation capacity, making it a cornerstone of sustainable ruminant production, while reducing the need for synthetic nitrogen fertilizers [3], abundant flavonoids and saponins [4], and robust symbiotic nitrogen fixation capacity (100–300 kg N/ha per year) [5], making it a cornerstone of sustainable ruminant production while reducing the need for synthetic nitrogen fertilizers.

Soil salinization has become a growing threat to alfalfa production, especially in irrigated areas, where it impairs stand establishment and reduces yield stability (e.g., ~40% forage mass reduction under saline compared with non-stressed conditions) [6]. Compounding this challenge, alfalfa’s complex reproductive biology—namely, gametophytic self-incompatibility coupled with an autotetraploid genome—substantially constrains conventional breeding and slows the introgression of adaptive alleles, including those conferring salt tolerance [7]. These combined challenges highlight an urgent need for integrative research strategies that can connect natural genetic variation to the post-transcriptional regulatory networks driving alfalfa’s salt stress responses. Salt stress imposes two sequential and interrelated phases of damage on plants. The first is the rapid osmotic stress phase, which reduces soil water potential, limits plant water uptake, and inhibits vegetative growth. The second is the slower ionic stress phase, where excess Na⁺ and Cl⁻ accumulate in plant tissues, disrupting ion homeostasis, displacing essential cations like K⁺, which is critical for enzyme activity and membrane potential maintenance, and impairing membrane integrity and key metabolic enzyme activity [8].

A key downstream effect of ion imbalance is the overproduction of reactive oxygen species (ROS), alongside other cytotoxic agents including malondialdehyde (MDA), methylglyoxal (MG), and toxic levels of free Na⁺ [9,10]. ROS exhibit a dual and concentration-dependent biphasic function: at high cytotoxic concentrations, they drive oxidative damage to cellular macromolecules, while at low physiological concentrations, they act as key signal mediators, along with Ca²⁺, nitric oxide (NO), and phytohormones, to fine-tune plant stress responses. These early signaling events are decoded and amplified by a network of downstream signal transduction pathways, including the Salt Overly Sensitive (SOS) pathway for ionic homeostasis, the mitogen-activated protein kinase (MAPK) cascades, and the abscisic acid (ABA) signaling network [11]. The rapid activation of pre-existing antioxidant enzymes and the sustained induction of their gene expression are both essential for salt tolerance. To cope with the ion imbalance stress, plants have evolved a set of coordinated adaptive mechanisms, including root ion exclusion, vacuolar Na⁺ sequestration, and enhanced antioxidant defense systems, all regulated by complex gene regulatory networks [12]. So far, most studies on alfalfa salt tolerance have focused on transcriptional regulatory mechanisms, such as ABA hormone signaling pathways, MAP kinase cascades, and various transcription factors [13,14]. However, transcriptional regulation alone cannot fully explain the rapid, cell-type-specific, and reversible changes in gene expression that occur under fluctuating salt stress. This gap highlights the critical role of post-transcriptional regulatory mechanisms—including mRNA stability control, processing, and translation—in shaping the stress-responsive proteome [15].

A core layer of post-transcriptional stress regulation is mediated by the 3′ untranslated region (3′UTR), which acts as a dynamic hub for RNA metabolism. Emerging evidence indicates that cellular stress reshapes the 3′UTR landscape via alternative polyadenylation (APA), generating transcript isoforms with distinct regulatory properties and stability [16,17]. These changes in 3′UTR structure affect mRNA degradation rate, translation efficiency, and subcellular localization by altering the accessibility of cis-regulatory elements and the binding sites for RNA-binding proteins (RBPs). RBPs are a structurally diverse and functionally versatile class of regulators that govern post-transcriptional gene expression through sequence- or structure-specific RNA interactions [18]. These proteins usually contain one or more conserved RNA-binding domains (RBDs), such as RNA recognition motifs (RRM), K homology (KH) domains, and zinc-finger motifs. Different domains give RBPs their target binding specificity and allow them to regulate RNA metabolism at every sequential step of the post-transcriptional process, which we elaborate below in the context of plant salt stress response [18,19]. RBPs can modulate spliceosome activity to regulate pre-mRNA alternative splicing, generating distinct transcript isoforms from the same gene, especially for salt stress-responsive genes involved in ion homeostasis and antioxidant defense [20]. In turn, RBPs also mediate the nucleocytoplasmic export and subcellular targeting of mature mRNA, enabling cell-type-specific and spatially resolved translation of salt tolerance-related transcripts in root tissues, the primary site of salt stress perception [21]. Furthermore, RBPs bind to cis-regulatory elements in the 3′UTR (whose structure is reshaped by APA under salt stress) to directly regulate mRNA half-life, either stabilizing transcripts of salt-tolerance genes or promoting the degradation of negative regulators [16,18]. Additionally, RBPs interact with the translation initiation complex via 3′UTR binding to modulate translation efficiency of target mRNAs, enabling rapid and reversible protein synthesis in response to fluctuating salt stress, which cannot be fully explained by transcriptional regulation alone [22]. Together, these regulatory events support the formation of dynamic ribonucleoprotein complexes, which coordinate the entire post-transcriptional RNA metabolism cascade [23].

