The sequences of the nine identified AtCSase genes were obtained from the NCBI database.¹ The M. sativa Xinjiangdaye genome sequence was downloaded from a website.² MsCSases were identified by two rounds of BLASTP. A hidden Markov model (HMM) was used by Pfam 31.1³ to ensure that the PF00291 domain was retained, and DNAMAN was used for sequence comparison searches for the PLP-binding site (PXXSVKDR) in alfalfa.

A phylogenetic tree was generated by MEGA 5 using the NJ method, with 1,000 bootstrap replicates. Multiple sequence alignments of CSases were created with ClustalX.

The conserved motif structures within the CSase sequences were identified by MEME Suite Version 5.2.0⁴ with the following parameters: zero or one occurrence per sequence of site distribution, a maximum of 10 misfits and a maximum width of motif between 6 and 50. NCBI Batch CD-Search⁵ was used to analyze the conserved domains of the CSase proteins, after which the domains were visualized by TBtools.

We used the CSase gene of M. truncatula (Medtr7g078070.1) as a reference sequence to clone the CSase gene of M. sativa. Transient expression vectors for tobacco and overexpression vectors for alfalfa were constructed by the one-step cloning method. Then, the expression vectors were transformed into Agrobacterium rhizogenes by using the freeze-thaw method. The CSase gene was transformed into alfalfa via Agrobacterium mediation using the cotyledon method, and regenerated alfalfa plants were obtained. The bar gene detection method and fluorescence quantitative analysis technology were used to screen overexpression plants. The sequences of the primers used are shown in Supplementary Table 1.

Nicotiana benthamiana plants were grown in plastic pots filled with vermiculite. Approximately 1-month-old seedlings were used for transient expression. M. sativa Longmu 801 was used in this study. Softwood cuttings from the CK line and overexpression lines OV#L11, OV#L12, and OV#L13 were transplanted into plastic pots containing vermiculite (one plant per pot). All the plants were grown under a 16 h light/8 h dark photoperiod under a day/night temperature cycle of 22°C/18°C. Hoagland solution (1/10 strength) was applied to the plants every 3 days.

For NaHCO₃ treatment, 150 mM NaHCO₃ was applied for 5 d, and a 5 d recovery period was selected as the best condition for identifying stress phenotypes. Samples were taken at 0, 1, 6, 12, 24, 48 h, and 5 d after the beginning of the treatment. Three biological replicates were included per line.

To explore the subcellular localization of CSase proteins, we constructed a transient expression vector. CSase gene was inserted downstream from the double CaMV 35S promoter in the pCAMBIA1300 vector. The pCAMBIA1300 vector carries GFP gene. The sequences of the primers used are listed in Supplementary Table 1. The resulting vector was introduced into the Agrobacterium tumefaciens strain GV3101. We used a syringe to infiltrate Agrobacterium tumefaciens strain GV3101 containing a tobacco transient expression vector into 1-month-old tobacco leaves. After infiltration, the plants were cultivated for 72 h under dark conditions. The fluorescence signal in the infested tobacco leaves was subsequently observed by confocal microscopy.

Physiological traits including GSH, cysteine, Pro, and MDA contents and SOD and POD activities were measured using reagent kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China). The instructions of the kits were followed for specific test procedures.

For qRT-PCR, total RNA was isolated from alfalfa samples using an RNeasy Plant Mini Kit (CWBIO, Jiangsu, China), and cDNA was synthesized using a kit (Vazyme). qRT-PCR was used to analyze the relative expression levels of alfalfa CSase-responsive genes. The GADPH gene was used as a reference. The sequences of the primers used are shown in Supplementary Table 1.

To determine significance, all statistical analyses were performed by using Microsoft Excel.

First, a total of 39 MsCSase gene sequences were retrieved from alfalfa using BLAST, PF00291 domain and PLP-binding site (PXXSVKDR) searches and named MsCSase01 to MsCSase39 according to their chromosome locations (Supplementary Table 2). They were unevenly mapped onto chromosomes 1, 4, 5, and 7, which contained 14, 9, 7, and 9 genes, respectively (Supplementary Table 2). Information about their coding DNA sequences (CDS) and resulting protein sequences are presented in Supplementary Table 3. Then, characterization of the proteins revealed that the predicted isoelectric points (pIs) of the MsCSase proteins ranged from 5.17 to 9.08 (Supplementary Table 2). Except for MsCSase01, MsCSase31 and MsCSase36, the length and molecular mass did not widely vary (Supplementary Table 2). The phylogenetic tree results demonstrated that the MsCSase proteins could be divided into 6 subfamilies according to the clades and classification from Arabidopsis, including 14, 9, 5, 4, and 4 members in the CysA subfamily, CysB subfamily, SCS subfamily, CysD subfamily and CysC subfamily, respectively (Figure 1). Similar to that which occurred in a study in tomato (Liu et al., 2018), MsCSase25, MsCSase27, and MsCSase28 were separated into a separate family and did not belong to the other five subfamilies. These results indicated that the characteristics and patterns of evolution in various species are more likely to differ.

