Temporal multi-omics analysis of HSF-HSP and flavonoid pathways under heat stress in Iris

Abstract

Heat stress poses a significant threat to the productivity of ornamental plants. Iris tectorum (Iris) is a highly valued ornamental species known for its diverse flower colors and architectural forms. However, heat‑induced physiological damage can substantially reduce its economic value. To investigate the temporal dynamics of the heat stress response in I. tectorum 'Hot Spicy', we conducted a time-series experiment at 40 °C for 0, 1, 3, 6, 9, and 24 h, integrating physiological measurements, transcriptomics, and metabolomics. Physiologically analyses showed that heat stress induced progressive membrane lipid peroxidation, as indicated by MDA accumulation, transient activation of antioxidant enzymes, with SOD and POD peaking at 6 h, and sustained proline accumulation, reflecting a phased stress response. Transcriptomic analysis showed temporally coordinated expression of heat stress factors—heat shock proteins (HSF-HSP) co-expression networks. Small HSPs genes, including HSP26.7, HSP16.0 and HSP22.0 were rapidly induced within 1–3 h, with increases of up to 39.46-fold. Broad activation of HSP genes peaked at 6–9 h, reaching up to 56.50-fold, whereas HSP70 expression persisted at 24 h, with increases of up to 7.77-fold. HSFs displayed hierarchical activation, with HSFA5 responded early, HSFA6b in the mid-phase, and HSFA3/HSFA4a in the late phase, suggesting a layered transcriptional temporal sequence. During the late phase (6–24 h), the ABA signaling pathway and flavonoid biosynthesis genes were upregulated, accompanied by the accumulation of flavonoids such as naringenin chalcone and (-)-epigallocatechin. Weighted Gene Co-expression co-expression Analysis (WGCNA) identified a core module (MEbrown4) enriched in phenylpropanoid biosynthesis and hormone signaling, with KIN10 and HOP2 as highly connected candidate hub genes. Together these results provide a time-resolved multi-omics landscape of heat adaptation in Iris and suggests a phased temporal pattern response from early HSF-HSP activation to sustained flavonoid accumulation. The identified candidates, including HSFA5, HSP70-16, ABF3, KIN10, HOP2, provide candidate genetic resources for future functional validation in Iris.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09422-9.

Introduction

Climate change-induced global warming, with intensifying heat island effects, increasingly threatens the productivity and ornamental value of horticultural plants [1]. Heat stress impairs plant growth, physiological performance, and metabolic homeostasis, thereby triggering complex adaptive responses [2]. In ornamental species, such responses directly affect aesthetic quality and economic viability [3, 4].

Iris tectorum Maxim.(Iris) is a high ornamental species, prized for its diverse flower colors and distinctive architectural forms [5]. The cultivar ‘Hot Spicy’ exhibits notable heat tolerance under field conditions, making it an ideal candidate for studying thermotolerance mechanisms in ornamental plants.

Plants activate conserved defense mechanisms under heat stress, including antioxidant enzyme systems, osmolyte accumulation, and the heat shock response [6–8]. The heat shock response is orchestrated by heat shock transcription factors (HSFs), which are known to induce the expression of heat shock proteins (HSPs) and other protective genes, thereby forming a critical hub in heat-stress adaption [6], [9]. However, how HSF-HSP associated regulatory co-expression networks are linked to downstream metabolic reprogramming, particularly the biosynthesis of stress-related secondary metabolites, remains poorly characterized in non-model ornamental species.

Multi-omics approaches have begun to decipher heat stress responses in ornamentals such as Rhododendron and Clematis, revealing both conserved and species-specific adaptive strategies [10]. In these species, flavonoid biosynthesis and HSP-related pathways are markedly activated under high temperature, contributing to the alleviating oxidative damage [9, 11–13]. Despite these advances, the systemic mechanisms underlying Iris species remain largely unexplored. Current research on Iris thermotolerance have mainly focused on physiological evaluations, with limited molecular insight [14, 15]. Integrated multi-omics analyses that connect gene expression with physiological and metabolic phenotypes still lacking.

To address this gap, we conducted an integrated time-series analysis (0, 1, 3, 6, 9, and 24 h at 40 °C) on Iris ‘Hot Spicy’, combining physiological assays, transcriptome sequencing, and metabolomics. This study provides the first time-resolved multi-omics landscape of heat stress response in Iris, revealing a sequential temporal transition from early stress perception and heat shock responses at 1–3 h to sustained metabolic defense at 9–24 h. Our results indicate that HSF-HSP-associated genes exhibit temporally coordinated expression patterns and are closely associated with flavonoid accumulation, suggesting a putative HSF-HSP-flavonoid regulatory axis involved in thermotolerance. Overall, this integrated multi-omics study establishes a systems-level framework for understanding heat adaptation in Iris and provides candidate genetic targets for developing thermotolerant ornamental cultivars.

