The grape suspension cell system was established from Vitis vinifera L. Cabernet Sauvignon berry skin-derived callus described in our previous study (Cheng et al., 2022). The grape cell suspensions were cultivated according to the method of Osuga et al. (1999) with slight modifications. Briefly, the calli were transferred to 250 mL flask containing 60 mL of B5 basic liquid medium (3.21 g L^-1, pH 5.8 to 6.0), supplemented with 20 g L^-1 of sucrose, 2.5 g L^-1 of acid-hydrolyzed casein, 0.2 mg L^-1 of kinetin (KT), and 0.1 mg L^-1 of naphthaleneacetic acid (NAA). The suspension cells were maintained at 25 ± 1°C in the dark with the a shaker speed of 95 rpm. Subcultures were performed every 14 days at a 1:2 ratio (v/v). For salinity treatment, NaCl was added to the fresh maintaining medium at the indicated concentration. Suspension cells were subjected to different NaCl concentrations (0 mM, 75 mM, and 150 mM) in 250 mL flasks, designated as the control group (Ctrl), low salinity group (L), and high salinity group (H), respectively. Each treatment was performed in three biological replicates. As the activation of anthocyanin pathway in grapevine is light-dependent, the experimental suspension cells were cultivated under the same condition as above, except for a 16 h/8 h light/dark cycle with a light intensity of 125 μmol m^-2 s^-1. Cells were harvested at the specified time intervals by vacuum filtration, washed with cold distilled water, and weighed. The samples for metabolite and gene expression analysis were frozen in liquid nitrogen and stored at -80°C until use.

Two mL of suspension cell cultures were subjected to gravity sedimentation for 10 min, and the supernatant was discarded. The cell pellets were extracted with 1 mL of 70% (v/v) acetone/water containing 0.5% (v/v) ascorbic acid by vortexing for 30 min. The resulting mixture was centrifuged at 8,000g for 5 min, and the supernatant was collected. The extraction was repeated twice, and the combined supernatants were evaporated by nitrogen flow. The dry extracts were re-dissolved in 400 μL of 50% (v/v) methanol/water to obtain soluble PAs. The remained cell pellets after soluble PAs extraction contained insoluble PAs. Soluble and insoluble PAs were quantified using DMACA method and Butanol-HCl lysis assay, respectively, as described in our previous study (Cheng et al., 2022).

To extract total anthocyanins from two mL of suspension cell cultures, the medium was removed and the cells were sonicated in 1 mL of 50% (v/v) methanol/water with an ice water bath for 20 min. The resulting homogenate was centrifuged at 8,000g for 5 min, and the supernatant was evaporated under the nitrogen flow. The dry extracts were re-dissolved in 400 μL of 50% (v/v) methanol/water. The quantification of total anthocyanins followed the pH-differential method optimized for grape cell cultures (Cheng et al., 2021).

To analyze the composition of PAs and anthocyanins, suspension cells were ground under liquid nitrogen and freeze-dried. For the cleavage of both soluble and insoluble PAs, 30 mg of dry powder was suspended in phloroglucinolysis buffer (0.5% (w/v) ascorbic acid, 0.3 N HCl, and 50 g/L phloroglucinol in methanol) and incubated at 50°C for 20 min. The reaction was stopped by adding 200 mM sodium acetate. After centrifugation at 8,000g for 5 min, the supernatant was evaporated under the nitrogen flow and re-dissolved in 400 μL of 50% (v/v) methanol/water. For the extraction of anthocyanins, 30 mg of dry powder was sonicated in 1 mL of 50% (v/v) methanol/water with an ice water bath for 20 min. The mixture was then centrifuged at 8,000g for 5 min and the supernatant was evaporated under the nitrogen flow and re-dissolved in 200 μL of 50% (v/v) methanol/water. Both PA building blocks and anthocyanins were detected using an Agilent 1200 HPLC system coupled with an Agilent 6410 triple quadrupole (QqQ) mass spectrometer. A Poroshell 120 EC-C18 column (150 × 2.1 mm, 2.7 μm, Agilent Technologies) was used for metabolite separation. Targeted LC/MS analysis of PA building blocks and anthocyanins with multiple reaction monitoring (MRM) followed the method established in our previous study (Li et al., 2017).

