CdWRKY2‐mediated sucrose biosynthesis and CBF‐signalling pathways coordinately contribute to cold tolerance in bermudagrass

Summary Bermudagrass (Cynodon dactylon) is one of the most widely cultivated warm‐season turfgrass species around the world. Cold stress has been a key environmental factor that adversely affects the growth, development, and geographical distribution of bermudagrass; however, the underlying mechanism of bermudagrass responsive to cold stress remains largely unexplored. Here, we identified a cold‐induced WRKY transcription factor CdWRKY2 from bermudagrass and demonstrated its function in cold stress response. Overexpression of CdWRKY2 enhanced cold tolerance in transgenic Arabidopsis and bermudagrass hairy roots, while knocking down CdWRKY2 expression via virus‐induced gene silencing increased cold susceptibility. RNA sequencing showed that overexpression of CdWRKY2 in Arabidopsis activated the expression of genes involved in sucrose synthesis and metabolism, including sucrose synthase 1 (AtSUS1) and sucrose phosphate synthase 2F (AtSPS2F). CdSPS1, the homology gene of AtSPS2F in bermudagrass, was subsequently proven to be the direct target of CdWRKY2 by yeast one‐hybrid, electrophoretic mobility shift assay, and transient expression analysis. As expected, overexpression of CdSPS1 conferred cold tolerance in transgenic Arabidopsis plants, whereas silencing CdSPS1 expression enhanced cold sensitivity in bermudagrass. Besides, CdCBF1 whose expression was dramatically up‐regulated in CdWRKY2‐overexpressing bermudagrass hairy roots but down‐regulated in CdWRKY2‐silencing bermudagrass both under normal and cold stress conditions was confirmed as another target of CdWRKY2. Collectively, this study reveals that CdWRKY2 is a positive regulator in cold stress by targeting CdSPS1 and CdCBF1 promoters and activating their expression to coordinately mediate sucrose biosynthesis and CBF‐signalling pathway, which provides valuable information for breeding cold‐resistant bermudagrass through gene manipulation.


Introduction
Bermudagrass (Cynodon dactylon) is one of the most widely used turfgrass species throughout the world for lawns, sports fields, parks, and slope protection (Fan et al., 2014). However, as a representative perennial warm-season grass, bermudagrass is challenged by low temperature including chilling (0-15°C) and freezing (<0°C). Cold stress has been a key environmental factor that adversely influences its growth, development, turf quality, green period, chlorophyll content, and distribution (Fan et al., 2014;Liu et al., 2016). Therefore, improving cold tolerance is recognised as an important and long-term target for bermudagrass breeding. To this end, identifying potential genes and uncovering the underlying regulatory networks involved in cold response are imperative.
Cold stress causes several damages at the cellular level, including membrane injury, generation of reactive oxygen species (ROS), and protein denaturation (Ruelland et al., 2009). Among them, the plasma membrane injury, which is indicated by electrolyte leakage (EL) and malondialdehyde (MDA) content is thought to be primary adverse effect imposed by cold stress because of its central role in the regulation of various cellular processes (Lyons, 1973;Premachandra et al., 1992). As sessile organisms, plants have established sophisticated regulatory mechanisms to cope with cold stress. A myriad of transcription factors (TFs) has been identified to be involved in cold signalling networks, among which C-repeat binding factors/dehydration responsive element protein 1 (CBFs/DREB1) plays a central role in cold resistance by directly activating the expression of cold responsive (COR) genes (Chinnusamy et al., 2007). In Arabidopsis, three CBFs, including CBF1, CBF2, and CBF3 have been demonstrated to be essential for cold acclimation (CA) by which plants acquire freezing tolerance upon exposure to advanced low non-freezing temperatures (Gilmour et al., 1998;Jia et al., 2016;Park et al., 2015;Zhao et al., 2016). The CBFs, along with their transcriptional activator, inducers of CBF expression 1 (ICE1), constitute ICE1-CBF-COR transcriptional cascade which is the most well-known cold signalling pathway (Chinnusamy et al., 2007). In addition to ICE1, a series of TFs including MYB15, CAMTA3, PIF3, EIN3, BZR1 have been identified as upstream regulators of CBFs (Agarwal et al., 2006;Doherty et al., 2009;Jiang et al., 2017a;Kim et al., 2013;Li et al., 2017;Shi et al., 2012). However, investigations on regulatory networks of the cold stress response are mainly focussed on model plants. For bermudagrass, most studies were about the physiology and biochemical measurement of cold stress response rather than decipherment of molecular mechanism (Fan et al., 2014;Hu et al., 2016;Liu et al., 2016;Shi et al., 2014). In the previous studies, wholegenome differentially expressed mRNAs and miRNAs during cold stress response have been identified in bermudagrass (Chen et al., 2015;Hu et al., 2018); however, due to the lack of an efficient transformation system, there is little progress on functional analysis of cold-stress responsive genes. Recently, we demonstrated that an ethylene responsive factor CdERF1 from bermudagrass positively regulates cold tolerance through ectopic overexpression of CdERF1 in Arabidopsis plants (Hu et al., 2020). Therefore, how to fill the gap in molecular mechanisms of cold stress response in bermudagrass is particularly important.
WRKYs are plant-specific TFs that have been demonstrated to play a crucial role in growth and development processes such as seed germination (Jiang and Yu, 2009a), flowering (Ma et al., 2020;Zhang et al., 2018a), anthocyanin biosynthesis , and leaf senescence (Jiang et al., 2014;Niu et al., 2020). Recent studies have also revealed pivotal roles of WRKY TFs in various biotic and abiotic stresses including cold stress (Jiang et al., 2017b;Kim et al., 2016;Rushton et al., 2010;Sun et al., 2019;Yokotani et al., 2013;Zhang et al., 2019;Zou et al., 2010). Most WRKYs have been reported to play a positive role in cold tolerance. For example, the rice WRKYs, including OsWRKY71 and OsWRKY76 enhance cold resistance (Kim et al., 2016;Yokotani et al., 2013). Consistently, overexpression of VaWRKY12 and VaWRKY33 confer cold resistance in transgenic Arabidopsis and grapevine callus Zhang et al., 2019). Niu et al. (2012) reported that transgenic Arabidopsis plants overexpressing TaWRKY19 results in improved tolerance to freezing stress as well. On the contrary, AtWRKY34 which might be involved in the CBF signal cascade negatively modulates the cold response of mature pollen (Zou et al., 2010). In bermudagrass, a total of 23 WRKY TFs displayed cold-induced expression patterns (Chen et al., 2015). However, the functions of CdWRKYs in cold stress response remain largely unknown.
Soluble sugars are considered to play protective roles in cold stress, given that they not only maintain osmotic pressure but also function as signalling molecules regulating the expression of coldresponsive genes during cold stress (Klotke et al., 2004;Rekarte-Cowie et al., 2008). As an important soluble sugar, sucrose accumulates noticeably under cold stress, which is accompanied by the increased transcription of genes encoding sucrose biosynthesis enzymes, such as sucrose synthase (SS), sucrose phosphate synthase (SPS), and invertase (Ruan, 2014). The sucrose content is reported to have a positive relationship with cold tolerance (Jitsuyama et al., 2002), which is further demonstrated by the finding that overexpression of SPS improved freezing tolerance after CA in transgenic Arabidopsis (Strand et al., 2003). However, the upstream regulatory TFs important for SPS activation under cold stress remain to be identified either in the model or nonmodel plants.
CdWRKY2 whose expression was significantly induced by both chilling and freezing with CA was identified according to our previous transcriptome data (Chen et al., 2015). To further investigate the role of CdWRKY2 in cold stress, function analysis of CdWRKY2 was performed in this study. We found that CdWRKY2 positively regulates cold tolerance through coordinately activating CdSPS1-involved sucrose synthesis and CdCBF1dependent signalling pathways. Our results unravel the mechanism of CdWRKY2-mediated cold resistance, and provide new genetic resources for enhancing cold tolerance in bermudagrass breeding.