Material and methods

Plant materials and heat stress treatment

The plant material used in this study was Iris tectorum Maxim. cv. ‘Hot Spicy’. Two-month-old clonal ramets of Iris ‘Hot Spicy’ were obtained from a single mother plant originally from Xiaoshan Aquatic Plant Base, Hangzhou, China. Biological replicates were independent clonal ramets grown under identical conditions and treated separately. The plants were propagated by division in herbarium of Foshan Key Laboratory. A voucher specimen (No. Iris-HS-2025–001) has been deposited in the Foshan Key Laboratory of Agriculture and Biological Information. The plants were grown in a growth chamber under a 16/8 h light/dark cycle (25/20 °C, 70% relative humidity, 200 μmol m⁻² s⁻¹). After 30 days, uniform plants were subjected to 40 ± 0.5 °C for 0 (control), 1, 3, 6, 9, and 24 h. Leaf samples from 3 biological replicates per time point were collected, immediately frozen in liquid nitrogen, and stored at − 80 °C for subsequent analyses.

Physiological and biochemical assays

Physiological assays were performed with 3 technical replicates per sample following established protocols. MDA content was determined by the thiobarbituric acid reaction [16], Planta). Superoxide dismutase (SOD) and peroxidase (POD) activities were assayed using the nitroblue tetrazolium and guaiacol methods, respectively [17, 18]. Pro and soluble sugar (SS) contents were measured by the acid ninhydrin and anthrone-sulfuric acid methods, respectively [19, 20]. All chemical reagents were of analytical grade. SOD, POD, MDA, Pro, and SS assay kits were purchased from Keming Biotechnology Co., Ltd (Suzhou, China). All experiments (including the below qRT-PCR experiments) were performed with three independent biological replicates, each consisting of three technical replicates. For each biological replicate, the three technical replicates were averaged to obtain a single value prior to statistical analysis. Data are presented as mean ± standard deviation (SD) of the three biological replicates. One-way ANOVA followed by Tukey's HSD post-hoc test (p < 0.05) was performed using the biological replicate means.

Transcriptome sequencing and analysis

Total RNA was extracted using TRIzol reagent. Library construction and sequencing (PE150) were performed on the Illumina NovaSeq 6000 platform (Illumiana, LingEn Bio, Shanghai, China). Raw RNA-seq data have been deposited in the NCBI Sequence Read Archive under BioProject ID: PRJNA1449163. The SRA accession numbers are SRR12345678–SRR12345695. BioSample accession numbers are SRX32813288-SRX32813305.

Clean reads were obtained by removing adapter sequences and low-quality reads using Fastp v0.23.0 (default parameters). A de novo transcriptome assembly was generated by pooling all clean reads from all time points and assembling them with Trinity v2.8.5 using default parameters [21]. BUSCO completeness assessment results and basic QC metrics (including sequencing depth, Q30 proportion, mapping rate, and sample correlation/clustering) can be detailed in Appendix 8.

Clean reads from each sample were aligned to the de novo assembled transcriptome using Salmon v1.10.1. Trinity-assembled transcripts were clustered into gene-level groups using the trinity_gene_level_counts.pl script provided with Trinity, and raw counts were summed per gene. All subsequent analyses were performed at the gene level. Raw read counts per transcript were obtained using featureCounts v2.0.1. Differential expression analysis was performed with DESeq2 (|log₂FC|> 1, FDR < 0.05) based on raw counts. For visualization purposes, FPKM (Fragments Per Kilobase of transcript per Million mapped reads) values were calculated separately using a custom R script based on the raw counts and transcript lengths.

Functional enrichment of Differential expressed genes (DEGs) was performed using Gene Ontology (GO) and further Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses (clusterProfiler, FDR < 0.05).

For transcriptome sequencing, 3 biological replicates per time point were collected independently. RNA extraction, library preparation, and sequencing were performed separately for each replicate. No pooling was applied prior to analysis.

WGCNA was performed using the WGCNA R package (v1.71) following established protocol (Langfelder and Horvath., 2008). For WGCNA, 15,543 genes (the top 75% DEGs across all 18 samples) were used to construct a co-expression network. A soft-thresholding power of β = 12 was selected as the smallest value achieving a scale-free topology fit index R² > 0.85 [22]. Modules were identified by dynamic tree cutting (min module size = 30, deepSplit = 2). Module eigengenes were calculated, and Pearson correlation coefficients between module eigengenes and physiological traits were computed with associated p-values. Hub genes were defined as those with intramodular connectivity (kWithin) > 0.8 and gene significance (GS) > 0.7. Given the limited sample size (n = 18), these hub genes are considered exploratory candidates requiring independent validation.

Quantitative Real-Time PCR (qRT-PCR) Validation: To validate the RNA-seq expression profiles, 10 core DEGs were selected for qRT-PCR analysis. Primer sequences used for qRT-PCR are listed in Appendix 6. The thermal cycling program was: 95℃ for 30 s, followed by 40 cycles of 95℃ for 5 s and 60℃ for 30 s. RNA extraction used the TB Green® Premix EX Taq™ (Takara, Shiga, Japan), 3 independent biological replicates were analyzed for each sample, and each reaction was performed in 3 technical replicates. The Iris actin gene was used as the internal reference for normalization [15]. Relative expression levels were calculated using the 2^−ΔΔCt method. The expression values were presented as mean ± standard deviation (SD).