The total RNA of grape suspension cells was isolated using the Universal Plant Total RNA Extraction Kit (BioTeke, Beijing, China). The quality and quantity of the RNA samples were assessed by an Agilent Bioanalyzer 2100 system (Agilent, CA, USA). The cDNA library construction and Illumina Novaseq sequencing were performed by Novogene Biotech Co., Ltd (Beijing, China) following the standard procedures.

To obtain clean reads, raw data were filtered by removing low-quality reads, adapter sequences, and poly-N sequences. The clean reads were then aligned to the Vitis vinifera reference genome (version V1) using HISAT2 v2.0.5 with default parameters. Reference-based transcript assembly and novel transcript prediction were performed by StringTie v1.3.3b. Read counts for each gene were obtained by FeatureCounts v1.5.0-p3. Gene expression levels were estimated by calculating the fragments per kilobase of transcript per million mapped reads (FPKM) values. The heatmaps for visualizing gene expression levels were plotted using the Heatmap tool on the Galaxy web-based platform (https://usegalaxy.org/).

To validate the RNA-seq results, the same RNA-seq samples were used for qRT-PCR analysis. The RNA was reverse transcribed using HiScript® II Q RT SuperMix for qPCR (+gDNAwiper) (Vazyme, Nanjing, China) following the manufacturer’s instructions. The qRT-PCR reactions were performed using SYBR qPCR Master Mix (Vazyme, Nanjing, China) on an ABI 7300 real-time system (Thermo Fisher, USA). The qRT-PCR program consisted of an initial denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 10 s and annealing/extension at 60°C for 31 s. VvUbiquitin1 was used as a reference gene to normalize the expression levels of the target genes (Bogs et al., 2005). The relative expression levels were calculated using the 2^-ΔΔCT method. The gene-specific primers used in the study are listed in Supplementary Table 1 .

To identify differentially expressed genes (DEGs) between two experimental groups, the R package DESeq2 1.20.0 was used with the default parameters. The Benjamini and Hochberg method was applied to compute the adjusted P-value (padj), which was then used together with the Fold change (FC) to detect significant differences in gene expression. The criteria for significant differential expression were padj < 0.05 and |Log2(FC)| > 1. The DEGs between two groups were visualized by volcano plots using Hiplot (https://hiplot.org), an online platform for interactive data visualization. The DEGs were subjected to GO and KEGG pathway enrichment analysis using TBtools, a software suite for biological data analysis (Chen et al., 2020). The enrichment results were visualized by bubble plots, rose charts, and bee plots using Hiplot. The functionally grouped GO analysis was performed using ClueGO, a Cytoscape plug-in that integrates GO terms and creates functionally organized GO networks (Bindea et al., 2009). The parameters for ClueGO analysis were set as follows: overlap threshold 50%, kappa score 0.5, and the default settings for the other options.

Gene co-expression analysis was performed using the R package WGCNA 1.72 with FPKM value matrix as the input. The soft threshold was selected as 6 based on the scale-free topology criterion and the mean connectivity of the network. The minimum number of genes allowed to be included in the module was set to 50. Other parameters for the analysis were kept as the defaults. The co-expression network was visualized by Cytoscape V3.6.1.

Transcription factor genes in DEGs were identified using BLASTN against Vitis vinifera database in Plant Transcription Factor Database (PlantTFDB) V5.0. The identified transcription factor genes were further annotated using BLASTX against a combined database of NCBI nr with the sequences of experimentally characterized flavonoid transcription factors retrieved manually based on the published literatures. NCBI BLAST standalone version 2.13.0+ was used in the above analysis with default parameters.

The statistical analysis of the data was conducted using the SPSS software version 21.0. Data were plotted using GraphPad software version 8.0.2.

