Integrated BSA-Seq and RNA-Seq reveals genes associated with salt tolerance at the germination stage in foxtail millet (Setaria italica (L.) P. Beauv.)

Abstract

Background

Salt stress has a significantly negative impact on growth, development, and yield of foxtail millet. Identifying and utilizing genes involved in salt tolerance for breeding is one of the effective ways to address this issue. However, no quantitative trait loci (QTLs) or genes have been identified in foxtail millet using a forward genetics approach. Therefore, it is essential to mine salt tolerance QTLs and candidate genes to further clarify the genetic basis of this complex trait in foxtail millet.

Results

In this study, an F_8:9 recombinant inbred line (RIL) population derived from a cross between Hongmiaozhangu (HMZG, salt-tolerant) and Changnong 35 (CN35, salt-sensitive) was utilized to identify salt tolerance QTLs at the germination stage via bulk segregant analysis sequencing (BSA-seq) using 20 extreme salt-tolerant and 20 extreme salt-sensitive RILs. Two QTLs related to relative germination potential (RGP) under salt stress, qRGP6-1 (19,520,000–20,590,000) and qRGP6-2 (20,610,000–20,900,000), were detected on chromosome 6 (Chr. 6) by ∆SNP-index/∆InDel-index and Euclidean Distance (ED) algorithms. In the mapping region, 46 candidate genes were identified. Two of these genes, Seita.6G119500 and Seita.6G119700, were differentially expressed between the parental lines under salt stress, as revealed by RNA-sequencing (RNA-seq) and validated by quantitative real-time PCR (qRT-PCR). Among them, Seita.6G119500 encodes a BURP domain-containing protein that has been reported to be involved in abiotic stress responses. Furthermore, a G/A SNP was detected between HMZG and CN35, leading to a nonsynonymous amino acid substitution from arginine (R) to histidine (H) in the salt-sensitive parent CN35. Association analysis using the Kompetitive Allele-Specific PCR (KASP) marker developed from the SNP variation within Seita.6G119500 revealed that the marker genotype was associated with RGP across the RIL population.

Conclusions

This study identifies a candidate gene Seita.6G119500 associated with salt tolerance in foxtail millet and provides a molecular marker for breeding salt-tolerant cultivars via marker-assisted selection (MAS).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-09046-z.

Background

Soil salinity is a major abiotic stress that threatens agricultural production and sustainable development [1]. It inhibits plant growth and development, particularly during the germination stage, which ultimately compromises seedling establishment and reduces crop yield [2]. According to a recent FAO estimate, the global area of salt-affected soils amounts to 1.381 × 10⁹ ha, accounting for approximately 10.7% of the total land area [3]. Identifying salt-tolerant genes and studying the genetic mechanisms of salt tolerance are promising strategies for breeding crop cultivars adapted to saline soils.

Salt tolerance of plants is a quantitative trait controlled by multiple genes [4]. Salt stress induces a series of secondary responses, such as calcium ion (Ca²⁺) signaling, reactive oxygen species (ROS) accumulation, and hormone signaling cascades. Ultimately, these early signaling pathways activate cellular homeostatic responses, such as ion transport, osmotic adjustment, and other salt tolerance mechanisms [5, 6]. These regulatory pathways mutually interact, forming complex regulatory networks [7]. Currently, numerous salt tolerance-related genes have been identified in plants. For example, genes associated with ion transport include OsHKT1;5 [8–10], ZmHKT1;2 [11], TaHKT7-A2 [12], ZmSOS1 [13], OsHAK21 [14], ZmHAK4 [15], etc. Genes associated with osmotic adjustment include ZmSOTs [16], AtP5CS1 [17], etc. Other salt tolerance genes include OsWRKY53 [18], OsBBX11 [19], OsCYBDOMG1 [20], OsSTH1 [21], TaSRO1 [22], and GmHSFB2b [23].

Foxtail millet (Setaria italica (L.) P. Beauv.) is a climate-resilient species capable of growing in harsh environments including saline soils [24, 25]. Abundant germplasm resources with genetic and phenotypic diversity could facilitate gene mapping and elite allelic variation discovery [26]. Via reverse genetics approach, to date, some genes associated with salt stress response have been identified in foxtail millet. Overexpression of SiCBL5, SiRAV1, SiDi19-3 and SiLEA14 enhances salt tolerance, whereas SiPLATZ12 negatively regulates salt tolerance [27–31]. Additionally, heterologous expression of SiMYB19 in rice and of SiNF-YC2 in Arabidopsis improves resistance to salt stress in the corresponding transgenic plants [32, 33]. By forward genetics approach, however, no QTLs/genes conferring salt tolerance have been successfully identified in foxtail millet.

