Seeds of K. pentacarpos were obtained from the Halophyte Research Laboratory of Nanjing University, which introduced K. pentacarpos from the Halophyte Biotechnology Center, University of Delaware, United States, in 1993.

Uniform and plump K. pentacarpos seeds were selected and soaked in concentrated sulfuric acid for 30 min, followed by rinsing with clean water and soaking for 24 h. Next, the seeds were placed on a wet towel and covered to induce germination. When one-third of the germinated seeds exhibited approximately 1 mm-long sprouts, they were sown in plastic cultivation bowls (diameter: 11 cm; height: 10 cm), with five seeds per bowl. Common garden soil (0.6 kg per bowl) was used for cultivation. A tray was arranged at the bottom of each bowl, and the bowls were placed in a greenhouse with day/night temperatures of 28/25°C. Then, 120 mL of water, based on specialized experimental determination, was added to each bowl per week. After all the seeds germinated, 120 mL of 25% Hoagland’s nutrient solution was added to provide nutrition once every 2 weeks. Furthermore, the water and the nutrient solution evenly permeated throughout the cultivation soil from the tray in this experiment.

According to the classification of China’s saline–alkali soil, the salt content of severe saline–alkali soil is 0.4–0.6% (Zhang, 2019). Therefore, in this study, salt concentration of the cultivation soil was set at 0.05 and 0.5%. Before the seedlings reached the age of 90 days, they were separately subjected to saline–alkali stress treatments for 1 and 7 days. Using the amount of cultivation soil in the bowls as the basis, NaCl and NaHCO₃ were accurately weighed to a mass ratio of 4:1 to obtain total concentrations of 0.5 g/kg (0.05%) and 5 g/kg (0.5%). The saline–alkali mixture was dissolved in 120 mL of distilled water, placed in the tray at the base of each bowl, and allowed to permeate evenly throughout the cultivation soil. All seedlings were sampled and measured at 90 days of age. In this experiment, seedlings cultivated using ordinary garden soil served as the control (CK). The treatment groups were as follows: (i) Tr1: 0.05% saline–alkali solution for 1 day; (ii) Tr2: 0.05% saline–alkali solution for 7 days; (iii) Tr3: 0.5% saline–alkali solution for 1 day; and (iv) Tr4: 0.5% saline–alkali solution for 7 days. Each treatment group consisted of six cultivation bowls.

The chlorophyll fluorescence parameters were measured using a YAXIN 1161G chlorophyll fluorometer (Beijing Yaxinliyi Science and Technology Co., Ltd., Beijing, China). Intact leaves from the middle–upper section of the seedlings were selected and darkened for 30 min using clamping blade clips before testing. The leaves were treated with saturated pulsed light at 3,000 μmol⋅m^–2⋅s^–1 for 1 s followed by actinic light at 1,000 μmol⋅m^–2⋅s^–1 for 9 s. The light-induced curve was then used to measure the initial fluorescence (F₀) and other indicators of chlorophyll fluorescence. From each treatment group, three cultivation bowls were randomly selected, and each bowl was tested five times to obtain the average value. Indicators were measured thrice.

Leaves from the middle–upper section of the seedlings and some tender stems were collected and immediately frozen using liquid nitrogen at −80°C for storage. From each treatment group, three cultivation bowls were selected for analyses. After extracting total RNA using a Takara RNA Preparation Kit (Takara Bio, Dalian, China), RNA concentration and quality were determined using a Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, DE, United States) and Agilent 2100 Bioanalyzer system (Agilent Technologies, CA, United States), respectively.

Full-length (FL) cDNAs were synthesized using a SMARTer™ PCR cDNA Synthesis Kit (Takara Bio, Dalian, China), and cDNA length (1–6 kb) was determined and screened using a BluePippin™ Size-Selection System (Sage Science, Beverly, MA, United States). Next, a DNA Template Prep Kit 2.0 (Pacific Biosciences, Menlo Park, California, United States) was used to establish the SMRTbell library before performing single-molecule real-time (SMRT) sequencing on the PacBio RSII platform (Pacific Biosciences, Menlo Park, California, United States).

