The class B heat shock factor HSFB1 regulates heat tolerance in grapevine

Abstract Grape is a widely cultivated crop with high economic value. Most cultivars derived from mild or cooler climates may not withstand increasing heat stress. Therefore, dissecting the mechanisms of heat tolerance in grapes is of particular significance. Here, we performed comparative transcriptome analysis of Vitis davidii ‘Tangwei’ (heat tolerant) and Vitis vinifera ‘Jingxiu’ (heat sensitive) grapevines after exposure to 25°C, 40°C, or 45°C for 2 h. More differentially expressed genes (DEGs) were detected in ‘Tangwei’ than in ‘Jingxiu’ in response to heat stress, and the number of DEGs increased with increasing treatment temperatures. We identified a class B Heat Shock Factor, HSFB1, which was significantly upregulated in ‘Tangwei’, but not in ‘Jingxiu’, at high temperature. VdHSFB1 from ‘Tangwei’ and VvHSFB1 from ‘Jingxiu’ differ in only one amino acid, and both showed similar transcriptional repression activities. Overexpression and RNA interference of HSFB1 in grape indicated that HSFB1 positively regulates the heat tolerance. Moreover, the heat tolerance of HSFB1-overexpressing plants was positively correlated to HSFB1 expression level. The activity of the VdHSFB1 promoter is higher than that of VvHSFB1 under both normal and high temperatures. Promoter analysis showed that more TATA-box and AT~TATA-box cis-elements are present in the VdHSFB1 promoter than the VvHSFB1 promoter. The promoter sequence variations between VdHSFB1 and VvHSFB1 likely determine the HSFB1 expression levels that influence heat tolerance of the two grape germplasms with contrasting thermotolerance. Collectively, we validated the role of HSFB1 in heat tolerance, and the knowledge gained will advance our ability to breed heat-tolerant grape cultivars.


Introduction
Heat stress is broadly considered a major determining factor affecting crop growth and development, limiting production and quality [1]. Photosynthesis is a sensitive physiological process affected by high temperature. High temperature rapidly inhibits photosynthesis by changing chloroplast structure, inactivating ribulose-1,5-bisphosphate carboxylase-oxygenase (Rubisco), decreasing the abundance of photosynthetic pigments, and damaging photosystem II [2,3]. In addition, high temperature inf luences the stability of proteins and membranes, induces the accumulation of reactive oxygen species (ROS), and changes the production and signal transduction of plant hormones, resulting in transcriptomic reprogramming and metabolomic changes [4,5]. Nonetheless, plants have also evolved complex and interconnected signaling pathways and response mechanisms to high temperature [6].
Grapevines are valuable crops widely cultivated throughout the world [7][8][9]. However, during the growing season, grapevines often suffer from heat stress, that affects development and fruit metabolism, thus limiting grape yield and quality [10]. In the past, our research on grape responses to high temperature mainly focused on physiological and morphological changes [11][12][13][14]. With advances in high-throughput RNA sequencing technology, we now possess a powerful tool for investigating the global response of plants to heat stress. Liu et al. characterized some stress-related genes that encode transcription factors (TFs), antioxidant enzymes, heat shock proteins (HSPs), and glycolytic enzymes by analysing the transcriptomics and proteomics of Vitis vinifera 'Cabernet Sauvignon' under high temperature [15,16]. Jiang et al. analysed the transcriptomics and proteomics of V. vinifera 'Jingxiangyu' grapevines exposed to 35 • C, 40 • C, and 45 • C, and revealed that alternative splicing is an important post-transcriptional regulatory event during grapevine responses to heat stress [17]. Based on phosphoproteomic and acetylproteomic analyses of the 'Jingxiangyu' leaves under elevated temperature, phosphorylation of serine/arginine-rich splicing factors is involved in heat response [18]. By investigating the effects of high temperature on grape berries based on proteomic and metabolomics analyses, Lecourieux et al. found 592 differentially accumulated proteins in Cabernet Sauvignon plants [19]. However, previous research has mostly focused on heat-sensitive cultivars of V. vinifera, resulting in a limited understanding of grapevine responses to high temperatures. By analysing the transcriptional profiles of heat-tolerant and sensitive tea cultivars, 78 differential expressed genes (DEGs) have been identified [20], validating a rational approach to identify the key genes using two varieties with contrasting heat tolerance. In addition, none of those genes reported in the previous studies have been experimentally verified to regulate heat tolerance in grapevines by transgenic methods. Therefore, it is essential to include heat-tolerant grapevine germplasms to acquire global insight into heat tolerance mechanisms and verify functionally of key heat response genes in grapevine.
Heat shock factors (HSFs) play critical, conserved roles in the plant transcriptional network that regulates thermo-responsive gene expression [21]. Based on the length of the f lexible linker between DBD and HR-A/B regions and the number of amino acid residues in the HR-A/B regions, plant HSFs are classified into three classes: A, B, and C [22,23]. Among class A HSFs, HSFA1s play critical roles in heat stress response and are regarded as indispensable regulators in the transcriptional network. In tomato and Arabidopsis, knockdown or knockout of HSFA1 genes, downregulate many heat stress-responsive genes and induce heat stress-sensitive phenotypes [24][25][26]. HSFA1s directly regulate the expression of genes involved in heat stress response, including DEHYDRATION-RESPONSIVE ELEMENT BINDING PROTEIN 2A (DREB2A), HSFA2, HSFA7a, HSFBs, and MULTIPROTEIN BRIDGING FACTOR 1C (MBF1C) [25]. In addition to HSFA1s, other members of the class A family, such as HSFA2 and HSFA3, also play critical roles in plant response to heat stress. HSFA2 is essential for heat stress response in plants. The hsfa2-knockout mutant exhibits high sensitivity to heat stress, together with reduced expression of many heat stress-inducible genes [27]. Knockout or knockdown of HSFA3 also reduced the expression of HSP genes during heat stress [28][29]. Compared with the known activator function of several HSFAs in model plants, the roles of HSFBs are less well understood [30]. Moreover, compared to the model plants, much less research on HSFs has been carried out for fruit crops. HSFAs or HSFBs are significantly upregulated under heat stress in grape, apple, citrus, and strawberry [31][32][33][34][35]; however, only a few reports describe the roles of grape HSFs in heat stress with genetic evidence [36,37].
In this study, we presented a high-resolution view of the transcriptional changes in heat tolerant (Vitis davidii 'Tangwei') and sensitive (V. vinifera 'Jingxiu') grapevines under varying degrees of high temperatures. The grape HSFB1 was identified and functionally characterized because it was differentially expressed only in 'Tangwei' and upregulated under heat stress. HSFB1 positively regulates heat tolerance in grapevine as a transcriptional repressor. Moreover, VdHSFB1 from 'Tangwei' and VvHSFB1 from 'Jingxiu', which differ only one amino acid, exhibit similar repression activity; however, the promoter variations of the two orthologues likely determine the difference of HSFB1 gene expression in the two grape germplasms with contrasting thermotolerance, contributing to the difference in heat tolerance.