The effect of salinity stress on flavonoid metabolism in grapes was investigated using a suspension cell system derived from berry skins of Vitis vinifera L. Cabernet Sauvignon. To determine the optimal salt concentration for inducing stress response, we treated the suspension cells with different concentrations of NaCl for two weeks, and measured the fresh weight as an indicator of cell growth. We found that 75 mM NaCl inhibited cell proliferation, and 150 mM NaCl further reduced cell growth ( Figure 1A ). Higher concentrations of NaCl (200 mM, 300 mM, and 400 mM) did not cause additional decrease in fresh weight compared to 150 mM NaCl ( Figure 1B ). Based on these results, we inferred that 75 mM NaCl was sufficient to induce a significant stress response in grape suspension cells, and that 150 mM NaCl triggered a more severe one. Therefore, it was worth studying the response of PA and anthocyanin biosynthesis under both treatments. For convenience, we used low salinity (L) and high salinity (H) to represent 75 mM and 150 mM of NaCl, respectively.

To investigate the effect of NaCl stress on PA and anthocyanin biosynthesis in grape suspension cells, we performed time-series assays to monitor the content of these metabolites throughout a culture cycle. We found that soluble PAs and insoluble PAs remained relatively stable under both L and H treatments for the first 6 days, and then increased from day 6 onwards ( Figure 2 ). Anthocyanins in H group also accumulated mainly from day 6 to day 14. In contrast, L stress did not induce significant amounts of anthocyanins in suspension cells, similar to the control group (Ctrl) ( Figure 2 ). The color of the suspension cells reflected the changes in anthocyanin content. Correspondingly, suspension cells under H at day 4 and day 7 (H-4d and H-7d) showed a deep red color, which was not observed in Ctrl at either day 4 or day 7 (Ctrl-4d and Ctrl-7d) ( Figure 3A ). The cells treated with L for 7 days (L-7d) had a slight pink hue, which was more obvious than that of L samples at day 4 (L-4d) ( Figure 3A ). In addition, L-4d and H-4d samples exhibited growth inhibition phenotype, as indicated by their lower fresh weight than Ctrl-4d ( Figure 3B ). Unlike L treatment group, the fresh weight of H samples did not increase from day 4 to day 7 ( Figure 3B ). These results suggest that the degree of salinity stress experienced by the L and H samples differed, at least from day 4 onwards.

To gain the insight into the composition of PA building blocks in the current suspension cell system, we directly subjected the ground cell samples to phloroglucinolysis followed by LC/MS analysis. The flavan-3-ol monomers detected by this technique consist of free flavan-3-ols and PA starter units, while phloroglucinol-flavan-3-ol adducts are derived from PA extension units. Quantification results revealed that (-)-epicatechin extension unit was the predominant PA building block in grape suspension cells with or without salinity treatments, while (-)-epigallocatechin-, galloyl-(-)-epicatechin- and (+)-catechin-type subunits together accounted for only one-tenth of the total flavan-3-ols ( Figure 4A ). This suggested that the flux to 2,3-cis flavan-3-ols was preferred in the suspension cells, which was the common case in grape berry skins. However, the low levels of (-)-epigallocatechin reflected the fact that C5’-hydroxylation was somewhat blocked in the suspension cell system under NaCl stress. Compared with L treatment, H stress was more effective in inducing total flavan-3-ol accumulation at both day 4 and day 7 ( Figure 4A ). The total flavan-3-ol level of H-7d was twice that of H-4d, while there was no further increase in total flavan-3-ol content in either L or Ctrl groups from day 4 to day 7 ( Figure 4A ).

To examine anthocyanin composition, we subjected 50% methanol/H₂O (v/v) soluble extracts to LC/MS analysis following the method we developed previously (Li et al., 2017). Consistent with the cell color shown in Figure 3A , LC/MS data revealed that anthocyanins were mainly accumulated in H-4d and H-7d samples. The profiles of anthocyanins in these two samples were nearly identical, although total anthocyanin concentration in H-7d was one-third higher than in H-4d ( Figure 4B ). Anthocyanins induced in suspension cells were predominantly cyanidin-based. More than half of these anthocyanins were 3’-O-methylated cyanidin-3-O-glucoside (also known as peonidin-3-O-glucoside), of which over forty percent were further acetylated and coumaroylated ( Figure 4B ). Thus, in despite of a deficiency of 3’5’-hydroxylated anthocyanins, the common types of anthocyanin modifications in grape berry skins existed in the suspension cells under NaCl stress.

