Integrated Transcriptomic and Physiological Analysis Reveals the Drought Adaptation Strategies of Sapindus delavayi, an Important Tree for Industrial Use in the Dry–Hot Valleys

Abstract

Sapindus delavayi is a drought-resistant tree species endemic to the dry–hot valleys of Southwestern China and is of great significance for soil and water conservation and ecological restoration. In this study, we sequenced the transcriptome of its leaves using the Illumina HiSeq 4000 platform and obtained 96.12 Gb of high-quality data (Q20 = 98.68%, Q30 = 95.62%), which were de novo assembled to obtain 89,228 unigenes (N50 = 1538 bp), of which 63,005 (70.61%) were successfully annotated to at least one database (NR, NT, SwissProt, KOG, KEGG, GO, Pfam). Overall, 53.96% of the unigenes in the S. delavayi leaves were annotated to Acer yangbiense, which belongs to the same family as S. delavayi. A total of 42,870 CDSs and 21,488 SSR loci were detected, with the highest mononucleotide repeat rate at 42.72% of the total number. Drought stress experiments identified 669–1203 differentially expressed genes (DEGs). Through our research, the first high-quality transcriptome database of S. delavayi has been constructed and its drought-resistance-related gene features have been analyzed, laying an important foundation for future functional gene mining, molecular marker development, molecular diversity studies, molecular breeding, and ecological adaptation research.

1. Introduction

Sapindus delavayi (Franch.) Radlk. is a large deciduous tree endemic to China, belonging to the genus Sapindus of the family Sapindaceae, and is distributed in the Yunnan, Sichuan, Guizhou, and Western Hubei provinces. It is commonly found in the dense forests of Central and Northwestern Yunnan and Southwestern Sichuan at an altitude of 1200–2600 m, and it is also a relatively common cultivated plant throughout the southwest of China [1]. It is an important forest tree species with a wide range of applications [2]. Due to the high oil content of its seeds, it can be used as a biodiesel feedstock that meets European and US standards, and its indices as an oilseed are higher than those of other tree species of the same genus, e.g., S. mukorossi [2,3]. Meanwhile, a variety of saponins and terpenoids can be isolated from the pericarp extract of S. delavayi, and researchers have noted that some extracts have high surface activity; thus, they have been used as a natural source of surfactants and exported from China to Japan as a substitute for S. mukurossi. Further research revealed that these compounds also have fungicidal, deworming, and medicinal properties [3,4]. S. delavayi is also known for its ornamental qualities and excellent wood, which shows resistance to pests and diseases and grows rapidly, and thus it can be used as an ornamental horticultural tree or for making furniture [1,5]. Aside from these properties, due to its straight trunk, well-developed root system, and remarkable resilience in acidic environments, it is considered to be a typical barren mountain and pollution-resistant tree species, showing remarkable resilience and potential in retaining soil and water [5]. Therefore, S. delavayi is an important multi-purpose cash crop in China with significant development value.

Drought is one of the major abiotic stress factors for plants and has been compared to plant “cancer” [6]. In recent years, due to a gradual increase in the global temperature, arid and semi-arid regions in mid-latitude areas have expanded in size. Concurrently, the problems of rocky desertification and soil erosion in the dry–hot valleys of Yunnan have become prominent [7]. These regions in Yunnan Province are in a mountainous area in the valley basin, with latitudes of 23°00′–27°21′ N and longitudes of 98°49′–103°23′ E. Due to its geographical latitude and valley topography, the region has a semi-arid climate, featuring persistent hot and dry weather and low precipitation [8]. All of these issues have constrained the socioeconomic development of local areas and the construction of the ecological environment [7]. At present, S. delavayi, as an indigenous tree species with outstanding drought resistance, has been planted in the rocky desertification area of Jianshui County, Yunnan, to better promote large-scale plantation and the reversal of rocky desertification [7,8].

The scientific investigation into S. delavayi has concentrated on the domains of cultivation and breeding, natural energy resources, interspecific variation, the impact of environmental factors, the composition of chloroplast genome sequences, and the assessment of genetic diversity [9,10,11,12]. Drought constitutes a substantial challenge to vegetation growth, thereby exerting a pivotal influence on the ecological rehabilitation and cultivation of S. delavayi. Thus, in this study, S. delavayi was considered as the research object, and Illumina HiSeq 4000 high-throughput sequencing technology was used to perform transcriptome sequencing, data filtering, and assembling; then, a large number of unigenes obtained with bioinformatics methods, such as functional annotations, metabolic pathways, CDS prediction, and SSR detection, were analyzed to understand the gene expression and overall functional characteristics of S. delavayi under drought stress, such as gene groups and metabolic pathways, and to provide a reference for future rational development and utilization of this precious native tree species.

