Seeds of bermudagrass (cv. ‘common’) were sown in pot filled with soil: sand: perlite (1:1:1, v/v/v) and maintained in a growth chamber at 30/25°C (day/night) with a light intensity of 400 µmol m^-2 s^-1 and 60% relative humidity under a 12-h photoperiod. To analyze gene expression across developmental stages and tissues, samples were collected as follows: seedlings were harvested at 20 days post-germination; mature leaves (the second fully expanded leaf from the top) and roots were collected during the vigorous growth stage (approximately 10 weeks after sowing); and spica were sampled at the heading stage (emergence of the inflorescence). For cold stress, plants were exposed to 5°C for 7 d and leaves were sampled at 0, 3, 6, 12, 24, and 48 h for gene expression analysis.

Total RNA was extracted from bermudagrass leaves, and first-strand cDNA synthesized for CdHsfA9 open reading frame (ORF) amplification with primers CdHsfA9-F and CdHsfA9-R (Supplementary Table 1) designed from the assembled sequence and genome of bermudagrass (Zhang et al., 2022). The amplicon was sequenced to confirm identity, and the CdHsfA9 sequence was deposited in GenBank (MG257788).

CdHsfA9 homologs sequence were collected using the BLASTp program and were aligned using the ClustalW program with standard parameters. A phylogenetic tree was constructed in MEGA5 using the neighbor-joining method with 1,000 bootstrap replicates (Sohpal et al., 2010). Conserved domains were analyzed by quering the deduced amino acid sequences against the NCBI Conserved Domain Database (CDD).

For subcellular localization, the CdHsfA9 coding sequence was fused in-frame to enhanced green fluorescent protein (eGFP) in the pEZR_K_LN vector. Rice protoplast isolation, purification, and PEG-mediated transfection were performed as described previously (Datta and Datta, 1999). DNA concentrations in the PEG solution were adjusted to 10 µg/mL, and cells were incubated for 8–10 min at room temperature. Fluorescence signals were observed with a FV1000 confocal laser-scanning microscope (Olympus, Tokyo, Japan). DAPI staining was used to visualize nuclei.

The full-length CdHsfA9 coding sequence was cloned into pGADT7 vector for expression in yeast strain Y187. It was found that no yeast colonies grew on the defective plate of SD/-Trp/-Leu/-His (SD-TLH) medium supplemented with 50 mM 3-amino-1,2,4-triazole (3-AT). Based on the self-activation test (Supplementary Figure 1), 50 mM 3AT was therefore used for subsequent screening. A motif library plasmid was transferred into Y187 containing the pGADT7-CdHsfA9 plasmid, and transformants were selected on SD-TLH + 50 mM 3-AT. Positive yeast clones were identified by DNA sequencing, and CdHsfA9-interacting motifs were compiled.

All experimentally derived motifs and standard HSE were converted into the format of position-weight matrices (PWMs). The heatmap of motif similarity clustering was drawn based on the pairwise comparison of these 39 motifs and the HSE using the Pearson Correlation Coefficient (PCC) method (Gupta et al., 2007). All experimental motifs were scanned across the promoter set with FIMO to obtain the matched genes for KEGG enrichment analysis. Furthermore, experimentally derived motifs were compared against JASPAR and PlantTFDB database using TOMTOM tool (Bailey et al., 2015; Gupta et al., 2007) to identify matches with known transcription factor (TF), specifically targeting HSF families. For genome-wide target prediction, the PWMs of candidate motifs and HSE were used to scan the 2.0 kb upstream promoter regions of the bermudagrass genome via FIMO (Grant et al., 2011). The predicted binding sites were then cross-referenced with canonical HSE motif. Genes containing promoter regions where the candidate motif binding sites spatially overlapped with HSE and annotated genes mapped to enriched pathways were designated as putative direct targets.

Transient expression in Nicotiana benthamiana (5-leaf stage) was performed by Agrobacterium-mediated transient transformation. The CdHsfA9 coding sequence and 620-bp promoter fragment containing the target motif-34 were cloned into pBinGFP2 (effector) and pGreenII_0800_LUC (reporter), respectively (KpnI/BamHI restriction sites). Agrobacterial suspensions carrying effector and reporter were infiltrated into leaves with a needle-less syringe. After injection, the plants were cultured for 24–48 h prior to imaging on a ChemiDoc system, and exposure settings were optimized for luminescence visualization. The injected leaves were extracted and tested for the activity of firefly luciferase (Fluc) reporter gene in TECAN infinite M200 PRO (Tecan, Mannedorf, Switzerland) according to the instructions of double-luciferase reporter assay kit (Transgen Biotechnology, Beijing, China). The Renilla luciferase (Rluc) gene, driven by the CaMV 35S promoter in the pGreenII_0800_LUC vector, was used as an internal control to normalize transformation efficiency. The relative luciferase activity (Fluc/Rluc) was calculated and compared between the experimental group (pBinGFP2-CdHsfA9 + pGreenII_0800_LUC-motif-34) and negative control groups (including pBinGFP2 + pGreenII_0800_LUC-motif-34, pBinGFP2-CdHsfA9 + pGreenII_0800_LUC, and the double empty vectors).

