Arabidopsis thaliana ecotype Columbia (Col-0) was used in this study. Arabidopsis seeds were sterilized with 75% (v/v) ethanol and then washed with sterile distilled water. The seeds were then sown on half-strength Murashige and Skoog (MS) medium. After stratification for 3 days in a 4°C refrigerator, the plates were transferred to growth chambers. Eight days after seed germination, Arabidopsis seedlings were transferred into 9 cm diameter pots containing soil, perlite, and vermiculite (2:1:1), with irrigation of half-strength Hoagland’s nutrient solution. The growth condition in the growth chambers was a 16 h light/8 h dark photoperiod with a day/night thermoperiod of 22°C/18°C, a relative humidity of 70%, and irradiance of 110 μmol m^–2 s^–1.

Four-week-old seedlings were randomly divided into four groups, and the seedlings in each group were subjected to treatment as follows: pretreatment–stressed (pretreated with H₂O₂ and salt-stressed, HN); non-pretreatment–stressed (pretreated with water and salt-stressed, WN); pretreatment–non-stressed (pretreated with H₂O₂ and not salt-stressed, HW); and non-pretreatment–non-stressed (i.e., control; pretreated with water and not salt-stressed, WW). For pretreatment–stressed and pretreatment–non-stressed plants, leaves were sprayed with 20 μM of H₂O₂ solution four times at 4-h intervals, while non-pretreatment–stressed and non-pretreatment–non-stressed plants were sprayed with water. Twenty-four hours after foliar spraying, Arabidopsis seedlings from pretreatment–stressed and non-pretreatment–stressed groups were subsequently watered with 150 mM NaCl every day, whereas pretreatment–non-stressed and control plants were treated with water. Twelve hours after treatment with 150 mM NaCl, Arabidopsis seedlings were collected for transcriptome profiling analysis. Four days after treatment with 150 mM NaCl, Arabidopsis seedlings were collected for the determination of various physiological parameters.

The fresh weight of the shoots from each treatment was determined immediately after harvesting, and samples were dried in an oven at 70°C for 24 h to obtain dry weights. Twenty individual plants were collected for each replicate and triplicates were analyzed in parallel.

The relative membrane permeability (RMP) of the seedlings was determined following the method of Yang et al. (1996). Excised fresh leaves (0.5 g) were immediately put into test tubes containing 10 mL of deionized distilled water and briefly vortexed. The solution was used to measure initial electrical conductivity (EC0). The test tubes containing leaves in distilled water were kept at 4°C for 24 h and EC1 was determined. The test tubes were then placed in a boiling water bath for 10 min, cooled to room temperature, and the boiled leachate was filtered and measured for EC2. RMP was computed using the following formula: RMP (%) = [(EC1 − EC0)/(EC2 − EC0)] × 100.

The level of lipid peroxidation in the leaf tissue was measured in terms of MDA (a product of lipid peroxidation) content, detected by the thiobarbituric acid reaction using the method of Dhindsa et al. (1981). Fresh leaf samples (0.4 g) were homogenized in 5 mL of 0.1% trichloroacetic acid, vortexed, and then 4 mL of 0.5% thiobarbituric acid was added. The mixture was heated to 95°C for 30 min and was quickly cooled in an ice bath. Afterward, the mixture was centrifuged at 3000 rpm for 10 min. The supernatant fraction was collected, and the absorbance of the supernatant at 532 and 600 nm was read. The value for the non-specific absorption at 600 nm was subtracted from that at 532 nm (Zhang and Kirkham, 1996). The concentration of MDA was calculated using the extinction coefficient of MDA (155 mM^–1 cm^–1) (Heath and Packer, 1968) and expressed as nmol MDA g^–1 fresh weight. Each treatment was carried out in triplicate.

Fifteen dry plants were pooled together and ground into fine powder; 0.02 g of dry powder was ashed in a muffle furnace at 300°C for 2 h, then 550°C for 10 h. The ash was resolved into small amounts of concentrated nitric acid and adjusted to a final volume of 10 mL. The ion content of samples was determined with a flame photometer (2655-00 Digital Flame Analyzer, Cole-Parmer Instrument Company, Chicago, IL, United States). Three independent determinations were performed for each treatment (Song et al., 2005).