Among plant RBPs, CCCH-type zinc-finger proteins are a specialized subclass that preferentially binds single-stranded RNA. They interact with cis-regulatory elements to adjust mRNA stability and translation efficiency, thus fine-tuning gene expression under stress conditions [24,25]. The cotton CCCH protein GhZFP1 was one of the first functionally characterized members of this family; its ectopic expression improved both salt tolerance and pathogen resistance in plants, supporting a conserved role of these proteins in stress resilience [26]. In fact, functional studies in a wide range of legume species, including soybean (Glycine max L.), common bean (Phaseolus vulgaris L.), and chickpea (Cicer arietinum L.), have repeatedly shown that CCCH zinc-finger proteins play key roles in tolerance to abiotic stress like salinity, drought, and oxidative stress [27,28,29]. At the mechanistic level, these proteins regulate the stability and translation efficiency of stress-related transcripts by interacting with the mRNA decay machinery and related co-factors that control transcript turnover and persistence within the cell.

Genome-wide analyses have confirmed that CCCH proteins form an evolutionarily conserved yet functionally diversified gene family in plants, acting as key nodes that connect post-transcriptional RNA regulation to environmental adaptation [30,31,32]. Emerging evidence indicates that some CCCH/TZF proteins can directly bind salt stress-responsive transcripts in a sequence-specific manner, thereby modulating mRNA stability and translational efficiency and linking RNA turnover with abiotic stress adaptation. Consistent with this regulatory role, PeC3H74 from Phyllostachys edulis was recently reported to enhance salinity tolerance and activate ABA/stress-responsive genes in transgenic plants, further supporting the conserved involvement of CCCH proteins in plant stress adaptation [33].

Despite these advances, the target binding specificity, RNA recognition rules, and regulatory networks of CCCH-type RBPs are still poorly understood, especially in polyploid crop species, where gene redundancy can obscure functional differences between family members. In alfalfa, genome-wide association studies (GWAS) are now widely used to map loci for salt tolerance and other key agronomic traits and have repeatedly demonstrated the complex polygenic basis of these traits [34,35]. Integrative strategies that combine GWAS with transcriptome profiling have improved candidate gene prioritization, by connecting genetic variation to gene expression changes under stress. For example, combined GWAS and RNA-seq studies in alfalfa have identified several salt stress regulators, including transcription factors and hormone signaling-related genes [36]. However, most of these studies have only reached the level of statistical association or co-expression inference and have not established a complete mechanistic chain from candidate genes to its direct regulatory targets, and its binding specificity.

In this study, we aimed to systematically identify salt stress-responsive CCCH-type RNA-binding proteins (RBPs) in alfalfa and characterize their potential post-transcriptional regulatory mechanisms by integrating GWAS, transcriptome profiling, co-expression network analysis, and conserved RNA motif discovery. Through this multi-layered integrative framework, we further sought to provide mechanistic insight into RNA-mediated salt stress adaptation in polyploid forage crops.