Metabolomics sequencing and analysis

Metabolites were extracted from 100 mg frozen leaf powder with 1 mL cold methanol/acetonitrile/water (2:2:1, v/v). After vortexing (30 s), sonication (15 min, 4℃), and incubation (1 h, −20 °C), the mixture was centrifuged (14,000 × g, 15 min, 4℃), and the supernatant was filtered (0.22 μm PTFE) for LC–MS analysis. Chromatographic separation was performed on a Waters ACQUITY HSS T3 column (2.1 × 100 mm, 1.8 μm) at 40 °C using a mobile phase system of water (A) and acetonitrile (B), both containing 0.1% formic acid, with a gradient elution: 0–1 min, 5% B; 1–12 min, 5–95% B; 12–14 min, 95% B; 14–14.1 min, 95–5% B; 14.1–16 min, 5% B, at a flow rate of 0.35 mL/min. The injection volume was 5 μL. The autosampler temperature was maintained at 4℃. Needle wash was performed with methanol/water (50:50, v/v) before each injection. Mass spectrometry was performed using a Q Exactive HFX mass spectrometer (Thermo Fisher Scientific) equipped with an electrospray ionization (ESI) source operating in both positive and negative ion modes. Full-scan range was m/z 70–1050 at a resolution of 70,000 (MS1) and 17,500 (MS2). MS/MS spectra were acquired in data-dependent acquisition (DDA) mode, with the top 10 most intense precursor ions selected for fragmentation using stepped normalized collision energies (20, 40, and 60 eV). Dynamic exclusion was set to 10 s.

Quality control (QC) samples, prepared by pooling equal volumes of all experimental samples, were injected every 10 analytical runs to monitor instrument stability. Features with coefficient of variation (CV) > 30% in QC samples were excluded. Features detected only in QC samples but not in biological samples were removed. All samples were analyzed in randomized order to minimize batch effects and the entire analysis was completed within a single batch.

Raw data were processed using Progenesis QI v2.4 for peak alignment, feature detection, and normalization using the total ion intensity method (abundance of each feature was divided by the total ion count of the respective sample). Missing values were imputed using the k-nearest neighbor method (k = 5) implemented in the R package VIM, applied within each time point group separately. Data were log₁₀ transformed prior to statistical analysis.

Metabolites were annotated by matching accurate m/z, retention time, and MS/MS spectra against the HMDB (v5.0), KEGG (v2022), and an in-house database, with confidence levels assigned according to Metabolomics Standards Initiative (MSI): Level 1 (authentic standard, where available), Level 2 (putatively annotated based on spectral similarity), and Level 3 (putatively characterized compound class). Key flavonoids, including naringenin chalcone and (-)-epigallocatechin, were annotated at MSI Level 2.

Differential accumulation metabolites (DAMs) were identified for each time point (1, 3, 6, 9, 24 h) compared to the control (0 h) using a two-tailed unpaired Student’s t-test. Raw p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure, with FDR < 0.05 and |log₂FC|> 1 considered significant. DAMs were particularly enriched in the phenylpropanoid and flavonoid biosynthesis pathways (map00941, map00944). For metabolomics, 3 biological replicates per time point were analyzed.

To elucidate the coordinated regulation of gene expression and metabolite accumulation, Pearson correlation coefficients were calculated between DEGs in the flavonoid pathway and their corresponding metabolites (|r|> 0.8, p < 0.05). Pathway-level integration was performed by mapping DEGs and differentially accumulated metabolites to KEGG pathways, with particular focus on the phenylpropanoid and flavonoid biosynthesis pathways. Co-enriched pathways were visualized using the clusterProfiler package.

Results

Physiological and biochemical responses of Iris to heat stress

Heat stress induced progressive morphological damage in Iris leaves (Fig. 1). Leaves began to show slight curling at 6 h, pronounced wilting and chlorosis at 9 h, and severe desiccation by 24 h. Heat stress induced progressive physiological changes in Iris leaves, with distinct temporal patterns across different stress indicators (Fig. 2). Membrane lipid peroxidation, measured as MDA content, increased continuously throughout the stress period, with a sharp rise observed at 9–24 h (p < 0.01), indicating cumulative oxidative damage to cellular membranes (Fig. 2A).

Fig. 1.

Phenotypic response of Iris under heat stress

Fig. 2.

Dynamic changes in physiological and biochemical parameters in Iris plants under heat stress. A MDA content, B SOD activity, C POD activity, D Pro content, E SS content. Note: Data are presented as mean ± SD (n = 3 biological replicates; technical replicates were averaged per biological replicate)

The antioxidant enzyme system exhibited a transient activation followed by decline. Both SOD and POD activities increased significantly during early to mid-stress (1–6 h), peaking at 6 h, suggesting an initial proactive defense against reactive oxygen species (ROS) accumulation (Fig. 2B-C). However, enzyme activities declined markedly under prolonged stress (9–24 h), coinciding with the period of drastic MDA accumulation, indicating potential collapse of the enzymatic antioxidant system under severe stress [23, 24].

