Comparative Analysis of the Transcriptomes of Three Varieties Provides Insights Into the Diversity of the Heat Response Mechanisms in Clematis Species

Background: Clematis species are commonly grown in western and Japanese gardens. Heat stress can inhibit many physiological processes mediating plant growth and development. The mechanism regulating responses to heat has been well characterized in Arabidopsis thaliana and some crops, but not in horticultural plants, including Clematis species. Results: In this study, we identied a heat-sensitive Clematis variety (Clematis alpina ‘Stolwijk Gold’) and two heat-tolerant Clematis varieties (Clematis vitalba and Clematis viticella ‘Polish Spirit’) based on heat-related physiological indices. The leaf transcriptomes under normal and heat stress conditions were analyzed by RNA sequencing. Additionally, heat tolerance-related genes (HTGs) were identied and their expression levels were analyzed. Following heat treatments, 41.67% of the differentially expressed HTGs in Stolwijk Gold had down-regulated expression levels, whereas only 9.80% and 21.36% of the differentially expressed HTGs in C. vitalba and Polish Spirit, respectively, had the same trend. The HTGs’ co-expression and protein–protein interaction networks revealed that the hub genes regulating Clematis resistance to heat stress encode heat shock transcription factors (HSFs) and heat shock proteins (HSPs). Moreover, the sensitivity of Stolwijk Gold to heat is mainly due to the heat-induced down-regulated expression of these genes. On the basis of phylogenetic and expression analyses, the differentially expressed HSF and HSP genes in the three examined varieties were divided into three and four clades, respectively, with similar expression proles common among orthologous family members. Furthermore, we identied two HSF classes in C. vitalba that may have diverse functions inuencing heat resistance. Conclusions: Our study provides insights into the diversity of the heat response mechanisms among Clematis species and may be useful for breeding new heat-resistant ornamental Clematis varieties.

mechanism underlying heat resistance in horticultural plant species, including those belonging to the genus Clematis, has not been thoroughly characterized.
In this study, by comparing the primary heat-related physiological indices before and after a high-temperature treatment, we identi ed a heat-sensitive Clematis variety (Clematis alpina 'Stolwijk Gold') and two heat-tolerant Clematis varieties (Clematis vitalba and Clematis viticella 'Polish Spirit'). We also analyzed the transcriptomes of these varieties under normal and heat stress conditions. Moreover, we compared the varieties regarding their responses to heat to clarify the differences in their heat resistance. Furthermore, to characterize the considerable heat resistance of C. vitalba, we identi ed two HSF classes with various functions related to heat resistance. The results of this study provide insights into the diversity of the heat response mechanisms among Clematis species and may be useful for breeding new heat-resistant ornamental Clematis varieties.
To compare the heat resistance of three Clematis varieties, speci c leaf physiological indices were analyzed before and after a high-temperature (HT) treatment (Supplementary Table 1). The resulting data indicated that the relative conductivity and malondialdehyde content were signi cantly higher in Stolwijk Gold than in C. vitalba and Polish Spirit, whereas the opposite trend was observed for the relative water content, soluble protein content, and superoxide dismutase activity (SOD) (Fig. 1B).
These observations indicated that the heat-induced membrane damage and peroxidation were greater in Stolwijk Gold than in C. vitalba and Polish Spirit. Accordingly, C. vitalba and Polish Spirit appeared to be more heat resistant than Stolwijk Gold. Moreover, nitroblue tetrazolium and diaminobenzidine staining revealed the substantial accumulation of ROS in Stolwijk Gold leaves (Fig. 1C) following the heat treatment, which was in contrast to the relatively unchanged ROS contents in C. vitalba and Polish Spirit. Thus, the antioxidant systems of Polish Spirit and C. vitalba remained active under heat stress conditions (Fig. 1D). These results re ected the heat resistance of C. vitalba and Polish Spirit as well as the sensitivity of Stolwijk Gold to heat stress.