Results
Identification and characterisation of cold-responsive CdWRKY2 in bermudagrass According to our previous transcriptome analysis (Chen et al., 2015), a CdWRKY (Comp160681_c0) whose expression was significantly induced after both chilling and freezing treatments was screened. Subsequently, a 1683-bp coding sequence (CDS) was isolated from cold-resistant bermudagrass by rapid amplification of cDNA ends (RACE), encoding 560 amino acids with a highly conserved WRKY domain followed by a C 2 H 2 -zincfinger motif in the N-terminus, which belonged to group II ( Figure S1a). Phylogenetic analysis suggested that Com-p160681_c0 exhibited the highest homology (82.89%) with Sorghum bicolour WRKY2 (SbWRKY2) ( Figure S1b). Hereafter, Comp160681_c0 was named C. dactylon WRKY2 (CdWRKY2, Genbank accession number: OL472363). To confirm whether CdWRKY2 is truly a cold-responsive gene, the expression pattern of CdWRKY2 was further analysed by quantitative real time-PCR (qRT-PCR). Consistent with the RNA sequencing (RNAseq) data, the transcript levels of CdWRKY2 were dramatically up-regulated in both genotypes after exposure to 1, 3, and 6 h of low temperature, especially in the cold-resistant genotype. Notably, the CdWRKY2 expression in cold-resistant bermudagrass was 2.7-fold higher than that in the cold-sensitive one after 3 h of cold treatment (Figure 1a). To further reveal these differentially expressed profiles, the promoter regions, a total of 902-bp and 904-bp in length, were cloned from cold-resistant and -sensitive genotypes, respectively. A total of 5 SNP variations were detected between two promoter regions ( Figure S2), which may contribute to altered expression levels of CdWRKY2 between the two genotypes. Apart from cold stress, the transcript levels of CdWRKY2 were increased by abscisic acid (ABA) and other abiotic stresses, including salt and dehydration in the cold-resistant bermudagrass as well (Figure 1b). To further verify the expression pattern of CdWRKY2, we obtained transgenic Arabidopsis plants containing a GUS reporter gene driven by the CdWRKY2 promoter which was cloned from a cold-resistant genotype. The GUS staining analysis showed that GUS activity was activated by cold stress, as well as NaCl, PEG6000, and ABA treatments (Figure 1c-g), which was consistent with the above qRT-PCR results (Figure 1a,b).
Subsequently, to determine the subcellular localisation of CdWRKY2 protein, the 35S:CdWRKY2-eGFP fusion expression vector was transformed into Arabidopsis, and the GFP signal was examined using confocal microscopy. The GFP signal was located in the nucleus, which was further validated by DAPI staining, indicating that the CdWRKY2 is exclusively located in the nucleus (Figure 1h).

Overexpression of CdWRKY2 enhances cold tolerance in transgenic Arabidopsis
To understand the biological function of CdWRKY2 in regulating cold tolerance, two CdWRKY2-overexpressing Arabidopsis lines (OE3 and OE4) with relative high expression were selected for assessing cold tolerance ( Figure S3a). No conspicuous difference in phenotype was observed between transgenic Arabidopsis plants and WT under normal conditions. However, 10-day-old CdWRKY2-overexpressing Arabidopsis improved cold tolerance after freezing treatments regardless of CA (Figure 2a). With CA, the average survival rates of OE3 and OE4 increased by 16% and 17% compared with WT, whereas the average survival rates of transgenic plants overexpressing CdWRKY2 increased by 39% and 36% without CA ( Figure 2b). Analogously, the phenotypes and average survival rates of 3-week-old transgenic Arabidopsis also indicated that overexpression of CdWRKY2 significantly enhances cold tolerance in transgenic plants (Figure 2c,d).
Subsequently, physiological responses to cold stress were compared between WT and CdWRKY2-overexpressing Arabidopsis plants. After 7 days of 4°C treatment, EL and MDA content which are two representative indicators of cell membrane stability in the transgenic lines were significantly lower than those in WT (Figure 2e,f). Chlorophyll fluorescence intensity can reflect the performance of photosystem II (PSII) in plants' response to cold stress. Before 4°C treatment, the chlorophyll a fluorescence transient (OJIP) curves were indistinguishable between transgenic CdACTIN2 was used as normalisation controls for quantitative real time-PCR (qRT-PCR). The error bars indicate the standard deviation (SD) values while different letters indicate significant statistical differences at among the treatments in two phenotypes according to Duncan's multiple range tests, respectively (n = 3, P < 0.05). Asterisks indicate significant differences between the R and S genotypes under the same treatment time according to Student's t-test (n = 3, **P < 0.01, *** P < 0.001).