Transcriptome analysis was performed with Illumina RNA-Seq to investigate the response of gene expression to NaCl stress in the grape suspension cell system. Transcripts were assembled from pair-end reads based on the grape PN40024.V1 genome, and their expression levels were normalized by Fragments Per Kilobase Million (FPKM). Differentially expressed genes (DEGs) between two given conditions were identified using DESeq2 (|Log2(FC)| > 1, padj < 0.05). qPCR data and FRKM values of the selected flavonoid pathway genes correlated well ( Supplementary Figure 1 ), suggesting that the RNA-Seq experiment was reliable.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis showed that flavonoid biosynthesis was the most significantly enriched pathway by H treatments, followed by phenylpropanoid biosynthesis and plant hormone signal transduction; DEGs between samples of L and Ctrl groups were not enriched in any KEGG terms, except for a few genes of mitogen-activated protein kinase (MAPK) signaling pathway in L-4d ( Figure 5A ). Correspondingly, the expression levels of the structural genes in flavonoid pathway were generally increased with the increasing NaCl concentration, which was consistent with the salt dose-dependent accumulation of PAs and anthocyanins ( Figure 5B ). Compared with other flavonoid structural genes, the response of F3’5’H genes to salinity stress was more complex in the suspension cells. Two of five isoforms of F3’5’Hs detected by the RNA-Seq assay were mainly expressed in H-4d and H-7d samples, while the transcription of the other three was activated only in H-4d ( Figure 5B ). F3’5’H competes with F3’H for naringenin or dihydrokaempferol to produce 3’5’-hydroxylated backbones of PAs and anthocyanins (Bogs et al., 2006). Similar to the results from a previous study on the expansion and sub-functionalization of F3’5’H (Falginella et al., 2010), the unsustainable expression of F3’5’H copies was likely responsible for the negligible levels of delphinidin-derived anthocyanins and epigallocatechin-based PAs in the suspension cells.

The PA and anthocyanin biosynthesis related transcription factors, including MYB, bHLH/MYB and WD40/WDR family proteins, have been extensively studied in grapes. However, these works have mostly been carried out in the context of berry development, rather than under stress conditions. VvMYB5a and VvMYB5b activate the general flavonoid pathway in grapevine (Deluc et al., 2006, 2008), but their gene expression levels were not affected by salinity stress in the grape suspension cells ( Figure 5C ). This suggested that additional regulators were required for channeling the flux to flavonoid pathway in grape cells under salinity stress. Two functional isoforms of MYBA genes, namely VvMYBA1 and VvMYBA2, are essential for anthocyanin synthesis in red grape cultivars (Walker et al., 2007), and their expression levels were up-regulated for at least 60 times under H treatment, compared to the controls ( Figure 5C ). This corresponded to the high expression levels of anthocyanin genes, including DFR, ANS, 3GT, AOMT and AAT, under H treatment at both day 4 and day 7 ( Figure 5B ). The expression level of VvMYBPA1, a gene that activates PA synthesis, increased under H treatment, but not under L treatment. This was in line with the expression patterns of VvANR and VvLARs. Whereas the gene encoding VvMYBPA2, another key positive transcription factor in PA pathway (Terrier et al., 2009), was not responsive to the salinity stress, the gene encoding VvMYBPAR, phylogenetically clustered with VvMYBPA2 (Koyama et al., 2014), was up-regulated in H-4d sample. Although two of three MYBC2 PA pathway repressor genes, namely VvMYBC2-L1 and VvMYBC2-L2 (Cavallini et al., 2015), were also activated by salinity stress, their inhibition of PA accumulation was not significant in the grape suspension cell system as indicated by the metabolite data. VvMYBC2 and VvMYBPA1 are the targets of VvbHLH93, which inhibits the expression of VvLARs in PA pathway and promotes the transcription of anthocyanin genes (Cheng et al., 2022). VvbHLH93 expression level was down-regulated in both H-4d and H-7d samples, which partially explained the similar expression patterns of VvMYBPA1 and VvMYBC2 genes under salinity stress ( Figure 5C ). In addition, VvMYB86, a gene oppositely regulating PA and anthocyanin biosynthesis (Cheng et al., 2021), was neither responsive to L nor H salinity stress. This was consistent with the similar accumulation patterns of PA and anthocyanin in cells under salt stress conditions.