In this study, BSA-seq was utilized to identify QTLs for salt tolerance at the germination stage using an F_8:9 RIL population derived from a cross between HMZG (salt-tolerant) and CN35 (salt-sensitive) in foxtail millet. Two QTLs associated with RGP under salt stress, qRGP6-1 and qRGP6-2, were detected on Chr. 6. Within the candidate region, two differentially expressed genes (DEGs) were identified between the parental lines under salt stress by RNA-seq. The qRT-PCR results showed the same expression patterns as the RNA-seq. Genomic sequence alignment revealed there was a SNP (G/A) in Seita.6G119500 but no SNPs or InDels in Seita.6G119700 between the parental lines. Notably, Seita.6G119500 encodes a BURP domain-containing protein involved in abiotic stress responses. Accordingly, Seita.6G119500 is identified as a candidate gene. Furthermore, we developed a KASP marker based on this SNP within Seita.6G119500, and validated that this marker was associated with RGP phenotypes in the RIL population. These results laid a foundation for studying the mechanism of salt tolerance regulation, as well as for developing salt-tolerant cultivars through molecular breeding.

Results

Salt-tolerant traits evaluation

In the present study, HMZG (salt-tolerant) and CN35 (salt-sensitive) were crossed to develop an F_8:9 RIL population containing 234 lines. The germination potential (GP), germination rate (GR), shoot length (SL), root length (RL), and their relative values were evaluated under salt stress and control conditions at the germination stage. Only the relative germination potential (RGP) followed a normal distribution, whereas the relative germination rate (RGR), relative shoot length (RSL), and relative root length (RRL) showed skewed distribution (Fig. 1c). Notably, RGP demonstrated an extremely significant difference (P < 0.001) between CN35 (56.83%) and HMZG (89.69%) (Fig. 1a and Additional file 1: Table S1). Furthermore, the RIL population displayed transgressive segregation for this trait, with RGP values ranging from 16.28% to 98.37% and a coefficient of variation of 0.28, indicating substantial phenotypic diversity (Additional file 1: Table S1). The absolute values of skewness and kurtosis were both less than 1, indicating that the data of RGP were suitable for subsequent QTL analysis (Additional file 1: Table S1). Thus, based on the RGP distribution, 20 lines with the highest RGP values (73.33%-93.27%) were selected as the tolerant pool (T-pool), and 20 lines with the lowest RGP values (24.14%-46.27%) were selected as the sensitive pool (S-pool) (Figs. 1b and 2a). To further validate the accuracy of the T-pool and S-pool, we compared activities of superoxide dismutase (SOD), peroxidase (POD), and catalase (CAT) between them. The results demonstrated that antioxidant enzyme activities of the T-pool were significantly higher than those of the S-pool under salt stress (Fig. 2b, c, and d).

Fig. 1.

Phenotypic evaluation of salt tolerance in the RIL population at the germination stage. A Comparisons of relative germination potential (RGP), relative germination rate (RGR), relative shoot length (RSL), and relative root length (RRL) between salt-tolerant parent HMZG and salt-sensitive parent CN35. Two-tailed Student’s t-test was used for significance analysis. *, **, *** indicate P < 0.05, P < 0.01, P < 0.001, respectively; ns indicates no significant difference. B Dot plots of RGP values in T-pool and S-pool. *** indicate P < 0.001 via two-tailed Student’s t-test. C Frequency distribution of four traits among 234 RILs. Arrows represent phenotypic values of two parents. Fitted normal distribution curves and Shapiro-Wilk normality test results are displayed. P > 0.05 denotes normal distribution of phenotypic data

Fig. 2.

Validation of two pools for BSA-seq. A Germination phenotypes of five randomly selected lines from two pools after 4 d of treatment with 180 mmol/L NaCl and under control conditions. Red dots highlight germinated seeds. Scale bars: 1 cm. B-D Activities of antioxidant enzymes, including (b) superoxide dismutase (SOD), (c) peroxidase (POD), and (d) catalase (CAT), in T-pool and S-pool. *, **, ***, **** indicate P < 0.05, P < 0.01, P < 0.001, and P < 0.0001, respectively, according to two-tailed Student’s t-test

Whole-genome resequencing

A total of 59.29 Gbp clean data were obtained from two parents and two pools, and all of the Q30 reached over 85%. The average sequencing depths were 41× for CN35, 23× for HMZG, 38× for the T-pool, and 38× for the S-pool. The average mapped ratio and genome coverage of the four samples were 98.40% and 97.08%, respectively (Additional file 1: Table S2). A total of 1,015,504 SNPs and 165,660 InDels were detected between CN35 and HMZG; 327,828 SNPs and 55,572 InDels were detected between S-pool and T-pool (Additional file 1: Table S3 and Table S4).