The polymerase reads that the length is less than 50 bp, and the accuracy is less than 0.90, were filtered according to the standard procedures of the SMRT Analysis Software package, and sub-sequences shorter than 50 bp were removed to obtain insert reads. The Iso-Seq module of the SMRT Link software was used to iteratively cluster similar full-length (FL) non-chimeric (FLNC) sequences. Consensus isoforms were obtained and further corrected to obtain high-quality transcriptomes with accuracies above 99%. Subsequently, the corresponding Illumina RNA-seq data were input in the Proovread 2.13.841 software to correct for low-quality consensus sequences, thereby increasing sequence accuracy. Finally, the CD-HIT 4.6.142 software was used to eliminate redundant sequences (Li and Godzik, 2006), resulting in a high-quality transcriptome database.

The operating instructions of the NEBNext^® Ultra™ RNA Library Preparation Kit (NEB, Beverly, MA, United States) were followed to generate a second-generation sequencing cDNA library. After purification of the cDNA fragments using the AMPure XP system, the Agilent 2100 Bioanalyzer was used to evaluate the quality of the library. After the quality was ascertained, cDNA library sequencing was performed on the Illumina HiSeq 2500 platform (Illumina, San Diego, CA, United States) to derive paired-end reads.

The raw data were processed to eliminate the sequencing adapters and primer sequences to obtain clean reads before the value of fragments per kilobase of exon per million fragments mapped (FPKM) was used to measure the level of gene expression. The DESeq R software package of the Bioconductor platform was then run to analyze the differential expression between the transcriptomes of the various treatment groups (Anders and Huber, 2010). Differentially expressed genes (DEGs) were screened using fold change ≥2 and false discovery rate (FDR) <0.01 as the standards.

The identified DEGs were clustered using k-means method, and then used for KEGG enrichment analysis. The KOBAS software was used to test the statistical enrichment of DEGs in KEGG pathways (Mao et al., 2005). The hypergeometric test was used to analyze pathway enrichment based on the KEGG pathway database as the unit. The results were compared with the transcriptome background to identify enriched pathways from the differentially expressed transcriptomes.

Using the NCBI database,¹ a homology search and comparison (E-value ≤ 1e-5) of the key genes (FL sequences) selected from the DEGs was performed. Based on query coverage, identity percentage, and E-value of matched nucleobases, the comparison result ranked first in the database were then screened.

The WGCNA R software package (Langfelder and Horvath, 2008) was used to construct a weighted gene co-expression network. The WGCNA analysis was performed on the DEGs with FPKM values ≥1 and coefficient of variation between treatments ≥0.5 for a total of 15 transcriptome samples (5 treatments, each with 3 replicates). After threshold screening and determination of the weighting coefficient β, the original scaled relationship matrix was subjected to power processing to obtain an unscaled adjacency matrix. Considering the correlation of expression patterns between a gene and other genes in WGCNA analysis, the adjacency matrix was further transformed into a topological overlap matrix (TOM). Based on topological dissimilarity matrix (diss TOM = 1-TOM), dynamic shearing algorithm was used for gene clustering and module division. Furthermore, the minimum number of genes in a module was 30 (min Module Size = 30), the threshold for merging similar modules was 0.1327 (minimum Height for Merging Modules = 0.1327), and the network type was “Unsigned” in this analysis.

The genes were selected as module members according to the kME value > 0.7. Some modules, which exhibited high correlations with sample traits, were selected from the heatmap, and their gene co-expression visualization network diagrams were constructed using the Cytoscape 3.7.2 software.

Leaves from the middle–upper section and tender stems were mixed following the aforementioned experimental design. Next, a SteadyPure Plant RNA Extraction Kit (Hunan Accurate Bio-Medical Co., Ltd., Changsha, China) was used to extract RNA for quality inspection according to the manufacturer’s instructions. After quality testing, a PrimeScript™ RT reagent kit with gDNA Eraser (Perfect Real Time) (Takara Bio, Dalian, China) was used to synthesize cDNA by reverse transcription.

A CFX96 real-time fluorescence quantitative PCR system (Bio-Rad Laboratories, Inc., California, United States) was used for qRT-PCR analysis. The reagent test kit used was the TB Green^® Premix EX Taq™ II (Tli RNase H Plus) (Takara Bio, Dalian, China), the dye was TB Green, and the internal reference gene was β-actin. The primer designing tool of NCBI was used to design the fluorescence quantitative PCR primers. Relative gene expression was analyzed using the 2^–ΔΔCT method (Livak and Schmittgen, 2001) with three replicates.

SPSS 21.0 was used to perform Duncan’s multiple range test at a significance level (α) of 0.05; SIMCA 14.1 was used to perform OPLS-RA.