We used a Venn diagram to compare the number of DEGs among the four conditions: Ctrl-4d vs. L-4d, Ctrl-7d vs. L-7d, Ctrl-4d vs. H-4d and Ctrl-7d vs. H-7d. The diagram showed that H treatment had a much stronger effect on the transcriptome of the suspension cells than L at both time points ( Figure 6A ). Only eight genes were commonly regulated by both L and H treatments: VvCORA-like, protein phosphatase 2C (VvPP2C), Kunitz trypsin inhibitor 2 (VvKTI2), VvMYBA3, VvAOMT, glutathione S-transferase 4 (VvGST4), and two isoforms of polygalacturonases (VvPG1 and VvPG2). These eight genes were all up-regulated by L treatment, and their transcriptions were further enhanced by H treatment ( Figure 6B ). This suggested that they were essential for the response to NaCl stress.

To investigate the potential associations among the eight genes, weighted correlation network analysis (WGCNA) was performed, and 16 modules were identified in total. Among them, VvCORA-like, VvMYBA3, VvAOMT, VvGST4, VvPG1 and VvPG2 were clustered in the brown module, while VvPP2C and VvKTI2 were assigned to the purple and red modules, respectively ( Supplementary Table 2 ). Eigengene-based clustering analysis revealed that the brown, red and purple modules were closely related ( Figure 6C ), indicating that they might share common biological functions. The flavonoid pathway of KEGG term was significantly enriched combining the genes from these three modules ( Figure 6D ). In particular, the majority of flavonoid genes were included in the brown module, while the red and purple modules contained only around 15% of the flavonoid genes, such as a VvCHI isoform (VIT_13s0067g03820), a VvDFR isoform (VIT_18s0001g12820), VvANS, and VvANR. This suggested that the brown module was more relevant to salinity-induced anthocyanin and PA biosynthesis.

Within the brown module, VvAOMT catalyzes anthocyanin methylation (Hugueney et al., 2009), and VvGST4 is a PA transporter and the final enzyme to product anthocyanidin (Pérez-Díaz et al., 2016; Eichenberger et al., 2023). Although the SNPs within the VvMYBA3 locus were tightly associated with grape berry color, the overexpression of VvMYBA3 in the grape suspension cells did not induce the expression of Vv3GT and anthocyanin accumulation (Walker et al., 2007; Fournier-Level et al., 2009; Guo et al., 2019). Moreover, the functions of VvCORA-like, VvPG1 and VvPG2 in vivo were still unknown. However, since they co-existed with flavonoid genes in the same module, we speculated that they might have a role in anthocyanin and PA biosynthesis under salinity stress. We retrieved the top 20 co-expression genes for each of these six genes based on the weight values of the adjacency matrix from WGCNA. The gene co-expression network was constructed accordingly, and only the nodes with more than one edge were kept for visualization. The genes in this network were mainly related to flavonoid biosynthesis and cell wall metabolism ( Figure 6E ). The hub genes of this network included the ones encoding shikimate kinase and acidic class III chitinase, which were required for the synthesis of flavonoid precursors and the defense against pathogens and oxidative stress in grapevine, respectively (Almagro et al., 2014; Gonçalves et al., 2019). However, the mechanism of how cell wall composition influences flavonoid biosynthesis under salinity stress was still unclear. Within the network, VvMYBA3 was directly linked with the well-characterized anthocyanin genes, VvAOMT, VvF3H1 and VvMYBA2, as well as a cellulose synthase-like coding gene (VvCSLH1) and two lignin synthesis genes, caffeoyl-coenzyme A O-methyltransferase (VvCCoAOMT) and hydroxycinnamoyltransferase (VvHCT). VvAAT was co-expressed with VvAMOT and VvGST4, which were also directly connected with each other. VvPG2 was linked to VvAMOT and VvGST4 through VvCYTb5 and VvF3H2, respectively. VvLAR1 and VvLAR2 were both the co-expression partners of VvPG1. The two VvPGs were also associated with the aid of VvUGT84A13, whose homolog in pedunculate oak was identified as a candidate gene involved in gallotannin biosynthesis (Mittasch et al., 2014). CORA-like proteins are believed to play roles in magnesium uptake and storage in plants, and their gene expression is enhanced by stress conditions (Chen and Ma, 2013; Jha et al., 2021). Our data showed that VvCORA-like was mainly co-expressed with VvLARs, VvF3H2, VvHCT and two aminotransferase coding genes, suggesting that it was potentially involved in phenolic metabolism under salinity stress.