Twelve hours after treatment with 150 mM NaCl, Arabidopsis seedlings from the four experimental groups (WW, WN, HW, and HN) were collected to extract total RNA using Biozol reagent (Bio Flux, Beijing, China) according to the manufacturer’s instructions. The integrity and quality of the isolated RNA were monitored by agarose gel electrophoresis. RNA concentration was quantified by a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Massachusetts, United States). The qualified RNA samples were sent to the Annoroad Gene Technology Corporation (Beijing, China), and the libraries were sequenced on an Illumina platform and 150 bp paired-end reads were generated. RNA-Seq data of the four experimental samples were obtained from three biological replicates, respectively. RNA-Seq data were deposited into the NCBI’s Sequence Read Archive (the accession number is PRJNA612654).

Low-quality reads were trimmed using Trimmomatic (Bolger et al., 2014) (v 0.36) with the settings “LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36”. Clean reads were mapped to the Arabidopsis TAIR10 release obtained from TAIR¹ using TopHat (v 2.1.1) with settings “-N 1 –num-threads 6”. Count data were generated by Cufflinks (v 2.2.1) and FPKM (fragments per kilobase per million mapped reads) was used to estimate the expression levels of individual genes.

Differentially expressed genes (DEGs) were identified by DESeq2 Moderated estimation of fold change and dispersion (Love et al., 2014) using the Bioconductor software², based on a comparison across all samples under control or different experimental conditions with false discovery rate (FDR) less than 0.05. The Goatools (v 0.8.9) python package was used for GO term enrichment (Klopfenstein et al., 2018) with the Arabidopsis association files downloaded from TAIR10. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of genes was performed using clusterProfiler package (v 4.2.2, Wu et al., 2021).

Transcriptome profiling results were validated and verified by quantitative real-time PCR experiments, in which 2 μg of total RNA was used for reverse transcription to obtain the cDNA using FastQuant RT Kit (with gDNase, TIANGEN, Beijing, China). The SuperReal PreMix Plus Kit (SYBR Green, TIANGEN) was used along with the cDNA for quantitative real-time PCR experiments using a real-time fluorescence quantitative PCR instrument (LightCycler^® 96, Roche, Basel, Switzerland). All reactions were assayed using three replicates. Actin2 was used as an endogenous control. The relative expression levels are presented as values relative to that of the corresponding control sample at the indicated time after normalization to actin transcript levels. Primer sequences are shown in the Supplementary Table 1.

All the above experiments involved three biological replicates, and each experiment (except RNA-Seq) was carried out twice at different times. All data are expressed as means ± standard deviation and the significance of differences between datasets was evaluated by one-way ANOVA following SPSS. P-values of <0.05 were considered to be significantly different.

In order to evaluate the effects of H₂O₂ pretreatment on Arabidopsis growth under salinity, we sprayed the leaves of Arabidopsis seedling with 20 μM H₂O₂ and subsequently exposed them to 150 mM NaCl. We then determined some physiological indices as shown below.

In order to analyze the molecular mechanisms of salt tolerance improvement induced by H₂O₂ pretreatment, we collected the leaves of 4-week-old Arabidopsis seedlings treated with HN, WN, HW, and WW (with WW being the control) and the total RNA was extracted for genome-wide transcriptome analysis. RNA-Seq data were analyzed from a total of twelve samples comprising three biological replicates for each treatment.

In total, 19,391 genes were detected in the leaves of Arabidopsis. We further obtained 1493 DEGs in HW, compared to WW, with at least twofold change of gene expression at P-value < 0.05. Among these, 993 genes were up-regulated and 500 genes were down-regulated (Figure 2A and Supplementary Table 2). Similarly, of the 2467 DEGs specifically responding to HN treatment in comparison to WW, 1212 genes were up-regulated whereas 1255 genes were down-regulated (Figure 2A and Supplementary Table 3). Among the 1533 DEGs in WN compared to WW, 922 genes were up-regulated while 604 genes were down-regulated (Figure 2A and Supplementary Table 4). We performed a preliminary analysis of up-regulated (Figure 2B) and down-regulated (Figure 2C) genes through Venn diagrams of HN vs. WW, HW vs. WW, and WN vs. WW. We found 602 unique DEGs in HW vs. WW (Supplementary Table 5), 455 unique DEGs in HN vs. WW (Supplementary Table 6), and 364 unique DEGs in WN vs. WW (Supplementary Table 7). There were an additional 361 DEGs that were common to both HW vs. WW and HN vs. WW (Supplementary Table 8), 169 DEGs common to both HW vs. WW and WN vs. WW (Supplementary Table 9), and 535 DEGs common to both HN vs. WW and WN vs. WW (Supplementary Table 10). Relevant data on down-regulated genes are also presented (Figure 2C and Supplementary Tables S11–16).