BSA-seq analysis

We utilized two algorithms, ∆SNP-index/∆InDel-index and Euclidean Distance (ED), to detect genomic regions associated with RGP. A total of 11 SNP-associated regions (Fig. 3a and c) by ∆SNP-index and ED and 7 InDel-associated regions (Fig. 3b and d) by ∆InDel-index and ED were detected on Chr. 6. Two QTLs, temporarily named qRGP6-1 and qRGP6-2, were identified by analyzing overlapping regions. They spanned physical intervals of 1.07 Mb (19,520,000–20,590,000 bp) and 0.29 Mb (20,610,000–20,900,000 bp), respectively (Table 1). Through sequence analysis within the mapping interval, 1,174 SNPs and 126 InDels were identified between CN35 and HMZG, of which 22 SNPs and 0 InDels resulted in amino acid changes. Within the qRGP6-1 and qRGP6-2 regions, 38 and 8 genes were predicted, respectively, using the Yugu1 reference genome (https://phytozome-next.jgi.doe.gov/info/Sitalica_v2_2) (Additional file 1: Table S5).

Fig. 3.

QTL mapping. A ΔSNP-index algorithm. The colored dots represent the calculated SNP-index (or ΔSNP-index) values, and the black line represents the fitted SNP-index (or ΔSNP-index) values. B ΔInDel-index algorithm. The colored dots represent the calculated InDel-index (or ΔInDel-index) values, and the black line represents the fitted InDel-index (or ΔInDel-index) values. Red, blue, and green lines denote the 99%, 95%, and 90% confidence thresholds in both (a) and (b). C ED algorithm for SNP. The colored dots indicate the ED value for SNP and the black line shows the fitted ED value. D ED algorithm for InDel. The colored dots indicate the ED value for InDel and the black line shows the fitted ED value. Red dashed lines in (c) and (d) indicate the association threshold (median + 3 SD)

Table 1.

Distribution of QTLs associated with salt tolerance

QTLs	Chromosome	Start (bp)	End (bp)	Size (Mb)
qRGP6-1	6	19,520,000	20,590,000	1.07
qRGP6-2	6	20,610,000	20,900,000	0.29

RNA-seq statistics

To identify DEGs between the two parental lines, we performed RNA-seq on a total of 12 samples, including HMZG and CN35 seedlings collected at 4 d after salt stress and control conditions, with three biological replicates per treatment. A total of 80.20 Gb of clean data were generated, with each sample yielding at least 5.70 Gb. The clean reads per sample ranged from 38.22 to 55.58 million, with sequencing depths ranging from 14× to 20×. The Q30 score, which indicates a 99.9% base call accuracy, was above 97.46% for all libraries. The GC content ranged from 51.65% to 54.47% (Additional file 1: Table S6). The clean reads were mapped to the Yugu1 reference genome, with mapping rates ranging from 89.11% to 95.41%. The uniquely mapped reads ranged from 85.86% to 91.65% (Additional file 1: Table S7). In summary, the sequencing data were of high quality and suitable for subsequent differential expression analysis.

DEGs identification

Next, we constructed four comparisons: CN35 vs. HMZG, TCN35 vs. THMZG, CN35 vs. TCN35, and HMZG vs. THMZG. The first two comparisons aimed to identify varietal differences under the same condition, while the latter two aimed to identify responses to salt stress within the same variety. By restricting |log₂(Fold Change)| ≥ 1 and FDR < 0.01, a total of 2208, 1563, 3912, and 3405 significant DEGs were identified in CN35 vs. HMZG, TCN35 vs. THMZG, CN35 vs. TCN35 and HMZG vs. THMZG, respectively. In CN35 vs. HMZG, 1434 DEGs were significantly up-regulated and 774 significantly down-regulated (Fig. 4a), while 904 DEGs up-regulated and 659 DEGs down-regulated in TCN35 vs. THMZG (Fig. 4b). Furthermore, in CN35 vs. TCN35, 2066 DEGs were up-regulated and 1846 down-regulated, while 1864 DEGs up-regulated and 1541 DEGs down-regulated in HMZG vs. THMZG (Fig. 4c and d). Apparently, more DEGs were identified in salt-sensitive line than in the salt-tolerant line under salt stress. In total, 7031 DEGs were detected in four comparisons. Furthermore, 9.39% (660/7031), 3.46% (243/7031), 23.04% (1620/7031), and 15.06% (1059/7031) of the DEGs were uniquely present in the comparisons of CN35 vs. HMZG, TCN35 vs. THMZG, CN35 vs. TCN35, and HMZG vs. THMZG, respectively (Fig. 4e). Obviously, both salt-tolerant and salt-sensitive varieties exhibited a large number of unique genes under salt stress.

Fig. 4.