2. Materials and Methods

2.1. Experimental Material

The experiments were conducted in the artificial climate chamber (BIC-250, Shanghai Boxun, Shanghai, China) at the Key Laboratory of Resource Insect Cultivation and Utilization of the State Forestry and Grassland Administration. The artificial climatic chamber was sterilized with 75% ethanol one week before the experiment, and the Petri dishes used for the seeds were autoclaved at 120 °C for 20 min prior to use. The deshelling process for S. delavayi seeds presented a significant challenge due to the hardness of the seed shells. To enhance the seed germination rate, a combination of mechanical deshelling and chemical treatment was necessary. The chemical treatment process involved initial immersion for 24 h in concentrated sulfuric acid, followed by submersion for 24 h in purified water. A total of 120 S. delavayi seeds were selected for the experiment and were collected from the wild populations in Yuanmou County, Yunnan Province, China. All the seeds were subjected to a sterilization process involving 0.1% KMnO₄ for 10 min. Following this treatment, the seeds were thoroughly washed and placed in Petri dishes containing a minimal amount of distilled water. The seeds were then transferred to an artificial climate chamber maintained at 25 °C, with a relative humidity of 75%, under a 12 h light/12 h dark cycle and with an intensity of 15,000 lx. Germination was induced under these conditions. After the germination of the seeds, the sprouted seedlings were transferred to polyethylene pots (20 cm × 20 cm × 25 cm) filled with a sterilized 3:1:1 mixture of nutrient soil, perlite, and vermiculite. Prior to cultivation, the substrate underwent a process of sterilization. Post-transplant, the plants were situated within an artificial climate chamber, with a temperature of 22–25 °C, 65% humidity, 16 h of light, 8 h of darkness, and a light intensity of 2000 lx. Following a 45 d period of standard growth, the weaker seedlings were removed to ensure that the height and growth of each pot remained essentially unchanged. The seedlings in the pots were treated after 90 d of growth, when they had reached approximately 10 cm in height and exhibited slight lignification of the stalks. Young leaves exhibiting comparable quality (5–6 leaves) were meticulously collected from each plant specimen and promptly stored in liquid nitrogen at −80 °C.

2.2. Drought Stress Experiment

S. delavayi specimens were cultivated under standard conditions prior to the stress treatment. These plants were irrigated at regular intervals of 1–2 d, and the pots were thoroughly hydrated prior to the experiment and weighed. The frequency of watering each time following the standard treatment (CK) was equivalent to the mitigation applied in the middle of the previous day’s spray. The mild drought (D1) treatment involved regular watering for the first 10 d and a cessation of watering for the subsequent 5 d. Moderate drought (D2) was characterized by regular watering for the first 5 d, followed by a 10 d period without watering. In the case of severe drought (D3), there was no watering treatment for the entire 15 d. The initial 10 d watering treatment corresponded to the mitigation applied in the middle of the previous day’s spray. Following the drought treatment, each drought treatment group underwent a 5 d rewetting process, as illustrated in Figure 1. Leaf samples from the same plant were collected on the 15th and 5th days of rewatering following the drought stress treatment. In addition, healthy leaves from mature branches were collected at 9:00 a.m. on a regular basis. Each treatment was replicated three times. After labeling, the leaves were immediately transferred to liquid nitrogen for quick freezing and stored at −80 °C for future use.

2.3. RNA Extraction and Transcriptome Sequencing of S. delavayi

S. delavayi leaf drought stress transcriptome (Supplementary Table S1) samples were subjected to total RNA extraction ( TRIzol RNA extraction kit, Invitrogen, Carlsbad, CA, USA), and the extracted total RNA was quality checked with Nanodrop 2000 Spectrophotometer (Thermo Fisher, Waltham, MA, USA) and Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The extracted total RNA was then stored at −80 °C as a backup. The construction of the library and the subsequent sequencing of the transcriptome were entrusted to Shenzhen Huada Gene Service Co., Ltd. (Shenzhen, China) using an Illumina HiSeqTM4000 platform, which generates 150 bp paired-end reads.

2.4. Transcriptomic Data Processing Analysis

The obtained raw data (raw reads) were filtered to remove low-quality splice contamination, splice sequences, fuzzy reads (N > 10%), and raw reads (more than 50% of bases with Q ≤ 20) to obtain high-quality clean reads. In this study, BGI Genomics’s self-developed filtering software SOAPnuke (version v1.4.0) was used to perform filtering. In the absence of a reference genome, the de novo assembly of clean reads was performed using Trinity (v2.0.6). The quality of the assembly was then assessed using BUSCO (v5.4.4). The obtained clean data were compared to the reference gene sequences using Bowtie2 software (v2.2.5) combined with RSEM software (v1.2.8) to calculate the expression levels of the genes and transcripts, which were normalized using FPKM. Candidate coding regions were identified in the unigenes with Transdecoder software (v3.0.1), and the coding regions were predicted using BLAST (v2.14.0) with SwissProt (http://www.ebi.ac.uk/uniprot/ (accessed on 17 April 2024)) and by searching for the homologous sequences of Pfam proteins with Hmmscan. Unigenes were detected using MISA (v1.0), followed by primer design for the detected SSRs using Primer3 (v2.2.2), after which expression was calculated on an All-Unigenes basis for each sample. Based on the NR annotation results, the Gene Ontology (GO) annotation of unigene data was performed using Blast2GO (v4.1.9 ), and GO functional classification was performed using WEGO (v2.0). All non-redundant unigenes obtained were compared to NR (ftp://ftp.ncbi.nlm.nih.gov/blast/db (accessed on 17 April 2024)), NT (ftp://ftp.ncbi.nlm.nih.gov/blast/db), KEGG (http://www.genome.jp/kegg (accessed on 23 April 2024)), KOG (https://www.ncbi.nlm.nih.gov/COG/ (accessed on 17 April 2024)), SwissProt, GO (http://geneontology.org (accessed on 23 April 2024)), and Pfam (http://pfam.xfam.org (accessed on 17 April 2024)) for a total of seven protein databases.

2.5. Differentially Expressed Genes

The expression level of the unigenes was assessed using the reads per kb per million reads (FPKM) method, and the differential expression analysis of S. delavayi samples was performed using DESeq2. If the FPKM value of the S. delavayi gene was ≥0.1, the gene was considered to be expressed in the sample, and all eligible All-Unigenes results of the analyzed genes were used for the identification of differentially expressed genes (DEGs). The KEGG database was searched using KOBAS (v3.0) to identify the significantly enriched KEGG metabolic pathways (p ≤ 0.05); GO enrichment analysis was performed using the TermFinder package, and time-series analysis was performed using Mfuzz software (v2.34.0).