Transcriptome pro les and annotations, differentially expressed genes, and GO and KEGG enrichment analyses To reveal the molecular basis of the differences in the heat resistance of the three examined Clematis varieties, the leaf transcriptomes under normal (control) and heat stress conditions were analyzed by RNA-sEq. Eighteen libraries corresponding to three biological replicates for the control and heat treatments of each variety were constructed and sequenced (Cv_NT_leaf1, Cv_NT_leaf2, Cv_NT_leaf3, Cv_HT_leaf1, Cv_HT_leaf2, Cv_HT_leaf3, PS_NT_leaf1, PS_NT_leaf2, PS_NT_leaf3, PS_HT_leaf1, PS_HT_leaf2, PS_HT_leaf3, SG_NT_leaf1, SG_NT_leaf2, SG_NT_leaf3, SG_HT_leaf1, SG_HT_leaf2, and SG_HT_leaf3). A total of approximately 895 million paired-end reads (raw reads) were generated, ltered, and trimmed, with 40-60 million reads per library (Supplementary Table 2). The raw data have been deposited in the NCBI Sequence Read Archive (PRJNA664279). For each Clematis variety, all clean reads were used for a de novo sequence assembly using Trinity (https://github.com/trinityrnaseq/trinityrnaseq) (Supplementary Table 3 Fig. 1), re ecting the close genetic relationship between Clematis species and A. coerulea. The clean reads were then mapped to the assembled sequence (Supplementary Table 5). The gene expression levels (i.e., transcripts per million reads) were analyzed using RSEM (http://deweylab.github.io/RSEM/). The hierarchical cluster analysis of gene expression among the different samples for each variety indicated the data for the biological replicates were reliable and the error was within the allowable range ( Fig. 2A). The differentially expressed genes (DEGs) among the three Clematis varieties were analyzed ( Supplementary Fig. 2). The number of DEGs and the ratio of the number of DEGs to the total number of genes were highest for Stolwijk Gold, and lowest for C. vitalba (Fig. 2C), implying that more biological processes were affected by heat stress in Stolwijk Gold than in C. vitalba and Polish Spirit. The GO functional annotation of the DEGs revealed that the heat treatment mainly altered membrane components, with some DEGs in C. vitalba and Polish Spirit related to heat responses ( Supplementary Fig. 3A). The KEGG pathway enrichment analysis indicated that the pathways affected by heat were mainly associated with secondary metabolism, with fewer pathways affected in C. vitalba than in the other examined varieties ( Supplementary Fig. 3B). Additionally, some signaling pathways in Stolwijk Gold were modulated by heat stress, including plant hormone signal transduction and the MAPK signaling pathway ( Supplementary Fig. 3B). These results suggested C. vitalba and Polish Spirit are more heat resistant than Stolwijk Gold.
Identi cation of heat tolerance-related genes and an analysis of their differential expression Regulatory processes in plants are affected by heat stress. On the basis of previous research [1,2], we divided the regulatory activities mediating plant responses to high temperatures into the following ve categories: heat signal transduction, transcriptional regulation, protein homeostasis, ROS homeostasis and RNA homeostasis. To elucidate the molecular mechanism underlying the responses of the three analyzed Clematis varieties to heat, we identi ed the heat tolerance-related genes (HTGs) associated with the ve categories (Supplementary Table 7). Speci cally, the HTGs were identi ed via a local blastp search using previously reported HTGs in other species as queries (Supplementary Table 6) and GO term annotations.