In contrast, osmoregulatory substances showed divergent accumulation patterns. Pro content accumulated steadily throughout the stress period, increasing progressively from 1–24 h (Fig. 2D), suggesting active osmotic adjustment as a sustained adaptive strategy. SS content exhibited a more transient response, peaking sharply at 3 h before declining to control levels by 6 h and remaining stable thereafter (Fig. 2E). This pattern implies rapid mobilization of soluble sugars as compatible solutes and energy sources during the initial shock phase, followed by metabolic shift as stress prolonged.

Collectively, these physiological dynamics reveal a temporally phased stress response: early activation of enzymatic antioxidants and soluble sugar accumulation (1–6 h), followed by sustained Pro accumulation and progressive membrane damage (9–24 h). This time-resolved physiological framework provides the phenotypic basis for subsequent multi-omics analysis.

Global transcriptomic dynamics under heat stress

Heat stress triggered a time-dependent increase in DEGs. Compared to control (0 h), 15,222 DEGs were identified at 1 h, rising to 33,668 at 24 h (Table 1). Notably, the transition from 9–24 h involved 20,967 DEGs, suggesting a critical phase shift in the molecular response.

Table 1.

Statistical Table of DEGs Number

Group	All	Up	Down
1 h vs 0 h	15,222	8569	6653
3 hvs 0 h	22,668	12,344	10,324
6 hvs 0 h	21,901	11,326	10,575
9 hvs 0 h	23,660	11,776	11,884
24 hvs 0 h	33,668	22,473	11,195
3 h vs 1 h	1952	320	1632
6 h vs 3 h	3732	2753	979
9 h vs 6 h	7068	3890	3178
24 h vs 9 h	20,967	14,507	6460

GO analysis revealed that the most significantly enriched terms included catalytic activity and binding in the molecular function category; cellular process, metabolic process, and response to stimulus in the biological process category; and cellular anatomical entity, intracellular, and protein‑containing complex in the cellular component category (Fig. 3A). These results indicate that heat stress activates genes involved in fundamental cellular metabolism, stimulus perception, and protein homeostasis. KEGG pathway enrichment analysis identified multiple stress—responsive pathways (Fig. 3B). Among the significantly enriched pathways were phenylpropanoid biosynthesis, MAPK signaling pathway, plant—pathogen interaction, flavone and flavonol biosynthesis, and flavonoid biosynthesis.

Fig. 3.

Functional enrichment of DEGs. A GO enrichment, B KEGG enrichment. Note: The horizontal axis represents the enrichment factor, which is the ratio of the number of differentially expressed genes enriched in a specific GO/KEGG pathway to the number of background genes obtained from sequencing; the vertical axis represents the functions to which the KEGG pathway is enriched. A larger circle indicates a relatively higher number of differentially expressed genes enriched in that function. The color gradient from blue to red represents the adjusted p-value

Temporal analysis of GO and KEGG enrichment across the time course (Appendix 1) revealed a phased transcriptional response. In the early phase (1–3 h), terms related to response to stress and response to stimulus were significantly enriched, indicating rapid activation of stress perception and signaling. During the mid—to—late phase (6–24 h), the enrichment shifted toward metabolic reprogramming, with biosynthesis of secondary metabolites and flavonoid biosynthesis becoming prominently enriched. This transition tentatively suggests that Iris may initially prioritize stress sensing and signal transduction, followed by the activation of secondary metabolic pathways, particularly flavonoid biosynthesis, as part of a potential sustained adaptive response.

Heat shock proteins: a temporally coordinated chaperone co-epression pattern

A total of 104 differentially expressed HSP genes were identified, including 34 small HSPs (sHSPs), 40 HSP70, and 30 HSP90 members, whose expression exhibited a tightly regulated temporal pattern (Fig. 4A-C and detailed in Appendix 2). Within 1–3 h of stress onset, multiple sHSP genes were strongly upregulated, indicating an immediate protective response. For example, HSP26.7 (TRINITY_DN62394_c2_g6) increased by 17.21-fold at 1 h and 39.46-fold at 3 h, while HSP16.0 (TRINITY_DN68816_c1_g1), and HSP22.0 (TRINITY_DN68971_c0_g1) also showed marked upregulation exceeding 4.32-fold. Most HSP70 and HSP90 members exhibited only modest induction at this stage, except HSP90-1 (TRINITY_DN87102_c0_g1, 5.87-fold at 3 h) and an HSP70 (TRINITY_DN72233_c1_g2, 4.40-fold at 3 h). During the mid-phase (6–9 h), most HSPs reached their maximum expression levels. HSP26.7 peaked at 6 h with a 56.50-fold increase, HSP16.0 reached 11.86-fold at 9 h, and an HSP70 (TRINITY_DN38515_c0_g1) increased by 5.98-fold at 9 h compared to 0 h. Notably, while most sHSPs began to decline after peaking, HSP70 and HSP90 members generally maintained elevated expression. By 24 h, most sHSPs and HSP90 members declined, whereas a subset of HSP70 genes continued to rise, such as HSP70-16 (TRINITY_DN67914_c2_g2) and another HSP70 (TRINITY_DN72233_c0_g1) exhibited 7.29-fold and 7.77-fold increases, respectively, indicating a shift toward HSP70-mediated long-term protein homeostasis.