2.6. Weighted Gene Co-Expression Network

Co-expression network modules were generated using 23,409 unigenes (the mean RPKM of all samples ≥ 1 and the coefficient of variation (CV) > 0.1). Data were analyzed using the R package WGCNA with the following parameter settings: softThreshold: 6; minModuleSize: 100; mergeCutHeight: 0.3; and default values were used for other parameters. Relevant genes identified in the WGCNA module of S. delavayi in response to drought stress were mapped to Arabidopsis thaliana’s protein sequences to construct a protein–protein interaction (PPI) network through the STRING (v11.5) online database (https://string-db.org/ (accessed on 12 May 2025)). Enrichment analysis was performed for each gene module significantly associated with the trait; then, the GO terms and KEGG pathways significantly enriched in the core module were found.

2.7. Quantitative Real-Time PCR

To validate the RNA-seq results, nine DEGs and Hub genes responsive to drought stress were randomly selected from the S. delavayi transcriptomic data for qRT-PCR analysis. The samples were sent to Shanghai Shenggong Biotechnology Co., Ltd., Shanghai, China, for testing, and the relative expression levels of the genes were calculated using the 2^−∆∆Ct method.

3. Results

3.1. Analysis of Sequencing and Assembly Results

The sequencing of all 15 sample transcriptomes yielded 686.83 × 106 raw reads, which were subsequently filtered to obtain a total of 640.77 × 106 raw reads and 96.12 Gb of high-quality clean reads. The average percentages of Q20, Q30, and the clean base were 98.68%, 95.62%, and 93.42%, respectively (Supplementary Table S2). After de novo assembly and redundancy removal, 89,228 unigenes were obtained, with a total length, average length, N50, and an average GC content of 80.59 × 106 bp, 903 bp, 1538 bp, and 41.57%, respectively (Supplementary Table S3). The distribution of the sequence lengths was highest in the 200–500 bp interval, after which the number of unigenes decreased as the lengths increased (Supplementary Figure S1).

3.2. Analysis of GO and KEGG Annotation Results

The unigenes were compared to seven major functional databases for annotation, and finally 59,509 (NR: 66.69%), 48,965 (NT: 54.88%), 41,391 (SwissProt: 46.39%), 36,342 (KOG: 40.73%), 42,591 (KEGG: 47.73%), 47,250 (GO: 52.95%), and 38,216 (Pfam: 42.83%) unigenes were functionally annotated, of which 63,005 (70.61%) unigenes were successfully annotated to at least one database (Supplementary Figure S2).

To enhance the understanding of the S. delavayi All-Unigenes function, its GO function was analyzed (Figure 2). Among the 89,228 unigenes of S. delavayi, 47,250 (52.95%) unigenes were annotated to the GO database. The functional categorization of GO entries was divided into three categories, i.e., biological process (BP), cellular component (CC), and molecular function (MF), with annotations accounting for 42.49%, 23.57%, and 33.94%, respectively. These categories were further divided into 45 subclasses, with BP representing 44.44% (23 subclasses), CC 4.44% (2 subclasses), and MF 51.11% (20 subclasses). Within BP, cellular process (34.09%) and metabolic process (27.35%) were the most represented, while biological adhesion (0.0045%) and viral process (0.0034%) were the least represented. The CC category consisted solely of the cellular anatomical entity (88.50%) and protein-containing complex (11.50%). For MF, binding (44.65%) and catalytic activity (41.17%) were the dominant subclasses; cargo receptor and toxin activities were both low, accounting for 0.0028%. In summary, S. delavayi had a high abundance of gene expression in cellular activities and binding, metabolic processes, and catalytic reactions, indicating that it has a high intrinsic metabolic capacity.

The S. delavayi unigene sequences were annotated on the KEGG database; the results showed that 42,591 unigenes (47.73%) were annotated, and the involved metabolic pathways could be categorized into five classes and nineteen subclasses (Supplementary Figure S3). The five metabolic pathway categories were metabolism (63.29%), organismal systems (5.95%), cellular processes (4.34%), genetic information processing (19.68%), and environmental information processing (6.73%). Metabolism was subdivided into 11 subcategories, with global and overview maps (41.62%) having the highest percentage and glycan biosynthesis and metabolism (3.34%), the metabolism of terpenoids and polyketides, (3.26%), and nucleotide metabolism (2.52%) having low percentages. Biological systems and cellular processes only had one subcategory: environmental adaptation and transport and catabolism. Genetic information processing was categorized into four subcategories: folding, sorting, and degradation (31.49%), replication and repair (12.50%), transcription (19.19%), and translation (36.82%). There were only two subclasses of environmental information processing: membrane transport (15.52%) and signal transduction (87.18%). In summary, the total KEGG annotation results indicated that S. delavayi metabolic activities and genetic information processing were higher in this period.

3.3. Analysis of the Annotation Results of Related Databases

The homologous species classification map of the S. delavayi leaf unigenes (Nr, E ≤ 1 × 10⁻⁵) (Supplementary Figure S4) showed that 32,212 (53.96%) unigenes were annotated to Acer yangbiense, followed by Pistacia vera (9.57%), Citrus sinensis (5.12%), C. clementina (2.90%), C. unshiu (2.90%), and others (26.19%). Due to the long-term lack of information on the genome and transcriptome of S. delavayi, a certain number of unigenes could not be annotated in the Nr library.