2. Materials and Methods

2.1. Genome-Wide Association Study (GWAS) Analysis

Whole-genome resequencing data and paired salt tolerance phenotypes for 220 global alfalfa accessions were obtained from a previously published dataset [37]. SNP calling was performed using bcftools v1.10.2 with the multiallelic calling model (-m), which improves variant detection performance compared with the original consensus caller model. SNP quality control and downstream GWAS analyses were subsequently conducted in PLINK v1.90b6.21 under a diploidized SNP framework, which is commonly adopted in alfalfa association studies to ensure compatibility with conventional GWAS models. We excluded SNPs with a genotype missing rate > 20% or minor allele frequency (MAF) < 0.05. These thresholds were selected to retain high-confidence variants for association analysis, while minimizing the impact of genotype uncertainty inherent to alfalfa’s autotetraploid genome. The remaining SNPs were phased and imputed with BEAGLE v5.4 (default parameters) to infer haplotypes, fill missing genotypes, and improve marker density and accuracy. Population structure was evaluated via principal component analysis (PCA) in PLINK, with the first three principal components included as fixed-effect covariates in the association model. GWAS was performed using the multi-locus random-SNP-effect mixed linear model implemented in Fast3VmrMLM v1.0 [38], with parameters set as follows: scan radius (svrad) = 2 × 10⁴, scan p-value threshold for pre-selection (svpal) = 1 × 10⁻⁵, scan minimum LOD score for final significance (svmlod) = 3, and nThreads = 25. No kinship matrix was included (fileKin = NULL), as the multi-locus model effectively controls false positives by simultaneously incorporating multiple associated loci. SNPs with LOD (−log₁₀P) ≥ 3 were considered significant, consistent with the default empirical setting of Fast3VmrMLM and related multi-locus GWAS methods. GWAS results were visualized with Manhattan plots using the CMplot v4.5.0 package in R [39]. To evaluate population stratification and potential genomic inflation, an additional single-locus linear model was performed in PLINK v1.9 using the first three principal components (PC1–PC3) as fixed-effect covariates. The genomic inflation factor (λGC) and quantile–quantile (QQ) plots were used to assess model calibration.

2.2. GO Enrichment Analysis

To identify candidate genes, we extracted all annotated genes within 200 kb of each significant GWAS locus, a window size commonly used to capture linkage disequilibrium in plant GWASs [40]. Gene annotation files (GFF and functional annotation tables) were retrieved from the MODMS database, and additional functional annotations were assigned using eggNOG-mapper v2.1.13 [41]. We further prioritized salt stress-responsive candidates via the PlantASRG database and set the prediction score threshold at 0.85 for this study, ensuring that potential salt stress regulatory genes were not excluded; all retained candidates were further validated through subsequent multi-omics integrative analysis. GO enrichment analysis was performed using topGO v2.62 with the go-basic.obo ontology file. The “elim” algorithm combined with Fisher’s exact test was applied to reduce redundancy among hierarchically related GO terms and to improve the specificity of enrichment signals. All annotated protein-coding genes in the diploid Medicago sativa ssp. caerulea reference genome served as the background set. p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method and GO terms with adjusted p ≤ 0.05 were considered significantly enriched. Enriched terms were classified into biological process (BP), molecular function (MF), and cellular component (CC). Results were visualized using ggplot2 and tidyverse packages in R.

2.3. Identification of Salt-Responsive Genes Using RNA-Seq Data

2.3.1. Transcriptome Data Analysis

Differentially expressed genes (DEGs) were identified by integrating three independent alfalfa salt-stress RNA-seq datasets (PRJNA1182104, PRJNA952537, and PRJNA685277). Raw sequencing reads were evaluated for quality using FastQC v0.11.9, and low-quality bases plus adapter contamination were removed with fastp v0.23.2 [42], applying a Phred score threshold ≥ 20 and a minimum retained read length of 36 bp. High-quality clean reads were pseudo-aligned and quantified against the alfalfa reference transcriptome using Salmon v0.12.0 [43]. Transcript expression abundance was calculated as transcripts per million (TPM). For standardized comparison and noise reduction, all TPM values were log₂-transformed as log₂ (TPM + 1), which served as the core expression metric in this study. Gene-level read counts, estimated by Salmon and aggregated using the tximport package, were used as input for differential expression testing via DESeq2 [44], with the false discovery rate (FDR) applied for multiple testing correction. Absolute log₂ fold change (|log₂FC|) was set as the minimum threshold for defining significantly differentially expressed genes. High-confidence salt-responsive DEGs were filtered according to stringent unified criteria: log₂ (TPM + 1) ≥ 1.5, FDR < 0.05, and |log₂FC| ≥ 1. DEGs were identified separately within each dataset, and the final salt-responsive gene set was defined as the union of DEGs across all datasets. The overall distribution of upregulated and downregulated salt-responsive transcripts was visualized in a volcano plot using the ggplot2 package in R.

2.3.2. Candidate Gene Integration and Identification of MsCCCH20

We intersected the salt-responsive gene set (from Section 2.3.1) with genes flanking GWAS-associated loci for candidate prioritization, then focused on CCCH-type zinc-finger proteins within this overlapping gene set, as these proteins have well-documented roles in stress-responsive post-transcriptional regulation. First, we performed functional annotation of protein-coding sequences using InterPro (https://www.ebi.ac.uk/interpro, accessed on 14 May 2026) to identify conserved CCCH zinc-finger domains [45]. To further validate sequence-level homology, BLASTP (version 2.17.0) searches were performed against the NCBI non-redundant (nr) protein database [46]. High-confidence homologs were retained only if they met all stringent filtering criteria: E-value ≤ 1 × 10⁻¹⁰, sequence identity ≥ 54%, and query coverage ≥ 98%. This process yielded a curated dataset of 77 CCCH20 protein sequences from representative plant species. Full-length amino acid sequences of these homologs were retrieved from public databases and aligned with ClustalW for phylogenetic analysis. The neighbor-joining (NJ) phylogenetic tree was constructed in MEGA 12 using pairwise p-distance, with branch reliability assessed via 1000 bootstrap replicates [47].