Photosynthetic response of different grapevine germplasms to heat stress
Photosynthesis is a sensitive physiological process affected by high temperature [38].The test of OJIP chlorophyll a fluorescence transient shows changes of photosystem II (PSII) electron transport chain. Heat injury under high temperature indirectly ref lects plant heat tolerance and can be quickly evaluated by the parameter F v /F m , which represents the potential maximum quantum yield of primary photochemistry [39]. Based on our previous work that evaluated grapevine heat tolerance [39], we selected nine representative grapevine germplasms for further heat tolerance evaluation: V. davidii cultivar 'Tangwei', V. quinquangularis cultivar 'Yeniang 2', wild species V. thunbergii, V. pseudoreticulata and V. f lexuosa; the V. vinifera cultivars 'Jingxiu', 'Jingfeng', and 'Xiangfei'; and the V. vinifera × V. labrusca hybrid 'Jingya'. Using the previously described method [39], the F v /F m values of detached grapevine leaves were measured after treatment at 25 • C and 47 • C for 40 min. The results showed that there was no significant difference in the F v /F m values among all grapevine germplasms at normal temperature treatment (Fig. 1A), However, the F v /F m values of the wild species were significantly higher than those of V. vinifera (Jingxiu, Jingfeng, and Xiangfei) and the V. vinifera × V. labrusca hybrid 'Jingya' after heat treatment (Fig. 1A), indicating that the wild species have relatively stronger heat tolerance than V. vinifera and its hybrid, consistent with previous reports [39]. 'Tangwei' and 'Yeniang 2' showed the highest heat tolerance, and 'Jingxiu' had the lowest (Fig. 1A). Therefore, we selected 'Tangwei' as a heattolerant germplasm and 'Jingxiu' as a heat-sensitive germplasm for comparative transcriptomic analysis. One-year-old 'Tangwei' and 'Jingxiu' grapevines were used to study the responses of grapevine germplasms with two heat treatments. There was no notable difference in the F v /F m values of 'Tangwei' and 'Jingxiu' leaves before different temperature treatments (Fig. 1B). After 40 • C or 45 • C treatment for 2 h, the F v /F m values of both 'Tangwei' and 'Jingxiu' declined significantly (P < 0.05) compared with the respective control groups (maintained at 25 • C). Moreover, the F v /F m values decreased with the increase of treatment temperature. However, the F v /F m values of 'Tangwei' were significantly higher than those of 'Jingxiu' after exposure to 40 • C or 45 • C for 2 h (Fig. 1B). These results suggested that 'Tangwei' and 'Jingxiu' were injured by high temperatures, but the extent of heat injury was significantly lower in 'Tangwei' than 'Jingxiu'.

Transcriptome and DEG analysis of 'Tangwei' and 'Jingxiu' grapevines exposed to high temperatures
The libraries were sequenced by the Illumina HiSeq 2500 platform with three replication each sample. After quality controls, we aligned the clean reads to the V. vinifera reference genome sequence (12X, PN40024) (Table S1, see online supplementary material).
DEGs (|log2(fold change)| >1 and false discovery rate (FDR) <0.05) were defined as significantly upregulated or downregulated in one sample compared with another sample. The numbers of DEGs were determined for six comparisons:

Kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis
KEGG annotations were used to identify DEG pathways, and 1400 DEGs from TW1 were assigned to 120 pathways (Additional file 1). For TW2, 2989 DEGs were assigned to 130 pathways (Additional file 1). The top four enriched pathways were metabolic pathway, biosynthesis of secondary metabolites, ribosome, and protein processing in endoplasmic reticulum in TW2 (Fig. S2C, see online supplementary material). For JX1, 1873 DEGs were assigned to 129 pathways, and for JX2, 1790 DEGs were assigned to 130 pathways (Additional file 1); the top four enriched pathways were metabolic pathway, biosynthesis of secondary metabolites, plant hormone signal transduction and protein processing in endoplasmic reticulum (Additional file 1). To validate the results of transcriptome analysis, qRT-PCR assays were performed on three DEGs (in Additional file 2) with gene-specific primers using RNA isolated from 'Tangwei' and 'Jingxiu' treated at 42 • C for different times. The results showed that the significantly higher upregulation of GOLS1 (galactinol synthase 1) and GPX (glutathione peroxidase) in 'Tangwei' compared to 'Jingxiu' ( Fig. S3A and C, see online supplementary material), while the upregulation of RS (raffinose synthase) in 'Jingxiu' is higher than that in 'Tangwei' (Fig. S3B, see online supplementary material) These results were consistent with the transcriptome analysis.