Fig. 4.

Heatmap of core genes under heat stress in Iris. A The expression of HSP70s; B HSP90s; and C sHSPs, D HSFs, and E ABA signaling pathways. Note:The expression data in the figure have all been normalized using Z-score

Heat stress factors: master regulators involved in the heat stress response

A total of 29 differentially expressed HSFs were identified, encompassing HSFA, HSFB, and HSFC families (Fig. 4D, detailed in Appendix 3). Among HSFA members, HSFA3 (TRINITY_DN66095_c1_g1) showed a progressive upregulation (2.54-fold at 24 h), HSFA4a (TRINITY_DN57479_c0_g1) peaked at 24 h (2.95-fold), and HSFA6b (TRINITY_DN16518_c0_g1) peaked at 6–9 h (up to 2.20-fold). Notably, HSFA5 (TRINITY_DN23603_c1_g2) exhibited a biphasic expression pattern: 4.46-fold at 1 h, 5.43-fold at 3 h, and 7.99-fold at 24 h, suggesting dual roles in signal initiation and late-stage reinforcement. In contrast, HSFA1a showed minor fluctuations, consistent with post-translational regulation. HSFB (specially HSFB1 and HSFB2c) showed varied responses, suggesting they may play auxiliary roles in fine-tuning the heat stress response.

Collectively, these temporal expression patterns indicate a hierarchical regulatory pattern: early-responsive HSFA5 initiates the temporal pattern, mid-phase HSFA6b coordinates with HSP accumulation, and late-phase HSFA3/HSFA4a may contribute to long-term adaptation.

Systematic activation of ABA signaling components

KEGG enrichment analysis of DEGs revealed that “plant hormone signal transduction” (map04075) was among the most significantly enriched pathways throughout the stress time course, with ABA signaling components exhibiting particularly pronounced and transcriptional changes (Fig. 4E, Appendix 4).

ABA receptors (Pyrabactin Resistance 1, PYLs), which act as negative regulators of ABA signaling, were induced under heat stress, with expression levels increasing over time. Multiple genes encoding ABA receptors of the PYL family were induced under heat stress, with expression levels increasing over time. PYL3 (TRINITY_DN58402_c0_g1) showed a sustained upward trend, reaching 1.73-fold at 3 h, 3.39-fold at 9 h, and 5.87-fold at 24 h. Another PYL3 member (TRINITY_DN70111_c2_g6) followed a similar pattern, peaking at 24 h with a 2.35-fold increase. PYL9 (TRINITY_DN30488_c0_g1) was induced only at 6 h, suggesting a transient role during mid-stress phase.

Protein phosphatases type 2 C (PP2C), which act as negative regulators of ABA signaling, showed complex expression patterns. PP2C (TRINITY_DN44551_c0_g3) displayed a biphasic response, with elevated expression at 6 h (1.12-fold) and a stronger peak at 24 h (1.27-fold). Another PP2C (TRINITY_DN83622_c0_g1) was moderately upregulated at 24 h (1.09-fold). The concurrent upregulation of both PYL receptors and certain PP2C members may reflect a finely tuned regulatory circuit that prevents overactivation of ABA signaling while maintaining responsiveness to persistent stress [25, 26].

Downstream transcription factors are activated at late stage. ABA-responsive element binding factors (ABFs), which mediate ABA-dependent gene expression, showed marked induction primarily at the late stress phase. ABF3 (TRINITY_DN33766_c0_g1) was progressively upregulated, increasing from 1 h (2.69-fold) to 3 h (4.30-fold), and remained elevated through 24 h (3.47-fold). In contrast, ABF2 (TRINITY_DN56838_c1_g1) showed a transient peak specifically at 6 h (2.59-fold), suggesting a distinct temporal role [27].

The late induction of ABF transcription factors temporally coincided with the upregulation of flavonoid biosynthesis genes (specially PAL, CHS, DFR) and the subsequent accumulation of flavonoid metabolites (Fig. 6). This temporal coupling suggests a potential regulatory link between ABA signaling and the activation of secondary metabolic pathways under heat stress. ABA has been implicated in the regulation of phenylpropanoid metabolism in other plant species, whether the systematic activation of the ABA pathway in Iris contributes to flavonoids accumulation remains a candidate hypothesis requiring further investigation.

Fig. 6.

Flavonoid pathway metabolic expression profile of Iris under heat stress. Note:The expression data in the figure have all been normalized using Z-score. F3H, flavanone 3-hydroxylase; DFR, dihydroflavonol 4-reductase; COMT, Caffeic acid 3-O-methyltransferase; C4H, Cinnamate 4-hydroxylase; 4CL, 4-coumarate-CoA ligase; Phe, 3-phenyl-L-alanine; HCT, Hydroxycinnamoyl Transferase; CYP 84 A, Cytochrome P450 84 A; F3′5’H, favanone 3’,5’-hydroxylase. Metabolite annotations are at MSI Level 2 (putatively annotated)

WGCNA identifies a core module linking gene expression to stress phenotypes

To systematically elucidate the regulatory relationships between gene expression and physiological phenotypes, WGCNA was performed. WGCNA constructed a gene co-expression correlation and identified 10 modules (Fig. 5A). Module-trait correlation analysis (Fig. 5A) revealed that the MEbrown4 module was positively correlated with SOD, POD, and MDA activities, and negatively correlated with Pro content.