Silencing of CdWRKY2 expression in bermudagrass increases hypersensitivity to cold stress
To elucidate the function of CdWRKY2 in regulating cold tolerance in bermudagrass, the virus-induced gene silencing (VIGS) system was employed. As shown in Figure S4a, the transcription level of CdWRKY2 was substantially decreased by about 50% in BSMV:CdWRKY2 compared with that of control Figure 2 Overexpression of CdWRKY2 confers cold tolerance in transgenic Arabidopsis. (a,b) Freezing phenotypes (a) and survival rates (b) of 10-day-old CdWRKY2-overexpressing and WT Arabidopsis seedlings on petri dishes with or without cold acclimation (CA, 7 days at 4°C). Ten-day-old seedlings were subjected to 1 h of freezing treatment at À6°C for non-acclimated (NA) and À7°C for CA, seedlings were transferred to 4°C for 24 h, then recovered at 22°C for an additional 3 days. (c,d) Freezing phenotypes (c) and survival rates (d) of CdWRKY2-overexpressing Arabidopsis and WT plants in soil with or without CA. Three-week-old plants were exposed to 8 h of freezing treatment at À8°C for NA and À9°C for CA followed by recovery at 22°C for 3 days. (e-g) Electrolyte leakage (EL) (e), malondialdehyde (MDA) content (f), and the chlorophyll a fluorescence transient (OJIP) curves (g) of WT and CdWRKY2-overexpressing Arabidopsis under normal (22°C for 7 days) and cold stress conditions (4°C for 7 days). Asterisks indicate significant differences between the transgenic lines and WT under the same growth conditions according to Student's t-test (n ≥ 3, *P < 0.05, **P < 0.01). plants (BSMV:00), which were only infiltrated with empty vectors. Under normal conditions, the BSMV:CdWRKY2 and BSMV:00 bermudagrass plants were morphologically indistinguishable. However, upon exposure to 21 days of 4°C cold stress treatment, leaves of BSMV:CdWRKY2 displayed more severely wilting and necrosis in comparison with those of BSMV:00 ( Figure S4b), which was supported by the results of histochemical staining with nitro blue tetrazolium (NBT; Figure 3a). Consistently, MDA content increased after cold treatment, to a greater degree in BSMV:CdWRKY2 plants ( Figure 3b). Besides, PSII efficiency which was indicated by OJIP curves, F v /F m ratios, PI ABS, and PI total values decreased significantly in BSMV:CdWRKY2 compared with that in control after cold treatment, although it was indistinguishable between BSMV:CdWRKY2 and BSMV:00 under normal conditions (Figure 3c and Figure S4c-e). Subsequently, expression levels of cold marker genes, such as CdABF1, CdCBF1, CdLEA3, and CdCOR440 were further analysed. As a result, CdABF1, CdCBF1, CdLEA3, and CdCOR440 were all down-regulated by knocking down the expression of CdWRKY2 with or without cold stress treatment in bermudagrass ( Figure S4f-i). Taken together, these results indicate that the knockdown of CdWRKY2 expression enhances cold susceptibility in bermudagrass.
Agrobacterium rhizogenes-mediated overexpression of CdWRKY2 confers cold tolerance in bermudagrass with transgenic hairy roots Agrobacterium tumefaciens-mediated genetic transformation in bermudagrass is still immature, which restricts gene functional analysis in vivo. Here, A. rhizogenes-mediated transformation was successfully developed in bermudagrass. The transgenic hairy roots were identified by genomic PCR, GFP signal detection, and qRT-PCR (Figure 3d-f). Transgenic hairy roots with relatively high expression of CdWRKY2 were selected for function characterisation, while transgenic hairy roots transferring an empty vector were used as a control ( Figure 3f). No phenotypic differences were observed between the plants with CdWRKY2-overexpressing hairy roots and control plants under normal conditions. After 10 days of cold stress (4°C day/night temperature) treatment, the leaves of control plants displayed slightly wilting and necrosis, while no obvious damages were observed in plants with CdWRKY2-overexpressing hairy roots ( Figure 3g). NBT staining result indicated that control hairy roots displayed more severe damages in comparison with transgenic hairy roots (Figure 3h), which was consistent with the result of O 2 À content ( Figure 3i).
Consistently, cold marker genes, such as CdABF1, CdCBF1, CdLEA3, and CdCOR440, exhibited significantly higher expression levels in three transgenic lines regardless of cold stress treatment ( Figure S5a-d). These results indicated that CdWRKY2 is indeed a positive regulator in bermudagrass against cold stress.

CdWRKY2 activates the expression of sucrose-related genes
To reveal the underlying regulatory mechanism of CdWRKY2 in plant cold response, RNA-Seq analysis was performed using 7day-old seedlings of WT and CdWRKY2-overexpressing transgenic (OE3) Arabidopsis before or after 6 h of 4°C treatments, named WTCK (7-day-old seedlings of WT before 6 h of 4°C cold stress), OE3CK (7-day-old seedlings of CdWRKY2-overexpressing transgenic Arabidopsis before 6 h of 4°C cold stress), WTLT (7day-old seedlings of WT after 6 h of 4°C cold stress), and OE3LT (7-day-old seedlings of CdWRKY2-overexpressing transgenic Arabidopsis after 6 h of 4°C cold stress), respectively. The hierarchical clustering analysis showed that cold stress dramatically altered gene expression profiles both in WT and transgenic plants ( The Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis showed that these DEGs were significantly enriched in starch and sucrose metabolism, photosynthesis-antenna proteins, carbon metabolism, and glyoxylate and dicarboxylate metabolism ( Figure S6). Gene ontology (GO) analysis displayed that these DEGs were highly involved in the biological processes including response to oxidative stress, response to cold, and cell wall organisation ( Figure S7). To further narrow down the scope of candidate genes involved in CdWRKY2-mediated cold stress, a Venn diagram analysis was performed using DEGs with a fourfold difference in comparisons OE3CK vs WTCK and WTLT vs WTCK. A total of 45 up-regulated and 507 down-regulated DEGs were overlapped between OE3CK vs WTCK and WTLT vs WTCK ( Figure 4c). Among the 45 upregulated DEGs, there were several genes responsive to oxidative stress, including two UDP-glycosyltransferases (UGT74E2 and UGT73B4), one Glutathione transferase (GSTF6), one Germin-like protein (GLP4), and two beta-glucosidases (BGLU34 and BGLU35). Most importantly, two up-regulated genes, AtSPS2F and AtSUS1 (Figure 4d), involved in sucrose synthesis and metabolism process that has been reported to be associated with cold tolerance were identified . The expression patterns of AtSPS2F and AtSUS1 were subsequently validated by qRT-PCR (Figure 4e,f). The AtSPS2F expression was significantly up-regulated in CdWRKY2-overexpressing Arabidopsis plants under normal and cold stress conditions ( Figure 4e). However, unlike with AtSPS2F, the AtSUS1 expression was significantly lower in CdWRKY2-overexpressing Arabidopsis plants under cold stress compared with that in WT ( Figure 4f). We thus further focussed on whether the expression of CdSPS1, which is the homology gene of AtSPS2F in bermudagrass is affected by CdWRKY2. The expression of CdSPS1 was performed in CdWRKY2-silencing bermudagrass and CdWRKY2overexpressing bermudagrass hairy roots under normal and lowtemperature conditions. As expected, mRNA abundance of CdSPS1 was significantly increased in the bermudagrass plants with CdWRKY2-overexpressing hairy roots relative to control roots ( Figure 4g). By contrast, transcript levels of CdSPS1 were dramatically down-regulated in BSMV:CdWRKY2 as compared with BSMV:00 with or without cold stress (Figure 4h). Collectively, sucrose synthesis genes including AtSPS2F and CdSPS1 might be induced by CdWRKY2 directly or indirectly during cold stress.