2.4. Co-Expression Analysis

Co-expression analysis was performed to identify transcriptional associations between the core candidate gene MsCCCH20 and salt-responsive genes. Pearson and Spearman correlation coefficients between MsCCCH20 expression and the integrated DEGs were calculated using the Hmisc package in R, based on log₂ (TPM + 1) expression profiles. Statistical significance was evaluated using p-values, followed by Benjamini–Hochberg false discovery rate (FDR) correction to account for multiple hypothesis testing. Gene pairs with |r| ≥ 0.70 and FDR ≤ 0.05 were defined as significantly co-expressed [48]. Loess smoothing and binning-based mean calculation were subsequently applied to characterize the quantitative relationship between correlation strength and salt-induced expression changes, and the results were visualized together with co-expression patterns. The resulting co-expression network was visualized using ComplexHeatmap and circlize in R [49,50]. Weighted gene co-expression network analysis (WGCNA) was performed separately for leaf (44 samples, ZM4/Beaver) and root (18 samples, ZM1/AT) transcriptomes. The optimal soft-thresholding power for network construction was determined independently for the two datasets, with the distribution of scale-free topology fitting index (R²) and mean connectivity across a gradient of soft power values (1 to 20) presented in Figure S2. For the leaf dataset, the optimal soft-thresholding power was set to 6 (scale-free R² = 0.87), while the optimal softPower for the root dataset was set to 8 (scale-free R² = 0.86); both satisfied the accepted scale-free topology criterion (R² > 0.85). The topological overlap measure (TOM) was calculated to assess network interconnectedness, and module eigengenes were correlated with the salt stress traits to identify biologically relevant modules.

2.5. Gene Structure and MEME Conserved Motif Analysis

To identify conserved cis-regulatory elements within the 3′ untranslated regions (3′UTRs) of target genes, de novo motif discovery was performed using the MEME Suite (v5.5.0) [51]. The analysis was conducted on 3′UTR sequences derived from salt-responsive co-expression gene sets, applying the ZOOPS (Zero or One Occurrence Per Sequence) model with motif widths ranging from 6 to 18 bp and a maximum of 10 motifs reported. The resulting position weight matrices (PWMs) were subsequently used as input for FIMO (Find Individual Motif Occurrences) to scan for occurrences of the identified motifs across 3′UTR sequences, with a significance threshold of p ≤ 1 × 10⁻⁵. In parallel, canonical AU-rich element (ARE) motifs were searched using MAST with a p-value cutoff of 0.001 and a sequence E-value < 10, allowing for both strand orientations and correcting for nucleotide background bias. The core motif sequence ATTTAATT (8 bp) was adopted based on previously defined ARE consensus sequences shown to mediate CCCH/TZF-dependent mRNA decay in planta [52,53].

2.6. Three-Dimensional Structure Prediction of Core Protein–RNA Complexes via AlphaFold2

Three-dimensional structures of MsCCCH20 protein in complex with RNA motifs derived from target gene 3′UTRs were predicted using an AlphaFold2-derived protein–RNA modeling framework [54]. The predicted motifs included both the canonical AU-rich element (ARE) and related AU/AG-rich sequences identified through de novo motif discovery, consistent with the known RNA-binding preference of CCCH zinc-finger proteins while potentially extending beyond the classical ARE repertoire. The model was run with num_recycles = 10, num_models = 5, and random_seed = 100,000. Model confidence was assessed using the predicted interface TM-score (ipTM), and complexes with ipTM ≥ 0.70 were considered high-confidence. The top-ranked complex structures were visualized with UCSF ChimeraX v1.11 [55]. To further confirm the biological relevance of these predicted protein-RNA interactions, we analyzed the co-expression relationship between MsCCCH20 and its predicted target genes.