In order to rule out that some DEGs may have only been influenced by salt stress in HN vs. WW, we removed the DEGs from HN vs. WW that did not have a large difference in abundance with WN vs. WW; that is, those genes whose ratio of log₂FC (HN vs. WW)/log₂FC (WN vs. WW) was between 0.67–1.5. Thus, there were still 1766 DEGs mainly affected by both H₂O₂ and NaCl in HN vs. WW at this time, where 780 DEGs were up-regulated and 986 DEGs were down-regulated. All subsequent data analysis on HN vs. WW was mainly carried out for these 1766 DEGs (Supplementary Table 17).

Gene ontology (GO) enrichment analysis was performed on 1493 and 1766 DEGs according to the biological processes in HW vs. WW (Figure 2D) and HN vs. WW (Figure 2E), respectively. GO terms both in HW vs. WW and HN vs. WW mainly included “regulation of biological processes” (HW vs. WW, P = 1.67 × 10^–5; HN vs. WW, P = 7.54 × 10^–8), “response to chemicals” (HW vs. WW, P = 2.64 × 10^–20; HN vs. WW, P = 5.98 × 10^–25), “response to abiotic stresses” (HW vs. WW, P = 9.72 × 10^–19; HN vs. WW, P = 1.39 × 10^–18), and “regulation of transcription” (HW vs. WW, P = 1.56 × 10^–4; HN vs. WW, P = 8.3 × 10^–8). This implies that many genes up-regulated by individual H₂O₂ pretreatment alone or HN, or both, may be related to the enhanced salt tolerance of plants through their function in the abovementioned process. More interestingly, we found that the GO terms related to “cell cycle process” (P = 1.38 × 10^–19) and “cell division” (P = 1.21 × 10^–9) were specially enriched in HW vs. WW; therefore, we speculate that the up-regulation of these genes may affect plant growth to cope with the subsequent stresses.

To assess the major transcriptional dynamics associated with the responses to H₂O₂ pretreatment and/or both H₂O₂ and NaCl, we further clustered these DEGs from HW vs. WW and HN vs. WW into nine groups according to their expression trends under the four different combinations of treatments (Figures 2F,G). Of these clusters, we mainly focus on five clusters on the basis of their functional annotations and the expression profiles which were up-regulated either in HW vs. WW, HN vs. WW, or both. The K1 group clusters those genes which may be primed by H₂O₂ pretreatment and have up-regulated expression with subsequent salt stress. These genes are enriched in response to abiotic stimuli, illustrating their positive regulatory roles in increased plant salt tolerance. The genes in K2 were significantly up-regulated in HN vs. WW without obviously different expression in HW vs. WW and, so, they are enriched in oxidation–reduction processes, response to abiotic stimuli, and cell wall organization or biogenesis, which may mean that the expression of these genes is initiated during H₂O₂ pretreatment and mainly functions in subsequent salt stress. The K5 group comprises genes which mitigated the degree of up-regulation due to H₂O₂ pretreatment followed by NaCl, as compared with NaCl alone, which is characterized by an abundance of genes related to stress responses, especially osmotic stress and ABA, response to chemicals, and response to abiotic stimuli. The K7 group clusters those genes that were successively up-regulated by individual H₂O₂ and H₂O₂ plus NaCl, but the magnitude of the increase in the latter was less than in NaCl alone. Based on K5 and K7, we speculate that pretreatment with H₂O₂ can alleviate the oscillation of plant intracellular environment caused by subsequent NaCl exposure. The K8 group includes genes which were mainly up-regulated by H₂O₂ pretreatment, while only a few of these were up-regulated by combined H₂O₂ and NaCl treatment. Moreover, these genes are concentrated in cell cycle processes and cell division, implicating that after H₂O₂-induced expression, these genes may promote cell proliferation, resulting in plant growth under subsequent high-salt stress.