CdWRKY2 positively regulates sucrose accumulation
Given that the expression levels of genes involved in sucrose biosynthesis were regulated by CdWRKY2, sucrose contents were further measured in CdWRKY2-overexpressing, CdWRKY2-silencing, and control plants. Consequently, sucrose concentrations were induced after cold treatment, and sucrose contents in Arabidopsis lines OE3 and OE4 were significantly higher than those of WT with or without cold stress treatment (Figure 4i). Consistently, sucrose contents in CdWRKY2 transgenic hairy roots were fivefold and 3.5-fold higher than those of control roots under control and cold stress conditions, respectively ( Figure 4j). In contrast, the BSMV:CdWRKY2 bermudagrass exhibited significantly lower sucrose concentrations relative to BSMV:00 plants ( Figure 4k). Taken together, our results indicated that CdWRKY2 positively regulates cold stress likely through inducing sucrose accumulation.
CdWRKY2 activates CdSPS1 expression by directly binding to the CdSPS1 promoter To explore whether CdWRKY2 activates the expression of AtSPS2F and AtSUS1 directly in Arabidopsis, a yeast onehybrid assay (Y1H) was carried out. As a result, yeast strain Y1H Gold harbouring pAtSPS2F-AbAi and pGADT7-CdWRKY2 vectors could grow on SD/-Leu medium supplemented with 100 ng/mL aureobasidin A (AbA), which could suppress the basal expression in the Y1H Gold harbouring pAtSPS2F-AbAi, suggesting the interaction between CdWRKY2 and AtSPS2F promoter ( Figure 5a). However, CdWRKY2 failed to activate the AbAr reporter gene driven by the AtSUS1 promoter ( Figure 5b). Therefore, the homology gene of AtSPS2F in bermudagrass was chosen for further analysis. Because 1000 ng/mL AbA was still unable to suppress the basal expression in the Y1H Gold harbouring proCdSPS1-AbAi, another Y1H system (pB42AD/ pLacZi) was used alternatively. The result showed that GAD-CdWRKY2 fusion protein, instead of GAD (GAL1 transcriptional activation domain, AD), activated the LacZ reporter gene driven by the CdSPS1 promoter which was cloned from cold-resistant genotype (Figure 5c), indicating that CdWRKY2 can bind to CdSPS1 promoter directly. As we know, WRKY TFs activate or suppress target gene expression by binding to the W-box (T/ CTGACC/T) or its core motif (TGAC). Expectedly, a TGAC core motif was found in upstream of the CdSPS1 promoter region (À860~À863 bp). To verify whether CdWRKY2 binds to the W-box of the CdSPS1 promoter, EMSA was conducted in vitro. The formation of the protein-DNA complex was only observed when the fusion protein was incubated with a labelled probe containing a wild-type (WT) W-box element, which was reduced by adding the unlabelled competitor. Besides, no bind shift was detected when the mutated probe was incubated with the fusion protein. (Figure 5d). These results confirmed that CdWRKY2 binds to the Wbox element of the CdSPS1 promoter.
Subsequently, a dual-luciferase reporter assay was performed to further verify the binding of CdWRKY2 to the CdSPS1 promoter. The firefly luciferase (LUC) driven by the CdSPS1 promoter (proCdSPS1:LUC) and renilla luciferase (REN) driven by 35S promoter (35S:REN) were transformed transiently into tobacco leaves with 35S:CdWRKY2 or empty vector (pMD35S), respectively. As shown in Figure 5e, the LUC activity, which is indicated by LUC/REN ratio in leaves co-expressing 35S:CdWRKY2 and proCdSPS1-LUC were about twofold higher than those in the control leaves infiltrated with the pMD35S and proCdSPS1-LUC. Additionally, a transient GUS expression assay was performed as well. The GUS staining was stronger in the tobacco leaf cotransformed with the 35S:CdWRKY2 and proCdSPS1:GUS, which Figure 5 CdWRKY2 directly binds to the promoters of AtSPS2F and CdSPS1 and activates their expression levels. (a,b) Yeast one-hybrid analysis, using pGADT7-CdWRKY2 as the prey, pAtSUS1-AbAi (a) and pAtSPS2F-AbAi (b) as the baits, pGADT7-p53 and p53-AbAi as the positive controls. (c) Yeast onehybrid analysis, using pB42AD-CdWRKY2 as the prey, ProCdSPS1-pLacZ as the baits, pB42AD-p53 and p53-pLacZ as the positive controls. (d) Electrophoretic mobility shift assay (EMSA) of the interaction between fusion protein GST-CdWRKY2 and the CdSPS1 promoter. The purified GST-CdWRKY2 protein was incubated with the biotin-labelled probes containing WT or mutated W-box element. The unlabelled WT probe (509 and 1009) was used as a competitor. The bound DNA-protein complex is indicated by the arrows. +, presence; À, absence. (e) Luciferase activity analysis using 35S: CdWRKY2 as the effector and ProCdSPS1:LUC as a reporter. The REN and LUC are Renilla luciferase and firefly luciferase, respectively. LUC:REN ratio of the control (tobacco leaves co-transformed with the reporter and the empty effector vector) was taken as 1 for normalisation. (f) Transient b-glucuronidase expression analysis, using 35S:CdWRKY2 as the effector and ProCdSPS1:GUS as reporter. GUS activity of the control (tobacco leaves co-transformed with the reporter and the empty effector vector) was taken as one for normalisations. The error bars indicate the SD values from at least three repetitions of each treatment. Asterisks (n = 5, *P < 0.05, ***P < 0.001) indicate significant differences compared with the control, respectively (Student's t-test).
was further supported by the GUS activity assays (Figure 5f). Collectively, all data indicated that CdWRKY2 can directly bind to the CdSPS1 promoter.

CdSPS1 positively regulates cold tolerance in transgenic plants
To investigate whether CdSPS1 plays a role in cold tolerance, two CdSPS1-overexpressing Arabidopsis lines (#7 and #10) with relative high expression levels were selected for further analysis ( Figure S8a). Ten-day-old WT and transgenic Arabidopsis plants were subjected to freezing treatments (À6°C, 1 h) without CA. Expectedly, CdSPS1-overexpressing plants significantly improved cold tolerance of transgenic Arabidopsis seedlings (Figure 6a), as indicated by significantly higher survival rates (Figure 6b). The subsequent physiological responses to cold stress further demonstrated the prominent role of CdSPS1 in cold resistance. After 7 days of 4°C treatment, the EL in the CdSPS1-overexpressing lines was significantly lower than those in WT, although cold stress led to the increase of EL in all plants (Figure 6c). Meanwhile, the decreased PSII efficiency reflected by the OJIP In addition, the function of CdSPS1 in regulating cold tolerance in bermudagrass was further validated by knocking down CdSPS1 expression using the VIGS system (Figure 6e). After 4°C cold stress treatment, more severely wilting and necrosis leaves were observed in BSMV:CdSPS1 compared with BSMV:00 ( Figure S9a), which was evidenced by the results of MDA contents, OJIP curves, F v /F m , PI ABS, and PI total values (Figure 6f,g and Figure S9b-d). Subsequently, the sucrose contents were measured in BSMV:00 and BSMV:CdSPS1 plants, showing that sucrose concentrations were indeed decreased by knocking down CdSPS1 expression (Figure 6h). Taken together, these results indicated that CdSPS1 plays a positive role in cold tolerance both in Arabidopsis and bermudagrass.