3. Results

3.1. Genome-Wide Association Mapping Reveals Major Loci Underlying Salt Tolerance

We carried out genome-wide association analysis (GWAS) for salt tolerance traits using 28,874,693 high-quality biallelic SNPs retained after quality filtering. A total of 20 SNPs were significantly associated with salt tolerance, and these loci were unevenly distributed across chromosomes 2 to 8. Full association statistics for each significant locus—including LOD scores, p-values, and genetic effect estimates—are compiled in Table S1. Among these 20 loci, 12 (60%) were classified as high-confidence significant (SIG) signals with LOD scores ≥ 16, while the remaining 8 (40%) were designated as suggestive association signals, with LOD scores ranging from 3.42 to 7.32. The phenotypic variance explained (PVE) by a single locus ranged from 0.0796 to 0.7365. Notably, the SNP locus chr4_72075593 on chromosome 4 showed the strongest association with salt tolerance, attaining a LOD score of 40.24 and the lowest p-value of 5.71 × 10⁻⁴¹, and explained 46.45% of the phenotypic variance. In the multi-locus model, background genetic effects were accounted for by simultaneously incorporating multiple associated loci, thereby partially accounting for polygenic background effects without explicitly incorporating a kinship matrix. GWAS results are visualized in Figure 1. The QQ plot showed that observed p-values closely followed the expected null distribution, with only slight deviation in the extreme tail corresponding to true positive associations. The genomic inflation factor (λGC) was 1.038 (Figure S1), indicating effective control of population stratification and polygenic background effects.

3.2. Identification of Candidate Genes

We extracted a total of 454 annotated genes within the LD regions of the 20 significant salt tolerance-associated GWAS loci, the full gene list and genomic coordinates are provided in Table S2. To narrow down high-confidence salt stress-responsive candidates, the 454 GWAS-extracted genes were filtered using the PlantASRG framework (score > 0.85) and intersected with the salt-responsive DEG set, yielding 59 candidate genes with functional annotations directly related to plant abiotic stress responses (Figure 2). GO enrichment analysis revealed that the most significantly enriched biological process terms were associated with ion transport, including positive regulation of potassium ion transmembrane transport (GO:1901381) and positive regulation of calcium ion transmembrane transport (GO:1904427). Additional significantly enriched BP terms included regulation of calcium ion transport into the cytosol (GO:0010522), negative regulation of the ethylene-activated signaling pathway (GO:0010105), gibberellin biosynthetic process (GO:0009686), cellular response to redox state (GO:0071461), regulation of the stress-activated MAPK cascade (GO:0032872), and the jasmonic acid-mediated signaling pathway (GO:0009867). Among molecular function terms, pectinesterase inhibitor activity (GO:0046910) was significantly enriched, a function linked to the maintenance of cell wall stability under salt stress. All of the annotated genes were listed in Table S3. In the cellular component category, candidate genes were significantly enriched in the basolateral plasma membrane (GO:0016323), the primary site for ion transmembrane transport and stress signal perception in plant cells. Collectively, these enriched terms position ion transport, hormone signaling, and membrane-associated processes at the center of the salt tolerance response mediated by the identified candidate genes.

3.3. Screening of Core Salt-Stress Responsive Genes

To pinpoint core candidates within the GWAS-associated intervals, transcriptomic data from three independent alfalfa salt-stress RNA-seq datasets (PRJNA952537, PRJNA1182104, and PRJNA685277) were integrated for multi-omics screening. A total of 737 salt-responsive DEGs were identified based on unified differential expression criteria, covering both upregulated and downregulated transcripts (Figure 3). Detailed correlation statistics for these DEGs are listed in Table S4. Intersection analysis was then performed between the 59 functionally annotated GWAS candidate genes and this DEG set. Only one gene, MsaG034610, was shared across all three datasets. Functional annotation revealed that this gene participates in the jasmonic acid-mediated signaling pathway (GO:0009867) and encodes a CCCH-type zinc-finger protein homologous to Arabidopsis AtC3H14. Based on its domain architecture and sequence homology, the gene was designated MsCCCH20. Phylogenetic analysis further confirmed the classification, placing MsCCCH20 in a well-supported clade with orthologous CCCH20 proteins from closely related legume species (Figure S3). Accordingly, MsCCCH20 was defined as the core candidate for downstream regulatory and structural analyses.