To identify the metabolic pathways in which the DEGs were involved and enriched, KEGG analysis was also performed. The results revealed that in HW vs. WW, 993 up-regulated genes were enriched in six pathways, including ribosome biogenesis in eukaryotes (ath03008, P = 3.46 × 10^–10), DNA replication (ath03030, P = 0.00011), flavonoid biosynthesis (ath00941, P = 0.00030), homologous recombination (ath03440, P = 0.00082), cutin, suberine and wax biosynthesis (ath00073, P = 0.00275), and mismatch repair (ath03430, P = 0.00019); 500 down-regulated genes were assigned to plant hormone signal transduction (ath04075, P = 0.01571) and alpha-linolenic acid metabolism (ath00592, P = 0.000588). Likewise, in HN vs. WW, 780 up-regulated genes were enriched in starch and sucrose metabolism (ath00500, P = 0.00496), glucosinolate biosynthesis (ath00966, P = 0.00496), and cutin, suberine and wax biosynthesis (ath00073, P = 0.02513); while 986 down-regulated genes mainly participated in plant hormone signal transduction pathway (ath04075, P = 2.68 × 10^–18) (Figure 3).

Further mining the transcriptome data, we found that after low-concentration H₂O₂ pretreatment of seedlings, a large proportion of genes related to the cell cycle and cell division were up-regulated, and most were significantly induced only under HW vs. WW (Figure 4A, Table 1, and Supplementary Table 18). Even though the expression levels of a few genes were increased under HW vs. WW and HN vs. WW, the extent of increase in the former was higher than in the latter (Figure 4A, Tables 1, 2, and Supplementary Tables 18, 20). Among these are sixteen core cell cycle genes, including two A-type cyclins (CYCA1;1, CYCA2;4), seven B-type cyclins (CYCB1;1, CYCB1;2, CYCB1;3, CYCB1;4, CYCB2;2, CYCB2;3, CYCB2;4), two plant-specific B-type CDKs (CDKB1;2; CDKB2;1) and upstream regulator DEL1 and its target CDT1A, and minichromosome maintenance genes (MCM2, MCM3, MCM6). B-type CDKs are plant-specific and are divided into two subtypes: CDKB1 and CDKB2. CDKB1 is activated by A2-type and all B-type cyclins and functions in the late S-to-M phase, while B2-type CDKs exclusively associate with B1-type cyclins and have transcript levels peaking late in the M phase (Van Leene et al., 2010). In our data, the increased transcript levels of both CDKB1;2, CDKB2;1 and their corresponding partners CYCB2;4, CYCB1;1 implied that the CDKB1;2/CYCB2;4 and CDKB2;1/CYCB1;1 complex may promote cell cycle progression through late S-to-M or M phases (Van Leene et al., 2010).

Besides CDKs and their cyclin partners, E2F transcription factors also belong to the core cell cycle machinery. Upon H₂O₂ pretreatment, the atypical E2F DP-E2F-like 1 (DEL1) was up-regulated, indicating that DEL1 may enhance cell proliferation by repressing the transcription of CCS52A2, which is required for endocycle onset (Lammens et al., 2008). DEL1 can also restrain the stress-induced switch from mitosis to the endocycle in dividing cells exposed to osmotic stress (Cookson et al., 2006). Contradictory with DEL1 inhibiting CCS52A2, CCS52A2 was also markedly up-regulated only under HW; this inconsistency may have been due to using transcriptomic data from whole shoots of Arabidopsis instead of defined cells. Besides CCS52A2, other anaphase-promoting complex/cyclosome (APC/C) coactivators such as CCS52B and CDC20 (CDC20.1, CDC20.2) were significantly elevated in transcripts under HW compared to WW, and this may be responsible for facilitating the switch from mitosis to endoreduplication through targeting of mitotic cyclins for destruction, thus inactivating cyclin-dependent kinase (CDK) (Kevei et al., 2011; Yang et al., 2017). CDKs/cyclins and their regulators DEL1, CCS52A2, CCS52B, CDC20.1, and CDC20.2 coordinate to balance cell proliferation and cell differentiation/expansion and, thus, balance plant growth and development.