The F₀ of seedlings increased with prolonged treatment with 0.05 and 0.5% saline–alkali solutions. All treatments exhibited F₀ values greater than that of the control, and the F₀ value was 32.85% higher than that of the control, with a significant difference under the high-concentration condition (P = 0.001, see Figure 1A) on day 7. Compared to the control, the maximum photochemical efficiency (F_v/F_m) was relatively stable and changed slightly under saline–alkali conditions (see Figure 1B). However, F_v/F_m significantly decreased under prolonged high-concentration condition (P = 0.022), and the value on day 7 was 5.02% lower than that on day 1. The photochemical quenching coefficient (qP) and PSII quantum yield (ΦPSII) also presented similar patterns (see Figures 1C,D): under the 0.05 and 0.5% saline–alkali conditions, both parameters decreased with prolonged treatment. The variations in qP and ΦPSII were significant under the 0.5% saline–alkali condition after 7 days (P = 0.010, P = 0.000), and qP and ΦPSII values decreased by 68.94 and 33.80%, respectively, compared with the respective control groups.

The SMRT sequencing technique was used to determine the FL transcriptomes of K. pentacarpos seedlings. An SMRT cell was used to establish an FL cDNA library with a sequence length of 1–6 kb (Table 1). Subreads smaller than 50 bp in length were filtered, yielding 18.95 G of clean data. A total of 410,351 circular consensus sequences (CCS) were extracted based on the criteria of full passes ≥3 and sequencing accuracy >0.9, with sequence length distributed between 1 and 3 kb (Supplementary Figure 1A). After removing the cDNA primer and polyA sequences from the CCS, 383,234 FLNC sequences were obtained, most of which were 1–3 kb in length (Supplementary Figure 1B). Following iterative clustering, 96,419 consensus isoforms were obtained, with the majority of the transcriptomes being approximately 2-kb long (Supplementary Figure 1C). Further correction yielded 93,218 high-quality consensus isoforms, the accuracies of which were above 99%. Finally, highly similar sequences were merged, and redundancies were removed, leaving 54,601 non-redundant sequences.

In this experiment, differential expression in the transcriptomes of K. pentacarpos seedlings was not evident following 0.05% saline–alkali treatment (Figures 2A,B). The number of DEGs on day 1 and 7 were 185 and 203, respectively. Under 0.5% saline–alkali treatment, differential expression in their transcriptomes became evident, with 1,588 and 1,764 DEGs on days 1 and 7, respectively. Among these, downregulated DEGs were predominant (Figures 2C,D) and accounted for 89.11 and 88.38% of the total expression on days 1 and 7 of 0.5% saline–alkali treatment, respectively. These results revealed that saline–alkali concentrations considerably affected K. pentacarpos seedlings than treatment duration.

The top 20 pathways with the smallest q values are shown in Figures 2E–H for the four treatments. Under 0.05% saline–alkali treatment, the enrichment factors of each pathway were small, but the q value was larger on day 1. The pathways were mainly enriched in the biosynthesis and endocytosis of ubiquinone and terpenoid-ubiquinone (Figure 2E). When the seedlings were subjected to stress for 7 days, a small portion of the pathway enrichment factors increased, while the q value became smaller. Most pathways were similar to those on day 1 and were mainly enriched in pathways, such as phagocytosis and metabolism of fructose and mannose (Figure 2F). The pathway enrichment conditions on days 1 and 7 were similar with 0.5% saline–alkali treatment. The enrichment factors of the various pathways increased significantly compared with that of low-concentration treatment, but the q value was small. The number of enriched transcriptomes also increased significantly. Enrichment occurred in various pathways, including those of carbon metabolism, amino acid biosynthesis, and fructose and mannose metabolism (Figures 2G,H).

We used kME values to evaluate the existence of effective connectivity between key genes and identify module members. In this experiment, DEGs with kME >0.7 were selected as module members, and similar modules were merged after their eigenvectors were calculated, resulting in six gene co-expression modules (Figure 3A). The modules had 52 (Memagenta) to 1,370 (Meblue) DEGs. The expression patterns of DEGs in the same module were similar and downregulated.

The modules–traits correlation heatmap (Figure 3B) reflected the correlation between genes in samples with related traits and the modules to which they belonged. The greater the absolute value, the stronger the correlation. Red and blue colors indicate positive and negative correlations, respectively. In this experiment, five gene modules were highly correlated with the saline–alkali stress in K. pentacarpos, with all their correlation coefficients being > 0.80. Among them, Memagenta (r = 0.81), Mebrown (r = 0.87), and Meblack (r = 0.92) were positively correlated with CK, Tr2, and Tr3, respectively; Meyellow (r = −0.97) and Megreen (r = −0.99) were negatively correlated with Tr3 and Tr4, respectively. The WGCNA visualization diagrams for the five modules were generated (Figures 3C–G), and the top five genes with the highest kME values in each module were selected as key genes for that module (marked in red, see Supplementary Table 1).