Fig. 5.

WGCNA identifies a key module linked to physiological traits. A Module-trait relationships form 10 modules labeled with different colors, B GO enrichment of MEbrown4 module, C KEGG enrichment of MEbrown4 module. Note: The horizontal axis represents the enrichment factor, which is the ratio of the number of differentially expressed genes enriched in a specific GO/KEGG pathway to the number of background genes obtained from sequencing; the vertical axis represents the functions to which the KEGG pathway is enriched. A larger circle indicates a relatively higher number of differentially expressed genes enriched in that function. The color gradient from blue to red represents the adjusted p-value

Module‑trait correlation analysis (Fig. 5A) showed that the MEbrown4 module was significantly positively correlated with SOD activity (r = 0.68, p = 0.002), POD activity (r = 0.67, p = 0.003), MDA content (r = 0.61, p = 0.008), and SS content (r = 0.55, p = 0.02), and negatively correlated with Pro content (r = −0.57, p = 0.01). Complete correlation information is provided in Appendix 7

GO enrichment analysis of MEbrown4 genes (Fig. 5B) highlighted terms related to hormone signaling and regulation of RNA/protein metabolism (Specially “hormone‑mediated signaling pathway,” “regulation of organelle organization,” “regulation of cellular protein metabolic process,” “regulation of RNA metabolic process”). KEGG enrichment (Fig. 5C) revealed pathways including “Glycolysis/Gluconeogenesis” and “Arginine and proline metabolism” (p < 0.05). Hub genes within MEbrown4 were defined as those with intramodular connectivity (kWithin) > 0.8 and gene significance (GS) > 0.7 for MDA (see Methods). Among these, the top 20 genes (ranked by kWithin) included transcription factors (C3H, AP2/ERF, GARP), the energy sensor KIN10 (SnRK1), the chaperone organizer HOP2 (Hsp70‑Hsp90 organizing protein 2), the ascorbate‑glutathione cycle gene MDAR2 (monodehydroascorbate reductase 2), and the proline catabolism gene POX1. These are presented as highly connected candidate nodes within the co‑expression network, warranting future functional investigation.

Transcriptome and metabolome association analysis of flavonoid synthesis in late stress phase

Untargeted metabolomics identified 4,067 metabolites, with QC samples clustering tightly in PCA, confirming data reproducibility. Differential metabolite analysis revealed a progressive increase in DAMs. KEGG enrichment of DAMs showed that “flavone and flavonol biosynthesis” and “phenylpropanoid biosynthesis” were specifically enriched at late phase (9–24 h) (Appendix 5).

Flavonoid biosynthesis pathway genes, including PAL, CHS, CHI, F3H, and DFR, showed coordinated upregulation starting from mid-phase (6 h) and peaking at 9–24 h (Fig. 6, left panel). This temporal pattern closely followed the HSF-HSP activation (1–6 h), suggesting that HSF-HSP co-expression network may be temporally associated with flavonoid synthesis as part of the adaptive response [28, 29].

Focusing on the flavonoid pathway, key putatively annotated metabolites, including naringenin chalcone and (-)-epigallocatechin (MSI Level 2), as well as several quercetin derivatives, accumulated significantly from 6 h onward, with peak levels at 24 h (Fig. 6, right panel). The accumulation pattern mirrored the expression of flavonoid biosynthesis genes (Fig. 7), showing coordinated transcriptional and metabolic changes of this pathway under prolonged heat stress.

Fig. 7.

KEGG pathways co-enriched by DAMs and DEGs qRT-PCR validation

To verify the reliability of the transcriptome sequencing results of Iris leaves under heat stress, 10 heat-responsive DEGs were selected for qRT-PCR analysis (Fig. 8). These genes included HSFB2C, HSP70, HSP18.8, HSP90, PAL, CHS2, PYL3, PP2C, ABF2, and MDAR2 (primer design details are provided in Appendix 7). The expression trends of all 10 DEGs detected by qRT-PCR were consistent with the RNA-Seq data, confirming the reliability of the transcriptome sequencing results (Fig. 8).

Fig. 8.

The qRT-PCR validation result of 10 genes. Note: The values were presented as mean ± SD (n = 3 biological replicates; technical replicates were averaged per biological replicate)

Discussion

The HSF-HSP co-expression network exhibits a temporally phased chaperone strategy

Iris shows a temporally reprogrammed HSF-HSP co-expression network that transitions from rapid emergency response to sustained protein protection. In the early phase (1–3 h), small HSPs (sHSPs) are rapidly induced, acting as "first responders" that bind to partially denatured proteins to prevent irreversible aggregation. This rapid sHSP response is consistent with observations in lily (Lilium longiflorum), where multiple sHSPs are induced within 0.5–3 h of heat stress [30]. During the mid-phase (6–9 h), HSP70 and HSP90 families are broadly activated, dominating protein folding and repair. In Clematis florida, HSP70 members are prominently upregulated under heat stress, similar to the sustained HSP70 expression observed in Iris [9, 31]. By the late phase (24 h), HSP70 continues to rise while sHSPs decline, indicating a shift toward HSP70-mediated long-term protein homeostasis. A notable finding in Iris is the sustained elevation of HSP70 at 24 h, suggesting a unique role for HSP70 in protein repair and aggregation prevention under prolonged stress, a feature that may contribute to the thermotolerance of the ‘Hot Spicy’ cultivar [32].