Besides the several genes that were constitutively responsive to salinity stress, many genes were differentially expressed depending on the NaCl concentration. The GO enrichment analysis revealed that the higher level of NaCl suppressed the expression of genes involved in cell growth processes, such as DNA replication, cell cycle regulation, spindle assembly, RNA primer synthesis, nuclear division inhibition, and meiosis II ( Figure 7 ). This further supported the findings that high salt concentrations hindered grape suspension cell proliferation at the transcriptional level. Compared with L treatment, the H treatment enhanced the expression of genes related to phenylalanine and flavonoid metabolism in grape suspension cells at both day 4 and day 7 ( Figure 7 ). This suggested the key roles of phenylalanine and flavonoid pathways in coping with the increased salt stress, possibly by enhancing the accumulations of PA and anthocyanin in cells.

To gain insights into the regulators responsible for the NaCl-dose dependent activation of PA and anthocyanin pathways in grape suspension cells, we obtained differentially expressed transcription factor genes (|Log2(FC)| > 1, padj < 0.05) from L-4d vs. H-4d and L-7d vs. H-7d comparison groups, using BLASTN against Plant Transcription Factor Database (PlantTFDB) V5.0. Seventeen transcription factors were further annotated as flavonoid pathway relevant, using BLASTX against a combined database of NCBI nr and experimentally characterized flavonoid transcription factors from crops and model species ( Table 1 ). Five flavonoid-relevant transcription factors exhibited NaCl-dose dependent expression only at day 4 but not at day 7: VvMYBC2-L3 (VIT_14s0006g01620), which was down-regulated in H-4d compared with L-4d; and the homologs of Fragaria × ananassa FAV1 (VIT_11s0037g00010), Arabidopsis NAC32 (VIT_18s0001g02300), Prunus persica NAC25 (VIT_00s0375g00040) and VvNAC17 (VIT_19s0014g03290) were up-regulated in H-4d. Among these three NAC transcription factors, AtNAC32 represses the stress-induced anthocyanin synthesis (Mahmood et al., 2016), while PpNAC25 promotes anthocyanin accumulation (Geng et al., 2022). Moreover, Badim et al. (2022) demonstrated that overexpressing VvNAC17 in grape suspension cells induces the biosynthesis of both PA and anthocyanin. FaRAV1 has been shown to activate the general phenylpropanoid and flavonoid pathways, while its homolog in Malus domestica (MdRAV1) is found to repress PA biosynthesis (Li et al., 2020; Zhang et al., 2020). At both day 4 and day 7, the expression of VIT_01s0011g03070 (a homolog of MdRVA1), VIT_18s0001g09250 (a homolog of the anthocyanin pathway repressor AtLBD38) (Rubin et al., 2009), and VIT_08s0058g01390 (a flavonoid pathway repressor, VvWRKY70) were down-regulated in the H samples compared with the L groups. The expression levels of the homologs of anthocyanin pathway activators OsMYB3 from rice (VIT_04s0008g01800) (Zheng et al., 2021) and MdERF109 from apple (VIT_03s0063g00460) (Ma et al., 2021), the classical flavonoid activators VvMYBPA1 and VvMYBA1/2/3, and that of the well characterized PA negative regulators VvMYBC2-L1 and VvMYBC2-L2 were increased with the elevation of NaCl concentration at both time points. The up-regulation of VvMYBC2-L1 expression may partially associate with the down-regulation of VvMYBC2-L3 transcript level, as VvMYBC2-L3 is a VvMYBC2-L1 negative regulator in grapevine (Cavallini et al., 2015). This indicates that VvMYBC2-L family proteins can fine-tune each other to regulate flavonoid biosynthesis under salinity stress. A homolog of Arabidopsis AtERF4 (VIT_19s0014g02240), which is proposed to promote light-induced anthocyanin biosynthesis (Koyama and Sato, 2018), did not exhibit NaCl-dose-dependent expression at day 4 but showed lower expression level at day 7 as NaCl stress increased. These findings extended our understanding of the coordination of various of flavonoid/phenepropanoid activators and suppressors to cope with different levels of salinity stress.