HSFs differentially expressed in 'Tangwei' and 'Jingxiu' in response to different high temperatures
Although there are a number of TFs in the above DEGs, the majority of the DEGs in protein processing in endoplasmic are HSFs, therefore the study focused HSFs. In grape, 19 putative HSF genes were identified [31,32]. We further analyse HSFs respond to high temperature in grapevine. A total 11 HSFs (HSFA1d, HSFA2, HSFA3, HSFA4a, HSFA6b, HSFA8, HSFB1, HSFB2a, HSFB2b, HSFB3, and HSFC1) were identified in DEGs. The changes of these genes were shown in Fig. S4 (see online supplementary material) and Additional file 3. For example, VIT_204s0008g01110 encoding HSFA2 was markedly upregulated in both TW and JX comparisons. VIT_216s0100g00720 encoding HSFB2a and VIT_202s0025g04170 encoding HSFB2b showed greater upregulation in 'Jingxiu' than in 'Tangwei' under heat stress. The class A HSFs have been widely studied in plants [24][25][26][27][28][29], but function of class B and class C HSFs in heat tolerance are still unclear [30]. VIT_207s0031g00670 encoding HSFB1 was differentially expressed only in ' Tangwei' and was upregulated under heat stress (Additional file 3). In addition, VpHSFB1a (HSFB1 in the study) significantly responds to heat stress in V. pseudoreticulata cultivar 'Baihe-35-1' [31]; and the expression of HSF14 (HSFB1 in the study) increases after the treatment at 38 • C for 2 h followed by 47 • C for 40 min compared to the treatment at 25 • C for 2 h followed by 47 • C for 40 min in 'Tangwei' [32]. To date, only the roles of HSFB2a and HSFB2b have been thoroughly studied in plants in HSFBs [40,41]. Therefore, we chose HSFB1 for further study. The relative expression levels of HSFB1 in 'Tangwei' and 'Jingxiu' were measured by using qRT-PCR. As expected, the expression of HSFB1 was upregulated almost three-fold in 'Tangwei' at 45 • C compared with 25 • C, whereas its expression was upregulated only 1.5-fold in 'Jingxiu' under the same conditions ( Fig. 2A).
The HSFB1 expression level of 'Tangwei' was also significantly higher than that of 'Jingxiu' at both 25 • C and 40 • C ( Fig. 2A). In addition, the upregulation folds of HSFB1 expression in 'Tangwei' were higher than that in 'Jingxiu' under 42 • C for a different treatment time (Fig. 2B). This result prompted us to explore if HSFB1 from 'Tangwei' and 'Jingxiu' inf luences the heat tolerance difference between these two grapevine germplasms.

Nuclear localization and transcriptional repression activity of HSFB1
We cloned VdHSFB1 and VvHSFB1 from 'Tangwei' and 'Jingxiu', respectively. There are two SNPs (single nucleotide polymorphisms) difference between the coding regions of VdHSFB1 and VvHSFB1, resulting in an amino acid difference (proline in VdHSFB1 and a leucine in VvHSFB1) (Figs S5 and S6A, see online supplementary material).
The respective proline and leucine are not located in any of the known functional domains, including DNA binding domain, HR-A/B domain, and B3 repression domain of HSFB1 (Fig. S6A, see online supplementary material). It has been suggested that the HSFBs have no transcriptional activation activity owing to the absence of an activation domain [42]. To explore the transcriptional activity of HSFB1, yeast and dual luciferase assays were performed. Yeast strain containing BD-VdHSFB1 or BD-VvHSFB1 did not turn blue on SD/−Trp/X-α-Gal/AbA medium or survive on SD-Trp/-His/−Ade selection medium, in contrast to the positive control ( Fig. 3A), suggesting that neither proteins possess transcriptional activation activity. We then determined whether the two factors have transcriptional repression activity in a transactivation assay using Arabidopsis protoplasts. The relative luciferase activities of Gal4BD-VdHSFB1 and Gal4BD-VvHSFB1 were similar but significantly lower than that of Gal4BD (Fig. 3B), suggesting that VdHSFB1 and VvHSFB1 are repressors with similar transcriptional repression activities.
We then explored the possibility of the single amino acid change altering the subcellular localization of HSFB1. VdHSFB1 and VvHSFB1 were fused to eGFP, and the resulting fusion proteins VdHSFB1-eGFP and VvHSFB1-eGFP were separately co-transformed with P19 and the nuclear localization marker H2B-mCherry into tobacco leaves by agro-infiltration. After 3 days, the subcellular localization of VdHSFB1 and VvHSFB1 was determined by laser confocal microscopy. Both VdHSFB1 and VvHSFB1 co-localized in the nucleus with the nuclear localization marker (Fig. 3C), suggesting that the amino acid change did not alter HSFB1 subcellular localization. In addition, phylogenetic analysis suggested that both VdHSFB1 and VvHSFB1 exhibited the highest homology with Citrus clementina HSFB1 (CcHSFB1) (Fig. S6B, see online supplementary material).