The CdHsfA9 ORF (BamHI/SpeI) was inserted into a pCAMBIA1301 where the original 35S promoter was replaced by the Ubiquitin (Ubi) promoter, with a C-terminal FLAG tag. The destination plasmid in Agrobacterium tumefaciens strain GV3101 was transformed into Arabidopsis by floral dip. The positive transgenic lines of Arabidopsis were selected through hygromycin resistance and PCR confirmation.

Arabidopsis ecotype Columbia was used as the wild type (WT). The hsfA8 (AT1G67970) T-DNA insertion mutant SALK_070573C was obtained from the Arabidopsis Biological Resource Center, with homozygous mutants confirmed by PCR. Surface-sterilized T1/T2 seeds were germinated on Murashige and Skoog (MS) medium containing 100 mg/L hygromycin at 25/20°C (day/night) under a 12-h photoperiod. Homozygous T3, WT, and hsfA8 seeds were germinated on MS medium and transferred to pots containing peat: vermiculite (3:1, v/v) for maturation, cold-response assays, and gene-expression analysis.

To evaluate physiological responses to cold stress, transgenic plants of CdHsfA9 overexpression (OE) and empty-vector (EV), WT, and hsfA8 were exposed to 5°C for 7 d. The chlorophyll content of transgenic plants was extracted by soaking 50 mg of fresh leaves in 10 mL dimethyl sulfoxide in the dark for 72 h. Absorbance at 663.2 nm, 646.8 nm, and 470 nm was recorded, and chlorophyll content was calculated according to Hiscox and Israelstam (1979) and Lichtenthaler and Buschmann (2001). The extracted supernatant from the powder of transgenic leaves (about 80 mg) was collected for protein quality, enzyme assays, and malondialdehyde (MDA) content. Lipid peroxidation was estimated by measuring MDA content with some modifications (Dhindsa et al., 1981). Protein concentration was quantified using the Bradford method (Bradford, 1976). The activities of antioxidant enzymes, including superoxide dismutase (SOD), peroxidase (POD), catalase (CAT), and ascorbate peroxidase (APX), were determined according to the method of Zhang and Kirkham (1996).

To further verify the binding of CdHsfA9 to the target motif, leaves from OE lines were cross-linked, and chromatin was isolated and sheared by sonication. Anti-FLAG antibodies were used for immunoprecipitation; input chromatin served as control. Eluted DNA was purified and analyzed by qPCR to assess enrichment at candidate promoter regions.

Total RNA from bermudagrass or Arabidopsis was extracted with TRIzol reagent (Invitrogen, USA). First-strand cDNA was synthesized using PrimeScript RT reagent Kit with gDNA Eraser (Perfect RealTime) (TransGen, China). Quantitative real-time PCR (qRT-PCR) was performed in a total volume of 10 μL containing 1 μL cDNA, 0.5 μL of each primer and 5 μL of SYBR Green mix (TransGen, China) on a Roche LightCycler 96 Sequence Detection System (Roche, USA), with reaction for 5 min at 95°C followed by 40 amplification cycles of 30 s at 95°C, 30 s at 55°C, and 30 s at 72°C. Data were normalized according to the AtActin2 for Arabidopsis (NM_112764.4) or CdActin for bermudagrass (Chen et al., 2021) gene expression level and determined by 2^-ΔΔCT calculation methods. Primers for qRT-PCR are listed in Supplementary Table 1. The analysis included three biological replicates and three technical replicates for each sample.

Gene expression levels and cold stress effects and variations among wild type, mutants and transgenic plants for physiological parameters were statistically analyzed using the one-way analysis of variance (ANOVA) followed by Fisher’s LSD test using GraphPad Prism 9. Differences were considered statistically significant at p < 0.05.

Based on the assembled HsfA9 sequence from bermudagrass genome and transcriptome databases, a 1200-bp ORF sequence was amplified from the cDNA library of cultivar ‘common’. BLAST and phylogenetic analysis placed CdHsfA9 within the Hsf-A group (Supplementary Figure 2). Given the functional diversification within HsfAs and the ambigous classification of HsfA8 and HsfA9 between dicots and monocots, further analysis of evolutionary relationship was performed including orthologs from rice, Brachypodium, barley, Arabidopsis, and etc. (Figure 1a). CdHsfA9 showed the highest similarity to HsfA9 orthologs in Panicum hallii, Setaria italica, and Zea mays, and clustered on the same branch as Arabidopsis HsfA8 (AtHsfA8).

Conserved domain analysis of the CdHsfA9 polypeptide revealed an N-terminal DNA-binding domain (DBD), an oligomerization domain (HR-A/B), an activator motif (AHA), and a C-terminal nuclear export signal (NES) (Figure 1b). A predicted nuclear localization signal (NLS) located between the DBD and HR-A/B suggested nuclear targeting of CdHsfA9. To verify subcellular localization, a CdHsfA9-GFP fusion was transiently expressed in rice protoplasts. The overlap of GFP and DAPI fluorescence indicated nuclear localization of CdHsfA9 (Figure 1c).