Additionally, their expression levels in the three examined Clematis varieties were compared ( Table 1, Fig. 3, Supplementary Table 7). Some of the differentially expressed HTGs in each species had down-regulated expression levels. More speci cally, 41.67% of the differentially expressed HTGs in Stolwijk Gold were signi cantly down-regulated under heat stress conditions, whereas only 9.80% and 21.36% of the differentially expressed HTGs in C. vitalba and Polish Spirit, respectively, exhibited the same trend (Fig. 3A). Polish Spirit had the most up-regulated HTGs. Clematis vitalba had the fewest down-regulated HTGs (Fig. 3A). These results may help to explain the heat resistance of C. vitalba and Polish Spirit. There were considerable differences in the expression of HTGs in the above-mentioned regulatory categories among the three Clematis species. Transcriptional regulation is critical for responses to high temperatures. The HSF family members as well as the ERF/AP2 family transcription factor DREB2A and the NAC transcription factor NAC019 positively affect the heat-activated transcriptional regulatory network [17,18]. Genes encoding these transcription factors were identi ed in the Clematis transcriptomes (Fig. 3). In response to heat stress, the expression levels of all HSF genes identi ed in C. vitalba were signi cantly up-regulated, whereas only half of the HSF genes identi ed in Stolwijk Gold had up-regulated expression levels (the rest had down-regulated expression levels). Moreover, many of the DREB2A and NAC019 transcription factor genes were expressed at lower levels in Stolwijk Gold than in the other varieties. We speculated that the sensitivity of Stolwijk Gold to heat is primarily due to a weak heat-activated transcriptional regulatory network. Additionally, maintaining homeostasis, especially related to protein and ROS contents, is extremely important for stabilizing the biological activities of plants exposed to heat stress [2,14]. Heat shock proteins, which are molecular chaperones, are important for the stabilization, renaturation, and degradation of unfolded proteins. Following the heat treatment, an analysis of the differentially expressed HSP genes indicated that Polish Spirit had the most up-regulated HSP genes, whereas C. vitalba had the highest proportion of up-regulated HSP genes (Fig. 3B). These ndings may be related to the differences in the heat resistance mechanisms of the evaluated Clematis varieties. In plants, ROS accumulation is a major cellular response to heat stress. Reactive oxygen species contribute to the early plant response to heat; however, high ROS contents lead to the oxidative damage of many cellular components [19,20]. The HTGs related to ROS homeostasis were not differentially expressed in C. vitalba, but had down-regulated expression levels in Polish Spirit, following the high-temperature treatment ( Table 1, Fig. 3B). This may have been because the heat resistance mechanism prevented the excessive accumulation of ROS in C. vitalba and Polish Spirit. Although the expression of two HTGs related to ROS homeostasis was up-regulated in Stolwijk Gold, four other HTGs related to ROS homeostasis had down-regulated expression levels, resulting in ROS accumulation ( Fig. 1C and D, Table 1, Fig. 3B). Additionally, the expression of some HTGs involved in heat signal transduction, such as CaM1 and CDPK2, was down-regulated in Stolwijk Gold, which may adversely affect downstream regulatory processes.

Genetic regulatory networks in Clematis varieties
To determine the potential interactions or regulatory relationships among the differentially expressed HTGs in the three Clematis varieties and to identify hub genes regulating heat resistance, we constructed gene co-expression and protein-protein interaction (PPI) networks (Fig. 4). These networks revealed that HSFs and HSPs, such as HSF30, HSF24, HSP70, and HSP90, have major roles associated with the heat tolerance of the three Clematis varieties. We speculated that the down-regulated expression of many genes encoding HSFs and HSPs critical for plant heat tolerance (e.g., HSF70 and HSF90) may be the main cause of the sensitivity of Stolwijk Gold to heat stress. Although C. vitalba and Polish Spirit were both resistant to heat stress, their heat-related genetic regulatory networks varied. Speci cally, C. vitalba had a relatively small regulatory network, with almost no down-regulated HTGs, whereas Polish Spirit had a relatively large regulatory network that included down-regulated HTGs. Accordingly, there are at least two distinct heat resistance mechanisms in Clematis species. Furthermore, the gene coexpression network was used to reveal potential targets of heat-responsive transcription factors, including HSFs, DREB2A, and NAC019. The identi ed gene targets may be useful for future investigations of the heat resistance mechanism in Clematis species.