F_v/F_m reflects the potential maximum light conversion efficiency of plants, and can indicate their overall health status (Bjorkman and Demming, 1987). Therefore, it is an important indicator of the impact of environmental stress on photosynthetic performance. In this study, OPLS-RA was performed on the F_v/F_m (Y) of K. pentacarpos and the FPKM value (X) of the selected 25 key genes. The degree of influence of each factor over photosynthetic performance was analyzed using the VIP value, which was the basis for screening the key genes. After fitting the principal component analysis model (R²X = 0.504, Q² = 0.149), the score chart of the samples (Figure 4A) revealed that the 15 sample groups were normally distributed with no abnormalities. The regression model was established using OPLS-RA fitting (R²X = 0.625, R²Y = 0.921, Q² = 0.542).

The VIP value of the model indicated the degree of influence that the relevant factors exhibited on Y. The selection criterion, based on the requirements stipulated in the SIMCA user guide, was that the VIP value must be > 1. After evaluation, nine DEGs in the K. pentacarpos seedlings were found to have VIP values > 1 (Supplementary Table 2) and were selected as key genes that responded to saline–alkali treatments (Figure 4B).

The FL cDNA sequences (Supplementary Table 3) were used to perform homology comparisons with the NCBI database. Among them, the functions of three genes was unknown, while those of the remaining six were known. The IDs of their transcriptome sequence were F01_transcript_53932, F01_transcript_13312, F01_transcript_3631, F01_transcript_ 7879, F01_transcript_59507, and F01_transcript_25894. After comparison, these six genes were highly homologous to plants, such as Hibiscus syriacus and Gossypium hirsutum, both of which belong to the Malvaceae family. These genes were predicted to be KpCPSF3, KpLARP1C, KpAGO4, KpVPS35A, KpPUB33, and KpHIR1 (Figures 4C–H). The specific comparisons are given in Table 2.

Functional analysis revealed that the key genes were involved in regulating pathways, such as vesicular transport (KpVPS35A), programmed cell death (PCD; KpHIR1) induction, transcription levels (KpCPSF3 and KpAGO4), translation levels (KpLARP1C), and post-translational protein levels (KpPUB33) (see Table 2). Most genes exhibited negative regulatory effects.

Specific primers were designed according to the FL transcriptome sequences (Supplementary Table 4) for qRT-PCR analysis of the nine key genes. For most treatments, the expression levels of the key genes were significantly lower than those of the control and were downregulated (Figure 5); this was consistent with the transcriptome results.

Among the nine genes, the expression patterns of five genes—F01_transcript_53932, F01_transcript_7879, F01_transcript_59507, F01_transcript_25894, and F01_transcript _9571—were similar. Compared to the control, gene expression gradually decreased under Tr1 and Tr2 (low saline–alkali treatments). Nonetheless, gene expression initially decreased but recovered under Tr3 and Tr4 (high saline–alkali treatments), despite being lower than that of the control (Figures 5A,D–H). However, their expression levels under Tr2 was the lowest among all treatments, and significantly decreased by 56.65, 53.80, 67.16, and 87.51% compared with those of their corresponding controls (P = 0.000).

The expression patterns of F01_transcript_13312, F01_transcript_95488, and F01_transcript_4187 were similar; under prolonged saline–alkali treatments, the expression levels of these three genes decreased. The expression levels of these genes inf most treatment groups were lower than those in the control, and only few genes exhibited expression levels greater than the control for the treatment groups on day 1 (Figures 5B,G,I), which under the Tr2 treatment were the lowest and 95.54, 55.92, and 44.14% lower than those of their respective controls (P = 0.000). This anomaly might be caused by an emergency response to saline–alkali stress. Under Tr1 and Tr2, the expression of F01_transcript_3631 increased with time, and the value under Tr1 significantly decreased by 50.90% compared with that of the control (P = 0.000), whereas under Tr3 and Tr4, its expression levels were relatively stable but consistently lower than that of the control (Figure 5C).

In this study, the Fv/Fm of seashore mallow was stable under saline–alkali stress, and the Fv/Fm value of each treatment was not significantly different from that of control plants. However, F0, qP, and ΦPSII changed significantly in the later stages of high-concentration saline–alkali treatment compared with their respective controls, and the variations were relatively small in other treatments.