Identification of DEGs under salt stress. A-D The volcano plots for the following comparisons: (a) CN35 vs. HMZG, (b) TCN35 vs. THMZG, (c) CN35 vs. TCN35, and (d) HMZG vs. THMZG. The X-axis and Y-axis represent the log₂(Fold Change) and -log₁₀(FDR), respectively. Red, green, and black dots represent significantly up-regulated, down-regulated, and non-differentially expressed genes, respectively. E Venn diagram illustrating unique and overlapping DEGs across the four comparisons

GO enrichment analysis

GO enrichment analysis was performed to elucidate the enriched functions of the DEGs in four comparisons. A total of 1545 (69.97%), 1044 (66.79%), 2957 (75.59%), and 2538 (74.54%) DEGs were assigned GO terms in CN35 vs. HMZG, TCN35 vs. THMZG, CN35 vs. TCN35, and HMZG vs. THMZG, respectively. Notably, DEGs identified in CN35 vs. TCN35 were assigned more GO terms than those in HMZG vs. THMZG. The most significantly enriched GO terms in the biological process (BP) were cellular process and metabolic process, indicating that germinating foxtail millet seeds under salt stress exhibited broad metabolic activities. In the cellular component (CC) category, the cellular anatomical entity was the most significantly represented group, implying that diverse cellular structures and compartments were actively involved in the response to salt stress. The GO terms mostly annotated in the molecular function (MF) category were binding and catalytic activity, suggesting that germinating foxtail millet seeds under salt stress engaged in extensive molecular interactions and biochemical reactions (Additional file 2: Figure S1).

KEGG pathway enrichment analysis

To further study the enriched metabolic pathways of the DEGs, we performed KEGG pathway analysis on the DEGs shared by CN35 and HMZG in response to salt stress, i.e., the overlapping DEGs from CN35 vs. TCN35 and HMZG vs. THMZG. The results showed that pathways related to stress response, including oxidative phosphorylation, MAPK signaling pathway, starch and sucrose metabolism, glutathione metabolism, plant hormone signal transduction, and peroxisome, were significantly enriched in the two parental lines at 4 d after NaCl treatment (Fig. 5).

Fig. 5.

KEGG pathway enrichment analysis of the overlapping DEGs from CN35 and HMZG in response to salt stress. The top 20 enriched pathways are displayed

Candidate gene analysis

BSA-seq analysis revealed two QTLs, qRGP6-1 and qRGP6-2, with physical interval of 1.07 Mb and 0.29 Mb, respectively (Table 1). According to the Yugu1 reference genome (https://phytozome-next.jgi.doe.gov/info/Sitalica_v2_2), a total of 46 genes (38 for qRGP6-1 and 8 for qRGP6-2) were retrieved (Additional file 1: Table S5). To detect candidate genes, we conducted a comprehensive analysis based on BSA-seq and RNA-seq results and selected 7 salt-responsive DEGs within the candidate regions from the comparisons of HMZG vs. THMZG and CN35 vs. TCN35. Among these, two genes, Seita.6G119500 and Seita.6G119700, were differentially expressed between the two parents under salt stress (TCN35 vs. THMZG) (Additional file 1: Table S8). Therefore, the two DEGs were selected as candidate genes for further validation.

To confirm the reliability of the transcriptome data, we used qRT-PCR to further detect the expression levels of Seita.6G119500 and Seita.6G119700. Based on the RNA-seq results, Seita.6G119500 and Seita.6G119700 were both up-regulated in HMZG vs. THMZG and CN35 vs. TCN35, with a higher fold change in the salt sensitive parent (CN35) than in the salt tolerant parent (HMZG) (Fig. 6a). The relative trends in expression from qRT-PCR were all consistent with the RNA-seq data, but the absolute expression levels exhibited some differences (Fig. 6b). Furthermore, we randomly selected three salt-tolerant lines (high RGP) and three salt-sensitive lines (low RGP) to validate the expression of the two genes in the RIL population. Under the salt stress treatment, Seita.6G119500 and Seita.6G119700 were up-regulated in salt-sensitive lines, and no significantly differential expression was observed in salt-tolerant lines (Fig. 6c). These results demonstrated that Seita.6G119500 and Seita.6G119700 were significantly induced by salt stress, implying that they may be involved in salt response.

Fig. 6.

Identification and characterization of candidate genes. A, B Expression patterns of candidate genes Seita.6G119500 and Seita.6G119700 under salt stress, as determined by (a) RNA-seq and (b) qRT-PCR. Data are presented as log₂(Fold Change) for the comparisons HMZG vs. THMZG (black bars) and CN35 vs. TCN35 (gray bars). C Validation of candidate gene expression in three RILs with high RGP (RIL213, RIL257, RIL279) and three with low RGP (RIL41, RIL62, RIL70). D Gene structure of Seita.6G119500 and sequence alignment between HMZG and CN35. “Ref” indicates the Yugu1 reference genome. E Spatiotemporal expression profile of Seita.6G119500 across different tissues and developmental stages in Yugu1, obtained from Setaria-db

To further analyze genomic difference, we amplified and sequenced the genomic sequences (including approximately 2 kb promoter region) of both Seita.6G119500 and Seita.6G119700 from HMZG and CN35. Sequence alignment revealed that there were no differences in the cDNA sequence and promoter region of Seita.6G119700 between HMZG and CN35. For Seita.6G119500, one SNP (G to A) at + 398 bp was identified in CN35 compared with HMZG, resulting in a nonsynonymous substitution of arginine (R) to histidine (H) (Fig. 6d).