Phylogenetic relationships and expression-level differences among the genes encoding heat shock transcription factors and heat shock proteins in three Clematis varieties Considering the importance of HSFs and HSPs for plant heat resistance, we analyzed the phylogenetic relationships of HSF and HSP genes and compared their expression levels to further characterize the differentially expressed HSF and HSP genes among three Clematis varieties (Fig. 5). The differentially expressed HSF genes were divided into three clades. The heat treatment down-regulated the expression of Clade 1 genes and PSHSFA1a and SGHSFA5 of Clade 2, suggesting these genes do not induce heat tolerance. All of the CvHSF genes belonged to Clades 2 and 3 and had heat-induced up-regulated expression levels, which may be related to the considerable heat tolerance of C. vitalba. Among the HSF gene family members, those in Clade 3 were generally more highly expressed than those in Clade 2. Thus, heat stress differentially affected the expression of HSF genes in different phylogenetic clades. We divided the differentially expressed HSP genes into four clades. The 3, 8, and 13 down-regulated HSP genes in C. vitalba, Polish Spirit, and Stolwijk Gold, respectively, were mainly clustered in Clades 1, 2, and 3. The expression levels of most of the HSP gene family members in Clade 4 were up-regulated, indicating this clade is important for the heat tolerance of Clematis varieties. The down-regulated expression of many SGHSP genes following the heat treatment may be related to the sensitivity of Stolwijk Gold to high temperatures. Although the expression of a substantial proportion of the PSHSP genes was down-regulated by an exposure to heat, the HSP genes were generally more highly expressed in Polish Spirit than in the other two varieties. Furthermore, our analysis revealed a clear expansion of the HSP gene family members in Clade 4, possibly because of an adaptive evolution to heat.
Classi cation and characterization of heat shock transcription factors in Clematis vitalba Considering C. vitalba is an original Clematis species with a small and e cient heat resistance genetic regulatory network, we predicted it may be useful for breeding. To further classify and characterize the HSFs in C. vitalba, we analyzed the phylogenetic relationships, predicted motifs, and expression of CvHSF genes. A phylogenetic analysis revealed that Classes A and B each contained three CvHSF genes, which were closely related to the orthologous AtHSF genes (Fig. 6A). On the basis of the PPI network, gene co-expression network, and expression pro les, CvHSF30-1 and CvHSF30-2 were identi ed as hub genes critical for the heat tolerance of C. vitalba. Thus, the expression of both genes was analyzed in a qRT-PCR assay. Additionally, to clarify the differences between the HSFs in Classes A and B and to functionally characterize the Class B HSFs, the expression of CvHSFB2a, which belongs to Class B, was also analyzed. An examination of the predicted motifs revealed that CvHSF30-2 and CvHSFB2a have similar N-terminals, but diverse C-terminals, indicative of functional differences between these two HSFs. As representative HSFs of Classes A and B, CvHSF30-2 has two activator peptide motifs (AHA motifs) and a nuclear export signal at the C-terminal, whereas CvHSFB2a has a repressor domain (Fig. 6B). The qRT-PCR data revealed the increasing CvHSF30-1 and CvHSF30-2 expression levels in the rst 2 h after a high-temperature treatment (42 ℃). Moreover, both genes were more highly expressed than CvHSFB2a, suggesting the HSF genes in Class A are important for the heat tolerance of Clematis species (Fig. 6C).