The S. delavayi unigene sequences were annotated to the KOG database to obtain the categorization information of the gene homologs. The results showed that 36,342 unigenes (40.73%) were annotated into 25 KOG functional classifications (Supplementary Figure S5), and the gene expression enrichment of each functional classification varied, containing a total of 36,342 functional annotations, basically covering most of S. delavayi’s life activities. Among them, general function prediction was the largest category, containing 7160 genes worth of functional annotation data, accounting for 19.70% of the total functional annotation information in the KOG database. Signal transduction mechanisms (10.27%), post-translational modifications, protein turnover, chaperones (10.20%), and transcription (5.59%) followed this category. However, extracellular structures (0.26%), nuclear structure (0.085%), and cell motility (0.011%) were poorly represented. However, there were 2194 genes with unknown functions, accounting for 6.04% of the total functional annotation information in the KOG database.

In the classification of the structural domains of plant resistance genes (PRGs), the RLP component had the highest number of entries with 1785 unigenes (27.33%) (Supplementary Figure S6), followed by NL (18.78%), CNL (13.81%), N (13.30%), and TNL (11.83%).

In the classification of transcription factor (TF) families to which the genes belong (Figure 3), the MYB component had the most entries with 210 unigenes (11.24%), followed by bHLH (7.44%), AP2-EREBP (7.01%), NAC (6.58%), C3H (5.35%), WRKY (4.87%), and bZIP (1.39%).

Candidate coding regions in the unigenes were identified using TransDecoder software. The longest open reading frame was extracted, and the coding regions were predicted by searching for the homologous sequences of Pfam proteins through Blast comparison with the SwissProt database and Hmmscan, and a total of 42,870 coding sequences (CDSs) were obtained, with a total length of 380,665,535 nt (Figure 4). The usual length of the CDSs was 400–2000 nt, and their maximum length, minimum length, N50 value, N90 value, and GC content were 7086 nt, 255 nt, 1107 nt, 444 nt, and 44.08%, respectively.

From the S. delavayi transcriptome sequencing data, a total of 21,488 SSR loci were retrieved (Supplementary Table S4), of which 9179 were mononucleotide repeats, 6842 were dinucleotide repeats, 4703 were trinucleotide repeats, 195 were tetranucleotide repeats, 250 were pentanucleotide repeats, and 319 were hexanucleotide repeats. The base type with the highest repetition rate was mononucleotide with 42.72% of the total number, followed by mononucleotide with 31.84% of the total number.

3.4. Analysis of Differentially Expressed Genes (DEGs)

3.4.1. Quantitative Expression Analysis of S. delavayi Genes Under Different Drought Stresses

Gene expression was calculated using the FPKM method, and a gene was considered to be expressed in the sample if the FPKM was ≥0.1. A total of 47,231 genes were expressed in the CK, D1, D3, R1, and R3 treatment groups (Figure 5a), and 2090, 1481, 864, 1339, and 1692 genes were expressed uniquely in CK, D1, D3, R1, and R3, respectively. Among them, 406, 283, 612, and 532 genes were expressed in CK vs. D1, CK vs. D3, D1 vs. R1, and D3 vs. R3, respectively. Gene expression in each S. delavayi treatment group was mostly concentrated between an FPKM of 1 and 10 (Figure 5b), though there were still some highly expressed genes (FPKM ≥ 10).

3.4.2. DEGs of S. delavayi Under Different Drought Stresses

By comparing drought and rehydration treatments at the different periods of drought stress, 863 (CK vs. D1), 1203 (CK vs. D3), 799 (D1 vs. R1), and 669 (D3 vs. R3) DEGs were obtained (Figure 6). Under drought stress, a total of 374 downregulated genes and 488 upregulated genes were identified in CK vs. D1, and 598 downregulated genes and 585 upregulated genes were identified in CK vs. D3. After rehydration, a total of 358 downregulated genes and 382 upregulated genes were identified in D1 vs. R1, and 230 downregulated genes and 404 upregulated genes were identified in D3 vs. R3 (Supplementary Figure S8). Venn diagrams (Supplementary Figure S9) showed that there was only one DEG in the four treatments, and 402, 709, 644, and 487 stage-specific DEGs in CK vs. D1, CK vs. D3, D1 vs. R1, and D3 vs. R3, respectively. There was a total of 330 DEGs in the CK vs. D1 and CK vs. D3 treatments, and the genes under both treatments had similar expression patterns.

3.5. GO and KEGG Enrichment Analysis of DEGs

In order to gain insights into the functional classes of the drought stress-induced DEGs, the GO enrichment analysis of differentially expressed genes (DEGs) in the leaves of S. delavayi under drought stress treatments was performed. In the CK vs. D1 treatment group under drought stress (Figure 7A), the most abundant GO category was the integral component of the membrane, followed by nucleus, ATP binding, zinc ion binding, and cytoplasm. In CK vs. D3 (Figure 7B), significant changes were revealed in the integral component of membrane, followed by ATP binding, cytoplasm, DNA binding, and plasma membrane. Under drought stress, DEGs in both CK vs. D1 and CK vs. D3 drought treatment groups were significantly enriched in the pathways of the integral component of the membrane, ATP binding, and cytoplasm.

After rehydration, in D1 vs. R1 (Figure 7C), the most abundant GO category was nucleus, followed by ATP binding, cytoplasm, DNA binding, and cytosol. In D3 vs. R3 (Figure 7D), the most abundant GO category was nucleus, followed by ATP binding, metal ion binding, DNA binding, and protein kinase activity. After rehydration, DEGs in both rehydrated groups, D1 vs. R1 and D3 vs. R3, were significantly enriched in the pathways of the nucleus, DNA binding, and ATP binding.