Increases of heat tolerance in heat sensitive grape plants by HSFB1 overexpression
Due to the difficulty of obtaining the stable transgenic grape plants, HSFB1 was transiently overexpressed in 'Jingxiu' plantlets to determine its function in heat tolerance. The coding sequence of VdHSFB1 and VvHSFB1 were cloned into the overexpression vector pCAMBIA-2300 to create OE-VdHSFB1 and OE-VvHSFB1. The overexpression vectors and the empty vector (EV) were transformed into Agrobacterium tumefaciens EHA105. The transformed Agrobacteria were infiltrated into 'Jingxiu' leaves. After incubation for 3 days, the leaves of EV, OE-VdHSFB1, and OE-VvHSFB1 plantlets were sampled for qRT-PCR to measure HSFB1 expression. The relative expression levels of HSFB1 in OE-VdHSFB1 and OE-VvHSFB1 were 3-and 6-fold more than that in EV, respectively (Fig. 4A).
Then EV, OE-VdHSFB1, and OE-VvHSFB1 plantlets were placed in an incubator at 43 • C. After 5 h, the infiltrated leaves of EV became exsiccated compared to those of OE-VdHSFB1 and OE-VvHSFB1 plantlets, which remain normal (Fig. 4B). In addition, the relative electrolyte leakage of EV plantlets was significantly higher than that of OE-VdHSFB1 or OE-VvHSFB1 plantlets under high temperature (Fig. 4C), Correspondingly, the F v /F m values of EV were significantly lower than those of OE-VdHSFB1 or OE-VvHSFB1 leaves (Fig. 4D). Taken together, these results indicated that overexpression of HSFB1 enhances heat tolerance in grape plantlets.

Decrease of heat tolerance of grape plants by suppression of HSFB1
Due to the rooting difficulty of 'Tangwei', we were unable to generate transformants of 'Tangwei'. We thus chose 'Yeniang 2' (V. quinquangularis), exhibiting similar heat tolerance as 'Tangwei' (Fig. 1A), for generation of tissue culture plants. The coding sequences of HSFB1 are identical in 'Yeniang 2' and 'Tangwei' (Fig. S7, see online supplementary material).
For RNA interference, a 271-bp fragment of VqHSFB1 was inserted into the vector pFGC5941. The resultant SiVqHSFB1 and EV were mobilized into A. tumefaciens EHA105. The transformed Agrobacteria were infiltrated into the leaves of one-year-old 'Yeniang 2' plants which had germinated for 8 weeks. After 5 days, the leaves of EV and SiVqHSFB1 plants were sampled to determine the VqHSFB1 expression using qRT-PCR. The relative expression level of VqHSFB1 in SiVqHSFB1 plants was significantly lower than that in EV (Fig. 5A).
EV and SiVqHSFB1 plants were placed in an incubator with light at 42 • C. After 3.5 h, the leaves of SiVqHSFB1 plants became significantly more wilted compared with EV or SiVqHSFB1 plants at 25 • C (Fig. 5B). Moreover, the relative electrolyte leakage of SiVqHSFB1 was notably higher than that of EV plants after heat treatment (Fig. 5C). In addition, the F v /F m values of EV leaves were approximately 2-fold higher than that of SiVqHSFB1 leaves under high temperature (Fig. 5D). The results indicated that downregulation of HSFB1 expression decreases heat tolerance of grape plants.

Enhancement of heat tolerance of grape suspension cells stably overexpressing HSFB1
The one amino acid difference between VdHSFB1 and VvHSFB1 does not alter the basic characteristics and functions of the two proteins (Figs 3-5; Fig. S6, see online supplementary material), VdHSFB1 was thus chosen to represent HSFB1 for further study. To explore whether HSFB1 confers heat tolerance in a stable transgenic system, VdHSFB1 was overexpressed in suspension cells of '41B' (V. vinifera 'Chasselas' × V. berlandieri). '41B' is identical to VvHSFB1 (Fig. S8, see online supplementary material).
The suspension cells were transformed as described previously [43]. Transgenic status and VdHSFB1 expression were verified by PCR and qRT-PCR assays ( Fig. 6A and B). Combinations of pGADT7-T with pGBKT7-Lam and pGBKT7-p53 were used as negative and positive controls, respectively. B Validation of VdHSFB1 and VvHSFB1 transcriptional activity in Arabidopsis protoplasts. The coding sequence of VdHSFB1 and VvHSFB1 were fused with Gal4 binding domain of Gal4BD vector, respectively. Constructed vectors were co-transformed into Arabidopsis protoplasts with LUC reporter vector and the REN vector. Gal4BD-VP16 and Gal4BD were used as positive and negative controls, respectively. Data represent the mean ± SE of three biological replicates. Different letters indicate significant differences according to Duncan test (P < 0.05). C Subcellular localization of VdHSFB1 and VvHSFB1 in tobacco leaves. VdHSFB1 and VvHSFB1 were fused with enhanced green f luorescent protein (eGFP), respectively. Constructed vectors were co-transformed with nuclear marker (H2B-mCherry) into tobacco leaves. eGFP f luorescence (green) and RFP f luorescence (red) were observed using a confocal microscope. Scale bars are 25 μm.
The relative expression of VdHSFB1 in OE-VdHSFB1 was 20fold higher than that in EV (Fig. 6B). The heat tolerance of OE-VdHSFB1 was then assessed by measuring the critical electrical conductivity temperature (T COND ) as previously described [44]. The T COND of OE-VdHSFB1 was approximately 2 • C higher than that of EV (Fig. 6C), suggesting that VdHSFB1 confers grape cell heat tolerance.