Interestingly, Seita.6G119500 was specifically expressed in seeds, with a particularly high expression level during seed germination (Yugu1, Setaria-db), implying its potential association with germination processes (Fig. 6e). Moreover, Seita.6G119500 encodes a BURP domain-containing protein, which has been reported to be involved in abiotic stress responses, including salt and drought stress (Additional file 1: Table S10). However, the functional annotation of Seita.6G119700 was “component of root urea treatment-specific coexpression subnetwork”, which is less directly associated with salt stress responses (Additional file 1: Table S5). Therefore, Seita.6G119500 was identified as a promising candidate gene.

KASP development and association analysis

To investigate the association between the SNP variation within Seita.6G119500 and RGP, we developed a KASP marker designated as Si6G119500 to genotype the two parents and 234 RIL lines, of which 207 were successfully genotyped. The population was grouped into three genotypic groups: A/A (CN35 genotype, n = 23), G/A (heterozygous, n = 78), and G/G (HMZG genotype, n = 106) (Fig. 7a and Additional file 1: Table S12). Simple linear regression revealed that the number of G alleles was positively associated with RGP, explaining 7.6% of the phenotypic variance (R² = 0.076, P < 0.0001) (Fig. 7b). One-way ANOVA indicated that genotype had a highly significant impact on RGP in the RIL population (F _(2,204) = 9.356, P < 0.001). Statistical analysis using a post-hoc Tukey’s HSD test revealed that RILs with the G/G genotype had significantly greater RGP (65.85%) than both the G/A (59.08%, P = 0.001) and A/A (57.41%, P = 0.007) genotypes. In contrast, no significant difference in RGP was observed between the G/A and A/A genotypes (P = 0.828) (Additional file 1: Table S13 and Table S14). These results indicated that the SNP within Seita.6G119500 is associated with salt tolerance at the germination stage in the RIL population.

Fig. 7.

Development of the KASP marker Si6G119500 and genotype–phenotype association analysis. A Fluorescence intensity scatter plot of KASP genotyping with marker Si6G119500 in the RIL population. Each dot denotes one RIL line, clustered into three distinct genotypes corresponding to the target SNP: A/A (CN35 genotype), G/A (heterozygous), and G/G (HMZG genotype). B Linear regression analysis between G allele dosage of the target SNP and RGP. R² = 0.076, P < 0.0001

Discussion

Soil salinity inhibits crop seed germination and hinders seedling establishment, ultimately resulting in decreased yield [34]. Given that salt tolerance in crops is a quantitative trait controlled by multiple genes, uncovering the underlying QTLs is fundamental to elucidating this complex trait. To date, some QTLs and genes associated with salt tolerance during seed germination have been identified in numerous crop species, such as rice, maize, soybean, and barley [35–38]. However, no QTLs or genes associated with salt tolerance have been identified during seed germination in foxtail millet using forward genetics. Therefore, it is essential to mine salt-tolerant QTLs and further pinpoint the underlying candidate genes.

Conventional map-based cloning and QTL fine-mapping are limited by high-density genetic map and the development of multiple near-isogenic lines (NILs). For instance, using traditional mapping approaches, a previous study required 11 generations to isolate LOC_Os06g10650, a major gene conferring salt tolerance in rice [39]. The strategy of combining BSA-seq and RNA-seq enables rapid identification of candidate genes of interest. For example, the salt-tolerant gene OsSAP16 was successfully identified through an F_2:3 population by a combined BSA-Seq and RNA-Seq analysis, which only required 3 years [35]. In this study, we identified two QTLs (qRGP6-1 and qRGP6-2) by BSA-seq, and rapidly detected two DEGs located within the candidate interval through integrated BSA-seq and RNA-seq analysis. Finally, we proposed one candidate gene based on qRT-PCR validation, sequence alignment, and functional annotation. Therefore, the combined BSA-seq and RNA-seq approach is an effective strategy to rapidly identify salt-tolerant genes.

The BURP protein family is a plant-specific family that has been implicated in various stress responses, as demonstrated by members such as OsRUBP12, OsRUBP16, AtRD22, AtUSPL1, ZmRD22A, and ZmRD22B [40–42]. Numerous studies have shown that the majority of BURP genes in rice exhibited expression changes in response to salinity stress [40]. Subsequent functional investigations have confirmed that some of these genes directly regulated salt tolerance. For example, overexpression of OsBURP16 increased sensitivity to salt, thereby reducing seedling survival [43]. The knockout of OsBURP12 enhanced salt tolerance in rice seedlings, accompanied by significantly higher antioxidant enzyme activity (SOD, POD, and CAT) compared to WT [44]. Consistently, physiological assays showed that under salt stress, antioxidant enzyme activities were significantly higher in the T-pool than in the S-pool in our study. Correspondingly, GO and KEGG enrichment analyses revealed that DEGs were predominantly associated with metabolic processes, catalytic activity, and antioxidant-related pathways, including oxidative phosphorylation, glutathione metabolism, and peroxisome, all of which were significantly enriched in both parental lines under salt stress. In this study, we proposed Seita.6G119500 as a candidate gene for salt tolerance during seed germination in foxtail millet. This gene encodes a BURP domain-containing protein, but its role in salt tolerance in foxtail millet remains to be elucidated. Further studies including genetic transformation or gene editing are required for functional validation.