The KEGG enrichment analysis of differentially expressed genes (DEGs) in S. delavayi leaves under drought stress treatment was performed. The results showed that, under drought stress, DEGs in CK vs. D1 were mainly enriched in the MAPK signaling pathway—plant, carbon metabolism, the biosynthesis of cofactors, plant–pathogen interaction, and nucleocytoplasmic transport. DEGs in CK vs. D3 were mainly enriched in ribosome, plant hormone signal transduction, the biosynthesis of cofactors, and the MAPK signaling pathway—plant. Under drought stress, DEGs were significantly enriched in the MAPK signaling pathway—plant and the biosynthesis of cofactors pathway in both drought treatment groups, CK vs. D1 and CK vs. D3. The plant MAPK signaling pathway is activated by reactive oxygen species (ROS) produced by plants under drought stress, which in turn regulates gene expression to adapt to a stressful environment.

After rehydration, DEGs in D1 vs. R1 were mainly enriched in the biosynthesis of amino acids, starch and sucrose metabolism, carbon metabolism, the biosynthesis of cofactors, and protein processing in the endoplasmic reticulum. DEGs in D3 vs. R3 were mainly enriched in the biosynthesis of amino acids, the biosynthesis of cofactors, carbon metabolism, amino sugar and nucleotide sugar metabolism, and glycolysis/gluconeogenesis (Supplementary Figure S10). After rehydration, DEGs were significantly enriched in the pathways of carbon metabolism, the biosynthesis of cofactors, and the biosynthesis of amino acids in both rehydrated treatment groups, D1 vs. R1 and D3 vs. R3. The GO analysis of the treated group after rehydration showed that DEGs were significantly enriched in the nucleus, DNA-binding, and ATP-binding pathways. This indicated that, after rehydration, S. delavayi showed significant changes in gene expression related to gene expression regulation, DNA repair, and energy metabolism.

3.6. Cluster Analysis of DEGs

The dynamic molecular response of S. delavayi under drought stress was elucidated by categorizing the expression trends of S. delavayi DEGs. The DEGs were grouped into 12 clusters under drought treatment. The clusters of genes whose DEG expression levels initially increased under drought stress but then decreased after rehydration were Cluster1, Cluster3, Cluster6, Cluster7, and Cluster11. These genes are mainly involved in metabolic and biosynthesis-related processes such as lipoic acid metabolism, purine metabolism, pyrimidine metabolism, lysine biosynthesis, ubiquinone and other terpenoid–quinone biosynthesis, and N-glycan biosynthesis. The clusters of genes whose DEG expression levels were significantly upregulated under drought–rehydration treatments were Cluster4, Cluster10, and Cluster12. These genes are mainly involved in metabolic degradation and plant rhythm-related processes such as sphingolipid metabolism, arginine and proline metabolism, limonene and pinene degradation, and circadian rhythm—plant. Two clusters of genes whose DEG expression levels were initially downregulated under drought stress but then rebounded after rehydration were Cluster2 and Cluster9. These genes are mainly involved in cellular metabolic degradation and photosynthesis-related processes such as carbon metabolism; zeatin biosynthesis; valine, leucine, and isoleucine degradation; photosynthesis; carbon fixation in photosynthetic organisms; and photosynthesis antenna proteins. The clusters of genes whose DEG expression levels were significantly downregulated under drought–rehydration treatments were Cluster5 and Cluster8, which are mainly involved in biosynthesis and biometabolism-related processes, such as terpenoid backbone biosynthesis, flavonoid biosynthesis, and glycerophospholipid metabolism (Supplementary Figures S11 and S12).

3.7. Transcription Factors

Under drought stress–rehydration, there were 25 categories of gene families that were enriched, and the TF families with a high number of enrichments (≥7) were NAC (17), bHLH (11), AP2-EREBP (10), MYB (9), C3H (8), and SBP (7), while the rest of the TF families were enriched with less than or equal to three enrichments. TFs such as MYB, bHLH, NAC, and other family genes play key roles in plant resistance to adversity. In this case, the identification of the TFs involved in this regulation was important for analyzing the molecular mechanism of drought resistance in S. delavayi. From this purpose, we hypothesized that transcription factors such as NAC, bHLH, AP2-EREBP, MYB, C3H, and SBP may enhance the expression of some drought critical genes or the association between related proteins in response to drought stress in S. delavayi, thus improving its adversity defense ability (Figure 8).

3.8. Weighted Gene Co-Expression Network (WGCNA)

To further understand the regulatory mechanisms of S. delavayi in response to drought stress, physiological indicators were used as trait files, and 23,409 unigenes were analyzed using WGCNA for co-expression network analysis; the WGCNA soft threshold (β) was determined, and β = 5 was used for power treatment to create co-expression networks (Figure 9). A total of 13 co-expression modules were identified based on similar expression patterns of the genes. Most of the genes (4135) belonged to the turquoise module, followed by the blue module (2741), brown module (1896), and yellow module (1660), and the rest of the modules had less than 1500 genes (Figure 10).

To identify the key modules and related genes in S. delavayi in response to drought stress, we screened the related modules with |cor| ≥ 0.6 and p < 0.05 in the module physiological index correlation analysis. Supplementary Figure S13 demonstrates the correlation heatmap of the identified thirteen co-expression modules with 28 physiological indices of S. delavayi in the association analysis, and a total of three corresponding modules were screened, namely, the blue, red, and magenta modules. Analyzing the gene significance (GS) and module membership (MM) relationships between the genes and physiological indicators in the three candidate modules, we found that the blue module was positively correlated with chlorophyll b (Chlb) and intercellular CO₂ concentration (Ci), the magenta module was positively correlated with soluble proteins (SPs), and the red module was negatively correlated with transpiration rate (Tr).