Cloning and sequence analysis of the HSFB1 promoters
The same transcriptional repression activity between VdHSFB1 and VvHSFB1 prompted us to speculate that the differential expression of VdHSFB1 and VvHSFB1 is due to a difference in promoter. We thus cloned the promoters of VdHSFB1 and VvHSFB1. Sequence alignment revealed that there were 27 SNPs and 11 Indels in the promoters of VdHSFB1 and VvHSFB1 (Fig. S9, see online supplementary material).
To investigate whether the SNPs and Indels in the HSFB1 promoters alter the cis-elements, the Plant Cis-Acting Regulatory Element (Plant CARE) database was used to analyse the VdHSFB1 and VvHSFB1 promoters [45]. The types of cis-elements in the VdHSFB1 and VvHSFB1 promoters remain the same; however, the promoter sequence variations resulted in more ARE, AT∼TATA-box, MYB, and TATA-box cis-elements in the VdHSFB1 promoter than the VvHSFB1 promoter (S2, see online supplementary material).

Analysis of the HSFB1 promoter activity in grape plantlets and Arabidopsis protoplasts
The heat tolerance of transgenic plantlets was found to be associated with the expression level of HSFB1 (Fig. 4). We next tested the promoter activities of VdHSFB1 and VvHSFB1. Ciselements analysis showed that there were 10 more TATA-boxes in the VdHSFB1 promoter than the VvHSFB1 promoter, and TATAbox is associated with promoter strength [46]. We measured the promoter activities of VdHSFB1 and VvHSFB1 in grape plantlets and Arabidopsis protoplasts. The promoters of VdHSFB1 and VvHSFB1 were individual in fusion with the luciferase reporter gene (LUC) in the pGreenII-0800-LUC vector (Fig. 7A). The resultant vectors, proVdHSFB1::LUC and proVvHSFB1::LUC, were mobilized into A. tumefaciens GV3101 (pSoup), and then transformed into the 'Jingxiu' plantlets. After incubation for 3 days, the leaves of proVdHSFB1::LUC and proVvHSFB1::LUC plantlets were sampled to measure the luciferase activity. The relative luciferase activity of proVdHSFB1 was significantly higher than that of proVvHSFB1 under both normal and high temperatures compared to that of VvHSFB1 in grape plantlets (Fig. 7B). In addition, the promoter activities of both VdHSFB1 and VvHSFB1 were higher when assayed at 37 • C compared to 25 • C (Fig. 7B). We also detected the promoter activity of VdHSFB1 and VvHSFB1 in Arabidopsis protoplasts. The plasmids of proVdHSFB1::LUC and proVvHSFB1::LUC were transformed into prepared Arabidopsis protoplasts, respectively. After incubation at 23 • C for 16 h, the protoplasts transfected with proVdHSFB1::LUC and proVvHSFB1::LUC were placed in temperature-controlled (23 • C and 37 • C) water baths for 10 min before measurement of LUC and REN activities. The relative luciferase activity of proVdHSFB1 was significantly higher than that of proVvHSFB1 when assayed at 23 • C, and at elevated temperature (37 • C) the luciferase activities driven by both promoters increased, compared to those at 23 • C, while the proVdHSFB1 activity remained notably higher than that of proVvHSFB1 (Fig. 7C). These results showed that the promoter activity of VdHSFB1 is significantly higher than that of VvHSFB1.

Discussion
In the past decades, climate change significantly inf luenced grape production. Worldwide viticulture will encounter a serious threat in the near future if annual temperatures continue to rise [47]. Most global cultivars derived from mild climates may be unable to withstand the extreme heat stress [48,49]. Grapevine germplasms from warmer regions may therefore be important genetic resources. Exploring the molecular mechanisms underlying heat tolerance of these germplasms is particularly important for breeding new heat-tolerant cultivars, thereby promoting the sustainability of grape cultivation and the wine industry [50,51]. Here, we evaluated the heat tolerance of nine representative germplasms, and found that the heat tolerance between V. davidii 'Tangwei' or V. quinquangularis 'Yeniang 2' and V. vinifera 'Jingxiu' was the most different (Fig. 1A). The strong heat tolerance of V. davidii and V. quinquangularis may be closely associated with their distribution in southern China and their long history of growth in hot and humid conditions. Although showing many similarities in heat response as other crops, the perennial grapes possess distinct characteristics [52]. We used V. davidii 'Tangwei' (heat tolerant) and V. vinifera 'Jingxiu' (heat sensitive) as two contrasting germplasms to investigate the specific factors in the regulatory networks that control grapevine response to high temperature.

More DEGs are found in heat-tolerant grapevine germplasms in response to heat stress
Transcriptomics is useful to understand how plants respond to abiotic stresses and to characterize genes involved in these responses. Together with the physiological data, transcriptional profiling aided us to dissect the molecular programming that distinguishes the heat stress responses between the heat tolerant and sensitive germplasms. The numbers of DEGs in both 'Tangwei' and 'Jingxiu' increased with increasing treatment temperature; however, the DEG numbers in 'Tangwei' (757) were significantly greater than in 'Jingxiu' (304) (Fig. S1, see online supplementary material), suggesting that the heat-tolerant germplasm activates more genes under higher temperature and can better respond to heat stress compared with the heat-sensitive germplasm. Moreover, there were far more unique DEGs in 'Tangwei' than 'Jingxiu' when comparing each germplasm at elevated temperature to normal temperature, i.e. 4337 and 4122 DEGs in TW2 and TW3, respectively, versus 2437 and 748 in JX2 and JX3, respectively ( Fig. S1D and E, see online supplementary material), suggesting that more genes are affected in 'Tangwei' than in 'Jingxiu'. The heat-tolerant germplasms seem to evolve more larger regulatory networks to adapt heat stress.