The decrease in Fv/Fm of the stressed seedlings can be attributed to the inactivation of the PSII reaction center (Da̧browski et al., 2015) or blockage of the photosynthetic electron transport chain (Tuba et al., 2010). However, the difference between the F_v/F_m values of the treated plants and the control was not significant under Tr4, indicating that the photosynthetic performance of the K. pentacarpos seedlings was relatively stable under saline–alkali stress conditions. However, qP was used to reflect the photosystem pressure due to the excess excitation energy of PSII (Öquist and Huner, 1993). With increasing saline–alkali concentrations, the qP of the K. pentacarpos seedlings decreased with time, indicating that the pressure of excitation energy gradually increased on photosystem and the photosynthetic function was affected (Öquist and Huner, 1993). As for the electron transport chain, the ΦPSII reflected the working status of PSII (Li and Feng, 2004). In this study, the variation was similar to that of F_v/F_m, indicating that the PSII electron transport chain was relatively normal in the early stage, but electron transfer was blocked to weaken photosynthetic function in the later stage. Based on chlorophyll fluorescence characteristics, photosynthetic performance of the seedlings was relatively stable, and K. pentacarpos showed strong tolerance to saline–alkali stress.

Plants must finely regulate their gene expression in response to environmental stress. Although previous studies have focused on positive regulatory mechanisms (Cao et al., 2017; Pang et al., 2017; Lu et al., 2019), recent studies have paid increasing attention to negative regulation. In this study, downregulated DEGs accounted for 89.11 and 88.38% of the expression under Tr3 and Tr4, respectively, with negative regulation being predominant. Three negative regulatory pathways, involving the key genes of K. pentacarpos, were involved in responding to saline–alkali stress: (i) LARP1 inhibited the translation of 5′-terminal oligopyrimidine mRNAs (TOP mRNAs) (Philippe et al., 2018); (ii) AGO4 -mediated DNA methylation through siRNA interaction (Pikaard et al., 2012; Matzke and Mosher, 2014); and (iii) the plant U-box33 recognized and labeled target proteins for degradation by the 26S proteasome in the ubiquitin pathway (Jin et al., 2007).

The 5′-TOP mRNAs, a class of eukaryotic mRNA family, contains proteins that regulate cell growth (Philippe et al., 2018), whose translation is regulated by the eukaryotic promoter 4F (eiF4F). Its translational abilities can be inhibited by LARP1, which competes to bind with TOP mRNAs (Tcherkezian et al., 2014). Fonseca et al. (2015) used RNA interference techniques to reduce the levels of LARP1, thereby alleviating its inhibitory effects on TOP mRNA translation. However, target of rapamycin (TOR) specifically controls the translation of 5′-TOP mRNAs by the putative TOR substrate, LARP1. Furthermore, the regulatory pathway of TOR–LARP1–5′-TOP is conserved in plants (Scarpin et al., 2020). In this study, KpLARP1C expression decreased with prolonged saline–alkali treatment, and its expression in most treatment groups was lower than that of the control. It is speculated that the decreased expression of KpLARP1C may reduce competition and the inhibition of TOP mRNA translation and promote cell growth, thereby enhancing the tolerance of K. pentacarpos seedlings to saline–alkali stress.

AGO4 has been mainly reported in studies of plant resistance to diseases (Brosseau et al., 2016). AGO4 achieves transcriptional silencing of genes through DNA methylation (Raja et al., 2008; Duan et al., 2015), leading to the regulation of plant responses to biotic and abiotic stress (Pu et al., 2021). Arabidopsis thaliana mutant, which over-expresses AtAGO4, is more likely to be infected by Pseudomonas syringae (Agorio and Vera, 2007), while the double mutant of AtAGO4 and AtAGO2 is susceptible to the tobacco rattle virus (Ma et al., 2015). AGO4 induces nucleic chromatin modifications and prevents recessive transcription to maintain or activate the expression of stress-responsive genes (Al et al., 2017), which regulate physiological pathways, such as jasmonic acid signaling pathway (Prashanm et al., 2020). As for hypoxia, AGO1 in Arabidopsis coordinates AGO4, which represses the expression of HR4 by DNA methylation to regulate stress tolerance (Loreti et al., 2020). Under saline–alkali stress, the expression of KpAGO4 was lower than that of the control plants., indicating that the decreased expression of KpAGO4 may weaken the inhibition of DNA methylation and transcriptional gene silencing. Then, the function of related genes mediated by KpAGO4 could be activated to respond to stress (Al et al., 2017), thereby improving the tolerance of K. pentacarpos seedlings to saline–alkali stress. This, in turn, maintained the stability of their photosynthetic function.