Among the WGCNA modules of genes of drought stress-treated S. delavayi, the magenta, blue, and red modules were selected for GO and KEGG analyses. GO enrichment analysis showed that the integral component of the membrane, nucleus, ATP binding, cytoplasm, and metal ion binding were all the most significantly enriched biological processes in the magenta, blue, and red modules. The genes in the magenta, blue, and red modules regulate the response of S. delavayi to drought stress by participating in the metabolism of some secondary substances, binding metal ions, and controlling the structure and function of the cell membrane and nucleus. KEGG enrichment analysis revealed that a total of 14 of the top 20 KEGG pathways in the magenta, blue, and red modules were identical in terms of enrichment, which could be categorized into three major parts: plant metabolism, transport and synthesis functions; ribosome processing and regulation; and plant signaling regulation and transduction. Plant metabolism, transport, and synthesis functions encompass glycolysis/gluconeogenesis, carbon metabolism, the biosynthesis of amino acids, the biosynthesis of cofactors, ubiquitin-mediated proteolysis, protein processing in the endoplasmic reticulum, endocytosis, and other pathways. Ribosome processing and regulation encompasses the ribosome, nucleocytoplasmic transport, mRNA surveillance pathway, spliceosome, and other pathways. Plant signaling regulation and transduction encompasses the MAPK signaling pathway, plant hormone signal transduction, and plant–pathogen interaction pathways (Figure 11).

The co-expression networks between genes in the blue, red, and magenta modules were constructed using Cytoscape(v 3.9.0). A total of eleven Hub genes were screened in the red module, among which six Hub genes were annotated using KEGG and five Hub genes were not listed, among which DN1116_c0_g1_i7, DN1632_c0_g1_i3, DN3378_c0_g1_i2, DN2092_c0_g1_i2, and DN691_c0 _g1_i5 were associated with metabolic pathways and DN843_c0_g1_i3 with oxidative phosphorylation (Figure 12A). A total of five Hub genes were screened in the blue module, of which DN3636_c0_g1_i1, DN7926_c0_g1_i1, DN15593_c0_g1_i1, and DN1933_c1_g1_i1 were associated with ribosomes, and DN293_c0_g1_i1 was associated with endoplasmic reticulum protein processing and ubiquitin-mediated proteolysis (Figure 12C). A total of seven Hub genes were screened in the magenta module, of which two Hub genes were annotated by KEGG and five Hub genes were not listed. Among them, DN3608_c0_g1_i2 was associated with ubiquitin-mediated proteolysis, and DN7847_c0_g2_i1 was associated with metabolic pathways, secondary metabolite biosynthesis, carotenoid biosynthesis, and diterpene biosynthesis (Figure 12B).

A total of 23 Hub genes were identified in the blue, red and magenta modules. Among them, the gene expression of DN1363_c0_g1_i1, DN1632_c0_g1_i3, DN3608_c0_g1_i2, DN2259_c1_g1_i3, DN293_c0_g1_i1, DN843_c0_g1_i3, DN7847_c0_g2_i1, DN1033_c0_g1_i3, DN7926_c0_g1_i1, DN3636_c0_g1_i1, DN15593_c0_g1_i1, DN1933_c1_g1_i1, DN2092_c0_g1_i2, DN509_c0_g2_i4, DN3378_c0_g1_i2, DN567_c0_g1_i2, DN1116_c0_g1_i7, DN485_c0_g2_i4, and DN691_c0_g1_i5 were gradually downregulated with the prolongation of drought stress–rehydration time. The gene expression of DN2683_c0_g2_i4, DN2301_c0_g1_i1, and DN279_c1_g1_i12 showed a gradual downward and then an upward trend. The gene expression of DN2372_c0_g2_i3 showed a gradual trend of initial upregulation and then downregulation (Figure 12).

3.9. qRT-PCR Validation

To further test the reliability and validity of the transcriptomic data, qRT-PCR was performed to validate the following nine DEGs and Hub genes associated with the response to drought stress identified using RNA-seq: TRINITY_DN10244_c0_g2_i1, TRINITY_DN1757_c0_g1_i2, TRINITY_DN9895_c0_g1_i1, TRINITY_DN1933_c1_g1_i1, TRINITY_DN2301_c0_g1_i1, TRINITY _DN2372_c0_g2_i3, TRINITY_DN293_c0_g1_i1, TRINITY_DN509_c0_g2_i4, and TRINITY_DN567_c0_g1_i2. The results showed that the qRT-PCR expression patterns of the nine genes were generally consistent with the RNA-seq data, indicating that the transcriptomic data generated in this study were reliable and valid (Figure 13).