Furthermore, H₂O₂ pretreatment also induced the expression of CDT1A and minichromosome maintenance genes (MCM2, MCM3, MCM6), where only CDT1A and MCM6 were differentially expressed under HW and HN, while only MCM2 and MCM3 were differentially expressed under HW (Figure 4A). CDT1A, as a DNA replication licensing factor, can recruit the MCM complex to form the components of the pre-replicative complex at the G1 phase (Nishitani et al., 2001). Therefore, high transcript levels of these genes facilitate activation of the replication origin, which can ensure that genomic DNA is replicated completely and accurately only once during the S phase in a single cell cycle (Tuteja et al., 2011). Meanwhile, according to previous research results, the high H₂O₂-induced expression of MCM6 can presumably confer plant salt tolerance by preserving normal DNA replication under salinity stress conditions (Dang et al., 2011).

Except for the abovementioned cell cycle components, low levels of H₂O₂ also increased transcription of a large number of genes encoding spindle assembly factors. These include genes for mitosis kinases (AUR2, AUR3, and AtHaspin); microtubule-associated proteins (MAPs), including TPX2, MAP65-3, and members of the kinesin superfamily (Kin4/chromokinesin, Kin5, Kin7, Kin12 and Kin14 families); chromosome organization proteins (SMC1, SMC2, SMC3, SMC4, RAD21.2, and TOPII); kinetochore complex (Ndc80 and Nuf2); and spindle assembly checkpoint complex (Mad2, Mad3.1, and Mad3.2). Among these, AtHaspin can activate AUR3 and promote its centromeric localization on chromosomes by phosphorylating histone H3 at Thr3. Then, AtHaspin and AUR3 together regulate proper chromosome alignment in the spindle during prometaphase/metaphase and chromosome segregation (Kozgunova et al., 2016). In this process, the cohesin complex, containing SMC1, SMC3, and RAD21, can contribute to chromosome alignment, while TOPII can release these cohesins from chromosomes to allow for chromosome segregation (Higgins, 2010; Kamenz and Hauf, 2017). In addition, the condensin complex, comprising SMC2 and SMC4, also ensures chromosome condensation and proper segregation (Wang H. et al., 2019). Ndc80 and Nuf2, as components of the kinetochore complex, are localized at the outer kinetochore, connecting spindle fibers to the kinetochore as well as mediating chromosome segregation during cell division (Shin et al., 2018). In mitosis, the SAC core proteins Mad2 and Mad3.2 are recruited to the kinetochore that is unattached to the spindle; then, Mad2 and Mad3.2 together with Mad3.1 may bind CDC20 to form the mitotic checkpoint complex (MCC) to inhibit the activity of APC/C (Komaki and Schnittger, 2017). Until the kinetochore is correctly attached to the spindle, CDC20 is released, which then activates APC/C for the removal of cohesin, thus promoting entry into anaphase (Singh et al., 2014). Moreover, AUR3 is present at kinetochores and is involved in kinetochore assembly during mitosis (Lermontova et al., 2015). Therefore, during exposure to H₂O₂ or combined H₂O₂ and NaCl treatment, all these up-regulated genes may coordinate to control proper condensation and segregation of chromosomes for successful cell division.

AUR2 is another member of the Arabidopsis Aurora kinase family, which is associated with spindle assembly, phragmoplast organization, and cell plate orientation during mitotic division (Demidov et al., 2014). In this process, AUR2 activity may be controlled by its upstream regulators AtHaspin (Kozgunova et al., 2016) and TPX2 (Petrovska et al., 2012). TPX2 is a MAP with multiple functions in microtubule organization, and can activate and phosphorylate AUR2. The TPX2–AUR2 complex can colocalize on spindle microtubules during mitosis and thereby control cell division (Petrovska et al., 2012).