HSFB1 is a key factor determining the difference between heat tolerant and sensitive grapevines
HSFs are considered to be one of the most important TF families for plant heat stress response [53]. Here, 11 grape HSFs were also identified as DEGs in response to heat treatments. Similar to the previous studies, the expression of VIT_204s0008g01110 (HSFA2) showed marked upregulation in the TW1, TW2, JX1, and JX2 comparisons (Additional file 3), indicating that HSFA2 may contribute to enhancing grapevine heat tolerance. Interestingly, HSFB1 from the HSFB family was identified as a DEG in the TW2 and TW3 comparisons but not in any JX comparisons (Additional file 3), which was confirmed by qRT-PCR assays (Fig. 3A). In addition, the upregulation of HSFB1 in 'Tangwei' was higher than that in 'Jingxiu' under 42 • C for a different treatment time (Fig. 2B). The above results suggest that HSFB1 may determine the difference in heat tolerance between heat tolerant and sensitive grapevines.

HSFB1 positively regulates heat tolerance as a transcriptional repressor in grapevine
HSFAs are commonly transcriptional activators due to the acidic AHA domains with activator potential [23]. Some HSFBs possess transcriptional repression activities due to the presence of B3 repression domains [42,54]. Here, we showed that the HSFB, HSFB1, possesses a B3 repression domain and represses transcription in an Arabidopsis protoplast-based assay ( Fig. 3B; Fig. S6A, see online supplementary material). The VdHSFB1 and VvHSFB1 proteins only differ in one amino acid which does not seem to affect transcriptional repression activity (Fig. 3B). Some HSFBs possess transcriptional repression or activation activity in plants. In tomato, HSFB1 not only functions as a repressor to repress the expression of HS-inducible genes but also as a co-activator of HSFA1a and other non-HSF TFs [55]. In Arabidopsis, HSFB2b is a transcriptional repressor of several HSP genes [41]. However, chickpea HSFB2 has weaker transcription activation activity [56]. HSFBs expression are often induced by heat stress; however, their roles in plant heat tolerance can be postive or negative. For example, TaHSF3, a member of HSFBs with transcriptional repression activity, enhances heat tolerance of Arabidopsis [57]. In contrast, the Arabidopsis HSFB1 and HSFB2 is a transcriptional repressor [54], but the hsfb1hsfb2b mutants exhibit higher basal thermotolerance than the wild type plant [41]. We overexpressed and RNAi VdHSFB1 and VvHSFB1 in grapevines. We found that transient overexpression of VdHSFB1 and VvHSFB1 enhanced heat tolerance in 'Jingxiu' (Fig. 4), and RNAi of HSFB1 decreased heat tolerance in a heat tolerant germplasm (Fig. 5). Moreover, overexpression of VdHSFB1 significantly enhanced the heat tolerance of OE-VdHSFB1 compared with EV suspension cells (Fig. 6). Furthermore, HSFB1 overexpression was positively correlated to heat tolerance (Fig. 4). These results suggest that the grape HSFB1 positively regulates heat tolerance in grapevine.

Promoter variations between VdHSFB1 and VvHSFB1 contribute to heat tolerance difference between 'Tangwei' and 'Jingxiu'
The induction of HSFBs in plants under abiotic stress have been reported. The expression of OsHSF2b is highly induced by heat, salt, and polyethylene glycol (PEG) treatments in rice [58]. TaHSF3 is significantly upregulated under heat and cold stress in wheat seedlings [57]. The expression of CarHSFB2 is upregulated under the heat, salt, and drought stress [56]. Here, VdHSFB1 and VvHSFB1 was induced by high temperatures. Moreover, the extent of upregulation of VdHSFB1 was notably higher than that of VvHSFB1 under different high temperatures ( Fig. 2A) and under 42 • C for different treatment time (Fig. 2B). These results and the genetic evidence (Figs 4-6) indicated the HSFB1 expression level correlates positively with grape heat tolerance. In rice, the promoter sequence variations of CTB4a confer varying degrees of cold tolerance [59]. Similarly, the sequence variations in the SLG1 promoter increase the thiolated tRNA level, thus enhancing heat tolerance of indica rice varieties [60]. Therefore, we speculated that the expression difference between VdHSFB1 and VvHSFB1 is associated with difference in the promoter activity. In this study, the transient expression assays using plantlets and Arabidopsis protoplasts showed that the promoter activity of VdHSFB1 is notably higher than that of VvHSFB1 under both normal and high temperatures (Fig. 7). In addition, we found more TATA-box and AT∼TATA-box cis-elements in VdHSFB1 promoter than the VvHSFB1 promoter (Table S2, see online supplementary material). TATA-box has been shown to enhance transcription efficiency [61]. Therefore, promoter variations between VdHSFB1 and VvHSFB1 inf luence heat difference of grapevines. Further study will be needed to pinpoint the role of each cis-element in enhancement of the promoter strength.