3.4. Co-Expression Network Analysis

Correlation-based screening using the salt-responsive gene set identified 110 genes with expression patterns significantly correlated with MsCCCH20 (Figure 4). Of these, 35 genes showed consistent positive correlations and concordant salt-induced expression changes, suggesting coordinated transcriptional behavior with MsCCCH20 under salt stress. Trend fitting further confirmed the most robust association between co-expression strength and salt-induced upregulation amplitude occurred at Pearson correlation coefficients of 0.7 to 1.0 (Figure 5A,B). MsaG027898 exhibited the strongest association with MsCCCH20 (Pearson r = 0.9546, adjusted p = 6.58 × 10⁻¹⁷; Spearman r = 0.7995, adjusted p = 8.19 × 10⁻⁵), marking it as a prominent downstream component of the regulatory network. In parallel, we performed weighted gene co-expression network analysis (WGCNA) to resolve tissue-level network architecture. After dynamic tree cutting and module merging, we identified 12 co-expression modules from the leaf dataset (Figure 6A,B) and 9 modules from the root dataset (Figure 6C,D). We then performed module eigengene-trait correlation analysis to screen for salt stress-responsive modules. Results showed two modules in leaf tissue (ME#94B3AE and ME#A997A5) and one module in root tissue (ME#C1B398) were significantly and positively correlated with salt treatment (Figure 7A,B). To further narrow down high-confidence downstream targets of MsCCCH20, we cross-compared these salt-responsive modules with the 110 salt stress-upregulated core genes significantly co-expressed with MsCCCH20. This analysis identified 5 overlapping genes in the leaf modules and 36 hub genes in the root-specific ME#C1B398 module. We retained 35 high-confidence targets from the combined analyses as the prioritized gene set for downstream cis-element analysis and protein–RNA structural modeling.

3.5. 3′UTR Motif Analysis

De novo motif discovery across the 3′UTR sequences of the 35 core salt-responsive genes identified 10 conserved motifs (6–18 bp). Three motifs exhibited strong statistical enrichment: MEME-1 (18 bp, E = 2.4 × 10⁻⁴) was present in all 35 genes (100%) and displayed a prominent AU-rich composition, consistent with the known RNA-binding preference of CCCH-type proteins. MEME-2 and MEME-5 were also recovered at significant enrichment levels. FIMO scanning confirmed the broad distribution of these motifs, detecting 94 high-confidence occurrences across 34 of the 35 genes (97.1%). We found several motifs were significantly enriched in specific genes, most notably in MsaG028907 and MsaG028910 (Table S5, Figure S4). When we mapped the positional distribution of these motifs, a clear enrichment peak centered 73.8 bp upstream of the stop codon was found (Figure 8). This distribution pattern is a classic feature of functional 3′UTR elements that mediate RNA–protein interactions. In parallel, MAST scanning for canonical AU-rich elements (AREs) identified 8 genes (22.9%) carrying high-confidence ARE-like motifs (E < 0.05), including MsaG003381, MsaG003549, MsaG015210, and MsaG045786. All these genes had clustered AU-rich repeats, which further confirms that the genes may regulate by protein MsCCCH20 carry evolutionarily conserved elements for post-transcriptional control.

3.6. Structural Prediction of MsCCCH20–RNA Interactions

Three-dimensional (3D) structural models of MsCCCH20–mRNA complexes were constructed using the AlphaFold2 (AF2) pipeline to characterize protein–RNA binding specificity. A total of 26 high-confidence complexes were retained, with interface predicted TM (ipTM) scores ranging from 0.70 to 0.79 (Table S6). The predicted interfaces exhibited low global predicted distance errors (gPDE), ranging from 1.27 to 1.36 Å with a mean of 1.30 Å, indicating high structural convergence and supporting the reliability of the modeled binding poses. In the optimal conformations, the conserved CCCH zinc-finger domain of MsCCCH20 consistently contacted AU/AG-rich elements within the 3′UTRs of target transcripts. Representative bound motifs included AGAAAGAGAAG (MsaG003983), UGGCAGCAACUGUGGCUG (MsaG010372), and UAUGUUGCCCACCAC (MsaG003381). These structural findings align with the 3′UTR motif enrichment results, in which AU/AG-rich motifs were identified as the predominant conserved elements across the target gene set. We further prioritized 8 core target genes for visualization, all with a ipTM score > 0.75: MsIAA22D (MsaG003381) and MsDRM3 (MsaG003983); MsPPR96 (MsaG016168), MsDUF3774 (MsaG010372), and MsGRP5 (MsaG009317); MsFBK-like (MsaG009782) (Figure 9). To evaluate whether the AF2-predicted target genes are transcriptionally coordinated with MsCCCH20 beyond random expectation, we compared the observed co-expression strength against a randomized background. The median Pearson correlation coefficient between MsCCCH20 and the 26 predicted targets was calculated, then benchmarked against 100 randomly sampled gene sets of equal size drawn from the transcriptome. The observed target set exhibited significantly stronger co-expression than the random controls (Figure 10), providing independent transcriptome-level support for the AF2-predicted interactions.

4. Discussion

Zinc-finger proteins are one of the largest regulatory protein families in plants. Most well-characterized members in plant stress research belong to the C2H2-type subfamily, which act as sequence-specific transcription factors. These proteins have been widely reported to regulate salt stress responses by modulating ion homeostasis, osmotic balance, and ROS detoxification pathways. For example, TaZFP1 acts as a transcription factor by binding promoter cis-elements to activate stress-responsive genes, while HvZFP1 confers salt tolerance mainly by regulating ROS scavenging systems [56,57].