Besides TPX2, many other MAPs regulate microtubule dynamics for the proper formation of different MT arrays during the cell cycle. AtMAP65-3 begins to accumulate at the narrow midline of the spindle at metaphase and is involved in antiparallel MT bundling at the phragmoplast midline at telophase. Similar to AtMAP65-3, kinesin-5 interdigitates microtubules at both spindle and phragmoplast midline (Bannigan et al., 2007). The kinesin-7 family member NACK1 participates in phragmoplast organization by recruiting MAPKKK (ANP) to the phragmoplast midline and activating the MAP kinase cascade during the late mitosis phase, which is critical for cell plate formation (Sasabe et al., 2015). The kinesin-12 family members POK1 and POK2 are important for PPB function (Rasmussen et al., 2011), and PAKRP1 and PAKRP1L are involved in MT interdigitation at the phragmoplast midline (Lee et al., 2007). ATK1 and ATK5 are two minus-end-directed kinesin-14s which are essential in spindle assembly and function (Ambrose and Cyr, 2007). Therefore, the H₂O₂-induced expression of all these MAP genes contributes to the assembly of microtubule arrays and the progression of cell division.

The abovementioned cell cycle genes are associated with mitotic cell cycle, chromatin dynamics, and microtubule-related processes in promoting cell proliferation and maintaining the normal structure of chromosomes. As cell proliferation and cell expansion are the main driving forces in leaf growth, the up-regulation of these genes may maintain plant growth in cope with subsequent stresses; however, the mechanisms by which low levels of H₂O₂ promote plant cell cycle progression and growth remain unclear.

Plants first adopt a series of molecular mechanisms in response to osmotic stress when exposed to high salinity, such as regulating the expression of many genes involved in stomatal closure and synthesizing osmotically protective substances (Feng et al., 2016).

In HW vs. WW, some of the identified DEGs were osmotic stress-responsive (Table 1 and Supplementary Tables 18, 19), but relatively few of these types of genes were up-regulated. Of these, HB33 was induced mainly by H₂O₂, while the transcript increases under HN and WN were not distinctly different from WW (Figure 4B). Some studies have shown that HB33 is a positive regulator in ABA, mediating plant growth and development as well as response to different abiotic stresses, such as osmotic stress (Wang et al., 2011). In HN vs. WW, more osmotic stress-responsive DEGs were up-regulated (Table 2 and Supplementary Tables 20, 21). RD29A, RD29B, and P5CS1 are typical representatives, but the magnitude of their expression increases were less than in WN vs. WW (Figure 4B). We speculate that pretreatment with H₂O₂ perhaps mitigates the osmotic stress caused by subsequent salt stress. RD29A and RD29B are osmotic stress-related marker genes, and their encoding proteins RD29A and RD29B act as protective molecules in response to osmotic stress. P5CS1 encodes a key enzyme in proline biosynthesis and promotes proline accumulation to confer plant osmotic stress resistance (Feng et al., 2016). Comparing the DEGs in HW vs. WW with those in HN vs. WW, we found that OSM34 and GRDP2 were significantly up-regulated under both conditions (Tables 1, 2); however, their expression patterns were different. The expression level of the former in HN was higher than that in HW, while for the latter, the converse was observed (Figure 4B). Osmotin34 (OSM34) encodes osmotin to combat osmotic stress (Sharma et al., 2013), and AtGRDP2 encodes a short glycine-rich domain protein which may improve the growth of plants under osmotic stress (Ortega-Amaro et al., 2014).

To investigate which genes or biological processes are involved in the H₂O₂-primed oxidative stress tolerance of plants, we performed GO analysis on all DEGs in HW vs. WW and HN vs. WW, and found that some DEGs which may protect plants from damage during subsequent salt stress that were activated by H₂O₂ (Supplementary Tables 18–21).