Conclusions
This study provides abundant transcriptomic data on the response of heat-tolerant ('Tangwei') and heat-sensitive ('Jingxiu') grapevines responding to different high temperatures. The significantly enriched DEGs in the heat-tolerant germplasm provide many candidate genes for future studies of the molecular mechanisms governing heat-tolerance in grapevines. In addition, we demonstrated that the grape HSFB1 is a critical factor involved in thermotolerance. VdHSFB1 and VvHSFB1 from the two contrasting germplasms possess similar transcriptional repression activity although the two proteins only differ in one amino acid. In addition, HSFB1 expression levels are markedly different in 'Tangwei' and 'Jingxiu'. We propose that the difference in the expression levels is due to the promoter strengths, which can be explained by promoter sequence variation that results in change of the number of cis-elements. Promoter variations between VdHSFB1 and VvHSFB1 contribute to heat tolerance The electrical conductivity of the transgenic suspension cells was measured during continuously heating, which abruptly increases at a certain temperature which is called as T COND . The T COND ref lects the heat tolerance of organisms or cells. Data are means (± SE) from three independent biological replicates. Significant differences were determined using Student's t-test: * P < 0.05; * * P < 0.01. difference between the two grape germplasms with contrasting thermotolerance.
One-year-old 'Tangwei' and 'Jingxiu' grapevines were cultivated in pots. The culture conditions were previously reported [17]. When the sixth leaves (measured from the base) of the grape plants had matured (about 30 days old), all grapevines were transferred to a controlled-environment room with similar culture conditions as previously reported [17]. These grape plants were divided into six groups (three temperature treatments × two germplasms) and adapted for 2 d. Then the grapevines were exposed to 25 • C, 40 • C, or 45 • C for 2 h. The F v /F m values of the sixth leaves were recorded with a Handy Plant Efficiency Analyzer made by Hansatech (Norfolk, UK) at the end of the treatment period. These leaves were wrapped with tin foil and then put in liquid nitrogen. Three grapevines were used for each biological replicate, and three replicates were used for both treatments (40 • C and 45 • C) and controls (25 • C). In addition, as for 'Tangwei' and 'Jingxiu' grapevines treated under 42 • C for different times, when the sixth leaves (measured from the base) of the grape plants had matured (about 30 days old), these grapevines were divided into 12 groups (six time treatments × two germplasms) and adapted for 2 d. Then the grapevines were exposed to 42 • C for 0, 1, 2, 4, 8, and 12 h, respectively. The sixth leaves (measured from the base) of the grape plants were wrapped with tin foil and then put in liquid nitrogen. Three plants were used for each biological replicate, and three replicates were used for treatments (42 • C) and controls (25 • C).
Arabidopsis thaliana (Col-0) and Nicotiana benthamiana in potting nutrient soil were cultured under 23 • C with a 14-h photoperiod.
41B embryogenic calli was initiated as described previously [62]. The culture and sub-cultured methods were previously described [63].
Gene annotation, expression quantification, and DE analysis were performed using Cuff links software (v2.2.1) [68]. Gene expression normalization of every sample was performed using DESeq2, and DEGs were identified based on fold change >2 and false discovery rate (FDR) <0.05.

KEGG pathway enrichment analysis
The statistical enrichment of DEGs in KEGG pathways used KOBAS software [69].

Quantitative reverse transcription-PCR analysis
qRT-PCR was used to detect gene expression according to the method previously described [17]. The primers for qRT-PCR are shown in Table S3 (see online supplementary material).

Generation of transgenic plants and heat treatments
The coding sequence of VdHSFB1 and VvHSFB1 were cloned from leaves of 'Tangwei' and 'Jingxiu', respectively. Then VdHSFB1 and VvHSFB1 were subcloned into the overexpression vector pCAM-BIA2300 to generate OE-VdHSFB1 and OE-VvHSFB1. Then Empty Vector (EV), OE-VdHSFB1 and OE-VvHSFB1 were mobilized into A. tumefaciens EHA105 (ZOMANBIO Company, ZC142), respectively. Six-week-old 'Jingxiu' plantlets were used for transient overexpression assays. The plantlets were infiltrated by EV, OE-VdHSFB1, and OE-VvHSFB1 under vacuum (−90 kPa) for 20 min, respectively. Then, ddH 2 O was used to wash these plantlets and blotting paper was used to dry these plantlets, and the roots were inserted into 1 / 2 MS medium and put in the greenhouse under 25 • C with a 14h photoperiod. After 3 days, these plantlets were sampled and phenotype evaluated. Plantlets in bottles were treated without opening the lids at the following temperatures: (i) the control plantlets were maintained at 25 • C for 5 h; and (ii) the treatment plantlets were treated at 43 • C for 5 h. Meanwhile, The F v /F m values of the third leaves from the root were recorded with a Handy Plant Efficiency Analyzer made by Hansatech (Norfolk, UK). The primers for constructing OE-VdHSFB1 and OE-VvHSFB1 are in Table S3 (see online supplementary material).
To knock down the expression of HSFB1 in 'Yeniang 2', a 271bp fragment (587 to 858 bp) of VqHSFB1 coding sequence was fused with pFGC5941 in sense and antisense orientation using a previously described method [70,71]. The binary construct named SiVqHSFB1 and Empty Vector (EV) were mobilized into A. tumefaciens EHA105 (ZOMANBIO Company, ZC142). One-yearold 'Yeniang 2' plants that had germinated for 8 weeks were used for transient interference assays. These plants were vacuum (−90 kPa) infiltrated by EV and SiVqHSFB1 for 20 min, respectively. Then, ddH 2 O was used to wash these plants and blotting paper was used to dry these plants, and the plants were cultured in a small pot containing nutrient soil in a growth room under 25 • C with a 14-h photoperiod. After 5 days, these plants were used for phenotype evaluation. The plants were treated by different temperatures: (i) the control plants were maintained at 25 • C for 3.5 h; and (ii) the treatment plants were treated at 42 • C for 3.5 h. Meanwhile, the F v /F m values of the fifth leaves from the root were recorded with a Handy Plant Efficiency Analyzer made by Hansatech (Norfolk, UK). The primers for constructing SiVqHSFB1 are in Table S3 (see online supplementary material).
In addition, VdHSFB1 was then transformed into suspension cells of '41B' (V. vinifera 'Chasselas' × V. berlandieri). Transgenic suspension cells were generated using a previously described method [43]. After antibiotic screening, transgenic suspension cells were obtained. The primers for cloning and transformation are in Table S3 (see online supplementary material). Appropriate 41B suspension cells were put in a water bath, and water at room temperature was heated from 25 • C to 65 • C. The electrical conductivity of the transgenic suspension cells was determined using PlanTherm PT100 made by Photon Systems (Brno, Czech Republic)during continuous heating. It increased abruptly at a specific temperature called T COND , which ref lects the heat tolerance of organisms or cells [44].