KASP genotyping technology is widely employed for high-accuracy allele discrimination at target loci, greatly enhancing selection efficiency in crop breeding programs [45–47]. For instance, KASP markers were developed for efficient germplasm screening using salt tolerance-associated SNPs from a soybean GWAS [37]. Foxtail millet is an important crop cultivated in arid and semi-arid regions and exhibits higher salinity tolerance [48]. Therefore, there is an urgent need to develop molecular markers closely linked with salt tolerance genes for breeding. In this study, we developed a KASP marker, Si6G119500, based on the SNP in Seita.6G119500 and validated that it is associated with salt tolerance in the RIL population. Thus, this molecular marker has the potential for salt tolerance variety breeding via MAS in foxtail millet.

Conclusions

In this study, BSA-seq and RNA-seq were used to identify salt tolerance genes at the germination stage in a RIL population derived from a cross between the salt-tolerant parent HMZG and the salt-sensitive parent CN35. Two QTLs, qRGP6-1 and qRGP6-2, associated with salt tolerance at the germination stage in foxtail millet were mapped to intervals of 1.07 Mb and 0.29 Mb on Chr. 6, respectively. In the candidate region, there were two genes that showed significantly different expression between the two parents under salt stress. Based on the sequence analysis and functional annotation, Seita.6G119500 was identified as a candidate gene associated with salt tolerance trait RGP in foxtail millet. Additionally, we developed a KASP marker, Si6G119500, based on the SNP in Seita.6G119500, and validated its association with salt tolerance at the germination stage. Collectively, our findings provide a candidate gene and a molecular marker for MAS breeding of salt-tolerant varieties in foxtail millet.

Methods

Plant materials

Two extreme foxtail millet varieties, Hongmiaozhangu (salt-tolerant, introduced from the National Crop Germplasm Bank of China, with the bank accession number I1I18140) as the female parent and Changnong 35 (salt-sensitive, bred by the Shanxi Academy of Agricultural Sciences) as the male parent, were crossed to produce the F₁. An F₈ RIL population was subsequently generated by single-seed descent (SSD). All plants were grown in an experimental field in Changzhi (113° 08′ E, 36° 18′ N) in 2-meter-long rows with a row spacing of 0.33 m and a sowing density of 15 plants per meter. The leaves of 234 F₈ lines collected at the jointing stage were stored at -80℃ for DNA extraction, and the corresponding F_8:9 seeds were collected for further experiments.

Phenotypic evaluation of salt tolerance at the germination stage

To break seed dormancy, seeds were dried at 37℃ for 7 d. For each material, 1.5 g of seeds were randomly selected and surface-sterilized with 75% ethanol for 30 s, rinsed three times with sterile distilled water, then disinfected in 5% sodium hypochlorite (NaClO) for 20 min, followed by three final rinses with sterile distilled water. After sterilization, 50 seeds were placed in each Petri dish and treated with either 180 mmol/L NaCl solution or an equivalent volume of sterile distilled water (control). The Petri dishes were kept in a constant-temperature chamber at 28℃ in the dark for the first 2 d, and then transferred to a 14 h light /10 h dark photoperiod for an additional 5 d at the same temperature. Each treatment was performed with three biological replicates (each replicate consisting of 50 seeds in one Petri dish).

GP was recorded as the number of seeds germinated on day 4, and GR was recorded as the number of germinated seeds on day 7. Meanwhile, five seedlings were randomly selected from each replicate on the 7th day of treatment. Following the measurement of seedling length and root length, their average values were adopted to represent the SL and RL of each replicate. The relative values were calculated to evaluate salt tolerance at the germination stage as follows:

Significance analysis between parental lines for each salt tolerance trait was performed using a two-tailed Student’s t-test via GraphPad Prism 10, and the normality of trait distribution in the RIL population was assessed using the Shapiro-Wilk test via IBM SPSS Statistics 23.

Construction of T-pool and S-pool

Based on the frequency distribution of the RGP phenotypic data, we selected 20 extreme lines from each tail of the distribution to construct T-pool and S-pool. Seedlings of the selected extreme lines were harvested after 4 d exposure to 180 mmol/L NaCl or control conditions, flash-frozen in liquid nitrogen, and stored at -80℃. The activities of POD (Kit No. POD-2-Y), SOD (Kit No. SOD-2-W), and CAT (Kit No. CAT-2-W) were measured using commercial assay kits (Comin Biotechnology, Suzhou, China) following the manufacturer’s instructions.