Comparing the DEGs in HW vs. WW with those in HN vs. WW, we found that ACS6 and TT4 were jointly up-regulated under both conditions (Tables 1, 2); however, their expression patterns were distinct, with ACS6 expression higher in HW than in HN, whereas TT4 expression increased progressively with HW and HN (Figure 4B). ACS6 is one of the most important genes in ethylene biosynthesis, controlling the level of ethylene. Datta et al. (2015) reported that ACS6 was significantly up-regulated during the glutathione–ethylene interaction in response to salt stress. Our transcriptomic data implies that H₂O₂ pretreatment may elevate ethylene production by ACS6 transcription increase to activate the ROS-detoxifying system in defending against subsequent salt stress. The expression pattern of ACS6 further confirmed that ethylene may participate in H₂O₂-primed redox balance reconstruction under salt stress. TT4 encodes chalcone synthase (CHS), a key enzyme involved in the biosynthesis of flavonoids. It has been reported that the up-regulation of TT4 led to an increase in anthocyanin synthesis. Anthocyanins can function as antioxidants, helping plants to scavenge ROS and maintaining redox homeostasis during salt stress. The DEGs only up-regulated in HW vs. WW, such as CRWN2, CRWN3, CRWN4, SOS6, and RBOHD, were all clustered to K8 (Figures 2F, 4B and Table 1). CRWNs constitute a small gene family containing only CRWN1–4 members, three of which were detected in our transcription data. CRWN proteins have been reported to maintain the size and morphology of the nucleus in order to promote the normal growth of plants (Wang et al., 2013). We speculate that H₂O₂ induces the expression of CRWN2, CRWN3, and CRWN4, which may be positive regulators of oxidative stress tolerance, inhibiting ROS production and DNA damage during subsequent salt stress (Wang Q. et al., 2019). RBOHD is a key member of the RBOHs family, where RBOH-mediated spatiotemporal control of ROS production is required for appropriate cell elongation. We consider that H₂O₂ pretreatment promoted moderate expression of RBOHD, thereby promoting the growth and development of plants and improving salt tolerance (Marino et al., 2012; Suzuki et al., 2012). SOS6 has an important role in osmotic stress tolerance and may be involved in the regulation of ROS levels under oxidative stress (Yang et al., 2016). The DEGs only in HN vs. WW include VTC2, FSD3, GPX7, and GSTUs, all of which were up-regulated (Figure 4B and Table 2). VTC2 encodes GDP-L-galactose phosphorylase, catalyzing the conversion of GDP-L-Gal into L-Gal, which is considered to be a committed step in ascorbate biosynthesis (Koffler et al., 2014). Ascorbate, as a relatively abundant small-molecule antioxidant in plants, can detoxify ROS throughout the cell. FeSOD is one of the three major classes of SOD. Overexpression of FSD3 results in great tolerance to oxidative stress through scavenging of ROS (Myouga et al., 2008). GPXs are important ROS scavengers due to their broad substrate specificity and high affinity for H₂O₂. The up-regulated expression of GPX7 has been shown to be important for maintenance of redox balance in the cell (Chang et al., 2009). Glutathione S-transferases (GSTs) protect plants from oxidative damage and enhance the antioxidant capacity of plants. We also found that several GSTUs were up-regulated, such as GSTU24.

The plant cell wall is the first defense against external environmental stresses. To check whether H₂O₂ pretreatment invoked the expression of genes encoding for cell wall components, we further analyzed the transcription data and found that, based on GO analysis, some genes could be primed by H₂O₂ pretreatment (Supplementary Tables 18, 19), whereas some genes were regulated by subsequent salt stress (Supplementary Tables 20, 21). These genes are involved in the regulation of the synthesis of various components of the cell wall, causing the cell wall to harden. The formation of a physical barrier protects plant cells from further dehydration and death under salt stress, thereby resisting salt stress.

The identical DEGs in HW vs. WW and HN vs. WW, such as AtGH9C2 and GRP14, were up-regulated (Figure 4B and Tables 1, 2). AtGH9C2 is a class C endo-1, 4-β-glucanase (cellulase). Some studies have found that such endoglucanases affect cell wall development by promoting cell wall crystallization processes (Glass et al., 2015). GRP14 is an important structural protein which is widely found in plant cell walls, and the expression of GRP14 helps in cell wall remodeling when plants are exposed to salt stress (Le Gall et al., 2015). The DEGs up-regulated only in HW vs. WW, such as AtGH9B1, AtGH9B13, GRP19, and FUT4, were clustered to K7 (Figures 2F, 4B and Table 1). AtGH9B1 and AtGH9B13 are both class B endoglucanases that play an important role in cell wall relaxation during cell growth and expansion (Tsabary et al., 2003; Urbanowicz et al., 2007). GRP19 is a glycine-rich protein with similar function to GRP14 (Le Gall et al., 2015). FUT4 is an arabinogalactan (AG)-specific fructosyltransferase (FUT), which is responsible for the fructosylation of proteins glycosylated with arabinogalactan (AGPs) in leaves, which maintains proper cell expansion and root growth under salt stress conditions (Tryfona et al., 2014).