Genetic structural diversity supported the phylogenetic groupings to some extent (Wei et al., 2016). Therefore, we analyzed the relationship between gene structure and phylogenetic clustering to gain insight into the evolution of the MsCSase gene family in alfalfa (Figures 2A,B). Gene structure analysis showed that genes within the same subfamily presented similar structures; for instance, the CysA subfamily members contained 10 exons, and the exon distribution of genes on the same branch was largely similar. In addition, the protein motif analysis (the conserved motifs via sequence logo are shown in Supplementary Figure 1) by MEME found a similar pattern and the order and distribution of the motifs were roughly similar among the members of the subfamilies (Figure 2C). These results supported the close evolutionary relationship of the classification of these MsCSase subfamilies. Moreover, the MsCSase genes had five characterized domains, including PLN02565, the PLN02565 superfamily, Trp-synth-beta_II superfamily and PLN02556 domains, all of which are related to cysteine synthesis (Figure 2D). Overall, the domain similarity suggested that these genes may have similar functions, but the differences in the activity of their encoded enzymes may be related to differences in gene structure and motifs between subfamily members.

To elucidate the mechanism through which the MsCSase gene family members in alfalfa expanded, gene duplication events were identified. Two pairs of tandem duplication genes (MsCSase9/10, MsCSase22/23) and 35 groups of synteny gene pairs in which MsCSase07/24 were segmentally duplicated genes were identified by TBtools and MCScanX software (Supplementary Table 4). Duplicated genes were located on chromosomes 1, 5, and 10 (Supplementary Figure 2). Taken together, the results indicated that there was no obvious relationship between chromosome length and the number of genes. Some MsCSase genes may have been generated by gene duplications and tandem and segmental duplications contributed to the evolution of MsCSases in alfalfa.

Using the Medtr7g078070.1 gene of Medicago truncatula as a probe, we cloned the gene with accession number MK334208 named CSase from alfalfa (cultivar Longmu 801). Sequence analysis showed that the amino acid sequences of MsCSas32, MsCSase34, and CSase were nearly identical (Supplementary Figure 3); they were 99.91% similar at the nucleotide level, and only one nucleotide differed between MsCSase34 and CSase (Supplementary Figure 4). Combining the results of the phylogenetic evolutionary tree analysis with these results, we determined that the CSase gene belonged to the CYS-C1 subfamily and was highly conserved in two different alfalfa varieties.

qRT-PCR analysis of different alfalfa tissue parts revealed that the relative expression of CSase was higher in the leaves than in the other tissues (Figure 3A). Moreover, the relative expression in mature leaves was much higher than that in other tissues. To further clarify the location of gene activity, we evaluated the subcellular localization of CSase. We fused its ORF sequence without the terminal codon to GFP at the N-terminus under the control of the CaMV 35S promoter and ultimately transiently expressed it in tobacco (N. benthamiana) leaf epidermal cells. The results showed that the presence of the CSase–GFP fusion protein in the chloroplasts of the cells (Figure 3B). This finding also directly validates the accuracy of the qRT-PCR results.

qRT-PCR analysis under different abiotic stresses revealed that CSase could respond positively to salt, alkali and drought stresses in alfalfa, and the response to alkali stress was more pronounced than that to the other stressors (Figure 4A). To determine whether alfalfa alkalinity tolerance was altered by up-regulation of CSase, we constructed alfalfa overexpression vectors (Supplementary Figure 5) and transformed the CDS of CSase into the alfalfa cultivar Longmu 801 to obtain CSase-overexpressing transgenic lines. The glucosamine gene is a marker gene that is present only in the vector itself and not in alfalfa. Bar gene detection revealed successful infestation of alfalfa in response to a bacterial solution (Supplementary Figure 6). qRT-PCR analysis showed that the relative expression of the MsCSase gene in the control (CK) lines were much lower than that in the OV# L11-, OV# L12-, and OV#L13-overexpressed lines (Figure 4B). This indicates that the CSase gene was successfully overexpressed in alfalfa. We ultimately selected the CK line and overexpression lines OV#L11, OV#L12, and OV#L13 for subsequent experiments.