CdWRKY2 activates CdCBF1 expression by directly binding to CdCBF1 promoter
It is well known that CBF1 plays a central role in cold resistance, and CdCBF1 expression was induced in CdWRKY2-overexpressing but suppressed in CdWRKY2-silencing transgenic plants (Figures S4g and S5b), which raised the possibility that CdCBF1 may be involved in CdWRKY2-mediated cold stress response. To explore whether CdWRKY2 activates CdCBF1 expression directly, Y1H, EMSA, and LUC-based transactivation experiment were carried out. As shown in Figure 7a, the LacZ reporter gene driven by the CdCBF1 promoter was activated by GAD-CdWRKY2 fusion protein, indicating that CdWRKY2 can bind to CdCBF1 promoter directly. The subsequent EMSA assay demonstrated that CdWRKY2 specifically binds to the W-box (À131~À136 bp) in the CdCBF1 promoter ( Figure 7b). Consistently, the LUC activity in leaves co-expressing 35S:CdWRKY2 and proCdCBF1-LUC was about 3.5-fold higher than that in the control leaves infiltrated with the pMD35S and proCdCBF1-LUC (Figure 7c). All results indicated that CdWRKY2 can directly bind to the CdCBF1 promoter and activate its expression.

Discussion
Cold stress is one of the most common factors that limits the development and geographical distribution of bermudagrass. Increasing evidence has indicated that WRKY TFs play pivotal roles in response to cold stress in many species (Kim et al., 2016;Niu et al., 2020;Sun et al., 2019;Yokotani et al., 2013;Zhang et al., 2019). However, there is no information about CdWRKYs involved in cold stress response as yet. In this study, we identified a cold-induced CdWRKY2 based on transcriptome data and qRT-PCR expression analysis. We found that overexpression of CdWRKY2 enhanced cold tolerance, while silencing of CdWRKY2 impaired cold resistance in bermudagrass, indicating that CdWRKY2 functions as a positive regulator in cold stress response of bermudagrass.

The CdWRKY2 expression is induced by cold stress as well as other abiotic stresses
The expression of CdWRKY2 was significantly induced by cold stress both in cold-resistant and cold-sensitive bermudagrass genotypes, and its level was significantly higher in the coldresistant genotype than that in the cold-sensitive one during cold stress (Figure 1a). Interestingly, CdWRKY2 expression level was 2.6-fold higher in the cold-resistant genotype than that in the cold-sensitive one under normal conditions (Figure 1a), which raises the possibility of variations in CdWRKY2 promoters between two genotypes. Consequently, a total of 5 SNP variations were detected between two CdWRKY2 promoters Figure 7 CdWRKY2 directly binds to the CdCBF1 promoter and activates its expression level. (a) Yeast one-hybrid analysis, using pB42AD-CdWRKY2 as the prey, ProCdCBF1-pLacZ as the bait, and pB42AD-p53 and p53-pLacZ as the positive controls. (b) EMSA analysis of the interaction between fusion protein GST-CdWRKY2 and the CdCBF1 promoter. The purified GST-CdWRKY2 protein was incubated with the biotin-labelled probes containing WT or mutated W-box elements. The unlabelled probe (509 and 1009) was used as a competitor. The bound DNA-protein complex is indicated by the arrows. +, presence; À, absence. (c) Luciferase activity analysis using 35S:CdWRKY2 as the effector and ProCdCBF1:LUC as a reporter. The error bars indicate the SD values from at least three repetitions of each treatment. Asterisks (n = 5, **P < 0.01) indicate significant differences compared with the control (Student's t-test).
from two genotypes ( Figure S2). However, whether the SNP variations result in the differential expression of CdWRKY2 between two genotypes needs to be further investigated. Apart from cold stress, CdWRKY2 expression was also induced by salt, drought, and ABA treatments (Figure 1b). Similarly, the expression level of AtWRKY2 was also increased by NaCl and mannitol treatments (Jiang and Yu, 2009b). As we know, cold, drought, and salinity can lead to water deficit which generates osmotic stress (Zhu, 2016). Moreover, ABA has been suggested to regulate CA by activating osmotic responses (Shi et al., 2014). Therefore, we speculate that CdWRKY2 is likely to play a role in cold-induced osmotic stress, which may be influenced by ABA signalling.

CdWRKY2 improves cold tolerance by positively regulating CdSPS1 expression to accumulate sucrose concentrations
To investigate the regulatory mechanism of CdWRKY2 during cold stress, transcriptome analysis was conducted between WT and CdWRKY2-overexpressing Arabidopsis plants. The data displayed that genes participating in sucrose synthesis and metabolism, especially AtSPS2F, were significantly induced in CdWRKY2-overexpressing Arabidopsis plants both under normal and cold stress conditions (Figure 4e). Likewise, CdSPS1 which is the homology gene of AtSPS2F was remarkably up-regulated in CdWRKY2-overexpressing hairy roots but dramatically downregulated in CdWRKY2-silencing bermudagrass (Figure 4g-h). Corresponding to the changes in expression levels of AtSPS2F and CdSPS1, sucrose concentrations were substantially elevated in CdWRKY2-overexpressing transgenic Arabidopsis and hairy roots under normal and low-temperature conditions (Figure 4i,j), while the opposite trend was observed in CdWRKY2-silencing bermudagrass plants (Figure 4k). Recent studies have suggested that sucrose, an osmotic protectant, plays a vital role in abiotic stresses, including cold stress . Sucrose content prominently increases during CA both in the field and artificial conditions in overwintering evergreens . The increase of sucrose biosynthesis during CA is essential for developing freezing tolerance (Strand et al., 2003). Considering SPS mediates the rate-limiting step of sucrose synthesis (Castleden et al., 2004), the SPS-mediated sucrose biosynthesis is undoubtedly an important issue during cold stress (Almadanim et al., 2017;Strand et al., 2003). For example, OsSPS4 was phosphorylated by OsCPK17 to accumulate sucrose concentration in rice, which is required for cold stress response (Almadanim et al., 2017). However, which upstream TFs regulate SPS expression under cold stress has not been reported either in the model or non-model plants until now. Here, the Y1H, EMSA, LUC, and GUS transient expression assays consistently demonstrated that CdWRKY2 can bind to the CdSPS1 promoter and activate its expression (Figure 5c-f). Moreover, overexpression of sucrose synthetic gene CdSPS1 indeed improved cold tolerance in transgenic Arabidopsis (Figure 6a-d), while knockdown of CdSPS1 expression improved susceptibility to cold stress in bermudagrass (Figure 6e-h), further revealing that CdWRKY2 confers cold resistance by directly activating CdSPS1 expression to increase sucrose concentrations. Although the role of WRKYmediated sucrose synthesis in cold stress has not been reported yet, WRKY-regulated sucrose synthesis has been documented to participate in other abiotic stresses, such as drought stress. For example, VvWRKY30 improves drought tolerance partially by increasing the transcript level of the SS gene SS4 (Zhu et al., 2018). Raineri et al. (2015) reported that HaWRKY76overexpressing Arabidopsis exhibits tolerance to flood and drought with higher sucrose contents relative to control.
In addition, sugar is an energy source that is important for photosynthesis. It was reported that the recovery of photosynthetic capacity under cold stress conditions is strongly dependent on the activation of the sucrose biosynthetic pathway (Strand et al., 2003). Herein, it is noteworthy that the PSII efficiency was rescued in CdWRKY2-or CdSPS1-overexpressing Arabidopsis plants, but was suppressed more severely in CdWRKY2-or CdSPS1-silencing bermudagrass plants during cold stress. Therefore, in addition to osmotic pressure, the photosynthetic ability is also improved by CdWRKY2-mediated sucrose synthesis during cold stress of bermudagrass. Interestingly, there was no significant up-regulation of sucrose content in CdWRKY2-overexpressing hairy roots after cold stress although the CdSPS1 expression was significantly induced in transgenic hairy roots (Figure 4g,j). Plants may activate negative feedback to avoid excess sucrose which may adversely affect plant growth and development, and its underlying mechanism is waiting to be explored.
It is worth noting that CdSPS1 expression was only slightly activated by salt or drought stress ( Figure S10) compared with cold stress. Therefore, the CdWRKY2-CdSPS1 regulatory module is speculated to mainly play an important role in plants against cold stress rather than salt and drought stresses, although CdWRKY2 expression levels were largely induced by salt and drought stresses.