4. Discussion

High-throughput sequencing (HTS), also referred to as next-generation sequencing (NGS), is an advanced genome sequencing technology. Its advantage lies in its high efficiency, high resolution, and large data processing capability, which can lead to obtaining a large amount of genetic information in a relatively short period of time. With the rapid development of this technology, it has been widely used in the various fields of botany, deepening the understanding of plant genomes and promoting progress in plant genetic improvement, crop variety optimization, and environmental adaptation research [13]. Compared with the leaf transcriptome of the congeneric species S. mukorossi (45.39 Gb) [14], this study yielded a more substantial amount of information, generating a total of 96.12 Gb of data. The processing of the data resulted in the identification of 89,228 unigenes, which provided a robust foundation for subsequent functional annotation and classification of the unigenes, CDS prediction, favorable gene cloning, and functional research. The average length and N50 value of the unigenes obtained after de novo assembly were 903 bp and 1538 bp, respectively, indicating the high abundance of medium-length fragments in the sequences obtained in this study. Furthermore, the quality of the assembly was determined to be superior, with the percentages of Q20 and Q30 reaching 98.68% and 95.62%, respectively. Generally, an elevated Q30 value signifies enhanced accuracy and reliability in the process of sequencing. In practice, a Q30 value of 90% or more is generally regarded as representing extremely high-quality sequencing data. Conversely, a higher Q20 value is often indicative of superior quality sequencing data and an increased accuracy in base identification. The N50 value of this study (1538 bp) was significantly higher than that of the congeneric plant transcriptome studies based on NGS [14], indicating that high-throughput data combined with optimized quality control can improve assembly contiguity. However, compared with third-generation sequencing (TGS) technologies, NGS still has limitations in the resolution of long fragments [15]. In the future, it may be combined with HiFi long-read sequencing to further improve the assembly of complex regions [16].

A total of 63,005 sequences were annotated in seven major functional databases, including NR, GO, KOG, and KEGG, accounting for 70.61% of the total number of annotations. The results of the functional gene annotations obtained from the database comparisons can provide a reference for future research on the Sichuan and Yunnan Sapindales. The GO functional annotation showed that the gene expression of cellular processes and metabolic processes were more abundant in the broad category of biological processes, indicating its own high metabolic ability. KEGG functional annotation showed that metabolic pathways accounted for the highest proportion, followed by genetic information processing. Combining the results of GO and KEGG analyses, it was found that a high proportion of metabolic pathway-related genes could be derived from both databases, indicating that the metabolic activities of S. delavayi leaves were vigorous, and the results were the same as those of the S. mukorossi annotation [12]. The highest homology was found between S. delavayi and Acer yangbiense in the Nr database, which may be due to the fact that they both belong to Sapindaceae in the APG III classification system. However, some genes failed to be annotated to the Sapindus genus, likely due to the scarcity of genomic information for this genus in the existing NR database. Recently established databases, such as the N3 database, which integrates genomic information from 3517 plant species, including high-quality genomic data generated using TGS technologies, could be leveraged in the future to enhance annotation efficiency [17]. The results of KOG functional annotation basically covered most of the activities of S. delavayi during its lifecycle, but there were still a small number of genes with unknown functions, suggesting that their functions are not yet fully resolved and provides room for exploration in future studies. The TF annotation results of S. delavayi showed that MYB was the largest family, followed by bHLH, which was different from the S. mukorossi TF, in which bHLH was the largest family [12]. This discrepancy may be associated with the species’ habitat adaptation strategies. The MYB family is known to be involved in cold and drought responses in A. thaliana [18,19], while S. delavayi is predominantly found in the high-altitude regions of China, such as Sichuan and Yunnan. This suggests that S. delavayi may enhance its stress resistance by regulating the ROS signaling pathway (such as the MAPK pathway) through MYB. In addition, the sequencing results of this study were accurate and reliable, where the obtained CDSs can be directly used for full-length cloning and functional analysis of the genes, and the SSR sites detected by the transcriptome can provide a basis for the development of molecular marker technology in S. delavayi.

In the early stages of drought stress, the number of the upregulated genes was more than that of the downregulated genes, indicating that S. delavayi may respond to the stress by activating a large number of genes; after the stress lasted for a period of time, the number of downregulated genes did not differ much from the number of upregulated genes, which indicated that it may enter into a state of adapting to the drought. GO enrichment showed that DEGs were significantly enriched in the integral component of the membrane, ATP binding, and cytoplasm pathways under drought stress. This suggests that the membrane structure, energy metabolism, and the expression of genes related to intracellular substance transport in S. delavayi cells were significantly changed during the water deficit. The enrichment of genes related to the integral component of the membrane may be related to regulating intracellular ion homeostasis, protecting photosynthetic systems, sensing and responding to external signals, and regulating cellular metabolism in plant cells under stress [20], whereas the enrichment of ATP-binding-related genes suggested the importance of energy metabolism in drought response. It is possible that intracellular water balance is maintained under drought conditions through related proteins by regulating intracellular water and solute transport [21]. KEGG enrichment, on the other hand, revealed that the DEGs were significantly enriched in the plant MAPK signaling pathway and the biosynthesis of cofactors. This was consistent with the result that the largest gene family is the MYB family. The genes related to the plant MAPK signaling pathway can play a role in stress tolerance, most of which are involved in the response to drought and osmotic or salt stress [22]. This suggests that the pathway is closely related to drought stress in S. delavayi, which is consistent with the results of the drought-related studies in Acer truncatum [23]. When subjected to drought stress, ROS are produced in plants [24]. ROS can act as signaling molecules to activate the plant MAPK signaling pathway, and modules such as MKK4–MPK3/MPK6 and MKK5–MPK3/MPK6 regulate ROS signaling [25]. The enrichment of the biosynthesis of cofactors category may be related to the cofactors that may help cells cope with various environmental stresses by regulating metabolism, gene expression, and cell signaling [26,27]. The upregulated genes outnumbered the downregulated genes when water was provided after drought stress, suggesting that plants need to activate genes more aggressively during droughts to restore growth and metabolism. The enrichment of nucleus-associated genes may be related to the regulation of gene expression and the activation of DNA repair mechanisms in the nucleus, while the enrichment of DNA-binding-associated genes implied the importance of gene transcriptional regulation in response to rehydration. KEGG enrichment, on the other hand, revealed that DEGs were significantly enriched in the carbon metabolism, biosynthesis of cofactors, and amino acid biosynthesis pathways. This suggests that the metabolic activities of plants recovered rapidly after rewatering, especially as the expression of genes related to energy metabolism and amino acid synthesis changed significantly. The enrichment of the carbon metabolism pathway may be related to the resumption of normal photosynthesis and respiration after rehydration [28], while the enrichment of the biosynthesis of cofactors pathway implies the importance of coenzymes in metabolic recovery after rehydration. The enrichment of nucleus-related genes may be related to the regulation of gene expression and the activation of the DNA repair mechanism in the nucleus. There was only one overlapping DEG in all treatment groups and a large number of stage-specific DEGs with unique gene expression patterns at each stage, suggesting that the physiological response of S. delavayi at the different stages of drought stress is very complex, highly dynamic, and stage-specific. The clustering analyses of the DEGs also supported this view.