Salinity stress duration also had varying effects on PA content, anthocyanin accumulation, and cell fresh weight in the H and L groups ( Figures 3A , 4 ). We identified salinity stress duration-dependent DEGs from the RNA-Seq data by excluding the DEGs that were only affected by cultivation time in the control groups ( Figures 8A, D ). The Venn diagram analysis showed that the L group had fewer DEGs than the H group with increasing salinity stress duration. The GO enrichment analysis showed that the L group DEGs were mainly enriched in biological processes or molecular functions related to acyltransferase activity, oligosaccharide metabolism, response to stimulus, water deprivation, and hormone ( Figure 8B ). The KEGG enrichment analysis showed that the L group DEGs had down-regulated expression of genes involved in flavonoid, polyketide and other secondary metabolite biosynthesis, circadian rhythm, and environmental adaptation pathways ( Figure 8C ). This at least explained why the extended L treatment did not result in an increase in PA and anthocyanin.

Compared with H-4d samples, the H-7d samples showed significant changes in gene expression patterns in grape suspension cells. A total of 651 differentially expressed genes (DEGs) were identified, of which 180 were up-regulated and 471 were down-regulated ( Figure 8D ). The GO enrichment analysis revealed that the extended H treatment duration may affect several biological processes or molecular functions, such as metal ion transport, lignin metabolism, response to stimulus, and oxidation-reduction ( Figure 8E ). These processes were likely related to the altered gene expression levels of genes associated with plasma membrane and extracellular region ( Figure 8E ). Moreover, KEGG pathway enrichment showed that these genes mainly involved in plant-pathogen interaction, environmental adaptation, polyketide biosynthesis, circadian rhythm, flavonoid biosynthesis, photosynthesis pathway, galactose metabolism and other secondary metabolism pathways ( Figure 8F ). Although flavonoid pathway was generally less active in H-7d samples compared with that of H-4d, the levels of PAs and anthocyanins were still increased as the salinity treatment prolonged ( Figure 4B ). This phenomenon might be partially attributed to the up-regulation of genes within the flavonoid pathway, as well as the enhanced transcription of genes related to photosynthesis, which produces precursors for flavonoids (Shao et al., 2022). In addition to the biological processes and pathways mentioned above, both GO and KEGG analyses suggested that transcription factors (TFs) were also enriched in the DEGs whose expression were affected by the duration of H stress ( Figure 8E ). Annotated by PlantTFDB database, there were a total of 53 TF genes (belonged to 18 protein families) differentially expressed as the H treatment time extended, with 44 being down-regulated and 9 being up-regulated ( Figure 9A ). Among these TFs, the proteins from ERF, MYB and WRKY families accounted for nearly a half. According to the biochemical evidence and bioinformatic prediction, 6 and 2 TF genes were classified into flavonoid activators and repressors, respectively ( Figure 9A ). Compared with the H-4d, the expression level of PA pathway specific activator VvMYBPA1 was increased by one-fold in the H-7d sample, while the expression of other proposed PA or anthocyanin activators, including VvHY5, VvMYBF1, and the homologs of MdWRKY40, MdERF109 and MdRAP2-4 (VIT_09s0018g00240, VIT_03s0063g00460, and VIT_18s0001g05250) (Shin et al., 2013; An et al., 2019; Li et al., 2020; Wang et al., 2020; Ma et al., 2021) was down-regulated from day 4 to day 7 under H stress ( Figure 9B ). This means that VvMYBPA1 is crucial for maintaining the PA accumulation as H treatment time extended, even though the general flavonoid pathway was less active. Moreover, the expression levels of the proposed flavonoid repressor VIT_01s0011g03070 and VIT_02s0025g01280 (homologs of MdRAV1 and AtWRKY41, respectively) (Ding et al., 2014; Li et al., 2020) were lower in H-7d sample than in H-4d, indicating that they might also contribute to the increased level of anthocyanin, from day 4 to day 7 under H stress.