CdCBF1 participates in CdWRKY2-mediated cold stress response in bermudagrass
CBFs play a crucial role in cold stress, the expression of which was rapidly induced by cold temperature (Chinnusamy et al., 2007;Gilmour et al., 1998). Currently, more and more researches demonstrate that CBF expression is also regulated by multiple factors in addition to the most thoroughly understood ICE1-CBF-COR transcriptional cascade (Chinnusamy et al., 2007). For example, BZR1 positively regulates freezing tolerance via the CBFdependent pathway in Arabidopsis (Li et al., 2017), while VvWRKY34 might negatively mediate cold sensitivity through CBF signal cascade in Arabidopsis (Zou et al., 2010). However, there remain few reports on upstream regulators of CBFsdependent pathways in bermudagrass. In our study, CdCBF1 expression was significantly increased about 100-fold in CdWRKY2-overexpressing hairy roots relative to the control, while its expression was dramatically down-regulated in CdWRKY2-silencing bermudagrass plants (Figures S4g and S5b). The interaction between CdWRKY2 and CdCBF1 promoter was further validated by Y1H, EMSA, and LUC transient expression assays (Figure 7). Correspondingly, COR, which is the target gene of CBF showed the same expression pattern, indicating that the CdWRKY2 directly activates the CdCBF1-COR signalling pathway during cold stress.

Antioxidant genes may be involved in CdWRKY2mediated cold stress response in bermudagrass
It is well known that cold stress is always accompanied by the accumulation of ROS which stimulates membrane lipid peroxidation, resulting in cell membrane damage, even cell death (Ruelland et al., 2009). Plants have evolved a complex antioxidant system to cope with the oxidative injury induced by ROS (Miller et al., 2010). Increasing evidence suggests that antioxidant capacity is positively correlated with cold resistance (Hu et al.,  Sun et al., 2019). In this study, a class of genes that were co-induced by cold stress and CdWRKY2 were noticeably enriched in response to the oxidative stress process, especially UGT73B4 and UGT74B2 (Figure 4d). CsUGT78A14 has been reported to enhance cold tolerance by accumulating flavonol glycosides and improving ROS scavenging capacity . Down-regulating UGT91Q2 reduced ROS scavenging capacity and accumulation of nerolidol glucoside, thus leading to lower cold tolerance in tea plants . Correspondingly, the cell membrane stability was rescued by overexpressing CdWRKY2 as indicated by decreased EL and MDA under cold conditions (Figure 2e,f). Therefore, the fine-tuning of genes with antioxidant capacity may partially contribute to CdWYKY2-mediated cold tolerance.

A. rhizogenes-mediated transformation system successfully removes barriers of functional analysis in bermudagrass
The genetic transformation with high efficiency is a determinative factor for investigating gene function and improving germplasm in bermudagrass. However, the A. tumefaciens-mediated transformation system in bermudagrass has a very low transformation efficiency, which greatly hampers gene function analysis . For many plant species, a fast and efficient transformation technique with A. rhizogenes has been alternatively developed (Kereszt et al., 2007;Meng et al., 2019). In this study, A. rhizogenes-mediated transformation has been successfully established in bermudagrass (Figure 3d-f). A series of hairy roots of bermudagrass with high expression of CdWRKY2 were generated (Figure 3d,f). Although only transgenic hairy roots rather than the whole transgenic plant are obtained, A. rhizogenes-mediated transformation has been well employed in investigation of abiotic stress previously. For example, soybean plants with GmMYB118-overexpressing hairy roots generated via A. rhizogenes-mediated transformation increased salt and drought tolerance (Du et al., 2018). The soybean seedlings with WRKY54-RNAi hairy roots which are produced by A. rhizogenesmediated transformation show higher drought sensitivity (Wei et al., 2019). Here, the bermudagrass seedlings with CdWRKY2overexpressing hairy roots indeed enhanced cold tolerance.
In conclusion, our findings demonstrate that CdWYKY2 acts as a positive regulator in cold stress by coordinately activating CdSPS1-involved sucrose synthesis and CdCBF1-dependent signalling pathways (Figure 8). Moreover, we develop an efficient A. rhizogenes-mediated transformation system in bermudagrass for gene functional analysis. The study provides valuable clues for the genetic modification of bermudagrass in response to cold stress.

Plant materials and growth conditions
In this study, the Columbia ecotype of Arabidopsis thaliana was used as the WT background. The plant seeds were sown on Petri dishes containing half-strength Murashige and Skoog (MS) medium supplement with 3% sucrose and 0.8% agar after surface sterilisation and incubated at 4°C for 2 days. Five-day-old seedlings were transferred to soil and moved to an artificial chamber under control environmental conditions (22 AE 1°C under a 16 h light/8 h dark cycle with a photon flux of 240 lmol/m/s).
Analogously, the cold-resistant and cold-sensitive bermudagrass (C. dactylon) genotypes screened previously were used here (Chen et al., 2015). Uniform stolons propagated from one original plant were planted in the plastic pots (7.5 cm in diameter and 9.0 cm deep) that were filled with nutrient soil (Beilei, Wuhan, China), then maintained in the artificial chamber with conditions of 12-h-light (240 lmol/m/s, 30°C)/12-h-dark (28°C) for about 2 months. During the establishing period, the plants were watered three times each week and fertilised weekly with full-strength Hoagland's solution.