To assess spatial and development expression of CdHsfA9, multiple tissues across bermudagrass growth stage were examined. CdHsfA9 transcripts were accumulated predominantly in mature leaves, followed by seedlings, spica, and roots (Figure 2a). To evaluate cold responsiveness, plants were exposed to 5°C and the relative expression level of CdHsfA9 were determined in the leaves collected 0, 3, 6, 12, 24, and 48 h after treatment. The transcription levels increased and peaked by 2.5- to 5-fold within 12 h (Figure 2b), followed by a gradual decline with prolonged cold exposure (Figure 2b).

To test the role of CdHsfA9 in regulating cold tolerance, a FLAG-tagged CdHsfA9 was expressed constitutively in Arabidopsis. Given that CdHsfA9 clustered with AtHsfA8 (Figure 1a), the hsfA8 mutant (SALK_070573C) was included for comparison.

The transcripts of CdHsfA9 were detected in overexpression lines (OE) but not in wild type (WT), empty-vector lines (EV), or the hsfA8 mutant, indicating successful transformation (Figure 3a). Compared with WT and EV, AtHsfA8 gene expression was strongly reduced in the mutant, confirming the mutant genotype. Notably, AtHsfA8 gene expression was also reduced in OE plants, possibly reflecting functional overlap between CdHsfA9 and AtHsfA8. By contrast, AtHsfA9 expression levels were relatively low in WT and EV lines, they appeared further suppressed in OE plants. Most notably, the AtHsfA9 transcipt level was significantly upregulated in the hsfA8 mutant (Figure 3a), suggesting functional redundancy and compensation between AtHsfA9 and AtHsfA8. Moreover, IAA content was elevated in OE lines (Figure 3b).

Following 7 d at 5°C treatment, the cold stress caused extensive leaf purple/yellow discoloration in WT, EV, and HsfA8 mutant plants (Figures 3c, d). While EV and mutant lines showed similar chlorophyll reduction under cold conditions, analysis of MDA content confirmed that EV and WT suffered comparable levels of cellular damage, distinct from the OE lines (Figures 3d, e). In contrast, OE plants exhibited robust cold tolerance, characterized by reduced discoloration, higher chlorophyll retention, and lower MDA accumulation compared to all control groups (Figures 3c-e). Consistent with this, OE plants maintained higher antioxidative activities of SOD and POD (Supplementary Figure 3). These data collectively demonstrate that heterologous overexpression of CdHsfA9 enhances cold resistance in Arabidopsis by preserving photosynthetic capacity and membrane integrity.

To explore the regulatory mechanism underlying CdHsfA9-mediated cold tolerance, a TF-Y1H assay was conducted using the bait of pGADT7-CdHsfA9. The sequencing of positive clones identified 39 CdHsfA9-interacting motifs. To identify HSE-like targets, a motif similarity clustering analysis of these motifs and the canonical HSE was performed by PCC method. As shown in the heatmap, several candidates, like motif-21 and motif-34, clustered closely with the reference HSE motif, indicating high structural similarity (Figure 4a). Genome-wide promoter scanning of bermudagrass predicted putative CdHsfA9 targets. KEGG pathway analysis revealed that these target genes were enriched in metabolic pathways (including carbon metabolism, glycolysis/gluconeogenesis, and pyruvate metabolism) and signal transduction pathways such as MAPK signaling pathway and zeatin biosynthesis (Figure 4b). These data suggest that CdHsfA9 not only regulates the classic temperature-response modules but also is implicated in energy metabolism reprogramming and kinase-mediated signaling networks. Through alignment with public databases, motif-34 (GGTACTA) was identified as a high-confidence HSF-binding consensus. Then, based on the strict criteria of sequence similarity and spatial overlap of targeted promoters with canonical HSEs, the gene encoding serine/threonine-protein kinase D6PK-like (D6PKL) was prioritized as a primary candidate. The D6PKL promoter was found to contain motif-34 binding site, consistent with the kinase-related pathway enrichment.

To determine the binding ability of CdHsfA9 and motif-34 in the promoter of D6PKL in plants, a dual-luciferase assay in Nicotiana benthamiana leaves was performed. The results of three repeated experiments all showed that co-expression of pGreenII_0800_LUC-D6PKL with pBinGFP2-CdHsfA9 led to a significant increase in the Fluc/Rluc ratio compared with control combinations (Figure 4c), indicating promoter activation of D6PKL by CdHsfA9. To further validate in planta binding, ChIP-qPCR was performed on FLAG-tagged OE lines using FLAG antibodies. ChIP-western blotting confirmed the specific presence of the CdHsfA9-FLAG fusion protein in the Input and ChIP eluates of the OE lines, while no signal was detected in the WT and EV negative controls (Figure 4d). Subsequent ChIP-qPCR analysis showed a ~10-fold enrichment of the promoter of AT5G03640 (the Arabidopsis homolog of D6PKL) in the OE lines relative to the WT and EV controls (Figure 4e), supporting direct binding of CdHsfA9 to the D6PKL promoter region.