In contrast, CCCH-type zinc-finger proteins are increasingly recognized as RNA-binding proteins engaged in post-transcriptional regulation. Tandem CCCH zinc-finger proteins in Arabidopsis, exemplified by AtTZF1, have been shown to specifically bind AU-rich elements within the 3′ untranslated regions of target mRNAs and modulate their stability and turnover. These TZF proteins are predominantly related to cytoplasmic processing bodies and stress granules, where they are reported to regulate mRNA decay and translational repression [58,59,60]. Direct experimental evidence has linked this RNA-binding activity to stress tolerance: AtTZF1 enhances salt tolerance by binding ARE/URE motifs in the 3′UTR of target transcripts such as ACA11, promoting their degradation and thereby modulating Ca²⁺ signaling and downstream stress responses [61]. CCCH-mediated post-transcriptional control appears to be evolutionarily conserved across plant species and is closely integrated with phytohormone signaling. Cotton GhC3H20, for example, enhances salt tolerance through interactions with key components of the ABA pathway [26], illustrating how CCCH proteins couple hormonal cues with stress adaptation (Figure 11).

Consistent with prior observations that CCCH/TZF proteins participate in plant stress responses, our multilayer evidence converges on MsCCCH20—a CCCH-type zinc-finger protein homologous to Arabidopsis AtC3H14—as a core candidate regulator of salt tolerance in alfalfa. Compared with single-omics or partially integrated studies, the main strength of our work lies in a stepwise integrative pipeline that connects population-level genetic signals to mechanistically interpretable post-transcriptional regulation. Motif analysis revealed enrichment of AU/AG-rich elements within the 3′UTRs of co-expressed target genes, suggesting that MsCCCH20 may recognize non-canonical ARE-like motifs. This observation aligns with emerging evidence that plant RBPs can engage diverse 3′UTR cis-elements beyond classical AREs, expanding the regulatory repertoire of post-transcriptional control [58]. Together with our multi-omics evidence, motif data, and structural predictions, we propose a working model for MsCCCH20-mediated salt stress adaptation in alfalfa.

The predicted downstream target genes of MsCCCH20 are significantly enriched in three canonical salt stress response pathways, providing functional support for our proposed post-transcriptional regulatory model. Specifically, targets related to ion homeostasis and auxin signaling (e.g., IAA22D, DRM3) are likely to participate in the maintenance of K⁺/Na⁺ balance during salt stress [62]. Genes involved in antioxidant defense (e.g., PPR96, DUF3774, GRP5) may contribute to scavenging reactive oxygen species and alleviating oxidative damage [63]. Additionally, the ubiquitination-related gene FBK-like has been previously reported to regulate stress-induced proteostasis [64]. Collectively, this regulatory network extends the current understanding of alfalfa salt stress responses from the widely reported transcriptional level to the understudied post-transcriptional regulatory layer.

Genome-wide association studies (GWAS) have been widely applied in alfalfa to dissect the genetic architecture of agronomic traits, and this has led to the identification of numerous loci linked to yield, forage quality, and developmental characteristics over the past decade [35]. Despite this progress, the functional connection between genetic variation and abiotic stress tolerance—specifically, the causal molecular mechanisms that translate genotype into adaptive phenotype—remains poorly defined in this species. Current approaches for mining alfalfa salt-tolerance genes largely fall into four categories, each with distinctive strengths but also notable limitations. The first, single-marker GWAS, has successfully associated loci with root ion homeostasis and highlighted the central role of Na⁺/K⁺ balance [65]. High-quality genome assemblies and large-scale resequencing have further enabled trait dissection in this complex autotetraploid [35]. GWAS alone, however, only establishes statistical associations between genomic regions and phenotypes; it does not resolve causal links between loci, genes, and biological functions, a gap that is especially problematic in autotetraploid alfalfa where extended linkage disequilibrium produces large mapping intervals harboring dozens to hundreds of genes. Pure transcriptome-driven gene discovery, the most widely adopted alternative, has provided a genome-wide view of the salt-stress transcriptome and revealed core response pathways such as ion transport, ROS scavenging, and hormone signaling [66]. While this strategy has provided a genome-wide characterization of the alfalfa salt-stress transcriptome landscape, most identified DEGs represent downstream stress-response genes rather than upstream regulatory drivers, resulting in a low success rate for subsequent functional validation. To overcome this limitation, integrative strategies that combine GWAS with DEG filtering have emerged as the current mainstream paradigm, narrowing candidate lists by overlapping statistically associated loci with stress-responsive genes [36,66]. Even in recent integrative multi-omics studies, candidate genes are often prioritized based on statistical overlap or co-expression relationships, while their direct regulatory targets, binding specificity, and post-transcriptional effects remain largely uncharacterized. Consequently, a fragmented understanding of stress-response pathways persists. Gene-family-centered reverse genetics, often combined with conserved motif analysis in species such as Zea mays L., alfalfa, and Populus trichocarpa L. [67,68,69], has contributed valuable insights into the evolution and functional diversification of stress-responsive proteins, but this approach starts from known gene families and is therefore inherently constrained in discovering novel regulatory factors, particularly those that function at the post-transcriptional level outside of classical transcription factor families.