The ubiquitin system can selectively degrade proteins related to stress response, growth, and development of plants to adapt to environmental stress (Varshavsky, 1997). The plant U-box (PUB) protein is a type of ubiquitin-linked enzyme, E3, that specifically identifies target proteins (Zhou and Zeng, 2017), enabling plants to respond to stress. Sixty-six StPUB genes have been identified in potato, and 200 proteins are modified, including 25 differential ubiquitination modification sites under PEG-induced drought (Tang et al., 2022). Arabidopsis thaliana proteins, PUB22 and PUB23, act on RPN12a and cooperate to negatively regulate drought-stress responses through the drought signaling pathway (Cho et al., 2008; Seo et al., 2012). Similarly, AtPUB11 is a negative regulator of drought tolerance, which degrades LRR1 and KIN7 (Chen et al., 2021). Capsicum frutescens CaPUB1 gene, which was heterologously transferred into rice, negatively regulated rice response to drought-stress and decreased drought-tolerance of rice (Min et al., 2016). Under salinity stress, A. thaliana protein PUB30 degraded BKI1 through ubiquitination and negatively regulated the salinity tolerance of plants (Zhang et al., 2017). After the Pohlia nutans PnSAG1 gene was heterologously overexpressed in A. thaliana, the sensitivity of transformed plants to salinity stress increased, indicating negative regulation (Wang J. et al., 2019). In this study, KpPUB33 expression was significantly downregulated in stressed K. pentacarpos plants. This indicates that a decrease in KpPUB33 expression maybe alleviate the ubiquitin-mediated degradation of target proteins, and then maintain the normal functions of the target proteins, thereby improving saline–alkali tolerance of K. pentacarpos.

Plant PCD can be classified as apoptotic or autophagic (Huang and Fu, 2010). Apoptotic PCD often occurs in stress-induced hypersensitivity reaction (HR), such as heavy metal or salinity stress (Pan et al., 2001; Liu et al., 2007). Hypersensitivity-induced response (HIR) genes can induce HR responses and participate in the regulation of ion channels and cell death (Zhou et al., 2010). Overexpression of the C. frutescens CaHIR1 in A. thaliana led to tissue necrosis similar to HR and improved plant resistance to bacterial and fungal infections (Jung and Hwang, 2007). The expression of Arachis hypogaea AhHIR was significantly decreased under low-temperature stress, which increased with time (Liu et al., 2014). This observation was similar to that of K. pentacarpos KpHIR1 under saline–alkali stress. The expression of KpHIR1 decreased under Tr3 but increased to 66.91% compared to the control value under Tr4 (P = 0.000), whereas its expression continuously decreased under Tr1 and Tr2. Downregulation of the expression of HIR gene was conducive to reducing cell mortality (Liu et al., 2014), whereas the upregulation of its expression promoted apoptosis-like PCD to form a barrier of dead cells (Liu et al., 2007), which prevented further tissue damage by the salt ions (Liu et al., 2007). This is the potential mechanism by which K. pentacarpos seedlings increase tolerance to saline–alkali stress.

Autophagic PCD is induced by stress, such as drought, salinity, and nutrient deficiency, where the endoplasmic reticulum is involved in regulating and inducing cell death (Huang and Fu, 2010). During PCD, endoplasmic reticulum recycles nutrients of damaged cells to supply them to other cells for survival. Phagocytes, however, reuse these nutrients through autophagy and vesicular transport (Song et al., 2016). The VPS35 protein in the vesicular transport complex Retromer specifically identifies the cargo protein, transports it to the vesicles of the Golgi reverse membranes, and then packages and exports it (Song et al., 2016), thereby ensuring reuse of the protein. Therefore, the Retromer complex could regulate the identification of dead cells by phagocytes through the cargo protein CED-1, and to recycle more nutrients (Yamanaka and Ohno, 2008). Under high-concentration saline–alkali stress, KpVPS35A expression increased with time, indicating that the ability to identify and transport the cargo protein was improved by VPS35. This led to improved precise identification of the PCD cells, which facilitated the recycle and reuse of their nutrients and maintained the vitality of other cells to help K. pentacarpos seedlings survive saline–alkali conditions.