The transcriptional temporal pattern is reflected in the functional differentiation of HSF family members.. HSFA5 exhibits a biphasic expression pattern with peaks at both early and late phases, suggesting dual roles in signal initiation and late-stage reinforcement. HSFA6b peaks at mid-phase, coinciding with the peak expression of HSPs and antioxidant enzyme activities. HSFA3 and HSFA4a show sustained upregulation at the late phase, maintaining long-term adaptive responses. HSFA2b maintains constitutively high expression throughout stress, potentially serving as a signal integrator. This functional specialization is consistent with studies in Arabidopsis and rose (Rosa rugosa Thunb.) [33, 34], where HSFA1s act as master regulators and HSFA2s as amplifiers of the heat shock response. In rose, RcHsfA6 has been shown to enhance thermotolerance by modulating ABA signaling and HSP expression [11, 12, 35]. Collectively, the temporal expression patterns of HSFs in Iris suggest a layered temporal expression pattern, but direct regulatory relationships remain to be experimentally validated.

ABA signaling associated with the HSF-HSP correlation to activate downstream metabolism

The late-stage induction of ABF transcription factors temporally coincides with upregulation of flavonoid biosynthesis genes (PAL, CHS, DFR), suggesting a potential link. In this study, ABF transcription factors were significantly upregulated at the late stress phase, and their activation timing closely coincided with the upregulation of flavonoid biosynthesis genes and the accumulation of flavonoid metabolites. This temporal coupling suggests that ABFs could potentially serve as a bridge between ABA signaling and the HSF-HSP co-expression pattern, mediating the regulation of secondary metabolism. Consistent with findings in rose (Rose Chinese), where the heat shock factor RcHsfA6 modulates ABA signaling by regulating ABA biosynthesis genes (AtABI5, AtPYL12) and enhancing HSP expression, the synchronous activation of ABFs and flavonoid genes in Iris implies that ABA signaling may be involved in HSF-HSP-mediated regulation of secondary metabolism [34–36]. However, direct binding of HSFs to ABFs or cooperative regulation of shared target genes by HSFs and ABFs has not been tested in Iris. Therefore, the proposed “HSF-HSP-ABA-flavonoid” putative regulatory axis remains a candidate hypothesis requiring experimental validation, such as through ChIP-qPCR, dual-luciferase assays, or exogenous ABA application coupled with gene expression analysis [30, 37, 38].

Flavonoid biosynthesis shows coordinated activation as a candidate downstream metabolic output

The flavonoid pathway shows coordinated activation during the late stress phase, potentially establishing a sustained chemical defense. Flavonoid biosynthesis genes (PAL, CHS, DFR) were significantly upregulated from 6 to 24 h, following a clear temporal sequence after HSF-HSP co-expressiom pattern (1–6 h). At the metabolite level, key putatively identified flavonoids including naringenin chalcone and (-)-epigallocatechin accumulated during the same period, showing high consistency between transcriptional and metabolic activation. These findings align with studies in other ornamentals. In Rhododendron, high temperature markedly activates flavonoid biosynthesis, with quercetin accumulation contributing to oxidative damage mitigation [39, 40]. In chrysanthemum, heat stress significantly perturbs flavonoid metabolism, and altered flavonoid profiles are associated with flower color changes under high temperature [41]. In rose (Rosa hybrida), integrated transcriptomic and metabolomic analyses revealed that heat stress significantly affects flavonoid biosynthesis pathways [40, 42]. The novel contribution of this study lies in revealing the temporal coupling between flavonoid biosynthesis and HSF-HSP co-expressed activation [43]. As antioxidants, flavonoids scavenge accumulated ROS and stabilize membrane structures during late stress, which may contribute to oxidative stress mitigation in Iris heat adaptation.

WGCNA identifies a core module associated with integration of multiple adaptive pathways

WGCNA identified MEbrown4 as a core module significantly positively correlated with physiological indicators (SOD, POD, MDA), and negatively correlated with Pro content. GO and KEGG enrichment patterns provide correlative evidence that MEbrown4 may integrate energy metabolism, amino acid metabolism, and hormone signaling under heat stress. Hub genes within this module include the energy sensor KIN10 (SnRK1), the chaperone organizer HOP2 [32], the antioxidant gene MDAR2, and several transcription factors [44].