Discussion
Heat stress is a major abiotic factor that plants must adequately respond to [2,14]. The mechanism regulating the heat tolerance of Clematis species remains relatively uncharacterized, which is in contrast to the available information regarding the corresponding mechanisms in traditional model plant species, including A. thaliana and Oryza sativa. In this study, we rst elucidated the molecular basis for the differences in the heat tolerance of three Clematis varieties based on a transcriptomic analysis of plants under normal and heat stress conditions. We identi ed HTGs and compared their expression levels during various regulatory activities (heat signal transduction, transcription regulation, protein homeostasis, ROS homeostasis and RNA homeostasis) ( Table 1, Fig. 3B). Compared with Polish Spirit, there were fewer differentially expressed HTGs, but more downregulated HTGs, in Stolwijk Gold, which may help to explain the sensitivity of Stolwijk Gold to high temperatures. Although C. vitalba and Polish Spirit were both con rmed as heat-tolerant varieties, the underlying mechanisms differed. More speci cally, although C. vitalba had fewer differentially expressed HTGs than Polish Spirit, nearly all of the differentially expressed CvHTG genes were up-regulated. Although Polish Spirit had more differentially expressed HTGs, it also had a greater proportion of down-regulated differentially expressed HTGs compared with Clematis vitalba (Fig. 3A). Additionally, the differentially expressed HTGs in C. vitalba were associated with transcriptional regulation and protein homeostasis, but not heat signal transduction and ROS homeostasis (Fig. 3B). This suggests that C. vitalba may quickly respond to heat stress by modulating the activities of intracellular proteins. Moreover, heat stress does not substantially affect the ROS content of C. vitalba ( Fig. 1C and D), likely because of changes to transcriptional regulation and protein homeostasis. The C. vitalba characteristics related to heat resistance may be relevant for breeding new varieties of Clematis species.
Gene co-expression and PPI networks revealed the core HSFs and HSPs contributing to the heat stress resistance of Clematis species (Fig. 4). Considering the importance of HSFs and HSPs for heat resistance, we analyzed the phylogenetic relationships of the differentially expressed HSF and HSP genes in three Clematis varieties. The diversity in the phylogenetic relationships among the differentially expressed HSF and HSP genes (Fig. 5) may be associated with the observed differences in heat resistance among the Clematis varieties. Orthologous family members often had similar expression pro les (Fig. 5). Notably, we detected a clear expansion of the Clade 4 HSP gene cluster in Polish Spirit, which may have in uenced the heat resistance of this variety. The related genes should be further analyzed in future studies. It was unclear why the expression levels of some HSF and HSP genes were down-regulated in Polish Spirit and even more so in Stolwijk Gold following the heat treatment.
According to previous researches, calcium (Ca 2+ ) signaling, ROS signaling, NO signaling and their considerable crosstalk with each other make a difference in heat signal transduction of plant [20][21][22][23][24][25][26][27][28][29][30][31][32]. For example, CaM is one of the most important intracellular Ca 2+ receptors. Knocking out the expression of AtCaM3 made the resulting mutant more susceptible to HS stress, whereas the overexpression of AtCaM3 resulted in enhanced plant thermotolerance [33]. We speculated that the down-regulated expression of HTGs related to heat signal transduction (e.g., CAM1, CAM2, CPK1, CYPA, and CDPK2) may be an important reason. A more thorough functional characterization of these genes may further clarify the mechanism regulating the responses of Polish Spirit and Stolwijk Gold plants to heat stress.
Because of the obvious heat resistance of C. vitalba, the differentially expressed CvHSF genes were further classi ed and characterized. Plant HSF family members can be divided into three classes: HSFA, HSFB, and HSFC [12]. The differentially expressed CvHSF genes identi ed in this study are HSFA and HSFB genes. We examined CvHSF30-1, CvHSF30-2, and CvHSFB2a to predict the encoded motifs. Accordingly, we determined that CvHSF30-1 and CvHSF30-2 are A2-type HSFs, whereas CvHSFB2a is a B2-type HSF. Of the HSFA1s-targeted transcription factors/co-activators, HSFA2 is a key regulator of plant thermotolerance [34,35]. A qRT-PCR experiment revealed the considerable increase in the CvHSF30-1 and CvHSF30-2 expression levels in the rst 2 h following a high-temperature treatment, which is consistent with the results of previous research on other plants. Therefore, these two HSF genes are likely important for the heat resistance of C. vitalba. Additionally, a phylogenetic analysis indicated CvHSFB2a is an ortholog of AtHSFB2b. Earlier research proved that AtHSFB2b represses the expression of heat-inducible HSF genes, but positively regulates thermotolerance [7]. Hence, CvHSFB2a may be relevant for the molecular breeding of heat-resistant Clematis varieties.