Cloning and sequence analysis
A 1683 bp CDS of CdWRKY2 in cold-resistant bermudagrass was amplified based on annotated transcriptome analysis described by Chen et al. (2015) using RACE kits (Takara, Dalian, China). The amino acid sequences of CdWRKY2 were used to search homologous proteins by the BLASTp program in the GenBank database (http://www.ncbi.nlm.nih.gov/). The phylogenetic tree was constructed using the MEGA6.0 software. Similarly, the CDS sequence of CdSPS1 was cloned according to the full-length transcriptome of bermudagrass using reverse transcription PCR (Zhang et al., 2018b). The promoters of CdWRKY2, CdSPS1, and CdCBF1 were amplified with specific primers (Table S1) using a genome walking kit (Takara, Dalian, China).

Abiotic stresses treatments of bermudagrass
In order to investigate the changes of CdWRKY2 and CdSPS1 expression under cold stress, 2-month-old cold-resistant and cold-sensitive bermudagrass were transferred into the growth chamber with 4°C for chilling treatment. For salt, drought, and ABA treatments, 2-month-old cold-resistant bermudagrass were treated with 250 mM NaCl, 25% PEG6000, and 100 µM ABA, respectively. The leaf samples for qRT-PCR analysis were harvested at 0, 3, 6, 12, and 24 h after treatments.

RNA extraction and gene expression analysis
The total RNA from plants was extracted using Trizol-reagent (Invitrogen, Carlsbad) and was digested by DNase I (Beyotime, Figure 8 A proposed model for explaining the regulatory mechanism of CdWRKY2-meidated cold stress response. CdWRKY2 is significantly upregulated when bermudagrass is exposed to cold stress. On one hand, the cold-induced CdWRKY2 directly activates CdSPS1 expression to mediate sucrose biosynthesis, thus conferring cold tolerance of bermudagrass. On the other hand, CdWRKY2 directly binds to CdCBF1 promoter to activate CORs expression, contributing to cold tolerance of bermudagrass. Shanghai, China) before first-strand cDNA synthesis. The firststrand cDNA was obtained using the M-MLV cDNA synthesis kit (Promega, Shanghai, China). qRT-PCR was performed using a SYBR Green PCR Master Mix kit (Monad, Wuhan, China) on the Step One Plus Real-Time PCR Systems (Applied Biosystems, Thermo Fisher Scientific Corp. 5). AtACTIN2 and CdACTIN2 were used as reference genes for Arabidopsis and bermudagrass, respectively. The relative expression levels were calculated as previously described (Hu et al., 2020). Primers used for qRT-PCR are listed in Table S1.

Arabidopsis transformation
To generate CdWRKY2-and CdSPS1-overexpressing transgenic plants, both CdWRKY2 and CdSPS1 CDS sequences were amplified from cold-resistant bermudagrass and cloned into pMD35S vector. To generate 35S:CdWRKY2-eGFP transgenic plants, the CdWRKY2 CDS without the termination codon was amplified from cold-resistant bermudagrass and inserted into pBI121-eGFP vector. To generate proCdWRKY2:GUS transgenic plants, the CdWRKY2 promoter region amplified from coldresistant bermudagrass was cloned into pCAMBIA1300-GUS vector. Transgenic plants were obtained according to the floral dip method (Clough and Bent, 1998) and screened with 50 mg/L kanamycin or 25 mg/L hygromycin.

Subcellular localisation of CdWRKY2
Seeds of the 35S:CdWRKY2-eGFP transgenic Arabidopsis were germinated on 1/2 MS medium after surface sterilisation, roots of five-day-old seedlings were used for fluorescence detection. The fluorescence signal of the 35S:CdWRKY2-eGFP fusion protein was observed using a DMI6000 CS confocal laser scanning microscope (Leica, Wetzlar, Germany).

GUS staining
Five-day-old seedlings of proCdWRKY2:GUS transgenic Arabidopsis were treated with various abiotic stresses, including 4°C (cold stress), 125 mM NaCl (salt stress), 25% PEG6000 (osmotic stress), and 100 lM ABA. After 6 h of various treatments, the seedlings were incubated in the X-Gluc solution immediately for 2 h at 37°C. Then, chlorophyll was removed using 70% (v/v) ethanol and the seedlings were photographed .

A. rhizogenes-mediated transformation system of bermudagrass
The seeds of commercial bermudagrass (Bailv Landscape Design Co., Ltd., Xi'an, China) were sown on 1/2 MS medium after surface sterilisation under dark conditions. After 2 weeks, the elongated stems were cut into 1 cm segments for A. rhizogenes infection. A. rhizogene strain MSU440 harbouring 35S: CdWRKY2-eGFP or empty vector (pBI121-eGFP) was cultured in 50 mL LB liquid medium plus 20 mg/L streptomycin and 50 mg/L kanamycin until the OD 600 value reached 0.5-0.6. The MSU440 suspension was centrifuged at 7656 g for 8 min under room temperature and then re-suspended in a liquid infection medium (MS + 100 lM acetosyringone). The stem segments were incubated with A. rhizogenes suspension, vibrated in an ultrasonic cleaner for 5 min and then shake with 180 rpm at 28°C for 0.5 h. Then, the infected stem segments were cultured on a coculture medium with (MS + 100 lM acetosyringone) at 22°C under darkness. After 3 days, the segments were transferred into a screening medium (MS + 50 mg/L kanamycin + 300 mg/L cefotaxime) for about 1 month. Subsequently, the growing hairy roots were transferred into MS medium containing 50 mg/L kanamycin and subcultured for about 1 month. Finally, to exclude the potential contamination caused by A. rhizogenes as possible as we can, we chose the hairy roots without the outbreak of A. rhizogenes for the subsequent experiments.

Virus-induced gene silencing
To knock down the expression of CdWRKY2 or CdSPS1 in bermudagrass, the VIGS system was performed according to the method as previously described (Hu et al., 2020). Briefly, a specific fragment of CdWRKY2 or CdSPS1 was inserted into the pCa-cbLIC vector under the control of CaMV35S promoter. The constructed vector was transformed into A. tumefaciens EHA105 which was then infiltrated into tobacco leaves. After 7 days, the infiltrated leaves were harvested and ground in 20 mM phosphate buffer (pH 7.2). Subsequently, the mixed liquid was inoculated into the leaves of 2-month-old bermudagrass via mechanical friction. Two weeks after infection, fully expanded leaves were collected from the transfected bermudagrass for qRT-PCR to screen putative VIGS plants which were used for further analyses.

Freezing treatments of Arabidopsis plants
For Arabidopsis seedlings on Petri dishes with CA, 10-day-old Arabidopsis seedlings were kept at 4°C for 7 days, and then transferred into a freezing chamber with À7°C for 1 h of freezing treatment. For Arabidopsis seedlings on Petri dishes with non-acclimation (NA), 10-day-old seedlings were directly exposed to À6°C for 1 h. After freezing treatments with or without CA, seedlings were transferred to 4°C for 24 h, then recovered at 22°C for an additional 3 days for phenotype observation and detection of survival rates.
For Arabidopsis plants in soil with CA, 3-week-old Arabidopsis plants were treated with 4°C for 7 days, and then transferred into a freezing chamber with À9°C for 8 h. For Arabidopsis plants in soil with NA, the age-matched seedlings were directly transferred into 8°C for 8 h of freezing treatment. The phenotype and survival rates were scored after 3 days of recovery at 22°C.