TFs have been reported to function as molecular switches in plants in response to abiotic stress [29]. TF families such as NAC, bHLH, and AP2-EREBP may enhance the expression of some drought-critical genes or linkages between related proteins in S. delavayi in response to drought stress [30,31,32]. Among them, the NAC superfamily of transcription factors is one of the largest families in the plant kingdom. NAC transcription factors play a crucial role in the drought stress response of plants, e.g., the NAC protein RD26 regulates the plant’s stress response through an ABA-dependent pathway under drought stress [33]. In addition, through transcriptomic data analysis, researchers have found that genes in the C2f, C72b, and C100a groups in white pear (Pyrus bretschneideri) are related to drought and cold stress responses. The genes in these groups are significantly upregulated under drought and cold stress, indicating that they may be involved in the response mechanisms of white pear to these adverse conditions [34]. The NAC superfamily were highly enriched in the S. delavayi TF families found in this study, suggesting that they play a key role in the response of S. delavayi to drought stress, which was in agreement with the results obtained in the study of the drought-associated transcriptome in Rosa chinensis [35]. From the thirteen co-expression modules identified in the WGCNA analysis, three modules that were significantly associated with drought stress response were screened using the correlation analysis of the modules, i.e., the blue, red, and magenta modules, with physiological indicators. The blue module was related to chlorophyll b and intercellular CO₂ concentration, which may be involved in the regulation of photosynthesis. The magenta module was related to soluble proteins and may be involved in intracellular metabolic regulation. The red module was related to transpiration rate and may be involved in water management. GO enrichment analysis showed that genes in the magenta, blue, and red modules were significantly enriched in biological processes such as the integral component of the membrane, nucleus, ATP binding, cytoplasm, and metal ion binding. This suggests that these genes regulate the response of S. delavayi to drought stress by participating in the metabolism of secondary metabolites, the binding of metal ions, and the structural and functional regulation of the cell membrane and nucleus. KEGG enrichment, on the other hand, was found in the magenta, blue, and red modules, which were categorized into three major groups: plant metabolism, transport, and synthetic functions; ribosomal processing and regulation; and plant signaling regulation and transduction. The WGCNA analysis revealed a complex gene regulatory network in S. delavayi under drought stress, and the identified key modules and enriched pathways indicated that it was synergistically regulated through multiple biological processes and metabolic pathways to adapt to drought stress.

Although this study identified 89,228 unigenes of S. delavayi, their functional annotation was still limited by the completeness of the reference genome. Recent studies on plant pan-genomics have shown that integrating genomes from multiple species can reveal more comprehensive variations in gene families [36]. In the future, chromosome-level genomes can be constructed by combining Hi-C technology with third-generation sequencing, and the functions of candidate genes can be verified using CRISPR–Cas9 to clarify their drought resistance mechanisms.

5. Conclusions

In this study, we constructed a transcriptome database of S. delavayi using the Illumina HiSeq 4000 high-throughput sequencing platform, obtaining a total of 89,228 unigenes related to the growth and development of S. delavayi. Through bioinformatics analysis, these unigenes were subjected to several studies including functional annotation, metabolic pathway analysis, CDS prediction, TF prediction, SSR detection, and expression statistics. The annotation results from two major functional databases, GO and KEGG, revealed that S. delavayi showed high metabolic activity and the ability to process genetic information. Under drought stress, DEGs were enriched in the plant MAPK signaling pathway and cofactor biosynthesis pathways; after rehydration, DEGs were enriched in the carbon metabolism and amino acid biosynthesis pathways, among others. The transcription factor families of S. delavayi, such as NAC, bHLH, and AP2-EREBP, may have enhanced the expression of some of the drought-resistant genes in S. delavayi, as well as the links between related proteins, in response to drought stress. Three modules, blue, red, and magenta, were screened to regulate the response of S. delavayi to drought stress through plant metabolism, transport and synthesis functions, and plant signaling regulation and transduction. The EFR, MYC2, NPR1, and WRKY33 genes played important roles in the drought tolerance pathway of S. delavayi. These transcriptomic data supplement the genome sequence of Sapindus. They greatly enrich the knowledge of gene expression for this genus, and lay a data foundation for the gene cloning, functional analysis, and the development of SSR molecular markers for Sapindus and their relatives in S. delavayi. These transcriptomic data provide resources for the marker-assisted breeding and gene editing of the drought-resistant varieties of S. delavayi (such as by targeting NAC genes). In addition, by elucidating the drought resistance metabolic pathways, these data guide the optimization of cultivation strategies and promote the large-scale planting and ecological restoration of S. delavayi in arid regions. The unigene data of S. delavayi can be compared and analyzed with those of other closely related species, providing important information for the in-depth investigation of their species’ evolution process and breeding for resistance.