Conclusions
Taken together, in this research, a heat-sensitive Clematis variety (Clematis alpina 'Stolwijk Gold') and two heat-tolerant Clematis varieties (Clematis vitalba and Clematis viticella 'Polish Spirit') were identi ed according to primary heat-related physiological indices before and after a high-temperature treatment. Gene expression pro les of three Clematis varieties under normal and high temperature based on transcriptome data were rstly reported, which provided valuable resources for the research of Clematis species. Moreover, we compared the varieties regarding their responses to heat to clarify the differences in their heat resistance based on HTGs that we identi ed. Furthermore, to characterize the considerable heat resistance of C. vitalba, we identi ed two HSF classes with various functions related to heat resistance. Our study provided insights into the diversity of heat response mechanisms of Clematis species, with implications for the breeding of heat-resistant and ornamental Clematis varieties.

Plant material growth conditions and high temperature treatment conditions
Triennial potted plants of three Clematis varieties (Clematis vitalba, Polish Spirit and Stolwijk Gold) were used in this study. The seeds were derived from the botanical garden in Germany, which were sown into autoclaved nutrient soil and grown for 3 years in 12-inch pots in greenhouse of Shanghai Botanical Garden. For each variety, six plants in great and similar growth condition were used and placed in two constant temperature incubators (one is at 42℃, the other is at 22℃) for 2 hours respectively (three plants per constant temperature incubator). Treated potted plants were used subsequent physiological indexes analysis and RNA sequencing.
Measurement of physiological indexes before and after high temperature treatment Relative conductivity: 0.2 g leaves was placed in a 20 mL tube, vacuumized, and stood for 30 min at normal temperature (shaken gently every 5 min during the period). The conductivity was measured with a DDS-11A type conductivity meter. After boiling water bath for 10 min and cooling, the conductivity value was again measured and the relative conductivity was calculated. The relative conductivity per gram of fresh weight was used to represent relative conductivity. Relative water content: Weigh 0.2 g of chopped leaves and place them in a weighing bottle, put them in an oven for 30 min at 105℃ for dehumidi cation, then set them to 80℃ for drying to constant weight and calculate the relative water content lastly. Coomassie brilliant blue G 250 method was used to soluble protein quanti cation. SOD activity was measured by Nitrogen Blue Tetrazolium Photoreduction Method. Malondialdehyde (MDA) content was measured by thiobarbituric acid colorimetric method.
Nitro blue tetrazole (NBT) and Diaminobenzidine (DAB) staining Leaves of three Clematis varieties with normal and high temperature treatment were soaked in DAB staining solution at 25 °C for 24 hours or NBT staining solution at 25 °C for 12 h both in the dark, and then soaked in 95% ethanol to remove chlorophyll and took the photo.

RNA isolation and sequencing
Eighteen samples (Cv_NT_leaf1, Cv_NT_leaf2, Cv_NT_leaf3, Cv_HT_leaf1, Cv_HT_leaf2, Cv_HT_leaf3, PS_NT_leaf1, PS_NT_leaf2, PS_NT_leaf3, PS_HT_leaf1, PS_HT_leaf2, PS_HT_leaf3, SG_NT_leaf1, SG_NT_leaf2, SG_NT_leaf3, SG_HT_leaf1, SG_HT_leaf2 and SG_HT_leaf3) from three Clematis varieties were used for RNA sequencing. RNA was isolated from the leaves of three Clematis varieties (Clematis vitalba, Clematis "Polish Spirit", Clematis "Stolwijk Gold") treated for 2 hours at 42℃ and 22℃ using TRIzol reagent, respectively. The extracted RNA was quanti ed using Nanodrop2000, and the RNA was electrophoresed on an agarose gel to check its integrity. Around 0.4 µg of total RNA was used for library construction and sequencing on an Illumina Novaseq 6000 (Illumina, San Diego, CA, USA) at Shanghai Majorbio Bio-pharm Technology Co.,Ltd (Shanghai, China). Prior to library construction, an Agilent 2100 Bioanalyzer (Agilent, CA, USA) was used to con rm the quality and quantity of RNA such that the rRNA ratio (28 s/18 s) was > 1.5 and the RNA integrity number > 7. In brief, 0.1 ~ 0.4 µg total mRNA was puri ed and fragmented using PCR plates with a magnetic plate stand. Fragmented mRNA was reverse-transcribed to cDNA using random primers and Superscript II (Invitrogen, Carlsbad, CA). Blunt-ended cDNA was generated by end repair and then ligated to yield 30 adenine base overhangs. Oligonucleotide adapters with thymine overhangs were ligated to the cDNA and added to the adapter index for each library. The library fragments were enriched by PCR ampli cation and ~ 895 million raw pair-end reads were generated on an Illumina Novaseq 6000.