BSA-seq analysis

Genomic DNA was extracted from the leaves of parent plants, HMZG and CN35, and from extreme RILs (20 with extreme salt tolerance and 20 with extreme salt sensitivity) using the sodium dodecyl sulfate (SDS) method with minor modifications [49]. Genomic DNA was randomly fragmented by ultrasonication and used to construct DNA libraries. Qualified libraries were sequenced using the Illumina NovaSeq X Plus platform (Illumina, CA, USA) at Biomarker Technologies Co., Ltd. (Beijing, China). Clean reads were obtained by filtering out low-quality reads. Then, these high-quality reads from both parents and the two DNA pools were mapped to the Yugu1 reference genome (https://phytozome-next.jgi.doe.gov/info/Sitalica_v2_2) using BWA software (v2.2) [50]. Detection and annotation of SNPs and InDels were performed by GATK (v3.8) [51] and SnpEff (v3.6c) [52], respectively.

The ∆SNP-index algorithm [53] was used for calculating the significant differences of genotype frequency between two pools. The ∆SNP-index was calculated as follows:

where M_T−pool and P_T−pool represent the read depths of T-pool derived from the HMZG and CN35, respectively; M_S−pool and P_S−pool represent the read depths of S-pool derived from the HMZG and CN35, respectively.

∆SNP-index values were fitted using the DISTANCE method. A sliding window size of 2 Mb with a step size of 10 kb was used to assess variations. We determined the statistical significance thresholds for the ∆SNP-index through computer simulations. Under the null hypothesis of no QTL, 10,000 ∆SNP-index datasets were simulated. Based on these simulated data, we calculated the 90%, 95%, and 99% confidence intervals for each SNP locus, which were used as thresholds to declare QTL significance. Genomic regions exceeding the 99% confidence threshold were identified as significant QTL candidates. The same simulation procedure was applied to determine the thresholds for the ∆InDel-index.

The ED algorithm [54] was further performed to identify SNPs and InDels associated with the salt tolerance trait. The ED value was calculated as follows:

where A, G, C, and T represent the frequencies of the corresponding bases in the T-pool and S-pool.

The original ED was raised to the fifth power to eliminate background noise and then fitted using the DISTANCE method. A sliding window size of 2 Mb with a step size of 10 kb was used. The correlation threshold was defined as the median + 3 SD of the fitted values at all sites, which was calculated as 0.53 for SNPs, and regions above this threshold were considered associated. The same ED method was applied to InDel analysis, and the threshold was calculated as 0.52.

Library construction, RNA sequencing, and alignment

Seedlings of HMZG and CN35 were collected at 4 d after treatment under 180 mmol/L NaCl and control conditions. Three samples per treatment were taken and a total of 12 samples were used for RNA-seq. Total RNA was extracted from these samples using RNAiso Plus (TaKaRa, Shiga, Japan) following the manufacturer’s protocol. RNA purity and concentration were determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher, DE, USA), and RNA integrity was detected by an Agilent 2100 Bioanalyzer (Agilent, CA, USA). Qualified samples were used for cDNA library construction. The concentration and quality of libraries were assessed using Qubit 3.0 Fluorometer (Thermo Fisher) and Agilent Bioanalyzer 2100 system (Agilent), respectively. Qualified libraries were sequenced on the Illumina NovaSeq X Plus platform (Illumina) using the paired-end (PE150) method at the Biomarker Technologies Co., Ltd. (Beijing, China).

Clean reads were obtained by filtering raw reads and removing adapters and low-quality reads. Quality scores of 30 (Q30, indicating detection accuracy > 99.9%) and GC contents of the clean reads were recorded. The clean reads were aligned to Yugu1 reference genome using HISAT2 (v2.0.4) [55] and then assembled by StringTie (v2.2.1) [56] software.

Identification and statistics of DEGs

After mapping and assembling, transcript level was quantified using Fragments Per Kilobase of transcript per Million mapped fragments (FPKM). DEGs were identified using DESeq2 (v1.30.1) [57]. The P values were adjusted using the Benjamini-Hochberg method to control the false discovery rate (FDR). Genes with |log₂(Fold Change)| ≥ 1 and FDR < 0.01 were considered differentially expressed.

Enrichment analysis

GO enrichment analysis of the DEGs was implemented by the clusterProfiler R package based on the Wallenius non-central hyper-geometric distribution, which can adjust for gene length bias. KEGG pathway enrichment analysis was performed using KOBAS database [58] and clusterProfiler. The significance for GO and KEGG pathway enrichment was set at q < 0.05.

qRT-PCR verification

PrimeScript II 1st strand cDNA synthesis kit (TaKaRa, Shiga, Japan) was utilized to synthesize cDNA from total RNA. The synthesized cDNA was then used as template for qRT-PCR amplification performed on a CFX96 Real-Time PCR System (Bio-Rad, CA, USA). Reaction volume of 10 µL comprised 5 µL TB Green Premix Ex Taq II (TaKaRa, Shiga, Japan), 1 µL template cDNA, 2 µL primers (2 µmol/L), and 2 µL ddH₂O. Thermal cycling conditions were as follows: 95℃ for 2 min, 39 cycles of 95℃ for 10 s and 60℃ for 30 s; followed by a melt curve analysis from 65 °C to 95 °C with an increment of 0.5℃/s. Relative gene expression was quantified using the 2^−ΔΔCt method, with SiACTIN (Seita.7G294000) as the internal reference gene. Primers were designed using Primer3 (https://bioinfo.ut.ee/primer3-0.4.0/) and are listed in Additional file 1: Table S9.