KIN10, the catalytic subunit of SnRK1, functions as a central energy sensor and stress regulator, coordinating metabolic reprogramming under adverse conditions, especially energy stress [45, 46]. Importantly, recent work has shown that AtHOPs are involved in the folding and stabilization of HSFA1a and are required for the onset of the transcriptional response associated with thermomorphogenesis [47]. The same study demonstrated that HSFA1a accumulates at lower levels in the hop1 hop2 hop3 mutant, uncovering HSFA1a as a client of HOP complexes in plants and revealing the participation of HOPs in HSFA1a stability [47]. This is particularly relevant as HOP2 in Iris is a candidate hub gene, and its potential role in HSF stability warrants further investigation [32]. MDAR2 is involved in the ascorbate–glutathione cycle, directly participating in ROS scavenging [48, 49]. These hub genes are presented as highly connected candidate nodes within the co-expression network, warranting future functional investigation rather than being interpreted as validated integrators. Similar candidate discovery approaches have been successfully employed in other plant species, where hub genes identified through WGCNA served as starting points for functional validation via transgenic approaches [50]. It should be noted that WGCNA based on 18 samples is limited in statistical power, and the proposed hub genes (KIN10, HOP2, MDAR2) are correlative candidates rather than validated regulators.

Breeding implications and candidate genes for further study

This study identified a suite of candidate genes temporally associated with heat adaptation in Iris ‘Hot Spicy’, including early‑responsive HSFA5, late‑sustained HSFA3 and HSFA4a, persistently elevated HSP70‑16, ABA‑associated ABF3, and WGCNA hub genes (KIN10, HOP2, MDAR2). However, direct regulation of flavonoid biosynthesis genes by these HSFs or ABFs remains to be experimentally validated. Moreover, the study was conducted on a single cultivar under continuous heat stress without a recovery phase, and functional validation (such as promoter binding assays, VIGS/overexpression, exogenous ABA treatment) is required before these candidates can be considered validated genes for further research.

Conclusion

This study provides a time-resolved multi-omics landscape of heat stress response in Iris tectorum 'Hot Spicy' (Fig. 9). Heat stress is associated with a temporally phased HSF-HSP expression pattern: shifting from rapid induction of small HSP genes to sustained HSP70-mediated protection. The hierarchiacl activation of HSF genes, including early HSFA5, mid-phase HSFA6b, and late-phase HSFA3/HSFA4a, temporally coincides with late-stage induction of ABF transcription factors genes and the accumulation of flavonoids, including naringenin chalcone and (-)-epigallocatechin. WGCNA identifies MEbrown4 as a core module integrating hormone signaling and phenylpropanoid biosynthesis, with KIN10, HOP2, and MDAR2 emerging as candidate hub genes. These findings suggest a phased temporal sequence from early HSF-HSP activation to sustained flavonoid accumulation and provide candidate genes for future functional characterization in Iris.

Fig. 9.

Proposed integrative regulatory model in Iris in response to heat stress

Abbreviations

SS

Soluble sugar

Phe

3-Phenyl-L-alanine

4CL

4-Coumarate-CoA ligase

ABF

ABA-responsive element binding factors

ABA

Abscisic acid

COMT

Caffeic acid 3-O-methyltransferase

CHS

Chalcone Synthase

C4H

Cinnamate 4-hydroxylase

CYP84A

Cytochrome P450 84A

DAMs

Differential accumulation of metabolites

DEGs

Differentially expressed genes

DFR

Dihydroflavonol 4-Reductase

F3H

Flavanone 3-hydroxylase

GO

Gene Ontology

HSP

Heat Shock Proteins

WGCNA

Weighted Gene Co-expression Network Analysis

HSF

Heat stress factors

HCT

Hydroxycinnamoyl Transferase

Iris

Iris tectorum

KEGG

Kyoto Encyclopedia of Genes and Genomes

F3′5’H

Favanone 3’,5’-hydroxylase

MDA

Malondialdehyde

POD

Peroxidase

PAL

Phenylalanine Ammonia-Lyase

Pro

Proline

PP2C

Protein phosphatases type 2C

PYL

Pyrabactin Resistance 1

ROS

Reactive oxygen species

sHSPs

Small heat shock proteins

SnRK2

Sucrose Non-Fermenting 1 (SNF1)-Related Protein Kinase 2

SOD

Superoxide dismutase

Authors’ contributions

Jiayu Hu reports financial support was provided by Guangdong Basic and Applied Basic Research Foundation. Jiayu Hu reports a relationship with Guangdong Basic and Applied Basic Research Foundation that includes: funding grants. Jiayu Hu has patent pending to Jiayu Hu. Each member is aware of and adheres to the submission guidelines. All members participated and contributed to the research process. Jiayu Hu was responsible for the overall planning and ideas of the research, as well as providing funding support; Youli Li was responsible for the progress of research experiments and data analysis; Yang Lin, Nuoya Wu, and Jiaxu Xie participated in experiments and data work. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This work was supported by the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2022A1515110779).

Data availability

Raw RNA-seq data have been deposited in the NCBI Sequence Read Archive under BioProject ID: PRJNA1449163, the data can be accessed at https://www.ncbi.nlm.nih.gov/sra/PRJNA1449163. The BioSample accession numbers are SRX32813288–SRX32813305.

Declarations

Ethics approval and consent to participate

All authors have read and approved the final manuscript and confirm that the research described is original, accurate, and has not been published elsewhere.

Consent for publication

All authors agree to submit the manuscript to BMC Plant Biology.

Competing interests

The authors declare no competing interests.