Taken together, all these mainstream approaches share a critical limitation: they fail to establish a direct, causal link between genetic loci, regulatory molecules, and their downstream molecular interactions. This gap is especially pronounced in the field of RNA regulation, where genome-wide identification of RNA-binding protein (RBP) targets and their cognate binding motifs remains technically challenging, and the causal relationship between RNA binding activity and phenotypic outcomes is rarely fully resolved [61]. To bridge this gap, we designed a multi-layered integrative analysis pipeline that extends beyond current strategies. Unlike conventional GWAS or DEG-based screening strategies, which almost exclusively converge on transcription factors or downstream effector genes, our framework pinpointed a CCCH-type RBP, MsCCCH20, as the core candidate regulator. This class of post-transcriptional regulators has been largely overlooked by traditional gene-mining approaches in alfalfa. By combining co-expression network analysis with 3′UTR motif enrichment, our pipeline provides convergent computational evidence linking this genetically supported candidate regulator to putative RNA targets. Our work expands the current understanding of alfalfa salt stress regulation from a predominantly transcriptional paradigm to a more comprehensive framework that incorporates post-transcriptional control, ultimately helping to bridge the long-standing gap between genetic association and molecular mechanism. The findings of this study provide usable resources for alfalfa salt-tolerance molecular breeding. The identified salt-tolerance associated loci and core candidate regulator MsCCCH20 may develop into functional molecular markers for marker-assisted selection, effectively addressing the breeding bottlenecks caused by alfalfa’s complex autotetraploid genome. The multi-omics integrative pipeline established here also offers a methodological framework for functional gene mining in alfalfa and other polyploid forage legumes. Future research will first focus on the in vivo functional validation of MsCCCH20 via gene editing and overexpression assays to verify its regulatory role in alfalfa salt tolerance and its post-transcriptional regulatory mechanism on target transcripts. We will also validate the selection efficiency of the identified significant loci in large alfalfa breeding populations, to support the development of salt-tolerant alfalfa varieties.

It is important to acknowledge a key limitation of our study: the MsCCCH20-centered regulatory model we put forward is based on integrated multi-omics and structural prediction analyses and has not yet been validated by direct wet-lab experiments. The specific RNA targets we predicted (including R3 and ITR-region functional genes), the binding preference of MsCCCH20 for AG/AU-rich motifs, and its downstream regulatory effects on transcript stability all need to be verified with rigorous biochemical and molecular assays. These include targeted tests like RNA immunoprecipitation (RIP), genome-wide CLIP-seq to map in vivo binding sites, and dual-luciferase reporter assays to validate its post-transcriptional regulatory activity.

5. Conclusions

We developed and applied a reproducible, multi-layer integrative pipeline—combining GWAS, multi-study RNA-seq, co-expression/WGCNA, 3′UTR motif discovery, and AlphaFold2 protein–RNA modeling—to generate post-transcriptional regulatory hypotheses for salt tolerance in autotetraploid alfalfa. This framework nominated MsCCCH20, a CCCH-type zinc-finger protein, as a core candidate regulator. A set of 35 high-confidence putative targets was prioritized based on convergent evidence from GWAS overlap, co-expression, and enrichment of conserved AU/AG-rich motifs in 3′UTRs. AlphaFold2-derived structural models provided sequence-informed, testable predictions of MsCCCH20 binding interfaces that co-localize with de novo motifs enriched upstream of stop codons. All mechanistic inferences are presented as hypotheses requiring experimental validation. A prioritized validation roadmap—encompassing binding assays, 3′-end isoform mapping, and reporter-based perturbation experiments—is proposed to transition from association to causality. The candidate genes and regulatory motifs reported here offer a focused starting set for breeding and mechanistic studies aimed at improving alfalfa salt tolerance, and they may serve as input features for genomic prediction pending empirical assessment.