Transient luciferase (LUC) expression assays
The 2518-bp (VdHSFB1) and 2541-bp (VvHSFB1) promoter sequences were cloned from DNA extracted from the leaves of 'Tangwei' and 'Jingxiu', respectively. The VdHSFB1 and VvHSFB1 promoters were subcloned into the pGreenII-0800-LUC vector. Then proVdHSFB1::LUC and proVvHSFB1::LUC were mobilized into A. tumefaciens GV3101 (pSoup) (ZOMANBIO Company, ZC142), respectively. Six-week-old 'Jingxiu' plantlets were used for transient luciferase (LUC) expression assays. The plantlets were infiltrated by EV, OE-VdHSFB1, and OE-VvHSFB1 under vacuum (−90 kPa) for 20 min, respectively. Then, ddH 2 O was used to wash these plantlets and blotting paper was used to dry these plantlets, and the roots were inserted into 1 / 2 MS medium and put in the greenhouse under 25 • C with a 14-h photoperiod. After 3 days, the leaves of these plantlets were sampled after temperature treatments. Plantlets in bottles were treated without opening the lids at the following temperatures: (i) the control plantlets were maintained at 25 • C for 3 h; and (ii) the treatment plantlets were treated at 37 • C for 3 h. Then the LUC activity and REN activity of these leaves were determined by using the Dual-Luciferase ® Reporter Assay System (E1910, Promega). 400 μL protein lysis buffer was added into the 0.05 g leaves of grape plantlets for total protein extraction. Then the activities of LUC and REN were measured by using GloMax 20/20 luminometer made by Promega (Wisconsin, Madison, USA). The ratio of LUC/REN stands for the promoter activity of VdHSFB1 and VvHSFB1, respectively.
The plasmid of proVdHSFB1::LUC and proVvHSFB1::LUC were transformed into prepared Arabidopsis protoplasts. After incubation at 23 • C for 16 h, the protoplasts transfected with VdHSFB1 and VvHSFB1 were placed in temperature-controlled (23 • C and 37 • C) water baths for 10 min. The LUC activity and REN activity of these Arabidopsis protoplasts were determined by using the Dual-Luciferase ® Reporter Assay System (E1910, Promega). 100 μL protein lysis buffer was added into the Arabidopsis protoplasts for total protein extraction and then the activities of LUC and REN were measured by using GloMax 20/20 luminometer (made by Promega (Wisconsin, Madison, USA).
The coding sequences of VdHSFB1 and VvHSFB1 were fused with the 35S-GAL4 vector, and Gal4BD-VdHSFB1 and Gal4BD-VvHSFB1 were transformed into Arabidopsis protoplasts with LUC and REN. After incubation at 23 • C for 16 h, the LUC activity and REN activity of these Arabidopsis protoplasts were determined by using the Dual-Luciferase ® Reporter Assay System (E1910, Promega). The activities of LUC and REN were determined as described above. 35S-GAL4 (Gal4BD) and 35S-GAL4-VP16 (Gal4BD-VP16) were the negative and positive and control, respectively. The primers for luciferase expression assays are shown in Table S3 (see online supplementary material).

Yeast assays
VdHSFB1 and VvHSFB1 coding sequences were individually fused with the pGBKT7 vector. The BD-VdHSFB1 and BD-VvHSFB1 plasmids were transformed into yeast strain Y2Hgold. They were transferred onto SD/−Trp/X-α-Gal/AbA and SD-Trp/-His/−Ade selection media with negative and positive controls. The primers for yeast assays are in Table S3 (see online supplementary material).

Subcellular localization
VdHSFB1 and VvHSFB1 coding sequences were individually fused with the eGFP of the pCAMBIA2300 vector, and the resultants were co-expressed with the nuclear marker H2B-mCherry in tobacco leaves. The subcellular localization of VdHSFB1 and VvHSFB1 was detected with a Leica TCS SP5 Confocal Scanning Microscope made by Lecia (Hessen, Germany). The peak excitation wavelengths of eGFP and RFP were 488 nm and 532 nm, respectively. The primers for subcellular localization assays are in Table S3 (see online supplementary material).

Measurement of electrolyte leakage
Electrolyte leakage was measured by the method described by Xu et al. with modification [64]. Ten leaves discs with 0.5 cm in diameter were incubated in 5 ml of distilled water. After shaking at 100 rpm and 25 • C for 20 min, initial conductivity (C1) was determined with FE30 made by Mettler Toledo (Zurich, Switzerland). Then the leaves were boiled for 20 min. The conductivity was remeasured after shaking at 100 rpm and 25 • C for 20 min as C2. The electrolyte leakage was determined by using the equation EL (%) = C1/C2 × 100.

Statistical analysis
The statistical data analysis was performed using GraphPad Prism 8.0.1 and spss 19.0. Data was considered statistically significant at a P-value <0.05 using Duncan test or Student's t-test. Data represent the mean ± SE from three independent experiments.