The DEGs up-regulated only in HN vs. WW, such as AtGH9B8 and UGP, were the representatives of cluster K2 (Figures 2F, 4B and Table 2). Like AtGH9B1 and AtGH9B13, AtGH9B8 also belongs to class B endoglucanases and is involved in cell wall relaxation during cell growth and expansion (Mele et al., 2003; Urbanowicz et al., 2007; Tryfona et al., 2014). The up-regulated expression of UGP1 can promote the biosynthesis of the cell wall, maintaining plant growth under salt stress and, therefore, increasing the salt tolerance of plants, after H₂O₂ pretreatment.

Following H₂O₂ pretreatment, some transcription factors were accumulated in plants to defend against subsequent high-salt stress. After GO analysis, we identified some transcription factors encoding genes in HW vs. WW (Supplementary Tables 18, 19) and HN vs. WW (Supplementary Tables 20, 21), respectively. The gene number in the former accounts for a large proportion of the DEGs, with significantly more than the latter. These transcription factors can be divided into five categories, belonging to the ERF, MYB, WRKY, HSFA, and bHLH families.

The identical DEGs up-regulated in HW vs. WW and HN vs. WW included MYB112, ERF5, ERF15, and bHLH100 (Tables 1, 2); however, the expression levels of bHLH100 and ERF5 were higher in HW than in HN, while the converse was the case for ERF15 and MYB112 (Figure 4B). MYB112 is a member of the R2R3 MYBs. The up-regulated expression of MYB112 promotes the accumulation of anthocyanins, which can respond to different abiotic stresses, including oxidative stress, osmotic stress, and high-salt stress (Lotkowska et al., 2015). The DEGs up-regulated only in HW vs. WW, including WRKY33, ERF6, ERF104, bHLH101, and HSFA4A, clustered to K8 (Figures 2F, 4B and Table 1). Some studies have shown that WRKY33, ERF6, and ERF104 can regulate the expression of salt-tolerant genes in different signaling pathways, and that the overexpression of these genes can increase salt tolerance in plants (Jiang and Deyholos, 2009; Vogel et al., 2014; Van den Broeck et al., 2017). bHLH101 can increase the oxidative stress tolerance of plants (Noshi et al., 2018). HSFA4A encodes a member of heat stress transcription factors (Hsfs), certain members of which have been shown to function as ROS-dependent redox sensors, controlling gene expression during oxidative stress. Although we do not know whether HSFA4A is such a redox sensor, there has been evidence showing that it plays key roles in a variety of stress signaling pathways, and its overexpression enhances a variety of stress tolerances, including salt stress, osmotic stress, oxidative stress, and heavy metal stress (Perez-Salamo et al., 2014; Lin et al., 2018).

The DEGs up-regulated only in HN vs. WW include ERF4 and MYB29 (Figure 4B and Table 2). ERF4 are important molecules in the signaling pathways of ethylene and jasmonic acid, regulating the expression of a large number of genes involved in many plant defense mechanisms. Overexpression of ERF4 has been shown to increase salt and drought stress tolerance in plants (Seo et al., 2010).

To further validate whether the expression of DEGs was induced by only H₂O₂ pretreatment or both H₂O₂ pretreatment and subsequent salt stress, we selected 15 genes involved in signal transduction (Figure 5A), response to osmotic stress (Figure 5B), response to oxidation–reduction process (Figure 5C), cell wall organization (Figure 5D), and transcription factors (Figure 5E) for RT-qPCR. It was verified that the expression trends of these genes tested by RT-qPCR were highly consistent with the transcriptome data; therefore, the conclusions obtained from the transcriptome analysis are reliable. Primer information is presented in Supplementary Table 1.

Salinity-induced oxidative stress in plants is associated with ROS overproduction. Through transcriptome analysis, we found some genes involved in the scavenging of ROS, such as FSD3 and GPX7.

The total SOD, and GPX activities in leaves are shown in Figure 6. The results show that the SOD activities in the HN and WN groups were greatly higher than those in the HW and control (WW) groups, which exhibit consistency between SOD activity and FSD3 expression. Under stress treatment, SOD activity in the HN group increased greatly compared with that in the WN group (Figure 6A). GPX activities in the HN group were significantly higher than in the WN group and, similarly, markedly higher in the HW group than in the WW group (Figure 6B). These results for GPX activity coincided with GPX7 expression.