Candidate gene prediction

A total of 46 genes located in the two mapped QTL intervals (qRGP6-1 and qRGP6-2) were retrieved based on the Yugu1 reference genome from the Phytozome database (https://phytozome-next.jgi.doe.gov/info/Sitalica_v2_2). A filtering pipeline was constructed as follows:

1. Salt-responsive DEGs within the QTL intervals were screened from transcriptomic comparisons of HMZG vs. THMZG and CN35 vs. TCN35, yielding seven candidate DEGs.

2. Genes with significant differential expression between the two parental lines under salt stress (TCN35 vs. THMZG) were further prioritized, resulting in two candidate genes: Seita.6G119500 and Seita.6G119700.

3. Integrated analyses of genomic sequence variations, tissue-specific expression patterns, and functional annotation were performed for final candidate gene prioritization.

For sequence analysis, the candidate genes were amplified from HMZG and CN35 using TransStart^® FastPfu DNA Polymerase (TransGen, Beijing, China). Sequencing of genomic DNA was performed by Tianyi Huiyuan Biotechnology Co., Ltd. (Beijing, China). Sequence alignment was conducted using DNAMAN software. Functional annotation was retrieved from the Yugu1 reference genome. Tissue-specific expression patterns were obtained from the Setaria-db database (http://111.203.21.71:8000/multi-omics/transcriptome.html).

Ultimately, Seita.6G119700 was excluded due to the absence of genomic sequence variation between parental lines and limited functional association with salt stress response. Thus, Seita.6G119500 was identified as a candidate gene.

KASP marker development

The KASP marker, Si6G119500 (Additional file 1: Table S11), was developed based on the 1 bp SNP identified within Seita.6G119500 between HMZG and CN35. The marker was further used to genotype 236 accessions, including 234 lines from the RIL population and the two parental lines. Reaction volume contained 2.5 µL KASP master mix (Genepioneer, Nanjing, China), 1.25 µL template DNA, and 1.25 µL primers. Thermal cycling protocol consisted of 95℃ for 10 min; 10 cycles of 95℃ for 20 s and 61℃-55℃ (decreasing by 0.6 °C per cycle) for 60 s; 27 cycles of 95℃ for 20 s and 55℃ for 60 s; and 25℃ for 30 s.

Abbreviations

QTLs

Quantitative trait loci

RIL

Recombinant inbred line

HMZG

Hongmiaozhangu

CN35

Changnong 35

BSA-seq

Bulk segregant analysis sequencing

ED

Euclidean Distance

RNA-seq

RNA sequencing

qRT-PCR

Quantitative real-time PCR

KASP

Kompetitive Allele-Specific PCR

MAS

Marker-assisted selection

GP

Germination potential

GR

Germination rate

SL

Shoot length

RL

Root length

RGP

Relative germination potential

RGR

Relative germination rate

RSL

Relative shoot length

RRL

Relative root length

DEGs

Differentially expressed genes

ROS

Reactive oxygen species

SOD

Superoxide dismutase

POD

Peroxidase

CAT

Catalase

BP

Biological process

CC

Cellular component

MF

Molecular function

NILs

Near-isogenic lines

SSD

Single-seed descent

FPKM

Fragments Per Kilobase of transcript per Million mapped fragments

Authors' contributions

YX L and J W conceived and designed the research. YX L and L S performed sample preparation, conducted the experiments, and analyzed the data. XF D and ZL W performed material development. ML X, KN H, TT M, SC L, HP W, YF L, and YL Z participated in the discussion of the results. LY Z conducted field management. YX L drafted the manuscript. J W revised the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by the Shanxi Province Central-Guided Local Science and Technology Development Fund (YDZJSX2024D047), the earmarked fund for Modern Agro-industry Technology Research System (2025CYJSTX04-02), the Shanxi Province Basic Research Program (202203021212462, 202403021221083, 202203021222144), the Key Support Project of the National Natural Science Foundation of China (NSFC) Joint Fund (U24A20382), the Shanxi Province Agricultural Key Core Technology Project (NYGG19), the Self-established Project of Houji Laboratory (202404010930003-Y16), the Changzhi Key Laboratory Establishment and Support Program (2024sy005), the Changzhi Key Research and Development Program (2024ZDYF012).

Data availability

All datasets supporting the findings of this study are available within the article and its supplementary materials. Sequencing data have been deposited in the NCBI Sequence Read Archive (SRA). BSA-seq data are accessible through accession number PRJNA1314454 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1314454), and RNA-seq data through accession number PRJNA1321414 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1321414).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.