Measurement of physiological and histochemical staining
For physiological analyses, both Arabidopsis and bermudagrass plants were treated at 4°C for 7 days. The relative EL and MDA contents were measured as previously described (Hu et al., 2020). The chlorophyll fluorescence transient curve was recorded by pulse-amplitude modulation (PAM) fluorimeter (PAM2500, Walz, Germany) and the OJIP curves were analysed using the JIP-test method described by Yusuf et al. (2010). Histochemical staining of O 2 À was conducted with NBT as previously reported (Hu et al., 2020 10 min in the water bath, then fast-cooled to room temperature and centrifuged at 25°C with 1500 g for 10 min. The supernatant was decolourised using activated carbon at 80°C for 10 min, and then added into 1 mL extracting solution. The mixed solution was centrifuged at 25°C with 1500 g for 10 min and the supernatant was collected for further analysis. The absorbance of the supernatant at 408 nm was measured using a microplate reader (M200 PRO, TECAN, M€ annedorf, Switzerland).

RNA-Seq analysis
RNA-Seq analysis was performed using 7-day-old seedlings of WT and CdWRKY2-overexpressing transgenic Arabidopsis before or after 6 h of 4°C treatments, named WTCK, OE3CK, WTLT, and OE3LT, respectively. Total RNA was extracted using the Trizol kit (Invitrogen, Carlsbad, CA) from seedlings. Two biological replicates were used for each sample, and the RNA quality and concentration were detected by NanoDrop 2000 (Thermo, Waltham) and a 1.2% agarose gel. The cDNA libraries were constructed and sequenced in Novogene Company (Beijing, China) according to the standard procedure. The raw data were filtered to remove the low-quality reads (Q value ≤ 20) and then the clean date was mapped to the Arabidopsis thaliana genome. The gene expression levels of each sample were calculated by FPKM method. The DEGs were analysed using DESeq2 software. GO analysis and KEGG classifications by DEGs were performed using clusterProfile software.

Yeast one-hybrid assay
For Matchmaker Gold Y1H Library Screening System, the CDS of CdWRKY2 was fused to the GAL4 AD in the pGADT7 vector to generate the prey vector (pGADT7-CdWRKY2), while the promoters of AtSUS1 or AtSPS2F was inserted into the pAbAi vector to construct the baits (proAtSUS1/AtSPS2F-AbAi). The pGADT7-CdWRKY2 vector was transformed into the Y1H Gold yeast strain. After selecting the transformants on SD/ÀUra plates, BstBI-cut bait vector was introduced into the Y1H Gold yeast strain containing pGADT7-CdWRKY2. Positively cotransformed cells were screened on SD/ÀLeu medium supplemented with AbA and cultured at 30°C for 3 days. A positive (pGADT7-p53 + p53-AbAi) control was processed in the same manner.
For the EGY48 Y1H system, the CDS regions of CdWRKY2 were fused to the GAL1 AD in the pB42AD vector to generate the prey vector (pB42AD-CdWRKY2), while the promoter of CdSPS1 and CdCBF1 was fused to the pLacZi vector to construct the baits. pB42AD-CdWRKY2 vector was transformed into the EGY48 yeast strain. After selecting the transformants on SD/ÀTrp plates, the NcoI-cut bait vector was introduced into the EGY48 yeast strain harbouring pB42AD-CdWRKY2. Positively co-transformed cells were screened on SD/ÀTrp/-Ura medium and cultured at 30°C for 3 days. The resultant transformants were tested for bgalactosidase activity on selective media (SD/ÀTrp/-Ura/BU salt/ X-gal). A positive (pB42AD-p53 + p53-LacZi) control was processed in the same manner.

Electrophoretic mobility shift assay
The CDS of CdWRKY2 was cloned into pGEX-6p-1 entry vector, and then were transformed into Escherichia coli BL21 (DE3) competent cells. The GST-WRKY2 fusion protein was induced by 1 mM Isopropyl-b-D-thiogalactopyranoside at 30°C for 4 h and purified using the GSTSep Glutathione Agarose Resin (Yeasen, Shanghai, China) according to the manufacturer's instructions. The purified GST-CdWRKY2 fusion protein and the biotin-labelled DNA probes containing either WT or mutated W-box (Table S1) were used for EMSA. The EMSA was carried out using the Light Shift Chemiluminescent EMSA Kit (Beyotime, Shanghai, China) according to the manufacturer's instructions. Photos were obtained by a multifunctional imaging system (FluorChem R, Proteinsimple, America).

LUC reporter assay
For transient expression assays, the promoter region of CdSPS1 and CdCBF1 were ligated into pGreenII-0080-LUC to generate the reporter construct, proCdSPS1-LUC and proCdCBF1-LUC, respectively. The 35S:CdWRKY2 construct was served as an effector. The effector and reporter constructs were transformed into A. tumefaciens strain GV3101 harbouring pSoup helper vector, respectively, which were further co-infected Nicotiana benthamiana leaves. The injected tobacco plants were kept in the dark for 2 days and then 1 days in the normal condition. Transient expression was reflected by measuring firefly LUC and REN luciferase activities using the Dual-Luciferase Reporter Assay System according to the manufacturer's instructions (Promega, Wisconsin). Five biological replications were measured for each sample.

GUS reporter assay
The same 900-bp promoter region of CdSPS1 was ligated into pCAMBIA1300-GUS to generate the reporter construct, proCdSPS1-GUS. The 35S:CdWRKY2 construct was used as an effector. The effector and reporter constructs were transformed into A. tumefaciens strain GV3101, and then were co-infected into N. benthamiana leaves. The GUS activity was assayed according to a previous method . Five biological replications were measured for each sample.

Statistical analysis
All experiments in this study were performed with at least three repetitions. The significance of differences was determined by ANOVA or Student's t-test using IBM SPSS 20 software, as indicated in the figure legends. manuscript; LC and LWC revised the manuscript. XBH and LWC contributed equally.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1 Characteristic analysis of CdWRKY2. Figure S2 The differences of cis-acting elements in CdWRKY2 promoter region between cold-sensitive (S) and cold-resistance (R) bermudagrass genotypes. Figure S3 CdWRKY2 expression and photosynthesis indexes in CdWRKY2-overexpressing Arabidopsis plants. Figure S4 Silencing of CdWRKY2 by VIGS leads to impaired cold tolerance in bermudagrass. Figure S5 Expression patterns of cold marker genes in HREV and HROEs after cold treatment. Figure S6 The KEGG pathway analyses. Figure S7 GO analyses. Figure S8 CdSPS1 expression and photosynthesis indexes in CdSPS1-overexpressing Arabidopsis plants. Figure S9 Silencing of CdSPS1 by VIGS leads to impaired cold tolerance in bermudagrass. Figure S10 Expression patterns of CdSPS1 under abiotic stresses. Table S1 Primers used in the study.