Under normal growth conditions, there was no significant difference in phenotype between the CK and overexpression lines, and the differences between the aboveground biomass fresh weight and the relative chlorophyll content were not significant. However, after 5 d of alkali stress, the leaf wilting of the overexpression lines was much lower than that of the CK (Figure 4C). The alkali treatment led to degradation of the relative chlorophyll content (Figure 4D), but the chlorophyll contents were still higher in the overexpression lines than in the CK line. Moreover, the difference in aboveground biomass between the lines was not significant (Figure 4E). Overall, overexpressing CSase provided increased tolerance to alkali stress in alfalfa.

To clarify the reason for the increased alkali tolerance of the CSase overexpression lines, we also measured the contents of malondialdehyde (MDA), proline (Pro) and soluble sugars (SSs) in the CSase overexpression lines and CK plants grown under normal and alkaline conditions. However, after alkali stress, the contents of MDA, Pro and SSs increased within each line, and the increase in MDA content in the overexpression line was significantly lower than that in CK, while the contents of osmoregulatory substances containing Pro and SSs were significantly higher than those in CK (Figures 5A–C). It can be hypothesized that, compared with the CK plants, the overexpression lines are alkaline tolerant due to their lower degree of membrane damage and higher accumulation of osmoregulatory substances.

CSases have been shown to encode CSase proteins (Hatzfeld et al., 2000). The increase in CSase content promotes an increase in cysteine content in plants. In the present study, the cysteine content in the overexpression lines was significantly higher than that in CK under normal growth conditions, and the cysteine content under alkali stress was higher than that in the CK and tended to increase (Figure 6A). Accordingly, we speculated that cysteine plays a role in improving alkali tolerance in plants.

Based on our preliminary research, we found that cysteine is a precursor for the synthesis of antioxidant substances such as GSH (Droux, 2004; Wirtz and Droux, 2005; Kopriva, 2006) and that GSH plays an important role in enhancing antioxidant capacity and transducing redox-sensitive signals in plants (Cnubben et al., 2001; Matés et al., 2002; Pastori and Foyer, 2002). Therefore, we determined the GSH content and the relative expression of γ-glutamyl cysteine synthetase (γ-ECS) in each line. The results showed that the GSH content in the plants increased in response to alkali stress and that the overexpression lines contained more GSH (Figure 6B). The trends of the relative expression of γ-ECS in the overexpression and CK lines were similar, but the relative expression of γ-ECS was higher in the overexpression line than in the CK line (Figure 6C). Overexpression of CSase and SAT in tobacco also significantly increased the relative expression of γ-ECSs as well as the GSH content in the plants (Nakamura et al., 2014). Thus, the overexpression of the CSase gene could regulate downstream metabolic pathways, which led to an increase in the relative expression of the downstream γ-ECS gene, and the increase in cysteine content provided the possibility of an increase in GSH content. This also laid the foundation for the improvement in alkali tolerance in the overexpression line.

Studies on DES1 and OAS-A1 in Arabidopsis showed that cysteine is an important determinant of antioxidant capacity in the cytoplasm (Loìpez-Martiìn et al., 2008; Alvarez et al., 2010a), while SCS plays an important role in chloroplast redox (Bermúdez et al., 2010). In addition, cysteine and GSH are also associated with plant antioxidants. Therefore, we measured the superoxide anion (OFR) and hydrogen peroxide (H₂O₂) contents and the enzymatic activity of several antioxidant enzymes. Under normal growth conditions, although the H₂O₂ and OFR contents in the overexpression line were lower than those in the CK line, only the H₂O₂ content differed significantly (Figures 7A,B). The levels in both lines increased significantly after alkali stress, but the levels in the overexpression lines were significantly lower than those in the CK line. Thus, we speculated that the overexpression lines may be superior to the CK line in terms of the scavenging ability of reactive oxygen species (ROS).

Based on the above speculation, we also measured the activities of peroxidase (POD) and superoxide dismutase (SOD). The results showed that the activities of both enzymes were much higher in the overexpression lines than in the CK plants under alkali stress, and the activities of POD in the former were also significantly higher than those in the CK plants under normal growth conditions (Figures 7C,D). In conclusion, it was clear that the overexpression lines could reduce the accumulation of ROS by increasing the activity of antioxidant enzymes under alkali stress to enhance the alkali tolerance of the overexpression lines.