Transcriptome annotation, identi cation and enrichment analysis of differential expressed genes (DEGs)

Hierarchical cluster analysis
The hierarchical clustering and other statistical analyses were carried out using R software (http://www.r-project.org). Pearson correlation was used to calculate distance of different samples.

Identi cation of heat tolerance-related genes
On the one hand, we obtained protein sequences based on transcriptome assembly sequence using Transdecoder (https://github.com/TransDecoder) and construct local protein databases of three Clematis varieties. Then proteins homologous with previously reported heat tolerance-related genes (HTGs) (Supplementary Table 6) were identi ed using the local blastp program of BLAST + 2.9.0 (ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/) (E-value < 1e-5, identity > 60%) in three Clematis varieties. On the other hand, genes of GO term related to heat-resistance in GO enrichment analysis were also regarded as heat-tolerant genes (HTGs). We combined both data to de ne heat tolerance-related genes (HTGs) (Supplementary Table 7) in three Clematis varieties, respectively.
Differentially expressed analysis of heat tolerance-related genes TPM data of the overlap between DGEs and HTGs, that is, expression pro les of differentially expressed HTGs were visualized in R software (https://www.r-project.org/). Blue and red represented "down regulated" and "up regulated", respectively.

Construction of co-expression network and protein-protein interaction network
We used differentially expressed HTGs for further network construction. Co-expression correlation coe cients of differentially expressed HTGs were obtained through Spearman algorithm (corrected p value < 0.05) based on gene expression data and visualized in Cytoscape v3.5.1 [36]. PPI networks of differentially expressed HTGs were constructed based on PPIs of Aquilegia coerulea in STRING database (https://string-db.org) for three Clematis varieties. Networks were also visualized using Cytoscape v3.5.1. Number of edges directly connected with nodes were computed using Network Analyzer in Cytoscape v3.5.1 [37].

Phylogeny and expression analysis of HSFs and HSPs
Phylogenetic trees of HSFs and HSPs from three Clematis species were constructed in IQ-TREE v2.0.6 [38] with JTT + I + G4 and VT + R3 model respectively. Support for each node was assessed by performing a bootstrap analysis with 1000 replicates. The phylogenetic analysis of HSFs from Clematis vitalba and Arabidopsis thaliana was inferred by using the Maximum Likelihood method based on the JTT matrix-based model in MEGA-X [39,40]. The bootstrap consensus tree inferred from 1000 replicates is taken to represent the evolutionary history of the taxa analyzed [41]. Fewer than 5% alignment gaps, missing data, and ambiguous bases were allowed at any position (partial deletion option). Expression pro les (TPM data) of differential expressed HSFs and HSPs were visualized in iTOL (https://itol.embl.de/). Green and red represented "down regulated" and "up regulated" respectively.

Motif prediction and visualization of HSFs
Motif prediction and visualization of CvHSF30-1, CvHSF30-2 and CvHSFB2a were performed in HEATSTER Availability of data and materials The datasets supporting the conclusions of this article are included within the article and its supplementary information les.
The RNA-seq data are available from NCBI of sequence Read Archive (SRA project: PRJNA664279).